diff --git a/.nojekyll b/.nojekyll new file mode 100644 index 0000000..e69de29 diff --git a/404.html b/404.html new file mode 100644 index 0000000..0de0349 --- /dev/null +++ b/404.html @@ -0,0 +1,728 @@ + + + + + + + + + + + + + + + + + + + Speckcn2 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
+ +
+
+ +
+ + + + + + +
+ + +
+ +
+ + + + + + +
+
+ + + +
+
+
+ + + + + +
+
+
+ + + +
+
+
+ + + +
+
+
+ + + +
+
+ +

404 - Not found

+ +
+
+ + + +
+ +
+ + + +
+
+
+
+ + + + + + + + + + + + + + \ No newline at end of file diff --git a/CODE_OF_CONDUCT/index.html b/CODE_OF_CONDUCT/index.html new file mode 100644 index 0000000..f976724 --- /dev/null +++ b/CODE_OF_CONDUCT/index.html @@ -0,0 +1,1110 @@ + + + + + + + + + + + + + + + + + + + + + + + Code of Conduct - Speckcn2 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
+ + + + Skip to content + + +
+
+ +
+ + + + + + +
+ + +
+ +
+ + + + + + +
+
+ + + +
+
+
+ + + + + +
+
+
+ + + +
+
+
+ + + +
+
+
+ + + +
+
+ + + + + + + +

Contributor Covenant Code of Conduct

+

Our Pledge

+

We as members, contributors, and leaders pledge to make participation in our +community a harassment-free experience for everyone, regardless of age, body +size, visible or invisible disability, ethnicity, sex characteristics, gender +identity and expression, level of experience, education, socio-economic status, +nationality, personal appearance, race, religion, or sexual identity +and orientation.

+

We pledge to act and interact in ways that contribute to an open, welcoming, +diverse, inclusive, and healthy community.

+

Our Standards

+

Examples of behavior that contributes to a positive environment for our +community include:

+
    +
  • Demonstrating empathy and kindness toward other people
    • +
  • Being respectful of differing opinions, viewpoints, and experiences
    • +
  • Giving and gracefully accepting constructive feedback
    • +
  • Accepting responsibility and apologizing to those affected by our mistakes, + and learning from the experience
    • +
  • Focusing on what is best not just for us as individuals, but for the + overall community
    • +
+

Examples of unacceptable behavior include:

+
    +
  • The use of sexualized language or imagery, and sexual attention or + advances of any kind
    • +
  • Trolling, insulting or derogatory comments, and personal or political attacks
    • +
  • Public or private harassment
    • +
  • Publishing others' private information, such as a physical or email + address, without their explicit permission
    • +
  • Other conduct which could reasonably be considered inappropriate in a + professional setting
    • +
+

Enforcement Responsibilities

+

Community leaders are responsible for clarifying and enforcing our standards of +acceptable behavior and will take appropriate and fair corrective action in +response to any behavior that they deem inappropriate, threatening, offensive, +or harmful.

+

Community leaders have the right and responsibility to remove, edit, or reject +comments, commits, code, wiki edits, issues, and other contributions that are +not aligned to this Code of Conduct, and will communicate reasons for moderation +decisions when appropriate.

+

Scope

+

This Code of Conduct applies within all community spaces, and also applies when +an individual is officially representing the community in public spaces. +Examples of representing our community include using an official e-mail address, +posting via an official social media account, or acting as an appointed +representative at an online or offline event.

+

Enforcement

+

Instances of abusive, harassing, or otherwise unacceptable behavior may be +reported to the community leaders responsible for enforcement at +s.ciarella@esciencecenter.nl. +All complaints will be reviewed and investigated promptly and fairly.

+

All community leaders are obligated to respect the privacy and security of the +reporter of any incident.

+

Enforcement Guidelines

+

Community leaders will follow these Community Impact Guidelines in determining +the consequences for any action they deem in violation of this Code of Conduct:

+

1. Correction

+

Community Impact: Use of inappropriate language or other behavior deemed +unprofessional or unwelcome in the community.

+

Consequence: A private, written warning from community leaders, providing +clarity around the nature of the violation and an explanation of why the +behavior was inappropriate. A public apology may be requested.

+

2. Warning

+

Community Impact: A violation through a single incident or series +of actions.

+

Consequence: A warning with consequences for continued behavior. No +interaction with the people involved, including unsolicited interaction with +those enforcing the Code of Conduct, for a specified period of time. This +includes avoiding interactions in community spaces as well as external channels +like social media. Violating these terms may lead to a temporary or +permanent ban.

+

3. Temporary Ban

+

Community Impact: A serious violation of community standards, including +sustained inappropriate behavior.

+

Consequence: A temporary ban from any sort of interaction or public +communication with the community for a specified period of time. No public or +private interaction with the people involved, including unsolicited interaction +with those enforcing the Code of Conduct, is allowed during this period. +Violating these terms may lead to a permanent ban.

+

4. Permanent Ban

+

Community Impact: Demonstrating a pattern of violation of community +standards, including sustained inappropriate behavior, harassment of an +individual, or aggression toward or disparagement of classes of individuals.

+

Consequence: A permanent ban from any sort of public interaction within +the community.

+

Attribution

+

This Code of Conduct is adapted from the Contributor Covenant, +version 2.0, available at +https://www.contributor-covenant.org/version/2/0/code_of_conduct.html.

+

Community Impact Guidelines were inspired by Mozilla's code of conduct +enforcement ladder.

+

For answers to common questions about this code of conduct, see the FAQ at +https://www.contributor-covenant.org/faq. Translations are available at +https://www.contributor-covenant.org/translations.

+ + + + + + + + + + + + + +
+
+ + + +
+ +
+ + + +
+
+
+
+ + + + + + + + + + + + + + \ No newline at end of file diff --git a/CONTRIBUTING/index.html b/CONTRIBUTING/index.html new file mode 100644 index 0000000..641d606 --- /dev/null +++ b/CONTRIBUTING/index.html @@ -0,0 +1,1023 @@ + + + + + + + + + + + + + + + + + + + + + + + + + Contributing - Speckcn2 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
+ + + + Skip to content + + +
+
+ +
+ + + + + + +
+ + +
+ +
+ + + + + + +
+
+ + + +
+
+
+ + + + + +
+
+
+ + + +
+
+
+ + + +
+
+
+ + + +
+
+ + + + + + + +

Contributing guidelines

+

Welcome! SpeckCn2 is an open-source project for the analysis of speckle patterns. If you're trying SpeckCn2 with your data, your experience, questions, bugs you encountered, and suggestions for improvement are important to the success of the project.

+

We have a Code of Conduct, please follow it in all your interactions with the project.

+

Questions, feedback, bugs

+

Use the search function to see if someone else already ran across the same issue. Feel free to open a new issue here to ask a question, suggest improvements/new features, or report any bugs that you ran into.

+

Submitting changes

+

Even better than a good bug report is a fix for the bug or the implementation of a new feature. We welcome any contributions that help improve the code.

+

When contributing to this repository, please first discuss the change you wish to make via an issue with the owners of this repository before making a change.

+

Contributions can come in the form of:

+
    +
  • Bug fixes
    • +
  • New features
    • +
  • Improvement of existing code
    • +
  • Updates to the documentation
    • +
  • ... ?
    • +
+

We use the usual GitHub pull request flow. For more info see GitHub's own documentation.

+

Typically this means:

+
    +
  1. Forking the repository and/or make a new branch
    2. +
  2. Making your changes
    3. +
  3. Make sure that the tests pass and add your own
    4. +
  4. Update the documentation is updated for new features
    5. +
  5. Pushing the code back to Github
    6. +
  6. Create a new Pull Request
    7. +
+

One of the code owners will review your code and request changes if needed. Once your changes have been approved, your contributions will become part of SpeckCn2. 🎉

+

Getting started with development

+

Setup

+

SpeckCn2 targets Python 3.9 or newer.

+

Clone the repository into the speckcn2 directory:

+
git clone https://github.com/MALES-project/SpeckleCn2Profiler speckcn2
+cd speckcn2
+
+

Initialize all submodules: +

git submodule update --recursive --init
+

+

Install using virtualenv:

+
python3 -m venv env
+source env/bin/activate
+python3 -m pip install -e .[develop]
+
+

Alternatively, install using Conda:

+
conda create -n speckcn2 python=3.10
+conda activate speckcn2
+pip install -e .[develop]
+
+

Running tests

+

SpeckCn2 uses pytest to run the tests. You can run the tests for yourself using:

+
pytest
+
+

To check coverage:

+
coverage run -m pytest
+coverage report  # to output to terminal
+coverage html    # to generate html report
+
+

Building the documentation

+

The documentation is written in markdown, and uses mkdocs to generate the pages.

+

To build the documentation for yourself:

+
pip install -e .[docs]
+mkdocs serve
+
+

You can find the documentation source in the docs directory. +If you are adding new pages, make sure to update the listing in the mkdocs.yml under the nav entry.

+

Making a release

+
    +
  1. +

    Make a new release.

    +
    2. +
  2. +

    Under 'Choose a tag', set the tag to the new version. The versioning scheme we use is SemVer, so bump the version (major/minor/patch) as needed. Bumping the version is handled transparently by bumpversion in this workflow.

    +
    3. +
  3. +

    The upload to pypi is triggered when a release is published and handled by this workflow.

    +
    4. +
  4. +

    The upload to zenodo is triggered when a release is published.

    +
    5. +
+ + + + + + + + + + + + + +
+
+ + + +
+ +
+ + + +
+
+
+
+ + + + + + + + + + + + + + \ No newline at end of file diff --git a/LICENSE b/LICENSE new file mode 100644 index 0000000..cba65d5 --- /dev/null +++ b/LICENSE @@ -0,0 +1 @@ +{!../LICENSE!} diff --git a/api/api/index.html b/api/api/index.html new file mode 100644 index 0000000..6b0c081 --- /dev/null +++ b/api/api/index.html @@ -0,0 +1,767 @@ + + + + + + + + + + + + + + + + + + + + + + + + + speckcn2 - Speckcn2 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
+ +
+
+ +
+ + + + + + +
+ + +
+ +
+ + + + + + +
+
+ + + +
+
+
+ + + + + +
+
+
+ + + +
+
+
+ + + +
+
+
+ + + +
+ +
+ + + +
+ +
+ + + +
+
+
+
+ + + + + + + + + + + + + + \ No newline at end of file diff --git a/api/io/index.html b/api/io/index.html new file mode 100644 index 0000000..cffbd04 --- /dev/null +++ b/api/io/index.html @@ -0,0 +1,1364 @@ + + + + + + + + + + + + + + + + + + + + + + + + + I/O - Speckcn2 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
+ + + + Skip to content + + +
+
+ +
+ + + + + + +
+ + +
+ +
+ + + + + + +
+
+ + + +
+
+
+ + + + + +
+
+
+ + + +
+
+
+ + + +
+
+
+ + + +
+
+ + + + + + + +

I/O

+ +
+ + + + +
+ +

This module provides utility functions for loading and saving model +configurations and states.

+

It includes functions to load configuration files, save model states, +load model states, and load the latest model state from a directory.

+ + + + + + + + +
+ + + + + + + + + +
+ + +

+ load(model, datadirectory, epoch) + +

+ + +
+ +

Load the model state and the model itself from a specified directory and +epoch.

+

This function loads the model's state dictionary and other relevant information +such as epoch, loss, validation loss, and time from a file in the specified directory.

+ + +

Parameters:

+
    +
  • + model + (Module) + – +
    +

    The model to load

    +
    +
    • +
  • + datadirectory + (str) + – +
    +

    The directory where the data is stored

    +
    +
    • +
  • + epoch + (int) + – +
    +

    The epoch of the model

    +
    +
    • +
+ +
+ Source code in src/speckcn2/io.py + 
65
+66
+67
+68
+69
+70
+71
+72
+73
+74
+75
+76
+77
+78
+79
+80
+81
+82
+83
+84
+85
+86
+87
+88
+89
+90
+91
def load(model: torch.nn.Module, datadirectory: str, epoch: int) -> None:
+    """Load the model state and the model itself from a specified directory and
+    epoch.
+
+    This function loads the model's state dictionary and other relevant information
+    such as epoch, loss, validation loss, and time from a file in the specified directory.
+
+    Parameters
+    ----------
+    model : torch.nn.Module
+        The model to load
+    datadirectory : str
+        The directory where the data is stored
+    epoch : int
+        The epoch of the model
+    """
+    model_state = torch.load(
+        f'{datadirectory}/{model.name}_states/{model.name}_{epoch}.pth')
+
+    model.epoch = model_state['epoch']
+    model.loss = model_state['loss']
+    model.val_loss = model_state['val_loss']
+    model.time = model_state['time']
+    model.load_state_dict(model_state['model_state_dict'], strict=False)
+
+    assert model.epoch[
+        -1] == epoch, 'The epoch of the model is not the same as the one loaded'
+
+
+
+ +
+ +
+ + +

+ load_config(config_file_path) + +

+ + +
+ +

Load the configuration file from a given path.

+

This function reads a YAML configuration file and returns its contents as a dictionary.

+ + +

Parameters:

+
    +
  • + config_file_path + (str) + – +
    +

    Path to the .yaml configuration file

    +
    +
    • +
+ + +

Returns:

+
    +
  • +config ( dict +) – +
    +

    Dictionary containing the configuration

    +
    +
    • +
+ +
+ Source code in src/speckcn2/io.py + 
18
+19
+20
+21
+22
+23
+24
+25
+26
+27
+28
+29
+30
+31
+32
+33
+34
+35
def load_config(config_file_path: str) -> dict:
+    """Load the configuration file from a given path.
+
+    This function reads a YAML configuration file and returns its contents as a dictionary.
+
+    Parameters
+    ----------
+    config_file_path : str
+        Path to the .yaml configuration file
+
+    Returns
+    -------
+    config : dict
+        Dictionary containing the configuration
+    """
+    with open(config_file_path, 'r') as file:
+        config = yaml.safe_load(file)
+    return config
+
+
+
+ +
+ +
+ + +

+ load_model_state(model, datadirectory) + +

+ + +
+ +

Loads the latest model state from the given directory.

+

This function checks the specified directory for the latest model state file, +loads it, and updates the model with the loaded state. If no state is found, +it initializes the model state.

+ + +

Parameters:

+
    +
  • + model + (Module) + – +
    +

    The model to load the state into

    +
    +
    • +
  • + datadirectory + (str) + – +
    +

    The directory where the model states are stored

    +
    +
    • +
+ + +

Returns:

+
    +
  • +model ( Module +) – +
    +

    The model with the loaded state

    +
    +
    • +
  • +last_model_state ( int +) – +
    +

    The number of the last model state

    +
    +
    • +
+ +
+ Source code in src/speckcn2/io.py + 
 94
+ 95
+ 96
+ 97
+ 98
+ 99
+100
+101
+102
+103
+104
+105
+106
+107
+108
+109
+110
+111
+112
+113
+114
+115
+116
+117
+118
+119
+120
+121
+122
+123
+124
+125
+126
+127
+128
+129
+130
+131
+132
+133
+134
+135
+136
+137
+138
+139
+140
+141
+142
+143
+144
+145
+146
+147
def load_model_state(model: torch.nn.Module,
+                     datadirectory: str) -> tuple[torch.nn.Module, int]:
+    """Loads the latest model state from the given directory.
+
+    This function checks the specified directory for the latest model state file,
+    loads it, and updates the model with the loaded state. If no state is found,
+    it initializes the model state.
+
+    Parameters
+    ----------
+    model : torch.nn.Module
+        The model to load the state into
+    datadirectory : str
+        The directory where the model states are stored
+
+    Returns
+    -------
+    model : torch.nn.Module
+        The model with the loaded state
+    last_model_state : int
+        The number of the last model state
+    """
+    # Print model information
+    print(model)
+    model.nparams = sum(p.numel() for p in model.parameters())
+    print(f'\n--> Nparams = {model.nparams}')
+
+    ensure_directory(f'{datadirectory}/{model.name}_states')
+
+    # Check what is the last model state
+    try:
+        last_model_state = sorted([
+            int(file_name.split('.pth')[0].split('_')[-1])
+            for file_name in os.listdir(f'{datadirectory}/{model.name}_states')
+        ])[-1]
+    except Exception as e:
+        print(f'Warning: {e}')
+        last_model_state = 0
+
+    if last_model_state > 0:
+        print(
+            f'Loading model at epoch {last_model_state}, from {datadirectory}')
+        load(model, datadirectory, last_model_state)
+        return model, last_model_state
+    else:
+        print('No pretrained model to load')
+
+        # Initialize some model state measures
+        model.loss = []
+        model.val_loss = []
+        model.time = []
+        model.epoch = []
+
+        return model, 0
+
+
+
+ +
+ +
+ + +

+ save(model, datadirectory) + +

+ + +
+ +

Save the model state and the model itself to a specified directory.

+

This function saves the model's state dictionary and other relevant information +such as epoch, loss, validation loss, and time to a file in the specified directory.

+ + +

Parameters:

+
    +
  • + model + (Module) + – +
    +

    The model to save

    +
    +
    • +
  • + datadirectory + (str) + – +
    +

    The directory where the data is stored

    +
    +
    • +
+ +
+ Source code in src/speckcn2/io.py + 
38
+39
+40
+41
+42
+43
+44
+45
+46
+47
+48
+49
+50
+51
+52
+53
+54
+55
+56
+57
+58
+59
+60
+61
+62
def save(model: torch.nn.Module, datadirectory: str) -> None:
+    """Save the model state and the model itself to a specified directory.
+
+    This function saves the model's state dictionary and other relevant information
+    such as epoch, loss, validation loss, and time to a file in the specified directory.
+
+    Parameters
+    ----------
+    model : torch.nn.Module
+        The model to save
+    datadirectory : str
+        The directory where the data is stored
+    """
+    model_state = {
+        'epoch': model.epoch,
+        'loss': model.loss,
+        'val_loss': model.val_loss,
+        'time': model.time,
+        'model_state_dict': model.state_dict(),
+    }
+
+    torch.save(
+        model_state,
+        f'{datadirectory}/{model.name}_states/{model.name}_{model.epoch[-1]}.pth'
+    )
+
+
+
+ +
+ + + +
+ +
+ +
+ + + + + + + + + + + + + +
+
+ + + +
+ +
+ + + +
+
+
+
+ + + + + + + + + + + + + + \ No newline at end of file diff --git a/api/loss/index.html b/api/loss/index.html new file mode 100644 index 0000000..0dfefb9 --- /dev/null +++ b/api/loss/index.html @@ -0,0 +1,1614 @@ + + + + + + + + + + + + + + + + + + + + + + + + + Loss - Speckcn2 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
+ + + + Skip to content + + +
+
+ +
+ + + + + + +
+ + +
+ +
+ + + + + + +
+
+ + + +
+
+
+ + + + + +
+
+
+ + + +
+
+
+ + + +
+
+
+ + + +
+
+ + + + + + + +

Loss

+ +
+ + + + +
+ +

This module implements various loss functions for training machine learning +models.

+

It includes custom loss functions that extend PyTorch's nn.Module, +allowing for flexible and efficient computation of loss values during +training. The loss functions handle different scenarios such as +classification, regression, and segmentation tasks. They incorporate +techniques like weighted losses, focal losses, and smooth L1 losses to +address class imbalances and improve model performance. The module +ensures that the loss calculations are compatible with PyTorch's +autograd system, enabling seamless integration into training loops.

+ + + + + + + + +
+ + + + + + + + +
+ + + +

+ ComposableLoss(config, nz, device) + +

+ + +
+

+ Bases: Module

+ + +

Compose the loss function using several terms. The importance of each +term has to be specified in the configuration file. Each term with a >0 +weight will be added to the loss function.

+

The loss term available are: +- MSE: mean squared error between predicted and target normalized screen tags +- MAE: mean absolute error between predicted and target normalized screen tags +- JMSE: mean squared error between predicted and target J +- JMAE: mean absolute error between predicted and target J +- Pearson: Pearson correlation coefficient between predicted and target J +- Fried: Fried parameter r0 +- Isoplanatic: Isoplanatic angle theta0 +- Rytov: Rytov variance sigma_r^2 that will be computed on log averaged Cn2 +- Scintillation_w: scintillation index for weak turbulence +- Scintillation_m: scintillation index for moderate-strong turbulence

+ + +

Parameters:

+
    +
  • + config + (dict) + – +
    +

    Dictionary containing the configuration

    +
    +
    • +
  • + nz + (Normalizer) + – +
    +

    Normalizer object to be used to extract J in its original scale

    +
    +
    • +
+ + + + + + +
+ Source code in src/speckcn2/loss.py + 
49
+50
+51
+52
+53
+54
+55
+56
+57
+58
+59
+60
+61
+62
+63
+64
+65
+66
+67
+68
+69
+70
+71
+72
+73
+74
+75
+76
+77
+78
+79
+80
+81
+82
+83
+84
+85
+86
+87
def __init__(self, config: dict, nz: Normalizer, device: torch.device):
+    super(ComposableLoss, self).__init__()
+    self.device = device
+    self.loss_functions: dict[str, Callable] = {
+        'MSE': torch.nn.MSELoss(),
+        'MAE': torch.nn.L1Loss(),
+        'JMSE': self._MSELoss,
+        'JMAE': self._L1Loss,
+        'Cn2MSE': self._do_nothing,
+        'Cn2MAE': self._do_nothing,
+        'Pearson': self._PearsonCorrelationLoss,
+        'Fried': self._FriedLoss,
+        'Isoplanatic': self._IsoplanaticLoss,
+        'Rytov': self._RytovLoss,
+        'Scintillation_w': self._ScintillationWeakLoss,
+        'Scintillation_ms': self._ScintillationModerateStrongLoss,
+    }
+    self.loss_weights = {
+        loss_name: config['loss'].get(loss_name, 0)
+        for loss_name in self.loss_functions.keys()
+    }
+    self.total_weight = sum(self.loss_weights.values())
+    self._select_loss_needed()
+
+    # And get some useful parameters for the loss functions
+    # the parameters are explained in ...
+    self.h = torch.Tensor([float(x) for x in config['speckle']['hArray']])
+    self.k = 2 * torch.pi / (config['speckle'].get('lambda', 550) * 1e-9)
+    self.cosz = np.cos(np.deg2rad(config['speckle'].get('z', 0)))
+    self.secz = 1 / self.cosz
+    self.L = config['speckle']['L']
+    self.p_fr = 0.423 * self.k**2 * self.secz
+    self.p_iso = self.cosz**(8. / 5.) / ((2.91 * self.k**2)**(3. / 5.))
+    self.p_scw = 2.25 * self.k**(7. / 6.) * self.secz**(11. / 6.)
+
+    # We need to ba able to recover the tags
+    self.recover_tag = nz.recover_tag
+    # Move tensors to the device
+    self.h = self.h.to(self.device)
+
+
+ + + +
+ + + + + + + + + +
+ + +

+ forward(pred, target) + +

+ + +
+ +

Forward pass of the loss function.

+ + +

Parameters:

+
    +
  • + pred + (Tensor) + – +
    +

    The predicted screen tags

    +
    +
    • +
  • + target + (Tensor) + – +
    +

    The target screen tags

    +
    +
    • +
+ + +

Returns:

+
    +
  • +loss ( Tensor +) – +
    +

    The composed loss

    +
    +
    • +
  • +losses ( dict +) – +
    +

    Dictionary containing the individual losses

    +
    +
    • +
+ +
+ Source code in src/speckcn2/loss.py + 
107
+108
+109
+110
+111
+112
+113
+114
+115
+116
+117
+118
+119
+120
+121
+122
+123
+124
+125
+126
+127
+128
+129
+130
+131
+132
+133
+134
+135
+136
+137
+138
+139
+140
+141
+142
+143
+144
+145
+146
def forward(self, pred: torch.Tensor,
+            target: torch.Tensor) -> tuple[torch.Tensor, dict]:
+    """Forward pass of the loss function.
+
+    Parameters
+    ----------
+    pred : torch.Tensor
+        The predicted screen tags
+    target : torch.Tensor
+        The target screen tags
+
+    Returns
+    -------
+    loss : torch.Tensor
+        The composed loss
+    losses : dict
+        Dictionary containing the individual losses
+    """
+    total_loss = 0
+    losses = {}
+
+    if self.Cn2required:
+        Cn2_pred = self.reconstruct_cn2(pred)
+        Cn2_target = self.reconstruct_cn2(target)
+
+    for loss_name, loss_fn in self.loss_needed.items():
+        weight = self.loss_weights[loss_name]
+        if loss_name in ['MAE', 'MSE']:
+            #if loss_name == 'MAE':
+            #    normalizing_factor = torch.abs(target) + 1e-7
+            #else:
+            #    normalizing_factor = target * target + 1e-7
+            this_loss = loss_fn(pred, target)
+            #this_loss = (this_loss / normalizing_factor).mean()
+        else:
+            this_loss = loss_fn(pred, target, Cn2_pred, Cn2_target)
+        total_loss += weight * this_loss
+        losses[loss_name] = this_loss
+
+    return total_loss / self.total_weight, losses
+
+
+
+ +
+ +
+ + +

+ get_FriedParameter(Jnorm) + +

+ + +
+ +

Compute the Fried parameter r0 from the screen tags.

+ +
+ Source code in src/speckcn2/loss.py + 
299
+300
+301
+302
def get_FriedParameter(self, Jnorm: torch.Tensor) -> torch.Tensor:
+    """Compute the Fried parameter r0 from the screen tags."""
+    J = torch.Tensor(self.get_J(Jnorm))
+    return (self.p_fr * torch.sum(J))**(-3 / 5)
+
+
+
+ +
+ +
+ + +

+ get_IsoplanaticAngle(Cn2) + +

+ + +
+ +

Compute the isoplanatic angle theta0 from the screen tags.

+ +
+ Source code in src/speckcn2/loss.py + 
332
+333
+334
+335
+336
+337
+338
def get_IsoplanaticAngle(self, Cn2: torch.Tensor) -> torch.Tensor:
+    """Compute the isoplanatic angle theta0 from the screen tags."""
+    # Integrate Cn2*z^(5/3)
+    integral = torch.sum(
+        Cn2 * (self.h[1:]**(8 / 3) - self.h[:-1]**(8 / 3))) * 3 / 8
+    # Then I can compute theta0
+    return self.p_iso / (integral**(3 / 5))
+
+
+
+ +
+ +
+ + +

+ get_J(Jnorm) + +

+ + +
+ +

Recover J from the normalized tags. This needs to be done to compute +Cn2.

+ + +

Parameters:

+
    +
  • + Jnorm + (Tensor) + – +
    +

    The normalized screen tags between 0 and 1

    +
    +
    • +
+ + +

Returns:

+
    +
  • +J ( Tensor +) – +
    +

    The recovered screen tags

    +
    +
    • +
+ +
+ Source code in src/speckcn2/loss.py + 
151
+152
+153
+154
+155
+156
+157
+158
+159
+160
+161
+162
+163
+164
+165
+166
+167
+168
+169
+170
+171
+172
+173
+174
+175
+176
+177
+178
def get_J(self, Jnorm: torch.Tensor) -> torch.Tensor:
+    """Recover J from the normalized tags. This needs to be done to compute
+    Cn2.
+
+    Parameters
+    ----------
+    Jnorm : torch.Tensor
+        The normalized screen tags between 0 and 1
+
+    Returns
+    -------
+    J : torch.Tensor
+        The recovered screen tags
+    """
+
+    if Jnorm.ndim == 1:
+        Jnorm = Jnorm[None, :]
+
+    J = []
+    for i in range(Jnorm.shape[0]):
+        J.append(
+            torch.tensor([
+                10**self.recover_tag[j](Jnorm[i][j], i)
+                for j in range(len(Jnorm[i]))
+            ],
+                         requires_grad=True).to(Jnorm.device))
+    J = torch.stack(J)
+    return J
+
+
+
+ +
+ +
+ + +

+ get_ScintillationModerateStrong(x) + +

+ + +
+ +

Compute the scintillation index for moderate-strong turbulence +sigma^2 from the screen tags.

+ +
+ Source code in src/speckcn2/loss.py + 
417
+418
+419
+420
+421
+422
def get_ScintillationModerateStrong(self, x: torch.Tensor) -> torch.Tensor:
+    """Compute the scintillation index for moderate-strong turbulence
+    sigma^2 from the screen tags."""
+    wsigma2 = self.get_ScintillationWeak(x)
+    return torch.exp(wsigma2 * 0.49 / (1 + 1.11 * wsigma2**(6 / 5)) +
+                     0.51 * wsigma2 / (1 + 0.69 * wsigma2**(6 / 5)))
+
+
+
+ +
+ +
+ + +

+ get_ScintillationWeak(Cn) + +

+ + +
+ +

Compute the scintillation index for weak turbulence sigma^2 from the +screen tags.

+ +
+ Source code in src/speckcn2/loss.py + 
379
+380
+381
+382
+383
+384
+385
+386
def get_ScintillationWeak(self, Cn: torch.Tensor) -> torch.Tensor:
+    """Compute the scintillation index for weak turbulence sigma^2 from the
+    screen tags."""
+    # Integrate Cn2*z^(5/6)
+    integral = torch.sum(
+        Cn * (self.h[1:]**(11 / 6) - self.h[:-1]**(11 / 6))) * 6 / 11
+    # Then I can compute sigma^2
+    return self.p_scw * integral
+
+
+
+ +
+ +
+ + +

+ reconstruct_cn2(Jnorm) + +

+ + +
+ +

Reconstruct Cn2 from screen tags +c_i = J_i / (h[i+1] - h[i])

+ + +

Parameters:

+
    +
  • + Jnorm + (Tensor) + – +
    +

    The screen tags normalized between 0 and 1

    +
    +
    • +
+ + +

Returns:

+
    +
  • +Cn2 ( Tensor +) – +
    +

    The Cn2 reconstructed from the screen tags, assuming a uniform profile

    +
    +
    • +
+ +
+ Source code in src/speckcn2/loss.py + 
180
+181
+182
+183
+184
+185
+186
+187
+188
+189
+190
+191
+192
+193
+194
+195
+196
def reconstruct_cn2(self, Jnorm: torch.Tensor) -> torch.Tensor:
+    """ Reconstruct Cn2 from screen tags
+    c_i = J_i / (h[i+1] - h[i])
+
+    Parameters
+    ----------
+    Jnorm : torch.Tensor
+        The screen tags normalized between 0 and 1
+
+    Returns
+    -------
+    Cn2 : torch.Tensor
+        The Cn2 reconstructed from the screen tags, assuming a uniform profile
+    """
+    J = self.get_J(Jnorm)
+    Cn2 = J / (self.h[1:] - self.h[:-1])
+    return Cn2
+
+
+
+ +
+ + + +
+ +
+ +
+ + + + +
+ +
+ +
+ + + + + + + + + + + + + +
+
+ + + +
+ +
+ + + +
+
+
+
+ + + + + + + + + + + + + + \ No newline at end of file diff --git a/api/mlops/index.html b/api/mlops/index.html new file mode 100644 index 0000000..6184b43 --- /dev/null +++ b/api/mlops/index.html @@ -0,0 +1,1578 @@ + + + + + + + + + + + + + + + + + + + + + + + + + MLOps - Speckcn2 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
+ + + + Skip to content + + +
+
+ +
+ + + + + + +
+ + +
+ +
+ + + + + + +
+
+ + + +
+
+
+ + + + + +
+
+
+ + + +
+
+
+ + + +
+
+
+ + + +
+
+ + + + + + + +

MLOps

+ +
+ + + + +
+ + + + + + + + +
+ + + + + + + + + +
+ + +

+ score(model, test_set, device, criterion, normalizer, nimg_plot=100) + +

+ + +
+ +

Tests the model.

+ + +

Parameters:

+
    +
  • + model + (Module) + – +
    +

    The model to test

    +
    +
    • +
  • + test_set + (list) + – +
    +

    The testing set

    +
    +
    • +
  • + device + (device) + – +
    +

    The device to use

    +
    +
    • +
  • + criterion + (ComposableLoss) + – +
    +

    The composable loss function, where I can access useful parameters

    +
    +
    • +
  • + normalizer + (Normalizer) + – +
    +

    The normalizer used to recover the tags

    +
    +
    • +
  • + nimg_plot + (int, default: + 100 +) + – +
    +

    Number of images to plot

    +
    +
    • +
+ + +

Returns:

+
    +
  • +test_tags ( list +) – +
    +

    List of all the predicted tags of the test set

    +
    +
    • +
  • +test_losses ( list +) – +
    +

    List of all the losses of the test set

    +
    +
    • +
  • +test_measures ( list +) – +
    +

    List of all the measures of the test set

    +
    +
    • +
  • +test_cn2_pred ( list +) – +
    +

    List of all the predicted Cn2 profiles of the test set

    +
    +
    • +
  • +test_cn2_true ( list +) – +
    +

    List of all the true Cn2 profiles of the test set

    +
    +
    • +
  • +test_recovered_tag_pred ( list +) – +
    +

    List of all the recovered tags from the model prediction

    +
    +
    • +
  • +test_recovered_tag_true ( list +) – +
    +

    List of all the recovered tags

    +
    +
    • +
+ +
+ Source code in src/speckcn2/mlops.py + 
120
+121
+122
+123
+124
+125
+126
+127
+128
+129
+130
+131
+132
+133
+134
+135
+136
+137
+138
+139
+140
+141
+142
+143
+144
+145
+146
+147
+148
+149
+150
+151
+152
+153
+154
+155
+156
+157
+158
+159
+160
+161
+162
+163
+164
+165
+166
+167
+168
+169
+170
+171
+172
+173
+174
+175
+176
+177
+178
+179
+180
+181
+182
+183
+184
+185
+186
+187
+188
+189
+190
+191
+192
+193
+194
+195
+196
+197
+198
+199
+200
+201
+202
+203
+204
+205
+206
+207
+208
+209
+210
+211
+212
+213
+214
+215
+216
+217
+218
+219
+220
+221
+222
+223
+224
+225
+226
+227
+228
+229
+230
+231
+232
+233
+234
+235
+236
+237
+238
+239
+240
+241
+242
+243
+244
+245
+246
+247
+248
+249
+250
+251
+252
+253
def score(
+        model: nn.Module,
+        test_set: list,
+        device: torch.device,
+        criterion: ComposableLoss,
+        normalizer: Normalizer,
+        nimg_plot: int = 100
+) -> tuple[list, list, list, list, list, list, list]:
+    """Tests the model.
+
+    Parameters
+    ----------
+    model : torch.nn.Module
+        The model to test
+    test_set : list
+        The testing set
+    device : torch.device
+        The device to use
+    criterion : ComposableLoss
+        The composable loss function, where I can access useful parameters
+    normalizer : Normalizer
+        The normalizer used to recover the tags
+    nimg_plot : int
+        Number of images to plot
+
+    Returns
+    -------
+    test_tags : list
+        List of all the predicted tags of the test set
+    test_losses : list
+        List of all the losses of the test set
+    test_measures : list
+        List of all the measures of the test set
+    test_cn2_pred : list
+        List of all the predicted Cn2 profiles of the test set
+    test_cn2_true : list
+        List of all the true Cn2 profiles of the test set
+    test_recovered_tag_pred : list
+        List of all the recovered tags from the model prediction
+    test_recovered_tag_true : list
+        List of all the recovered tags
+    """
+    counter = 0
+    conf = normalizer.conf
+    data_dir = conf['speckle']['datadirectory']
+    batch_size = conf['hyppar']['batch_size']
+    # Setup the EnsembleModel wrapper
+    ensemble = EnsembleModel(conf, device)
+
+    # For scoring the model, you can compose the loss as you did during training
+    # so it is possible to include:
+    # 1. MAE on screen tags
+    # 2. Fried MAE
+    # 3. Isoplanatic angle MAE
+    # 4. Scintillation (weak) index MAE
+    # TODO: this should be controllable in the configuration file
+    criterion.loss_weights = {
+        'MSE': 1,
+        'MAE': 1,
+        'JMSE': 0,
+        'JMAE': 0,
+        'Cn2MSE': 0,
+        'Cn2MAE': 0,
+        'Pearson': 0,
+        'Fried': 1,
+        'Isoplanatic': 1,
+        'Rytov': 0,
+        'Scintillation_w': 1,
+        'Scintillation_ms': 0,
+    }
+    criterion._select_loss_needed()
+
+    with torch.no_grad():
+        # Put model in evaluation mode
+        model.eval()
+
+        test_tags = []
+        test_losses = []
+        test_measures = []
+        test_cn2_pred = []
+        test_cn2_true = []
+        test_recovered_tag_pred = []
+        test_recovered_tag_true = []
+        # Initialize the loss max and min. They are used to plot the images with the
+        # highest and lowest loss. We skip examples with a common average value of the loss.
+        loss_max = 0
+        loss_min = 1e6
+        # create the directory where the images will be stored
+        ensure_directory(f'{data_dir}/{model.name}_score')
+
+        for idx in range(0, len(test_set), batch_size):
+            batch = test_set[idx:idx + batch_size]
+
+            # Forward pass
+            outputs, targets, inputs = ensemble(model, batch)
+
+            # Loop each input separately
+            for i in range(len(outputs)):
+                loss, losses = criterion(outputs[i], targets[i])
+
+                # Get the Cn2 profile and the recovered tags
+                Cn2_pred = criterion.reconstruct_cn2(outputs[i])
+                Cn2_true = criterion.reconstruct_cn2(targets[i])
+                recovered_tag_pred = criterion.get_J(outputs[i])
+                recovered_tag_true = criterion.get_J(targets[i])
+                # and get all the measures
+                all_measures = criterion._get_all_measures(
+                    outputs[i], targets[i], Cn2_pred, Cn2_true)
+                this_loss = loss.item()
+
+                if counter < nimg_plot and (this_loss > loss_max
+                                            or this_loss < loss_min):
+                    loss_max = max(this_loss, loss_max)
+                    loss_min = min(this_loss, loss_min)
+                    print(f'Plotting item {counter} loss: {this_loss:.4f}')
+                    score_plot(conf, inputs, targets, loss, losses, i, counter,
+                               all_measures, Cn2_pred, Cn2_true,
+                               recovered_tag_pred, recovered_tag_true)
+                    counter += 1
+
+                # and get all the tags for statistic analysis
+                for tag in outputs:
+                    test_tags.append(tag)
+
+                # and get the other information
+                test_losses.append(losses)
+                test_measures.append(all_measures)
+                test_cn2_pred.append(Cn2_pred)
+                test_cn2_true.append(Cn2_true)
+                test_recovered_tag_pred.append(recovered_tag_pred)
+                test_recovered_tag_true.append(recovered_tag_true)
+
+    return (test_tags, test_losses, test_measures, test_cn2_pred,
+            test_cn2_true, test_recovered_tag_pred, test_recovered_tag_true)
+
+
+
+ +
+ +
+ + +

+ train(model, last_model_state, conf, train_set, test_set, device, optimizer, criterion) + +

+ + +
+ +

Trains the model for the given number of epochs.

+ + +

Parameters:

+
    +
  • + model + (Module) + – +
    +

    The model to train

    +
    +
    • +
  • + last_model_state + (int) + – +
    +

    The number of the last model state

    +
    +
    • +
  • + conf + (dict) + – +
    +

    Dictionary containing the configuration

    +
    +
    • +
  • + train_set + (list) + – +
    +

    The training set

    +
    +
    • +
  • + test_set + (list) + – +
    +

    The testing set

    +
    +
    • +
  • + device + (device) + – +
    +

    The device to use

    +
    +
    • +
  • + optimizer + (optim) + – +
    +

    The optimizer to use

    +
    +
    • +
  • + criterion + (ComposableLoss) + – +
    +

    The loss function to use

    +
    +
    • +
+ + +

Returns:

+
    +
  • +model ( Module +) – +
    +

    The trained model

    +
    +
    • +
  • +average_loss ( float +) – +
    +

    The average loss of the last epoch

    +
    +
    • +
+ +
+ Source code in src/speckcn2/mlops.py + 
 17
+ 18
+ 19
+ 20
+ 21
+ 22
+ 23
+ 24
+ 25
+ 26
+ 27
+ 28
+ 29
+ 30
+ 31
+ 32
+ 33
+ 34
+ 35
+ 36
+ 37
+ 38
+ 39
+ 40
+ 41
+ 42
+ 43
+ 44
+ 45
+ 46
+ 47
+ 48
+ 49
+ 50
+ 51
+ 52
+ 53
+ 54
+ 55
+ 56
+ 57
+ 58
+ 59
+ 60
+ 61
+ 62
+ 63
+ 64
+ 65
+ 66
+ 67
+ 68
+ 69
+ 70
+ 71
+ 72
+ 73
+ 74
+ 75
+ 76
+ 77
+ 78
+ 79
+ 80
+ 81
+ 82
+ 83
+ 84
+ 85
+ 86
+ 87
+ 88
+ 89
+ 90
+ 91
+ 92
+ 93
+ 94
+ 95
+ 96
+ 97
+ 98
+ 99
+100
+101
+102
+103
+104
+105
+106
+107
+108
+109
+110
+111
+112
+113
+114
+115
+116
+117
def train(model: nn.Module, last_model_state: int, conf: dict, train_set: list,
+          test_set: list, device: torch.device, optimizer: optim.Optimizer,
+          criterion: ComposableLoss) -> tuple[nn.Module, float]:
+    """Trains the model for the given number of epochs.
+
+    Parameters
+    ----------
+    model : torch.nn.Module
+        The model to train
+    last_model_state : int
+        The number of the last model state
+    conf : dict
+        Dictionary containing the configuration
+    train_set : list
+        The training set
+    test_set : list
+        The testing set
+    device : torch.device
+        The device to use
+    optimizer : torch.optim
+        The optimizer to use
+    criterion : ComposableLoss
+        The loss function to use
+
+    Returns
+    -------
+    model : torch.nn.Module
+        The trained model
+    average_loss : float
+        The average loss of the last epoch
+    """
+
+    final_epoch = conf['hyppar']['maxepochs']
+    save_every = conf['model']['save_every']
+    datadirectory = conf['speckle']['datadirectory']
+    batch_size = conf['hyppar']['batch_size']
+
+    # Setup the EnsembleModel wrapper
+    ensemble = EnsembleModel(conf, device)
+
+    print(f'Training the model from epoch {last_model_state} to {final_epoch}')
+    average_loss = 0.0
+    model.train()
+    for epoch in range(last_model_state, final_epoch):
+        total_loss = 0.0
+        model.train()
+        t_in = time.time()
+        for i in range(0, len(train_set), batch_size):
+            batch = train_set[i:i + batch_size]
+
+            # Zero out the optimizer
+            optimizer.zero_grad()
+
+            # Forward pass
+            outputs, targets, _ = ensemble(model, batch)
+            loss, _ = criterion(outputs, targets)
+
+            # Backward pass
+            loss.backward()
+            optimizer.step()
+
+            # Accumulate the loss
+            total_loss += loss.item()
+
+        # Shuffle the training set
+        random.shuffle(train_set)
+
+        # Calculate average loss for the epoch
+        average_loss = total_loss / len(train_set)
+
+        # Log the important information
+        t_fin = time.time() - t_in
+        model.loss.append(average_loss)
+        model.time.append(t_fin)
+        model.epoch.append(epoch + 1)
+
+        # And also the validation loss
+        val_loss = 0.0
+        model.eval()
+        with torch.no_grad():
+            for i in range(0, len(test_set), batch_size):
+                batch = test_set[i:i + batch_size]
+                # Forward pass
+                outputs, targets, _ = ensemble(model, batch)
+                loss, _ = criterion(outputs, targets)
+                # sum loss
+                val_loss += loss.item()
+        val_loss = val_loss / len(test_set)
+        model.val_loss.append(val_loss)
+
+        # Print the average loss for every epoch
+        message = (f'Epoch {epoch+1}/{final_epoch} '
+                   f'(in {t_fin:.3g}s),\tTrain-Loss: {average_loss:.5f},\t'
+                   f'Test-Loss: {val_loss:.5f}')
+        print(message, flush=True)
+
+        if (epoch + 1) % save_every == 0 or epoch == final_epoch - 1:
+            # Save the model state
+            save(model, datadirectory)
+
+    return model, average_loss
+
+
+
+ +
+ + + +
+ +
+ +
+ + + + + + + + + + + + + +
+
+ + + +
+ +
+ + + +
+
+
+
+ + + + + + + + + + + + + + \ No newline at end of file diff --git a/api/models/index.html b/api/models/index.html new file mode 100644 index 0000000..009f60d --- /dev/null +++ b/api/models/index.html @@ -0,0 +1,1896 @@ + + + + + + + + + + + + + + + + + + + + + + + + + Models - Speckcn2 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
+ + + + Skip to content + + +
+
+ +
+ + + + + + +
+ + +
+ +
+ + + + + + +
+
+ + + +
+
+
+ + + + + +
+
+
+ + + +
+
+
+ + + +
+
+
+ + + +
+
+ + + + + + + +

Models

+ +
+ + + + +
+ +

This module contains the definition of the EnsembleModel class and a +setup_model function.

+

The EnsembleModel class is a wrapper that allows any model to be used +for ensembled data. The setup_model function initializes and returns a +model based on the provided configuration.

+ + + + + + + + +
+ + + + + + + + +
+ + + +

+ EnsembleModel(conf, device) + +

+ + +
+

+ Bases: Module

+ + +

Wrapper that allows any model to be used for ensembled data.

+ + + +

Parameters:

+
    +
  • + conf + (dict) + – +
    +

    The global configuration containing the model parameters.

    +
    +
    • +
  • + device + (device) + – +
    +

    The device to use

    +
    +
    • +
+ + + + + + +
+ Source code in src/speckcn2/mlmodels.py + 
26
+27
+28
+29
+30
+31
+32
+33
+34
+35
+36
+37
+38
+39
+40
+41
+42
+43
+44
+45
+46
+47
+48
+49
+50
+51
+52
+53
+54
+55
+56
+57
def __init__(self, conf: dict, device: torch.device):
+    """Initializes the EnsembleModel.
+
+    Parameters
+    ----------
+    conf: dict
+        The global configuration containing the model parameters.
+    device : torch.device
+        The device to use
+    """
+    super(EnsembleModel, self).__init__()
+
+    self.ensemble_size = conf['preproc'].get('ensemble', 1)
+    self.device = device
+    self.uniform_ensemble = conf['preproc'].get('ensemble_unif', False)
+    resolution = conf['preproc']['resize']
+    self.D = conf['noise']['D']
+    self.t = conf['noise']['t']
+    self.snr = conf['noise']['snr']
+    self.dT = conf['noise']['dT']
+    self.dO = conf['noise']['dO']
+    self.rn = conf['noise']['rn']
+    self.fw = conf['noise']['fw']
+    self.bit = conf['noise']['bit']
+    self.discretize = conf['noise']['discretize']
+    self.rot_sym = conf['noise'].get('rotation_sym', 0)
+    if self.rot_sym > 0:
+        self.rot_fold = 360 // self.rot_sym
+    self.apply_masks = conf['noise'].get('apply_masks', False)
+    if self.apply_masks:
+        self.mask_D, self.mask_d, self.mask_X, self.mask_Y = self.create_masks(
+            resolution)
+
+
+ + + +
+ + + + + + + + + +
+ + +

+ apply_noise(image_tensor) + +

+ + +
+ +

Processes a tensor of 2D images.

+ + +

Parameters:

+
    +
  • + image_tensor + (Tensor) + – +
    +

    Tensor of 2D images with shape (batch, channels, width, height).

    +
    +
    • +
+ + +

Returns:

+
    +
  • +processed_tensor ( Tensor +) – +
    +

    Tensor of processed 2D images.

    +
    +
    • +
+ +
+ Source code in src/speckcn2/mlmodels.py + 
112
+113
+114
+115
+116
+117
+118
+119
+120
+121
+122
+123
+124
+125
+126
+127
+128
+129
+130
+131
+132
+133
+134
+135
+136
+137
+138
+139
+140
+141
+142
+143
+144
+145
+146
+147
+148
+149
+150
+151
+152
+153
+154
+155
+156
+157
+158
+159
+160
+161
+162
+163
+164
+165
+166
def apply_noise(self, image_tensor: torch.Tensor) -> torch.Tensor:
+    """Processes a tensor of 2D images.
+
+    Parameters
+    ----------
+    image_tensor : torch.Tensor
+        Tensor of 2D images with shape (batch, channels, width, height).
+
+    Returns
+    -------
+    processed_tensor : torch.Tensor
+        Tensor of processed 2D images.
+    """
+    batch, channels, height, width = image_tensor.shape
+    processed_tensor = torch.zeros_like(image_tensor)
+
+    # Apply rotation symmetry
+    if self.rot_sym > 0:
+        angle = random.randint(0, self.rot_fold)
+        image_tensor = torch.rot90(image_tensor, angle, (2, 3))
+
+    # Normalize wrt optical power
+    image_tensor = image_tensor / torch.mean(
+        image_tensor, dim=(2, 3), keepdim=True)
+
+    amp = self.rn * 10**(self.snr / 20)
+
+    for i in range(batch):
+        for j in range(channels):
+            B = image_tensor[i, j]
+
+            ## Apply masks
+            if self.apply_masks:
+                B[self.mask_D] = 0
+                B[self.mask_d] = 0
+                B[self.mask_X] = 0
+                B[self.mask_Y] = 0
+
+            # Add noise sources
+            A = self.rn + self.rn * torch.randn(
+                height, width, device=self.device) + amp * B + torch.sqrt(
+                    amp * B) * torch.randn(
+                        height, width, device=self.device)
+
+            # Make a discretized version
+            if self.discretize == 'on':
+                C = torch.round(A / self.fw * 2**self.bit)
+                C[A > self.fw] = self.fw
+                C[A < 0] = 0
+            else:
+                C = A
+
+            processed_tensor[i, j] = C
+
+    return processed_tensor
+
+
+
+ +
+ +
+ + +

+ create_masks(resolution) + +

+ + +
+ +

Creates the masks for the circular aperture and the spider.

+ + +

Parameters:

+
    +
  • + resolution + (int) + – +
    +

    Resolution of the images.

    +
    +
    • +
+ + +

Returns:

+
    +
  • +mask_D ( Tensor +) – +
    +

    Mask for the circular aperture.

    +
    +
    • +
  • +mask_d ( Tensor +) – +
    +

    Mask for the central obscuration.

    +
    +
    • +
  • +mask_X ( Tensor +) – +
    +

    Mask for the horizontal spider.

    +
    +
    • +
  • +mask_Y ( Tensor +) – +
    +

    Mask for the vertical spider.

    +
    +
    • +
+ +
+ Source code in src/speckcn2/mlmodels.py + 
168
+169
+170
+171
+172
+173
+174
+175
+176
+177
+178
+179
+180
+181
+182
+183
+184
+185
+186
+187
+188
+189
+190
+191
+192
+193
+194
+195
+196
+197
+198
+199
+200
+201
+202
def create_masks(
+    self, resolution: int
+) -> tuple[torch.Tensor, torch.Tensor, torch.Tensor, torch.Tensor]:
+    """Creates the masks for the circular aperture and the spider.
+
+    Parameters
+    ----------
+    resolution : int
+        Resolution of the images.
+
+    Returns
+    -------
+    mask_D : torch.Tensor
+        Mask for the circular aperture.
+    mask_d : torch.Tensor
+        Mask for the central obscuration.
+    mask_X : torch.Tensor
+        Mask for the horizontal spider.
+    mask_Y : torch.Tensor
+        Mask for the vertical spider.
+    """
+    # Coordinates
+    x = torch.linspace(-1, 1, resolution, device=self.device)
+    X, Y = torch.meshgrid(x, x, indexing='ij')  # XY grid
+    d = self.dO * self.D  # Diameter obscuration
+
+    R = torch.sqrt(X**2 + Y**2)
+
+    # Masking image
+    mask_D = R > self.D
+    mask_d = R < d
+    mask_X = torch.abs(X) < self.t / 2
+    mask_Y = torch.abs(Y) < self.t / 2
+
+    return mask_D, mask_d, mask_X, mask_Y
+
+
+
+ +
+ +
+ + +

+ forward(model, batch_ensemble) + +

+ + +
+ +

Forward pass through the model.

+ + +

Parameters:

+
    +
  • + model + (Module) + – +
    +

    The model to use

    +
    +
    • +
  • + batch_ensemble + (list) + – +
    +

    Each element is a batch of an ensemble of samples.

    +
    +
    • +
+ +
+ Source code in src/speckcn2/mlmodels.py + 
 59
+ 60
+ 61
+ 62
+ 63
+ 64
+ 65
+ 66
+ 67
+ 68
+ 69
+ 70
+ 71
+ 72
+ 73
+ 74
+ 75
+ 76
+ 77
+ 78
+ 79
+ 80
+ 81
+ 82
+ 83
+ 84
+ 85
+ 86
+ 87
+ 88
+ 89
+ 90
+ 91
+ 92
+ 93
+ 94
+ 95
+ 96
+ 97
+ 98
+ 99
+100
+101
+102
+103
+104
+105
+106
+107
+108
+109
+110
def forward(self, model, batch_ensemble):
+    """Forward pass through the model.
+
+    Parameters
+    ----------
+    model : torch.nn.Module
+        The model to use
+    batch_ensemble : list
+        Each element is a batch of an ensemble of samples.
+    """
+
+    if self.ensemble_size == 1:
+        batch = batch_ensemble
+        # If no ensembling, each element of the batch is a tuple (image, tag, ensemble_id)
+        images, tags, ensembles = zip(*batch)
+        images = torch.stack(images).to(self.device)
+        images = self.apply_noise(images)
+        tags = torch.tensor(np.stack(tags)).to(self.device)
+
+        return model(images), tags, images
+    else:
+        batch = list(itertools.chain(*batch_ensemble))
+        # Like the ensemble=1 case, I can process independently each element of the batch
+        images, tags, ensembles = zip(*batch)
+        images = torch.stack(images).to(self.device)
+        images = self.apply_noise(images)
+        tags = torch.tensor(np.stack(tags)).to(self.device)
+
+        model_output = model(images)
+
+        # To average the self.ensemble_size outputs of the model I extract the confidence weights
+        predictions = model_output[:, :-1]
+        weights = model_output[:, -1]
+        if self.uniform_ensemble:
+            weights = torch.ones_like(weights)
+        # multiply the prediction by the weights
+        weighted_predictions = predictions * weights.unsqueeze(-1)
+        # and sum over the ensembles
+        weighted_predictions = weighted_predictions.view(
+            model_output.size(0) // self.ensemble_size, self.ensemble_size,
+            -1).sum(dim=1)
+        # then normalize by the sum of the weights
+        sum_weights = weights.view(
+            weights.size(0) // self.ensemble_size,
+            self.ensemble_size).sum(dim=1)
+        ensemble_output = weighted_predictions / sum_weights.unsqueeze(-1)
+
+        # and get the tags and ensemble_id of the first element of the ensemble
+        tags = tags[::self.ensemble_size]
+        ensembles = ensembles[::self.ensemble_size]
+
+        return ensemble_output, tags, images
+
+
+
+ +
+ + + +
+ +
+ +
+ + +
+ + +

+ get_a_resnet(config) + +

+ + +
+ +

Returns a pretrained ResNet model, with the last layer corresponding to +the number of screens.

+ + +

Parameters:

+
    +
  • + config + (dict) + – +
    +

    Dictionary containing the configuration

    +
    +
    • +
+ + +

Returns:

+
    +
  • +model ( Module +) – +
    +

    The model with the loaded state

    +
    +
    • +
  • +last_model_state ( int +) – +
    +

    The number of the last model state

    +
    +
    • +
+ +
+ Source code in src/speckcn2/mlmodels.py + 
235
+236
+237
+238
+239
+240
+241
+242
+243
+244
+245
+246
+247
+248
+249
+250
+251
+252
+253
+254
+255
+256
+257
+258
+259
+260
+261
+262
+263
+264
+265
+266
+267
+268
+269
+270
+271
+272
+273
+274
+275
+276
+277
+278
+279
+280
+281
+282
+283
+284
+285
+286
+287
+288
+289
+290
+291
+292
+293
def get_a_resnet(config: dict) -> tuple[nn.Module, int]:
+    """Returns a pretrained ResNet model, with the last layer corresponding to
+    the number of screens.
+
+    Parameters
+    ----------
+    config : dict
+        Dictionary containing the configuration
+
+    Returns
+    -------
+    model : torch.nn.Module
+        The model with the loaded state
+    last_model_state : int
+        The number of the last model state
+    """
+
+    model_name = config['model']['name']
+    model_type = config['model']['type']
+    pretrained = config['model']['pretrained']
+    nscreens = config['speckle']['nscreens']
+    data_directory = config['speckle']['datadirectory']
+    ensemble = config['preproc'].get('ensemble', 1)
+
+    if model_type == 'resnet18':
+        model = torchvision.models.resnet18(
+            weights='IMAGENET1K_V1' if pretrained else None)
+        finaloutsize = 512
+    elif model_type == 'resnet50':
+        model = torchvision.models.resnet50(
+            weights='IMAGENET1K_V2' if pretrained else None)
+        finaloutsize = 2048
+    elif model_type == 'resnet152':
+        model = torchvision.models.resnet152(
+            weights='IMAGENET1K_V2' if pretrained else None)
+        finaloutsize = 2048
+    else:
+        raise ValueError(f'Unknown model {model_type}')
+
+    # If the model uses multiple images as input,
+    # add an extra channel as confidence weight
+    # to average the final prediction
+    if ensemble > 1:
+        nscreens = nscreens + 1
+
+    # Give it its name
+    model.name = model_name
+
+    # Change the model to process black and white input
+    model.conv1 = torch.nn.Conv2d(1,
+                                  64,
+                                  kernel_size=(7, 7),
+                                  stride=(2, 2),
+                                  padding=(3, 3),
+                                  bias=False)
+    # Add a final fully connected piece to predict the output
+    model.fc = create_final_block(config, finaloutsize, nscreens)
+
+    return load_model_state(model, data_directory)
+
+
+
+ +
+ +
+ + +

+ get_scnn(config) + +

+ + +
+ +

Returns a pretrained Spherical-CNN model, with the last layer +corresponding to the number of screens.

+ +
+ Source code in src/speckcn2/mlmodels.py + 
296
+297
+298
+299
+300
+301
+302
+303
+304
+305
+306
+307
+308
+309
+310
+311
+312
+313
+314
+315
+316
+317
+318
+319
def get_scnn(config: dict) -> tuple[nn.Module, int]:
+    """Returns a pretrained Spherical-CNN model, with the last layer
+    corresponding to the number of screens."""
+
+    model_name = config['model']['name']
+    model_type = config['model']['type']
+    datadirectory = config['speckle']['datadirectory']
+
+    model_map = {
+        'scnnC8': 'C8',
+        'scnnC16': 'C16',
+        'scnnC4': 'C4',
+        'scnnC6': 'C6',
+        'scnnC10': 'C10',
+        'scnnC12': 'C12',
+    }
+    try:
+        scnn_model = SteerableCNN(config, model_map[model_type])
+    except KeyError:
+        raise ValueError(f'Unknown model {model_type}')
+
+    scnn_model.name = model_name
+
+    return load_model_state(scnn_model, datadirectory)
+
+
+
+ +
+ +
+ + +

+ setup_model(config) + +

+ + +
+ +

Returns the model specified in the configuration file, with the last +layer corresponding to the number of screens.

+ + +

Parameters:

+
    +
  • + config + (dict) + – +
    +

    Dictionary containing the configuration

    +
    +
    • +
+ + +

Returns:

+
    +
  • +model ( Module +) – +
    +

    The model with the loaded state

    +
    +
    • +
  • +last_model_state ( int +) – +
    +

    The number of the last model state

    +
    +
    • +
+ +
+ Source code in src/speckcn2/mlmodels.py + 
205
+206
+207
+208
+209
+210
+211
+212
+213
+214
+215
+216
+217
+218
+219
+220
+221
+222
+223
+224
+225
+226
+227
+228
+229
+230
+231
+232
def setup_model(config: dict) -> tuple[nn.Module, int]:
+    """Returns the model specified in the configuration file, with the last
+    layer corresponding to the number of screens.
+
+    Parameters
+    ----------
+    config : dict
+        Dictionary containing the configuration
+
+    Returns
+    -------
+    model : torch.nn.Module
+        The model with the loaded state
+    last_model_state : int
+        The number of the last model state
+    """
+
+    model_name = config['model']['name']
+    model_type = config['model']['type']
+
+    print(f'^^^ Initializing model {model_name} of type {model_type}')
+
+    if model_type.startswith('resnet'):
+        return get_a_resnet(config)
+    elif model_type.startswith('scnnC'):
+        return get_scnn(config)
+    else:
+        raise ValueError(f'Unknown model {model_name}')
+
+
+
+ +
+ + + +
+ +
+ +
+ + + + + + + + + + + + + +
+
+ + + +
+ +
+ + + +
+
+
+
+ + + + + + + + + + + + + + \ No newline at end of file diff --git a/api/normalizer/index.html b/api/normalizer/index.html new file mode 100644 index 0000000..3ff1877 --- /dev/null +++ b/api/normalizer/index.html @@ -0,0 +1,1092 @@ + + + + + + + + + + + + + + + + + + + + + + + + + Normalizer - Speckcn2 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
+ + + + Skip to content + + +
+
+ +
+ + + + + + +
+ + +
+ +
+ + + + + + +
+
+ + + +
+
+
+ + + + + +
+
+
+ + + +
+
+
+ + + +
+
+
+ + + +
+
+ + + + + + + +

Normalizer

+ +
+ + + + +
+ +

This module defines the Normalizer class, which handles the normalization of +images and tags based on a given configuration.

+

The class includes methods to precompile normalization functions, +normalize images and tags, and define specific normalization strategies +such as Z-score and uniform normalization. The normalization process +involves replacing NaN values, creating masks, and scaling values to a +specified range. The module also provides functions to recover the +original values from the normalized data. The Normalizer class ensures +that both images and tags are consistently normalized according to the +specified configuration, facilitating further processing and analysis.

+ + + + + + + + +
+ + + + + + + + +
+ + + +

+ Normalizer(conf) + +

+ + +
+ + +

Class to handle the normalization of images and tags.

+ + +

Parameters:

+
    +
  • + conf + (dict) + – +
    +

    Dictionary containing the configuration

    +
    +
    • +
+ + + + + + +
+ Source code in src/speckcn2/normalizer.py + 
30
+31
def __init__(self, conf: dict):
+    self.conf = conf
+
+
+ + + +
+ + + + + + + + + +
+ + +

+ normalize_imgs_and_tags(all_images, all_tags, all_ensemble_ids) + +

+ + +
+ +

Normalize both the input images and the tags to be between 0 and 1.

+ + +

Parameters:

+
    +
  • + all_images + (list) + – +
    +

    List of all images

    +
    +
    • +
  • + all_tags + (list) + – +
    +

    List of all tags

    +
    +
    • +
  • + conf + (dict) + – +
    +

    Dictionary containing the configuration

    +
    +
    • +
+ + +

Returns:

+
    +
  • +dataset ( list +) – +
    +

    List of tuples (image, normalized_tag)

    +
    +
    • +
+ +
+ Source code in src/speckcn2/normalizer.py + 
 67
+ 68
+ 69
+ 70
+ 71
+ 72
+ 73
+ 74
+ 75
+ 76
+ 77
+ 78
+ 79
+ 80
+ 81
+ 82
+ 83
+ 84
+ 85
+ 86
+ 87
+ 88
+ 89
+ 90
+ 91
+ 92
+ 93
+ 94
+ 95
+ 96
+ 97
+ 98
+ 99
+100
+101
+102
+103
+104
def normalize_imgs_and_tags(
+    self,
+    all_images: list[torch.tensor],
+    all_tags: list[np.ndarray],
+    all_ensemble_ids: list[int],
+) -> list[tuple[torch.tensor, np.ndarray, int]]:
+    """Normalize both the input images and the tags to be between 0 and 1.
+
+    Parameters
+    ----------
+    all_images : list
+        List of all images
+    all_tags : list
+        List of all tags
+    conf : dict
+        Dictionary containing the configuration
+
+    Returns
+    -------
+    dataset : list
+        List of tuples (image, normalized_tag)
+    """
+    self._normalizing_functions(all_images, all_tags, all_ensemble_ids)
+
+    # Normalize the images
+    normalized_images = [self.normalize_img(image) for image in all_images]
+
+    # And normalize the tags
+    normalized_tags = np.array([[
+        self.normalize_tag[j](tag, tag_id) for j, tag in enumerate(tags)
+    ] for tag_id, tags in enumerate(all_tags)],
+                               dtype=np.float32)
+
+    # I can now create the dataset
+    dataset = [(image, tag, ensemble_id) for image, tag, ensemble_id in
+               zip(normalized_images, normalized_tags, all_ensemble_ids)]
+
+    return dataset
+
+
+
+ +
+ + + +
+ +
+ +
+ + + + +
+ +
+ +
+ + + + + + + + + + + + + +
+
+ + + +
+ +
+ + + +
+
+
+
+ + + + + + + + + + + + + + \ No newline at end of file diff --git a/api/plot/index.html b/api/plot/index.html new file mode 100644 index 0000000..c93b44f --- /dev/null +++ b/api/plot/index.html @@ -0,0 +1,3001 @@ + + + + + + + + + + + + + + + + + + + + + + + + + Plot - Speckcn2 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
+ + + + Skip to content + + +
+
+ +
+ + + + + + +
+ + +
+ +
+ + + + + + +
+
+ + + +
+
+
+ + + + + +
+
+
+ + + +
+
+
+ + + +
+
+
+ + + +
+
+ + + + + + + +

Plot

+ +
+ + + + +
+ + + + + + + + +
+ + + + + + + + + +
+ + +

+ plot_J_error_details(conf, tags_true, tags_pred, nbins=10, linear_bins=False) + +

+ + +
+ +

Function to plot the histograms per single bin of each single screen tag +to quantify the relative error as a function of J.

+ + +

Parameters:

+
    +
  • + conf + (dict) + – +
    +

    Dictionary containing the configuration

    +
    +
    • +
  • + tags_true + (list) + – +
    +

    The true tags of the validation set

    +
    +
    • +
  • + tags_pred + (list) + – +
    +

    The predicted tags of the validation set

    +
    +
    • +
  • + nbins + (int, default: + 10 +) + – +
    +

    The number of bins in which to partition the data

    +
    +
    • +
  • + linear_bins + (bool, default: + False +) + – +
    +

    If True, the bins are linearly spaced, otherwise they are log spaced

    +
    +
    • +
+ +
+ Source code in src/speckcn2/plots.py + 
484
+485
+486
+487
+488
+489
+490
+491
+492
+493
+494
+495
+496
+497
+498
+499
+500
+501
+502
+503
+504
+505
+506
+507
+508
+509
+510
+511
+512
+513
+514
+515
+516
+517
+518
+519
+520
+521
+522
+523
+524
+525
+526
+527
+528
+529
+530
+531
+532
+533
+534
+535
+536
+537
+538
+539
+540
+541
+542
+543
+544
+545
+546
+547
+548
+549
+550
+551
+552
+553
+554
+555
+556
+557
+558
+559
+560
+561
+562
+563
def plot_J_error_details(conf: dict,
+                         tags_true: list,
+                         tags_pred: list,
+                         nbins: int = 10,
+                         linear_bins: bool = False) -> None:
+    """Function to plot the histograms per single bin of each single screen tag
+    to quantify the relative error as a function of J.
+
+    Parameters
+    ----------
+    conf : dict
+        Dictionary containing the configuration
+    tags_true : list
+        The true tags of the validation set
+    tags_pred : list
+        The predicted tags of the validation set
+    nbins : int
+        The number of bins in which to partition the data
+    linear_bins : bool
+        If True, the bins are linearly spaced, otherwise they are log spaced
+    """
+
+    nscreens = conf['speckle']['nscreens']
+    data_dir = conf['speckle']['datadirectory']
+    model_name = conf['model']['name']
+
+    dirname = f'{data_dir}/{model_name}_score/J_bin_details'
+    ensure_directory(dirname)
+
+    if conf['preproc'].get('J_details', False):
+        for screen_id in range(nscreens):
+            print(f'\nComputing screen-{screen_id} details')
+
+            # Collect the data
+            params = []
+            loss = []
+            for i in range(len(tags_true)):
+                params.append(tags_true[i][0,
+                                           screen_id].detach().cpu().numpy())
+                loss.append((
+                    (tags_pred[i][0, screen_id] - tags_true[i][0, screen_id]) /
+                    (tags_true[i][0, screen_id])).detach().cpu().numpy())
+            params = np.array(params)
+            loss = np.array(loss)
+
+            if linear_bins:
+                bins = np.linspace(min(params), max(params), num=nbins)
+            else:
+                bins = np.logspace(np.log10(min(params)),
+                                   np.log10(max(params)),
+                                   num=nbins)
+            bin_indices = np.digitize(params, bins)
+            # get the average and std of the error per bin of J[screen_id]
+            bin_centers = 0.5 * (bins[:-1] + bins[1:])
+
+            for idx, single_bin in enumerate(bin_centers):
+                l_data = loss[bin_indices == idx]
+                if len(l_data) == 0:
+                    continue
+                fig, axs = plt.subplots(1, 1, figsize=(5, 5))
+                axs.hist(l_data, bins=50, alpha=0.5, density=True)
+                mu = np.mean(l_data)
+                sigma = np.std(l_data)
+                print(
+                    f'J-{screen_id} = {single_bin:.3g} -> mu = {mu:.3f}, sigma = {sigma:.3f}'
+                )
+                if sigma > 0:
+                    x = np.linspace(mu - 3 * sigma, mu + 3 * sigma, 100)
+                    x = x[x > min(l_data)]
+                    x = x[x < max(l_data)]
+                    axs.plot(x,
+                             stats.norm.pdf(x, mu, sigma),
+                             label=f'Average err: {mu:.3f}, Std: {sigma:.3f}')
+                axs.set_xlabel(f'Relative error J (screen-{screen_id})')
+                axs.set_ylabel('Frequency')
+                axs.legend()
+                plt.title(f'J (screen-{screen_id}) value = {single_bin:.3g}')
+                plt.tight_layout()
+                plt.savefig(f'{dirname}/Jscreen{screen_id}_bin{idx}.png')
+                plt.close()
+
+
+
+ +
+ +
+ + +

+ plot_histo_losses(conf, test_losses, data_dir) + +

+ + +
+ +

Plots the histogram of the losses.

+ + +

Parameters:

+
    +
  • + conf + (dict) + – +
    +

    Dictionary containing the configuration

    +
    +
    • +
  • + test_losses + (list[dict]) + – +
    +

    List of all the losses of the test set

    +
    +
    • +
  • + data_dir + (str) + – +
    +

    The directory where the data is stored

    +
    +
    • +
+ +
+ Source code in src/speckcn2/plots.py + 
200
+201
+202
+203
+204
+205
+206
+207
+208
+209
+210
+211
+212
+213
+214
+215
+216
+217
+218
+219
+220
+221
+222
+223
+224
+225
+226
+227
+228
+229
+230
+231
+232
+233
+234
def plot_histo_losses(conf: dict, test_losses: list[dict],
+                      data_dir: str) -> None:
+    """Plots the histogram of the losses.
+
+    Parameters
+    ----------
+    conf : dict
+        Dictionary containing the configuration
+    test_losses : list[dict]
+        List of all the losses of the test set
+    data_dir : str
+        The directory where the data is stored
+    """
+    model_name = conf['model']['name']
+    data_dir = conf['speckle']['datadirectory']
+
+    dirname = f'{data_dir}/{model_name}_score/histo_losses'
+    ensure_directory(dirname)
+
+    # plot the loss
+    fig, axs = plt.subplots(1, 1, figsize=(5, 5))
+    for key in ['MAE', 'Fried', 'Isoplanatic', 'Scintillation_w']:
+        loss = [d[key].detach().cpu() for d in test_losses]
+        bins = np.logspace(np.log10(min(loss)), np.log10(max(loss)),
+                           num=50).tolist()
+        axs.hist(loss, bins=bins, alpha=0.5, label=key, density=True)
+    axs.set_xlabel('Loss')
+    axs.set_ylabel('Frequency')
+    axs.set_yscale('log')
+    axs.set_xscale('log')
+    axs.legend()
+    plt.title(f'Model: {model_name}')
+    plt.tight_layout()
+    plt.savefig(f'{dirname}/histo_losses_{model_name}.png')
+    plt.close()
+
+
+
+ +
+ +
+ + +

+ plot_loss(conf, model, data_dir) + +

+ + +
+ +

Plots the loss of the model.

+ + +

Parameters:

+
    +
  • + conf + (dict) + – +
    +

    Dictionary containing the configuration

    +
    +
    • +
  • + model + (Module) + – +
    +

    The model to plot the loss of

    +
    +
    • +
  • + data_dir + (str) + – +
    +

    The directory where the data is stored

    +
    +
    • +
+ +
+ Source code in src/speckcn2/plots.py + 
136
+137
+138
+139
+140
+141
+142
+143
+144
+145
+146
+147
+148
+149
+150
+151
+152
+153
+154
+155
+156
+157
+158
+159
+160
+161
+162
+163
+164
+165
+166
def plot_loss(conf: dict, model, data_dir: str) -> None:
+    """Plots the loss of the model.
+
+    Parameters
+    ----------
+    conf : dict
+        Dictionary containing the configuration
+    model : torch.nn.Module
+        The model to plot the loss of
+    data_dir : str
+        The directory where the data is stored
+    """
+
+    model_name = conf['model']['name']
+    data_dir = conf['speckle']['datadirectory']
+
+    dirname = f'{data_dir}/{model_name}_score'
+    ensure_directory(dirname)
+
+    # plot the loss
+    fig, axs = plt.subplots(1, 1, figsize=(5, 5))
+    axs.plot(model.epoch, model.loss, label='Training loss')
+    axs.plot(model.epoch, model.val_loss, label='Validation loss')
+    axs.set_xlabel('Epoch')
+    axs.set_ylabel('Loss')
+    axs.set_yscale('log')
+    axs.legend()
+    plt.title(f'Model: {model_name}')
+    plt.tight_layout()
+    plt.savefig(f'{dirname}/loss_{model_name}.png')
+    plt.close()
+
+
+
+ +
+ +
+ + +

+ plot_param_histo(conf, test_losses, data_dir, measures) + +

+ + +
+ +

Plots the histograms of different parameters.

+ + +

Parameters:

+
    +
  • + conf + (dict) + – +
    +

    Dictionary containing the configuration

    +
    +
    • +
  • + test_losses + (list[dict]) + – +
    +

    List of all the losses of the test set

    +
    +
    • +
  • + data_dir + (str) + – +
    +

    The directory where the data is stored

    +
    +
    • +
  • + measures + (list) + – +
    +

    The measures of the model

    +
    +
    • +
+ +
+ Source code in src/speckcn2/plots.py + 
418
+419
+420
+421
+422
+423
+424
+425
+426
+427
+428
+429
+430
+431
+432
+433
+434
+435
+436
+437
+438
+439
+440
+441
+442
+443
+444
+445
+446
+447
+448
+449
+450
+451
+452
+453
+454
+455
+456
+457
+458
+459
+460
+461
+462
+463
+464
+465
+466
+467
+468
+469
+470
+471
+472
+473
+474
+475
+476
+477
+478
+479
+480
+481
def plot_param_histo(conf: dict, test_losses: list[dict], data_dir: str,
+                     measures: list) -> None:
+    """Plots the histograms of different parameters.
+
+    Parameters
+    ----------
+    conf : dict
+        Dictionary containing the configuration
+    test_losses : list[dict]
+        List of all the losses of the test set
+    data_dir : str
+        The directory where the data is stored
+    measures : list
+        The measures of the model
+    """
+    model_name = conf['model']['name']
+    data_dir = conf['speckle']['datadirectory']
+
+    dirname = f'{data_dir}/{model_name}_score'
+    ensure_directory(dirname)
+
+    for param_model, param_true, name, units in zip(
+        ['Fried_pred', 'Isoplanatic_pred', 'Scintillation_w_pred'],
+        ['Fried_true', 'Isoplanatic_true', 'Scintillation_w_true'],
+        ['Fried parameter', 'Isoplanatic angle', 'Rytov index'],
+        ['[m]', '[rad]', '[1]'],
+    ):
+        fig, axs = plt.subplots(1, 1, figsize=(5, 5))
+
+        params_model = [d[param_model].detach().cpu() for d in measures]
+        params_true = [d[param_true].detach().cpu() for d in measures]
+
+        pairs = sorted(zip(params_true, params_model))
+        params_true, params_model = zip(*pairs)
+        params_true = np.array(params_true)
+        params_model = np.array(params_model)
+
+        bins = np.logspace(np.log10(min(params_true)),
+                           np.log10(max(params_true)),
+                           num=50).tolist()
+        axs.hist(params_true,
+                 bins=bins,
+                 alpha=0.5,
+                 label=param_true,
+                 density=True)
+
+        bins = np.logspace(np.log10(min(params_model)),
+                           np.log10(max(params_model)),
+                           num=50).tolist()
+        axs.hist(params_model,
+                 bins=bins,
+                 alpha=0.5,
+                 label=param_model,
+                 density=True)
+
+        axs.set_xlabel(f'{name} {units}')
+        axs.set_xscale('log')
+        axs.set_yscale('log')
+        axs.set_ylabel('Frequency')
+        axs.legend()
+        plt.title(f'Model: {model_name}')
+        plt.tight_layout()
+        plt.savefig(f'{dirname}/histo_{param_true}_{model_name}.png')
+        plt.close()
+
+
+
+ +
+ +
+ + +

+ plot_param_vs_loss(conf, test_losses, data_dir, measures, no_sign=False, nbins=10, linear_bins=False) + +

+ + +
+ +

Plots the parameter vs the loss. Optionally, it also plots the detailed +histo for all the bins for the desired metrics.

+ + +

Parameters:

+
    +
  • + conf + (dict) + – +
    +

    Dictionary containing the configuration

    +
    +
    • +
  • + test_losses + (list[dict]) + – +
    +

    List of all the losses of the test set

    +
    +
    • +
  • + data_dir + (str) + – +
    +

    The directory where the data is stored

    +
    +
    • +
  • + measures + (list) + – +
    +

    The measures of the model

    +
    +
    • +
  • + no_sign + (bool, default: + False +) + – +
    +

    If True, it will plot the abs of the relative error

    +
    +
    • +
  • + nbins + (int, default: + 10 +) + – +
    +

    The number of bins in which to partition the data

    +
    +
    • +
  • + linear_bins + (bool, default: + False +) + – +
    +

    If True, the bins are linearly spaced, otherwise they are log spaced

    +
    +
    • +
+ +
+ Source code in src/speckcn2/plots.py + 
237
+238
+239
+240
+241
+242
+243
+244
+245
+246
+247
+248
+249
+250
+251
+252
+253
+254
+255
+256
+257
+258
+259
+260
+261
+262
+263
+264
+265
+266
+267
+268
+269
+270
+271
+272
+273
+274
+275
+276
+277
+278
+279
+280
+281
+282
+283
+284
+285
+286
+287
+288
+289
+290
+291
+292
+293
+294
+295
+296
+297
+298
+299
+300
+301
+302
+303
+304
+305
+306
+307
+308
+309
+310
+311
+312
+313
+314
+315
+316
+317
+318
+319
+320
+321
+322
+323
+324
+325
+326
+327
+328
+329
+330
+331
+332
+333
+334
+335
+336
+337
+338
+339
+340
+341
+342
+343
+344
+345
+346
+347
+348
+349
+350
+351
+352
+353
+354
+355
+356
+357
+358
+359
+360
+361
+362
+363
+364
+365
+366
+367
+368
+369
+370
+371
+372
+373
+374
+375
+376
+377
+378
+379
+380
+381
+382
+383
+384
+385
+386
+387
+388
+389
+390
+391
+392
+393
+394
+395
+396
+397
+398
+399
+400
+401
+402
+403
+404
+405
+406
+407
+408
+409
+410
+411
+412
+413
+414
+415
def plot_param_vs_loss(conf: dict,
+                       test_losses: list[dict],
+                       data_dir: str,
+                       measures: list,
+                       no_sign: bool = False,
+                       nbins: int = 10,
+                       linear_bins: bool = False) -> None:
+    """Plots the parameter vs the loss. Optionally, it also plots the detailed
+    histo for all the bins for the desired metrics.
+
+    Parameters
+    ----------
+    conf : dict
+        Dictionary containing the configuration
+    test_losses : list[dict]
+        List of all the losses of the test set
+    data_dir : str
+        The directory where the data is stored
+    measures : list
+        The measures of the model
+    no_sign : bool
+        If True, it will plot the abs of the relative error
+    nbins : int
+        The number of bins in which to partition the data
+    linear_bins : bool
+        If True, the bins are linearly spaced, otherwise they are log spaced
+    """
+    model_name = conf['model']['name']
+    data_dir = conf['speckle']['datadirectory']
+
+    dirname = f'{data_dir}/{model_name}_score'
+    ensure_directory(dirname)
+
+    for param, lname, name, units in zip(
+        ['Fried_true', 'Isoplanatic_true', 'Scintillation_w_true'],
+        ['Fried', 'Isoplanatic', 'Scintillation_w'],
+        ['Fried parameter', 'Isoplanatic angle', 'Rytov index'],
+        ['[m]', '[rad]', '[1]'],
+    ):
+
+        dirname = f'{data_dir}/{model_name}_score'
+        p_data = [d[param].detach().cpu() for d in measures]
+        if no_sign:
+            l_data = [d[lname].detach().cpu() for d in test_losses]
+        else:
+            pname = lname.split('_true')[0] + '_pred'
+            l_data = [((d[pname] - d[param]) / d[param]).detach().cpu()
+                      for d in measures]
+
+        pairs = sorted(zip(p_data, l_data))
+        params, loss = zip(*pairs)
+        params = np.array(params)
+        loss = np.array(loss)
+
+        if linear_bins:
+            bins = np.linspace(min(params), max(params), num=nbins)
+        else:
+            bins = np.logspace(np.log10(min(params)),
+                               np.log10(max(params)),
+                               num=nbins)
+        bin_indices = np.digitize(params, bins)
+        bin_means = [
+            loss[bin_indices == i].mean() if np.any(bin_indices == i) else 0
+            for i in range(1, len(bins))
+        ]
+        bin_stds = [
+            loss[bin_indices == i].std() if np.any(bin_indices == i) else 0
+            for i in range(1, len(bins))
+        ]
+        bin_centers = 0.5 * (bins[:-1] + bins[1:])
+
+        # Plotting the results
+        fig, axs = plt.subplots(1, 1, figsize=(5, 5))
+        axs.errorbar(bin_centers,
+                     bin_means,
+                     yerr=bin_stds,
+                     marker='o',
+                     linestyle='-',
+                     alpha=0.75)
+
+        # Plot error reference lines+shade
+        axs.axhline(y=1.0, linestyle='--', color='tab:red', label='100% error')
+        axs.axhline(y=0.5,
+                    linestyle='--',
+                    color='tab:orange',
+                    label='50% error')
+        axs.axhline(y=0.1,
+                    linestyle='--',
+                    color='tab:green',
+                    label='10% error')
+        axs.axhline(y=-1.0, linestyle='--', color='tab:red')
+        axs.axhline(
+            y=-0.5,
+            linestyle='--',
+            color='tab:orange',
+        )
+        axs.axhline(
+            y=-0.1,
+            linestyle='--',
+            color='tab:green',
+        )
+        axs.axhline(
+            y=0,
+            linestyle='--',
+            color='black',
+        )
+        plt.tight_layout()
+        x_min, x_max = axs.get_xlim()
+        axs.fill_between([x_min, x_max],
+                         -0.1,
+                         0.1,
+                         color='tab:green',
+                         alpha=0.1)
+        axs.fill_between([x_min, x_max],
+                         0.1,
+                         0.5,
+                         color='tab:orange',
+                         alpha=0.1)
+        axs.fill_between([x_min, x_max],
+                         -0.5,
+                         -0.1,
+                         color='tab:orange',
+                         alpha=0.1)
+        axs.fill_between([x_min, x_max], 0.5, 1.0, color='tab:red', alpha=0.1)
+        axs.fill_between([x_min, x_max],
+                         -1.0,
+                         -0.5,
+                         color='tab:red',
+                         alpha=0.1)
+        axs.set_xlabel(f'{name} {units}')
+        axs.set_xscale('log')
+        axs.set_yscale('symlog', linthresh=0.1)
+        axs.set_ylabel('Relative error')
+        yticks = [-1, -0.5, -0.1, 0, 0.1, 0.5, 1]
+        plt.yticks(yticks)
+        yticklabels = ['-100%', '-50%', '-10%', '0', '10%', '50%', '100%']
+        plt.gca().set_yticklabels(yticklabels)
+        plt.title(f'Model: {model_name}')
+        plt.tight_layout()
+        plt.savefig(f'{dirname}/{param}_vs_sum_{model_name}.png')
+        plt.close()
+
+        # If specified, plot the histogram per single bin
+        if conf['preproc'].get(lname + '_details', False):
+            print(f'\nComputing {lname} details')
+            dirname = f'{data_dir}/{model_name}_score/{lname}_bin_details'
+            ensure_directory(dirname)
+
+            for idx, single_bin in enumerate(bin_centers):
+                l_data = loss[bin_indices == idx]
+                if len(l_data) == 0:
+                    continue
+                fig, axs = plt.subplots(1, 1, figsize=(5, 5))
+                axs.hist(l_data, bins=50, alpha=0.5, density=True)
+                mu = np.mean(l_data)
+                sigma = np.std(l_data)
+                if no_sign:
+                    print(
+                        'Warning: you are requesting the analysis of absolute value using'
+                        + ' normal gaussian assumption.' +
+                        'This is not correct and the error will be overestimated.'
+                    )
+                print(
+                    f'{lname} = {single_bin:.3f} -> mu = {mu:.3f}, sigma = {sigma:.3f}'
+                )
+                if sigma > 0:
+                    x = np.linspace(mu - 3 * sigma, mu + 3 * sigma, 100)
+                    x = x[x > min(l_data)]
+                    x = x[x < max(l_data)]
+                    axs.plot(x,
+                             stats.norm.pdf(x, mu, sigma),
+                             label=f'Average err: {mu:.3f}, Std: {sigma:.3f}')
+                axs.set_xlabel(f'Relative error {lname}')
+                axs.set_ylabel('Frequency')
+                axs.legend()
+                plt.title(f'{lname} value = {single_bin:.3f} {units}')
+                plt.tight_layout()
+                plt.savefig(f'{dirname}/{lname}_bin{idx}.png')
+                plt.close()
+
+
+
+ +
+ +
+ + +

+ plot_samples_in_ensemble(conf, test_set, device, model, criterion, trimming=0.2, n_max_plots=100) + +

+ + +
+ +

Plot the prediction over a sample and compare it with the ones +from its ensemble.

+ + +

Parameters:

+
    +
  • + conf + (dict) + – +
    +

    Dictionary containing the configuration

    +
    +
    • +
  • + test_set + (list) + – +
    +

    The test set

    +
    +
    • +
  • + device + (device) + – +
    +

    The device to use

    +
    +
    • +
  • + model + (Torch) + – +
    +

    The trained model

    +
    +
    • +
  • + criterion + (ComposableLoss) + – +
    +

    The loss function

    +
    +
    • +
  • + trimming + (float, default: + 0.2 +) + – +
    +

    The trimming to use for the mean

    +
    +
    • +
  • + n_max_plots + (int, default: + 100 +) + – +
    +

    The maximum number of plots

    +
    +
    • +
+ +
+ Source code in src/speckcn2/plots.py + 
566
+567
+568
+569
+570
+571
+572
+573
+574
+575
+576
+577
+578
+579
+580
+581
+582
+583
+584
+585
+586
+587
+588
+589
+590
+591
+592
+593
+594
+595
+596
+597
+598
+599
+600
+601
+602
+603
+604
+605
+606
+607
+608
+609
+610
+611
+612
+613
+614
+615
+616
+617
+618
+619
+620
+621
+622
+623
+624
+625
+626
+627
+628
+629
+630
+631
+632
+633
+634
+635
+636
+637
+638
+639
+640
+641
+642
+643
+644
+645
+646
+647
+648
+649
+650
+651
+652
+653
+654
+655
+656
+657
+658
+659
+660
+661
+662
+663
+664
+665
+666
+667
+668
+669
+670
+671
+672
+673
+674
+675
+676
+677
+678
+679
+680
+681
+682
+683
+684
+685
+686
+687
+688
+689
+690
+691
+692
+693
+694
+695
+696
+697
+698
+699
+700
+701
+702
+703
+704
+705
+706
+707
+708
+709
+710
+711
+712
+713
+714
+715
+716
+717
+718
+719
+720
+721
+722
+723
+724
+725
+726
+727
+728
+729
+730
+731
+732
+733
+734
+735
+736
+737
+738
+739
+740
+741
+742
+743
+744
+745
+746
+747
+748
+749
+750
+751
+752
+753
+754
def plot_samples_in_ensemble(conf: dict,
+                             test_set: list,
+                             device: Device,
+                             model: nn.Torch,
+                             criterion: ComposableLoss,
+                             trimming: float = 0.2,
+                             n_max_plots: int = 100) -> None:
+    """Plot the prediction over a sample and compare it with the ones
+    from its ensemble.
+
+    Parameters
+    ----------
+    conf : dict
+        Dictionary containing the configuration
+    test_set : list
+        The test set
+    device : torch.device
+        The device to use
+    model : nn.Torch
+        The trained model
+    criterion : ComposableLoss
+        The loss function
+    trimming : float
+        The trimming to use for the mean
+    n_max_plots : int
+        The maximum number of plots
+    """
+
+    data_directory = conf['speckle']['datadirectory']
+    model_name = conf['model']['name']
+    n_screens = conf['speckle']['nscreens']
+
+    dirname = f'{data_directory}/{model_name}_score/single-shot_predictions'
+    ensure_directory(dirname)
+
+    # group the sets that have the same n[1]
+    grouped_test_set: Dict = {}
+    for n in test_set:
+        key = tuple(n[1])
+        if key not in grouped_test_set:
+            grouped_test_set[key] = []
+        grouped_test_set[key].append(n)
+    print('\nChecking if averaging speckle predictions improves results')
+    print(f'Number of samples: {len(test_set)}')
+    print(f'Number of speckle groups: {len(grouped_test_set)}')
+    # Define a random probability to plot each ensemble
+    p_plot = n_max_plots / len(grouped_test_set)
+
+    # In the end, we will plot groups that have uncommon values of loss
+    loss_min = 1e10
+    loss_max = 0
+    ensemble_count = 0
+
+    ensemble = EnsembleModel(conf, device)
+    with torch.no_grad():
+        model.eval()
+
+        for key, value in grouped_test_set.items():
+            _outputs = []
+            _all_tags_pred = []
+
+            for count, speckle in enumerate(value, 1):
+                output, target, _ = ensemble(model, [speckle])
+                _outputs.append(output.detach().cpu().numpy())
+                loss, losses = criterion(output, target)
+
+                # Get the Cn2 profile and the recovered tags
+                Cn2_pred = criterion.reconstruct_cn2(output)
+                Cn2_true = criterion.reconstruct_cn2(target)
+                recovered_tag_pred = criterion.get_J(output)
+                _all_tags_pred.append(recovered_tag_pred.detach().cpu())
+                # and get all the measures
+                all_measures = criterion._get_all_measures(
+                    output, target, Cn2_pred, Cn2_true)
+
+                if count == 1:
+                    fig, ax = plt.subplots(1, 3, figsize=(12, 4))
+                    loss_0 = loss
+
+                    # (0) Plot the speckle pattern
+                    ax[0].axis('off')  # Hide axis
+                    ax[0].imshow(speckle[0][0, :, :], cmap='bone')
+
+                    # (1) Plot J vs nscreens
+                    recovered_tag_true = criterion.get_J(target)
+                    ax[1].plot(recovered_tag_true.squeeze(0).detach().cpu(),
+                               '*',
+                               label='True',
+                               color='tab:green',
+                               markersize=10,
+                               markeredgecolor='black',
+                               zorder=100)
+                    ax[1].plot(recovered_tag_pred.squeeze(0).detach().cpu(),
+                               'o',
+                               label='This speckle',
+                               color='tab:red',
+                               markersize=7,
+                               markeredgecolor='black',
+                               zorder=90)
+
+                    # (2) Plot the parameters of this speckle prediction
+                    ax[2].axis('off')  # Hide axis
+                    recap_info = f'LOSS TERMS:\nTotal Loss: {loss.item():.4g}\n'
+                    # the individual losses
+                    for key, value in losses.items():
+                        recap_info += f'{key}: {value.item():.4g}\n'
+                    recap_info += '-------------------\nPARAMETERS:\n'
+                    # then the single parameters
+                    for key, value in all_measures.items():
+                        recap_info += f'{key}: {value:.4g}\n'
+                    ax[2].text(0.5,
+                               0.5,
+                               recap_info,
+                               horizontalalignment='center',
+                               verticalalignment='center',
+                               fontsize=10,
+                               color='black')
+
+            # Now at the end of the loop, we decide if this set needs to plotted or not
+            # by checking that the loss is uncommon, or via a random probability
+            if loss_0 > loss_max or loss_0 < loss_min or np.random.rand(
+            ) < p_plot:
+                avg_tags_trim = stats.trim_mean(_all_tags_pred,
+                                                trimming).squeeze()
+                percentiles_50 = np.percentile(_all_tags_pred, [25, 75],
+                                               axis=0).squeeze()
+                percentiles_68 = np.percentile(_all_tags_pred, [16, 84],
+                                               axis=0).squeeze()
+                percentiles_95 = np.percentile(_all_tags_pred, [2.5, 97.5],
+                                               axis=0).squeeze()
+
+                x_vals = np.arange(n_screens)
+                alp = 0.3
+                ax[1].plot(avg_tags_trim,
+                           label='Mean',
+                           color='tab:red',
+                           zorder=50)
+                ax[1].fill_between(x_vals,
+                                   percentiles_50[0],
+                                   percentiles_50[1],
+                                   color='gold',
+                                   alpha=alp,
+                                   label='50% CI',
+                                   zorder=5)
+                ax[1].fill_between(x_vals,
+                                   percentiles_68[0],
+                                   percentiles_50[0],
+                                   color='cadetblue',
+                                   alpha=alp,
+                                   label='68% CI',
+                                   zorder=4)
+                ax[1].fill_between(x_vals,
+                                   percentiles_50[1],
+                                   percentiles_68[1],
+                                   color='cadetblue',
+                                   alpha=alp,
+                                   zorder=4)
+                ax[1].fill_between(x_vals,
+                                   percentiles_95[0],
+                                   percentiles_68[0],
+                                   color='blue',
+                                   label='95% CI',
+                                   alpha=alp,
+                                   zorder=3)
+                ax[1].fill_between(x_vals,
+                                   percentiles_68[1],
+                                   percentiles_95[1],
+                                   color='blue',
+                                   alpha=alp,
+                                   zorder=3)
+
+                ax[1].set_yscale('log')
+                ax[1].set_ylabel('J')
+                ax[1].set_xlabel('# screen')
+                ax[1].legend()
+                fig.tight_layout()
+                plt.subplots_adjust(top=0.92)
+                plt.suptitle(
+                    'Prediction from a single speckle, compared to similar')
+                plt.savefig(
+                    f'{dirname}/single_speckle_loss{loss_0.item():.4g}.png')
+                loss_max = max(loss_0, loss_max)
+                loss_min = min(loss_0, loss_min)
+                ensemble_count += 1
+
+            plt.close()
+
+            if ensemble_count >= n_max_plots:
+                break
+
+
+
+ +
+ +
+ + +

+ plot_time(conf, model, data_dir) + +

+ + +
+ +

Plots the time per epoch of the model.

+ + +

Parameters:

+
    +
  • + conf + (dict) + – +
    +

    Dictionary containing the configuration

    +
    +
    • +
  • + model + (Module) + – +
    +

    The model to plot the loss of

    +
    +
    • +
  • + data_dir + (str) + – +
    +

    The directory where the data is stored

    +
    +
    • +
+ +
+ Source code in src/speckcn2/plots.py + 
169
+170
+171
+172
+173
+174
+175
+176
+177
+178
+179
+180
+181
+182
+183
+184
+185
+186
+187
+188
+189
+190
+191
+192
+193
+194
+195
+196
+197
def plot_time(conf: dict, model, data_dir: str) -> None:
+    """Plots the time per epoch of the model.
+
+    Parameters
+    ----------
+    conf : dict
+        Dictionary containing the configuration
+    model : torch.nn.Module
+        The model to plot the loss of
+    data_dir : str
+        The directory where the data is stored
+    """
+
+    model_name = conf['model']['name']
+    data_dir = conf['speckle']['datadirectory']
+
+    dirname = f'{data_dir}/{model_name}_score'
+    ensure_directory(dirname)
+
+    # plot the loss
+    fig, axs = plt.subplots(1, 1, figsize=(5, 5))
+    axs.plot(model.epoch, model.time, label='Time per epoch')
+    axs.set_xlabel('Epoch')
+    axs.set_ylabel('Time [s]')
+    axs.legend()
+    plt.title(f'Model: {model_name}')
+    plt.tight_layout()
+    plt.savefig(f'{dirname}/time_{model_name}.png')
+    plt.close()
+
+
+
+ +
+ +
+ + +

+ score_plot(conf, inputs, tags, loss, losses, i, counter, measures, Cn2_pred, Cn2_true, recovered_tag_pred, recovered_tag_true) + +

+ + +
+ +

Plots side by side: +- [0:Nensemble] the input images (single or ensemble) +- [-3] the predicted/exact tags J +- [-2] the Cn2 profile +- [-1] the different information of the loss +normalize value in model units.

+ + +

Parameters:

+
    +
  • + conf + (dict) + – +
    +

    Dictionary containing the configuration

    +
    +
    • +
  • + inputs + (Tensor) + – +
    +

    The input speckle patterns

    +
    +
    • +
  • + tags + (list) + – +
    +

    The exact tags of the data

    +
    +
    • +
  • + loss + (Tensor) + – +
    +

    The total loss of the model (for this prediction)

    +
    +
    • +
  • + losses + (dict) + – +
    +

    The individual losses of the model

    +
    +
    • +
  • + i + (int) + – +
    +

    The batch index of the image

    +
    +
    • +
  • + counter + (int) + – +
    +

    The global index of the image

    +
    +
    • +
  • + measures + (dict) + – +
    +

    The different measures of the model

    +
    +
    • +
  • + Cn2_pred + (Tensor) + – +
    +

    The predicted Cn2 profile

    +
    +
    • +
  • + Cn2_true + (Tensor) + – +
    +

    The true Cn2 profile

    +
    +
    • +
  • + recovered_tag_pred + (Tensor) + – +
    +

    The predicted tags

    +
    +
    • +
  • + recovered_tag_true + (Tensor) + – +
    +

    The true tags

    +
    +
    • +
+ +
+ Source code in src/speckcn2/plots.py + 
 17
+ 18
+ 19
+ 20
+ 21
+ 22
+ 23
+ 24
+ 25
+ 26
+ 27
+ 28
+ 29
+ 30
+ 31
+ 32
+ 33
+ 34
+ 35
+ 36
+ 37
+ 38
+ 39
+ 40
+ 41
+ 42
+ 43
+ 44
+ 45
+ 46
+ 47
+ 48
+ 49
+ 50
+ 51
+ 52
+ 53
+ 54
+ 55
+ 56
+ 57
+ 58
+ 59
+ 60
+ 61
+ 62
+ 63
+ 64
+ 65
+ 66
+ 67
+ 68
+ 69
+ 70
+ 71
+ 72
+ 73
+ 74
+ 75
+ 76
+ 77
+ 78
+ 79
+ 80
+ 81
+ 82
+ 83
+ 84
+ 85
+ 86
+ 87
+ 88
+ 89
+ 90
+ 91
+ 92
+ 93
+ 94
+ 95
+ 96
+ 97
+ 98
+ 99
+100
+101
+102
+103
+104
+105
+106
+107
+108
+109
+110
+111
+112
+113
+114
+115
+116
+117
+118
+119
+120
+121
+122
+123
+124
+125
+126
+127
+128
+129
+130
+131
+132
+133
def score_plot(
+    conf: dict,
+    inputs: torch.Tensor,
+    tags: list,
+    loss: torch.Tensor,
+    losses: dict,
+    i: int,
+    counter: int,
+    measures: dict,
+    Cn2_pred: torch.Tensor,
+    Cn2_true: torch.Tensor,
+    recovered_tag_pred: torch.Tensor,
+    recovered_tag_true: torch.Tensor,
+) -> None:
+    """Plots side by side:
+    - [0:Nensemble] the input images (single or ensemble)
+    - [-3] the predicted/exact tags J
+    - [-2] the Cn2 profile
+    - [-1] the different information of the loss
+    normalize value in model units.
+
+    Parameters
+    ----------
+    conf : dict
+        Dictionary containing the configuration
+    inputs : torch.Tensor
+        The input speckle patterns
+    tags : list
+        The exact tags of the data
+    loss : torch.Tensor
+        The total loss of the model (for this prediction)
+    losses : dict
+        The individual losses of the model
+    i : int
+        The batch index of the image
+    counter : int
+        The global index of the image
+    measures : dict
+        The different measures of the model
+    Cn2_pred : torch.Tensor
+        The predicted Cn2 profile
+    Cn2_true : torch.Tensor
+        The true Cn2 profile
+    recovered_tag_pred : torch.Tensor
+        The predicted tags
+    recovered_tag_true : torch.Tensor
+        The true tags
+    """
+    model_name = conf['model']['name']
+    data_dir = conf['speckle']['datadirectory']
+    ensemble = conf['preproc'].get('ensemble', 1)
+    hs = conf['speckle']['splits']
+    nscreens = conf['speckle']['nscreens']
+    if len(hs) != nscreens:
+        print(
+            'WARNING: The number of screens does not match the number of splits'
+        )
+        return
+
+    dirname = f'{data_dir}/{model_name}_score/single-shot_predictions'
+    ensure_directory(dirname)
+
+    fig, axs = plt.subplots(1, 3 + ensemble, figsize=(4 * (2 + ensemble), 3.5))
+
+    # (1) Plot the input images
+    for n in range(ensemble):
+        img = inputs[ensemble * i + n].detach().cpu().squeeze().abs()
+        axs[n].imshow(img, cmap='bone')
+    title_string = f'Input {ensemble} images' if ensemble > 1 else 'Input single image'
+    axs[1].set_title(title_string)
+
+    # (2) Plot J vs nscreens
+    axs[-3].plot(recovered_tag_true.squeeze(0).detach().cpu(),
+                 'o',
+                 label='True')
+    axs[-3].plot(recovered_tag_pred.squeeze(0).detach().cpu(),
+                 '.',
+                 color='tab:red',
+                 label='Predicted')
+    axs[-3].set_yscale('log')
+    axs[-3].set_ylabel('J')
+    axs[-3].set_xlabel('# screen')
+    axs[-3].legend()
+
+    # (3) Plot Cn2 vs altitude
+    axs[-2].plot(hs, Cn2_true.squeeze(0).detach().cpu(), 'o', label='True')
+    axs[-2].plot(hs,
+                 Cn2_pred.squeeze(0).detach().cpu(),
+                 '.',
+                 color='tab:red',
+                 label='Predicted')
+    axs[-2].set_xscale('log')
+    axs[-2].set_yscale('log')
+    axs[-2].set_ylabel(r'$Cn^2$')
+    axs[-2].set_xlabel('Altitude [m]')
+
+    # (4) Plot the recap information
+    axs[-1].axis('off')  # Hide axis
+    recap_info = f'LOSS TERMS:\nTotal Loss: {loss.item():.4g}\n'
+    # the individual losses
+    for key, value in losses.items():
+        recap_info += f'{key}: {value.item():.4g}\n'
+    recap_info += '-------------------\nPARAMETERS:\n'
+    # then the single parameters
+    for key, value in measures.items():
+        recap_info += f'{key}: {value:.4g}\n'
+    axs[-1].text(0.5,
+                 0.5,
+                 recap_info,
+                 horizontalalignment='center',
+                 verticalalignment='center',
+                 fontsize=10,
+                 color='black')
+
+    plt.tight_layout()
+    plt.savefig(f'{dirname}/single_speckle_loss{loss.item():.4g}.png')
+    plt.close()
+
+
+
+ +
+ + + +
+ +
+ +
+ + + + + + + + + + + + + +
+
+ + + +
+ +
+ + + +
+
+
+
+ + + + + + + + + + + + + + \ No newline at end of file diff --git a/api/postprocess/index.html b/api/postprocess/index.html new file mode 100644 index 0000000..4114ab4 --- /dev/null +++ b/api/postprocess/index.html @@ -0,0 +1,2428 @@ + + + + + + + + + + + + + + + + + + + + + + + + + Postprocess - Speckcn2 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
+ + + + Skip to content + + +
+
+ +
+ + + + + + +
+ + +
+ +
+ + + + + + +
+
+ + + +
+
+
+ + + + + +
+
+
+ + + +
+
+
+ + + +
+
+
+ + + +
+
+ + + + + + + +

Postprocess

+ +
+ + + + +
+ + + + + + + + +
+ + + + + + + + + +
+ + +

+ average_speckle_input(conf, test_set, device, model, criterion, n_ensembles_to_plot=100) + +

+ + +
+ +

Test to see if averaging the speckle patterns (before the prediction) +improves the results. This function is then going to plot the relative +error over the screen tags and the Fried parameter to make this evaluation.

+ + +

Parameters:

+
    +
  • + conf + (dict) + – +
    +

    Dictionary containing the configuration

    +
    +
    • +
  • + test_set + (list) + – +
    +

    The test set

    +
    +
    • +
  • + device + (device) + – +
    +

    The device to use

    +
    +
    • +
  • + model + (Torch) + – +
    +

    The trained model

    +
    +
    • +
  • + criterion + (ComposableLoss) + – +
    +

    The loss function

    +
    +
    • +
  • + n_ensembles_to_plot + (int, default: + 100 +) + – +
    +

    The number of ensembles to plot

    +
    +
    • +
+ +
+ Source code in src/speckcn2/postprocess.py + 
355
+356
+357
+358
+359
+360
+361
+362
+363
+364
+365
+366
+367
+368
+369
+370
+371
+372
+373
+374
+375
+376
+377
+378
+379
+380
+381
+382
+383
+384
+385
+386
+387
+388
+389
+390
+391
+392
+393
+394
+395
+396
+397
+398
+399
+400
+401
+402
+403
+404
+405
+406
+407
+408
+409
+410
+411
+412
+413
+414
+415
+416
+417
+418
+419
+420
+421
+422
+423
+424
+425
+426
+427
+428
+429
+430
+431
+432
+433
+434
+435
+436
+437
+438
+439
+440
+441
+442
+443
+444
+445
+446
+447
+448
+449
+450
+451
+452
+453
+454
+455
+456
+457
+458
+459
+460
+461
+462
+463
+464
+465
+466
+467
+468
+469
+470
+471
+472
+473
+474
+475
+476
+477
+478
+479
+480
+481
+482
+483
+484
def average_speckle_input(conf: dict,
+                          test_set: list,
+                          device: Device,
+                          model: nn.Torch,
+                          criterion: ComposableLoss,
+                          n_ensembles_to_plot: int = 100) -> None:
+    """Test to see if averaging the speckle patterns (before the prediction)
+    improves the results. This function is then going to plot the relative
+    error over the screen tags and the Fried parameter to make this evaluation.
+
+    Parameters
+    ----------
+    conf : dict
+        Dictionary containing the configuration
+    test_set : list
+        The test set
+    device : torch.device
+        The device to use
+    model : nn.Torch
+        The trained model
+    criterion : ComposableLoss
+        The loss function
+    n_ensembles_to_plot : int
+        The number of ensembles to plot
+    """
+
+    data_directory = conf['speckle']['datadirectory']
+    model_name = conf['model']['name']
+
+    dirname = f'{data_directory}/{model_name}_score/effect_averaging'
+    ensure_directory(dirname)
+
+    # group the sets that have the same n[1]
+    grouped_test_set: Dict = {}
+    for n in test_set:
+        key = tuple(n[1])
+        if key not in grouped_test_set:
+            grouped_test_set[key] = []
+        grouped_test_set[key].append(n)
+    print('\nChecking if averaging speckle patterns improves results')
+    print(f'Number of samples: {len(test_set)}')
+    print(f'Number of speckle groups: {len(grouped_test_set)}')
+
+    # For each group compare the model prediction to the exact tag
+    ensemble = EnsembleModel(conf, device)
+    with torch.no_grad():
+        model.eval()
+
+        for ensemble_count, (key,
+                             value) in enumerate(grouped_test_set.items()):
+            avg_speckle = None
+            cmap = plt.get_cmap('coolwarm')
+            norm = plt.Normalize(1, len(value))
+
+            if ensemble_count > n_ensembles_to_plot:
+                continue
+
+            for count, speckle in enumerate(value, 1):
+                color = cmap(norm(count))
+
+                if avg_speckle is None:
+                    avg_speckle = speckle
+                else:
+                    avg_speckle = (torch.add(avg_speckle[0],
+                                             speckle[0]), *avg_speckle[1:])
+
+                # Average only the speckle pattern (first element)
+                avg_speckle_divided = (torch.div(avg_speckle[0],
+                                                 count), *avg_speckle[1:])
+
+                output, target, _ = ensemble(model, [avg_speckle_divided])
+                loss, losses = criterion(output, target)
+
+                if count == 1:
+                    fig, ax = plt.subplots(1, 3, figsize=(12, 4))
+
+                    recovered_tag_true = criterion.get_J(target)
+                    ax[0].plot(recovered_tag_true.squeeze(0).detach().cpu(),
+                               '*',
+                               label='True',
+                               color='tab:green',
+                               zorder=100)
+                    blue_patch = mpatches.Patch(color=color,
+                                                label='One speckle')
+
+                ax[1].plot((torch.abs(output - target) /
+                            (target + 1e-7)).flatten().detach().cpu(),
+                           color=color)
+
+                # Get the Cn2 profile and the recovered tags
+                Cn2_pred = criterion.reconstruct_cn2(output)
+                Cn2_true = criterion.reconstruct_cn2(target)
+                recovered_tag_pred = criterion.get_J(output / count)
+                ax[0].plot(recovered_tag_pred.squeeze(0).detach().cpu(),
+                           '.',
+                           color=color)
+                # and get all the measures
+                all_measures = criterion._get_all_measures(
+                    output, target, Cn2_pred, Cn2_true)
+
+                Fried_err = torch.abs(
+                    all_measures['Fried_true'] -
+                    all_measures['Fried_pred']) / all_measures['Fried_true']
+                ax[2].scatter(count, Fried_err.detach().cpu(), color=color)
+                if count == 1:
+                    ax[2].plot(
+                        [], [],
+                        ' ',
+                        label='(True) Fried = {:.3f}'.format(
+                            all_measures['Fried_true'].detach().cpu().numpy()))
+
+            ax[0].set_yscale('log')
+            ax[0].set_ylabel('J')
+            ax[0].set_xlabel('# screen')
+            red_patch = mpatches.Patch(color=color, label='All speckles')
+            handles, labels = ax[0].get_legend_handles_labels()
+            handles.extend([blue_patch, red_patch])
+            ax[0].legend(handles=handles)
+            ax[1].set_xlabel('# screen')
+            ax[1].set_ylabel('J relative error ')
+            ax[2].set_xlabel('# averaged speckles')
+            ax[2].set_ylabel('Fried relative error')
+            ax[1].set_yscale('log')
+            ax[2].set_yscale('log')
+            ax[2].legend(frameon=False)
+            fig.tight_layout()
+            plt.subplots_adjust(top=0.92)
+            plt.suptitle('Effect of averaging speckle patterns')
+            plt.savefig(f'{dirname}/average_speckle_loss{loss.item():.4g}.png')
+            plt.close()
+
+
+
+ +
+ +
+ + +

+ average_speckle_output(conf, test_set, device, model, criterion, trimming=0.1, n_ensembles_to_plot=100) + +

+ + +
+ +

Test to see if averaging the prediction of multiple speckle patterns +improves the results. This function is then going to plot the relative +error over the screen tags and the Fried parameter to make this evaluation.

+ + +

Parameters:

+
    +
  • + conf + (dict) + – +
    +

    Dictionary containing the configuration

    +
    +
    • +
  • + test_set + (list) + – +
    +

    The test set

    +
    +
    • +
  • + device + (device) + – +
    +

    The device to use

    +
    +
    • +
  • + model + (Torch) + – +
    +

    The trained model

    +
    +
    • +
  • + criterion + (ComposableLoss) + – +
    +

    The loss function

    +
    +
    • +
  • + n_ensembles_to_plot + (int, default: + 100 +) + – +
    +

    The number of ensembles to plot

    +
    +
    • +
+ +
+ Source code in src/speckcn2/postprocess.py + 
139
+140
+141
+142
+143
+144
+145
+146
+147
+148
+149
+150
+151
+152
+153
+154
+155
+156
+157
+158
+159
+160
+161
+162
+163
+164
+165
+166
+167
+168
+169
+170
+171
+172
+173
+174
+175
+176
+177
+178
+179
+180
+181
+182
+183
+184
+185
+186
+187
+188
+189
+190
+191
+192
+193
+194
+195
+196
+197
+198
+199
+200
+201
+202
+203
+204
+205
+206
+207
+208
+209
+210
+211
+212
+213
+214
+215
+216
+217
+218
+219
+220
+221
+222
+223
+224
+225
+226
+227
+228
+229
+230
+231
+232
+233
+234
+235
+236
+237
+238
+239
+240
+241
+242
+243
+244
+245
+246
+247
+248
+249
+250
+251
+252
+253
+254
+255
+256
+257
+258
+259
+260
+261
+262
+263
+264
+265
+266
+267
+268
+269
+270
+271
+272
+273
+274
+275
+276
+277
+278
+279
+280
+281
+282
+283
+284
+285
+286
+287
+288
+289
+290
+291
+292
+293
+294
+295
+296
+297
+298
+299
+300
+301
+302
+303
+304
+305
+306
+307
+308
+309
+310
+311
+312
+313
+314
+315
+316
+317
+318
+319
+320
+321
+322
+323
+324
+325
+326
+327
+328
+329
+330
+331
+332
+333
+334
+335
+336
+337
+338
+339
+340
+341
+342
+343
+344
+345
+346
+347
+348
+349
+350
+351
+352
def average_speckle_output(conf: dict,
+                           test_set: list,
+                           device: Device,
+                           model: nn.Torch,
+                           criterion: ComposableLoss,
+                           trimming: float = 0.1,
+                           n_ensembles_to_plot: int = 100) -> None:
+    """Test to see if averaging the prediction of multiple speckle patterns
+    improves the results. This function is then going to plot the relative
+    error over the screen tags and the Fried parameter to make this evaluation.
+
+    Parameters
+    ----------
+    conf : dict
+        Dictionary containing the configuration
+    test_set : list
+        The test set
+    device : torch.device
+        The device to use
+    model : nn.Torch
+        The trained model
+    criterion : ComposableLoss
+        The loss function
+    n_ensembles_to_plot : int
+        The number of ensembles to plot
+    """
+
+    data_directory = conf['speckle']['datadirectory']
+    model_name = conf['model']['name']
+    n_screens = conf['speckle']['nscreens']
+
+    dirname = f'{data_directory}/{model_name}_score/effect_averaging'
+    ensure_directory(dirname)
+
+    # group the sets that have the same n[1]
+    grouped_test_set: Dict = {}
+    for n in test_set:
+        key = tuple(n[1])
+        if key not in grouped_test_set:
+            grouped_test_set[key] = []
+        grouped_test_set[key].append(n)
+    print('\nChecking if averaging speckle predictions improves results')
+    print(f'Number of samples: {len(test_set)}')
+    print(f'Number of speckle groups: {len(grouped_test_set)}')
+
+    # In the end, we will plot groups that have uncommon values of loss
+    loss_min = 1e10
+    loss_max = 0
+    ensemble_count = 0
+
+    ensemble = EnsembleModel(conf, device)
+    with torch.no_grad():
+        model.eval()
+
+        for key, value in grouped_test_set.items():
+            _outputs = []
+            _all_tags_pred = []
+            cmap = plt.get_cmap('coolwarm')
+            norm = plt.Normalize(1, len(value))
+
+            for count, speckle in enumerate(value, 1):
+                color = cmap(norm(count))
+                output, target, _ = ensemble(model, [speckle])
+
+                if count == 1:
+                    fig, ax = plt.subplots(1, 4, figsize=(16, 4))
+
+                    # (1) Plot J vs nscreens
+                    recovered_tag_true = criterion.get_J(target)
+                    ax[0].plot(recovered_tag_true.squeeze(0).detach().cpu(),
+                               '*',
+                               label='True',
+                               color='tab:green',
+                               markersize=10,
+                               markeredgecolor='black',
+                               zorder=100)
+                    ax[1].plot(recovered_tag_true.squeeze(0).detach().cpu(),
+                               '*',
+                               label='True',
+                               color='tab:green',
+                               markersize=10,
+                               markeredgecolor='black',
+                               zorder=100)
+                    blue_patch = mpatches.Patch(color=color,
+                                                label='One speckle')
+
+                # Use the trimmed mean to get the average output
+                # (trim_mean works only on cpu so you have to move back and forth)
+                _outputs.append(output.detach().cpu().numpy())
+                avg_output = torch.tensor(stats.trim_mean(_outputs,
+                                                          trimming)).to(device)
+
+                loss, losses = criterion(avg_output, target)
+
+                # Get the Cn2 profile and the recovered tags
+                Cn2_pred = criterion.reconstruct_cn2(avg_output)
+                Cn2_true = criterion.reconstruct_cn2(target)
+                recovered_tag_pred = criterion.get_J(avg_output)
+                _all_tags_pred.append(recovered_tag_pred.detach().cpu())
+                ax[1].plot(recovered_tag_pred.squeeze(0).detach().cpu(),
+                           'o',
+                           color=color)
+                # and get all the measures
+                all_measures = criterion._get_all_measures(
+                    avg_output, target, Cn2_pred, Cn2_true)
+
+                Fried_err = torch.abs(
+                    all_measures['Fried_true'] -
+                    all_measures['Fried_pred']) / all_measures['Fried_true']
+                ax[2].scatter(count, Fried_err.detach().cpu(), color=color)
+                if count == 1:
+                    ax[2].plot(
+                        [], [],
+                        ' ',
+                        label='(True) Fried = {:.3f}'.format(
+                            all_measures['Fried_true'].detach().cpu().numpy()))
+
+            # Now at the end of the loop, we decide if this set needs to plotted or not
+            # by checking that the loss
+            if loss > loss_max or loss < loss_min:
+                avg_tags_trim = stats.trim_mean(_all_tags_pred,
+                                                trimming).squeeze()
+                percentiles_50 = np.percentile(_all_tags_pred, [25, 75],
+                                               axis=0).squeeze()
+                percentiles_68 = np.percentile(_all_tags_pred, [16, 84],
+                                               axis=0).squeeze()
+                percentiles_95 = np.percentile(_all_tags_pred, [2.5, 97.5],
+                                               axis=0).squeeze()
+
+                x_vals = np.arange(n_screens)
+                alp = 1
+                ax[0].plot(avg_tags_trim,
+                           label='Mean',
+                           color='tab:red',
+                           zorder=50)
+                ax[0].fill_between(x_vals,
+                                   percentiles_50[0],
+                                   percentiles_50[1],
+                                   color='gold',
+                                   alpha=alp,
+                                   label='50% CI',
+                                   zorder=5)
+                ax[0].fill_between(x_vals,
+                                   percentiles_68[0],
+                                   percentiles_50[0],
+                                   color='cadetblue',
+                                   alpha=alp,
+                                   label='68% CI',
+                                   zorder=4)
+                ax[0].fill_between(x_vals,
+                                   percentiles_50[1],
+                                   percentiles_68[1],
+                                   color='cadetblue',
+                                   alpha=alp,
+                                   zorder=4)
+                ax[0].fill_between(x_vals,
+                                   percentiles_95[0],
+                                   percentiles_68[0],
+                                   color='blue',
+                                   label='95% CI',
+                                   alpha=alp,
+                                   zorder=3)
+                ax[0].fill_between(x_vals,
+                                   percentiles_68[1],
+                                   percentiles_95[1],
+                                   color='blue',
+                                   alpha=alp,
+                                   zorder=3)
+
+                ax[3].axis('off')  # Hide axis
+                recap_info = f'LOSS TERMS:\nTotal Loss: {loss.item():.4g}\n'
+                # the individual losses
+                for key, value in losses.items():
+                    recap_info += f'{key}: {value.item():.4g}\n'
+                recap_info += '-------------------\nPARAMETERS:\n'
+                # then the single parameters
+                for key, value in all_measures.items():
+                    recap_info += f'{key}: {value:.4g}\n'
+                ax[3].text(0.5,
+                           0.5,
+                           recap_info,
+                           horizontalalignment='center',
+                           verticalalignment='center',
+                           fontsize=10,
+                           color='black')
+
+                ax[0].set_yscale('log')
+                ax[0].set_ylabel('J')
+                ax[0].set_xlabel('# screen')
+                ax[0].legend()
+                ax[1].set_yscale('log')
+                ax[1].set_ylabel('J')
+                ax[1].set_xlabel('# screen')
+                red_patch = mpatches.Patch(color=color, label='All speckles')
+                handles, labels = ax[1].get_legend_handles_labels()
+                handles.extend([blue_patch, red_patch])
+                ax[1].legend(handles=handles)
+                ax[2].set_xlabel('# averaged speckles')
+                ax[2].set_ylabel('Fried relative error')
+                ax[2].legend(frameon=False)
+                ax[2].set_yscale('log')
+                fig.tight_layout()
+                plt.subplots_adjust(top=0.92)
+                plt.suptitle('Effect of averaging speckle predictions')
+                plt.savefig(
+                    f'{dirname}/average_ensemble_loss{loss.item():.4g}.png')
+                loss_max = max(loss, loss_max)
+                loss_min = min(loss, loss_min)
+                ensemble_count += 1
+
+            plt.close()
+
+            if ensemble_count > n_ensembles_to_plot:
+                break
+
+
+
+ +
+ +
+ + +

+ screen_errors(conf, device, J_pred, J_true, nbins=20, trimming=0.1) + +

+ + +
+ +

Plot the relative error of Cn2 for each screen.

+ + +

Parameters:

+
    +
  • + conf + (dict) + – +
    +

    Dictionary containing the configuration

    +
    +
    • +
  • + device + (device) + – +
    +

    The device on which the data are stored

    +
    +
    • +
  • + J_pred + (Tensor) + – +
    +

    The predicted J profile

    +
    +
    • +
  • + J_true + (Tensor) + – +
    +

    The true J profile

    +
    +
    • +
  • + nbins + (int, default: + 20 +) + – +
    +

    Number of bins to use for the histograms

    +
    +
    • +
  • + trimming + (float, default: + 0.1 +) + – +
    +

    The fraction of data to trim from each end of the distribution

    +
    +
    • +
+ +
+ Source code in src/speckcn2/postprocess.py + 
487
+488
+489
+490
+491
+492
+493
+494
+495
+496
+497
+498
+499
+500
+501
+502
+503
+504
+505
+506
+507
+508
+509
+510
+511
+512
+513
+514
+515
+516
+517
+518
+519
+520
+521
+522
+523
+524
+525
+526
+527
+528
+529
+530
+531
+532
+533
+534
+535
+536
+537
+538
+539
+540
+541
+542
+543
+544
+545
+546
+547
+548
+549
+550
+551
+552
+553
+554
+555
+556
+557
+558
+559
+560
+561
+562
+563
+564
+565
+566
+567
+568
+569
+570
+571
+572
+573
+574
+575
+576
+577
+578
+579
+580
+581
+582
+583
+584
+585
+586
+587
+588
+589
+590
+591
+592
+593
+594
+595
+596
+597
+598
+599
+600
+601
+602
+603
+604
+605
+606
+607
+608
+609
+610
+611
+612
+613
+614
+615
+616
+617
+618
+619
+620
def screen_errors(conf: dict,
+                  device: Device,
+                  J_pred: Tensor,
+                  J_true: Tensor,
+                  nbins: int = 20,
+                  trimming: float = 0.1) -> None:
+    """Plot the relative error of Cn2 for each screen.
+
+    Parameters
+    ----------
+    conf : dict
+        Dictionary containing the configuration
+    device : torch.device
+        The device on which the data are stored
+    J_pred : torch.Tensor
+        The predicted J profile
+    J_true : torch.Tensor
+        The true J profile
+    nbins : int, optional
+        Number of bins to use for the histograms
+    trimming : float, optional
+        The fraction of data to trim from each end of the distribution
+    """
+    data_directory = conf['speckle']['datadirectory']
+    model_name = conf['model']['name']
+    n_screens = conf['speckle']['nscreens']
+
+    dirname = f'{data_directory}/{model_name}_score'
+    ensure_directory(dirname)
+
+    # Plot the distribution of each tag element
+    fig, axs = plt.subplots(2, 4, figsize=(20, 10))
+    for si in range(n_screens):
+
+        if device == 'cpu':
+            data_pred = np.asarray(J_pred)[:, 0, si]
+            data_true = np.asarray(J_true)[:, 0, si]
+        else:
+            data_pred = np.asarray([d.detach().cpu().numpy()
+                                    for d in J_pred])[:, 0, si]
+            data_true = np.asarray([d.detach().cpu().numpy()
+                                    for d in J_true])[:, 0, si]
+        # Define the bins
+        bins = np.linspace(min(data_true), max(data_true), nbins + 1)
+
+        # Digitize the data to get bin indices
+        bin_indices = np.digitize(data_true, bins)
+
+        # Compute the relative error for each bin
+        relative_errors = []
+        percentiles_50 = []
+        percentiles_68 = []
+        percentiles_95 = []
+
+        for bi in range(1, len(bins)):
+            model_bin_values = data_pred[bin_indices == bi]
+            true_bin_values = data_true[bin_indices == bi]
+
+            if len(true_bin_values) > 0:
+                relative_error = np.abs(
+                    (model_bin_values - true_bin_values) / true_bin_values)
+                relative_errors.append(
+                    stats.trim_mean(relative_error, trimming))
+                percentiles_50.append(
+                    np.percentile(relative_error, [25, 75], axis=0))
+                percentiles_68.append(
+                    np.percentile(relative_error, [16, 84], axis=0))
+                percentiles_95.append(
+                    np.percentile(relative_error, [2.5, 97.5], axis=0))
+            else:
+                relative_errors.append(0)
+                percentiles_50.append([0, 0])
+                percentiles_68.append([0, 0])
+                percentiles_95.append([0, 0])
+
+        percentiles_50 = np.asarray(percentiles_50)
+        percentiles_68 = np.asarray(percentiles_68)
+        percentiles_95 = np.asarray(percentiles_95)
+        # Plot the relative errors
+        axs[si // 4, si % 4].plot(bins[:-1],
+                                  relative_errors,
+                                  label='Mean',
+                                  color='tab:red',
+                                  zorder=50)
+        axs[si // 4, si % 4].set_title(f'Screen {si}')
+        # and the percentiles
+        axs[si // 4, si % 4].fill_between(bins[:-1],
+                                          percentiles_50[:, 0],
+                                          percentiles_50[:, 1],
+                                          color='gold',
+                                          alpha=0.5,
+                                          label='50% CI',
+                                          zorder=5)
+        axs[si // 4, si % 4].fill_between(bins[:-1],
+                                          percentiles_68[:, 0],
+                                          percentiles_50[:, 0],
+                                          color='cadetblue',
+                                          alpha=0.5,
+                                          label='68% CI',
+                                          zorder=4)
+        axs[si // 4, si % 4].fill_between(bins[:-1],
+                                          percentiles_50[:, 1],
+                                          percentiles_68[:, 1],
+                                          color='cadetblue',
+                                          alpha=0.5,
+                                          zorder=4)
+        axs[si // 4, si % 4].fill_between(bins[:-1],
+                                          percentiles_95[:, 0],
+                                          percentiles_68[:, 0],
+                                          color='blue',
+                                          label='95% CI',
+                                          alpha=0.5,
+                                          zorder=3)
+        axs[si // 4, si % 4].fill_between(bins[:-1],
+                                          percentiles_68[:, 1],
+                                          percentiles_95[:, 1],
+                                          color='blue',
+                                          alpha=0.5,
+                                          zorder=3)
+        axs[si // 4, si % 4].set_yscale('symlog', linthresh=0.1)
+    axs[0, 1].legend()
+    for ax in axs.flatten():
+        ax.axhline(
+            y=0,
+            linestyle='--',
+            color='black',
+        )
+    for i in range(1, n_screens - 1):
+        axs.flat[i].sharey(axs.flat[0])
+    plt.suptitle('Relative Error of J')
+    plt.tight_layout()
+    figname = f'{dirname}/{model_name}_Jerrors'
+    plt.savefig(f'{figname}.png')
+    plt.close()
+
+
+
+ +
+ +
+ + +

+ tags_distribution(conf, train_set, test_tags, device, nbins=20, rescale=False, recover_tag=None) + +

+ + +
+ +

Function to plot the following: +- distribution of the tags for unscaled results +- distribution of the tags for rescaled results +- distribution of the sum of the tags.

+ + +

Parameters:

+
    +
  • + conf + (dict) + – +
    +

    Dictionary containing the configuration

    +
    +
    • +
  • + train_set + (list) + – +
    +

    The training set

    +
    +
    • +
  • + test_tags + (Tensor) + – +
    +

    The predicted tags for the test dataset

    +
    +
    • +
  • + device + (device) + – +
    +

    The device to use

    +
    +
    • +
  • + data_directory + (str) + – +
    +

    The directory where the data is stored

    +
    +
    • +
  • + nbins + (int, default: + 20 +) + – +
    +

    Number of bins to use for the histograms

    +
    +
    • +
  • + rescale + (bool, default: + False +) + – +
    +

    Whether to rescale the tags using recover_tag() or leave them between 0 and 1

    +
    +
    • +
  • + recover_tag + (list, default: + None +) + – +
    +

    List of functions to recover each tag

    +
    +
    • +
+ +
+ Source code in src/speckcn2/postprocess.py + 
 19
+ 20
+ 21
+ 22
+ 23
+ 24
+ 25
+ 26
+ 27
+ 28
+ 29
+ 30
+ 31
+ 32
+ 33
+ 34
+ 35
+ 36
+ 37
+ 38
+ 39
+ 40
+ 41
+ 42
+ 43
+ 44
+ 45
+ 46
+ 47
+ 48
+ 49
+ 50
+ 51
+ 52
+ 53
+ 54
+ 55
+ 56
+ 57
+ 58
+ 59
+ 60
+ 61
+ 62
+ 63
+ 64
+ 65
+ 66
+ 67
+ 68
+ 69
+ 70
+ 71
+ 72
+ 73
+ 74
+ 75
+ 76
+ 77
+ 78
+ 79
+ 80
+ 81
+ 82
+ 83
+ 84
+ 85
+ 86
+ 87
+ 88
+ 89
+ 90
+ 91
+ 92
+ 93
+ 94
+ 95
+ 96
+ 97
+ 98
+ 99
+100
+101
+102
+103
+104
+105
+106
+107
+108
+109
+110
+111
+112
+113
+114
+115
+116
+117
+118
+119
+120
+121
+122
+123
+124
+125
+126
+127
+128
+129
+130
+131
+132
+133
+134
+135
+136
def tags_distribution(conf: dict,
+                      train_set: list,
+                      test_tags: Tensor,
+                      device: Device,
+                      nbins: int = 20,
+                      rescale: bool = False,
+                      recover_tag: Optional[list[Callable]] = None) -> None:
+    """Function to plot the following:
+    - distribution of the tags for unscaled results
+    - distribution of the tags for rescaled results
+    - distribution of the sum of the tags.
+
+    Parameters
+    ----------
+    conf : dict
+        Dictionary containing the configuration
+    train_set : list
+        The training set
+    test_tags : torch.Tensor
+        The predicted tags for the test dataset
+    device : torch.device
+        The device to use
+    data_directory : str
+        The directory where the data is stored
+    nbins : int, optional
+        Number of bins to use for the histograms
+    rescale : bool, optional
+        Whether to rescale the tags using recover_tag() or leave them between 0 and 1
+    recover_tag : list, optional
+        List of functions to recover each tag
+    """
+
+    data_directory = conf['speckle']['datadirectory']
+    model_name = conf['model']['name']
+    ensemble = conf['preproc'].get('ensemble', 1)
+
+    dirname = f'{data_directory}/{model_name}_score'
+    ensure_directory(dirname)
+
+    # Get the tags from the training set
+    if ensemble > 1:
+        train_set = list(itertools.chain(*train_set))
+    _, tags, _ = zip(*train_set)
+    tags = np.stack(tags)
+    train_tags = np.array([n for n in tags])
+
+    # Get the tags from the test set
+    predic_tags = np.array([n.cpu().numpy() for n in test_tags])
+
+    # Keep track of J=sum(tags) for each sample
+    J_pred = np.zeros(predic_tags.shape[0])
+    J_true = np.zeros(train_tags.shape[0])
+
+    # Plot the distribution of each tag element
+    fig, axs = plt.subplots(2, 4, figsize=(20, 10))
+    for i in range(train_tags.shape[1]):
+        if rescale and recover_tag is not None:
+            recovered_tag_model = np.asarray(
+                [recover_tag[i](predic_tags[:, i], i)]).squeeze(0)
+            recovered_tag_true = np.asarray(
+                [recover_tag[i](train_tags[:, i], i)]).squeeze(0)
+            J_pred += 10**recovered_tag_model
+            J_true += 10**recovered_tag_true
+            axs[i // 4, i % 4].hist(recovered_tag_model,
+                                    bins=nbins,
+                                    color='tab:red',
+                                    density=True,
+                                    alpha=0.5,
+                                    label='Model prediction')
+            axs[i // 4, i % 4].hist(recovered_tag_true,
+                                    bins=nbins,
+                                    color='tab:blue',
+                                    density=True,
+                                    alpha=0.5,
+                                    label='Training data')
+        else:
+            axs[i // 4, i % 4].hist(predic_tags[:, i],
+                                    bins=nbins,
+                                    color='tab:red',
+                                    density=True,
+                                    alpha=0.5,
+                                    label='Model prediction')
+            axs[i // 4, i % 4].hist(train_tags[:, i],
+                                    bins=nbins,
+                                    color='tab:blue',
+                                    density=True,
+                                    alpha=0.5,
+                                    label='Training data')
+        axs[i // 4, i % 4].set_title(f'Screen {i}')
+    axs[0, 1].legend()
+    plt.tight_layout()
+    figname = f'{dirname}/{model_name}_tags'
+    if not rescale:
+        figname += '_unscaled'
+    plt.savefig(f'{figname}.png')
+    plt.close()
+
+    if rescale and recover_tag is not None:
+        # Also plot the distribution of the sum of the tags
+        fig, axs = plt.subplots(1, 1, figsize=(6, 6))
+        axs.hist(np.log10(J_pred),
+                 bins=nbins,
+                 color='tab:red',
+                 density=True,
+                 alpha=0.5,
+                 label='Model prediction')
+        axs.hist(np.log10(J_true),
+                 bins=nbins,
+                 color='tab:blue',
+                 density=True,
+                 alpha=0.5,
+                 label='Training data')
+        axs.set_title('Sum of J')
+        axs.legend()
+        plt.tight_layout()
+        plt.savefig(f'{dirname}/{model_name}_sumJ.png')
+
+        plt.close()
+
+
+
+ +
+ + + +
+ +
+ +
+ + + + + + + + + + + + + +
+
+ + + +
+ +
+ + + +
+
+
+
+ + + + + + + + + + + + + + \ No newline at end of file diff --git a/api/preprocess/index.html b/api/preprocess/index.html new file mode 100644 index 0000000..e9228fe --- /dev/null +++ b/api/preprocess/index.html @@ -0,0 +1,2753 @@ + + + + + + + + + + + + + + + + + + + + + + + + + Preprocess - Speckcn2 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
+ + + + Skip to content + + +
+
+ +
+ + + + + + +
+ + +
+ +
+ + + + + + +
+
+ + + +
+
+
+ + + + + +
+
+
+ + + +
+
+
+ + + +
+
+
+ + + +
+
+ + + + + + + +

Preprocess

+ +
+ + + + +
+ +

This module contains functions for training and evaluating a neural network +model using PyTorch. It includes the following key components:

+
    +
  1. train: Trains the model for a specified number of epochs, logs training + and validation losses, and saves the model state at specified intervals.
    2. +
  2. score: Evaluates the model on a test dataset, calculates various metrics, + and generates plots for a specified number of test samples.
    3. +
+

The module relies on several external utilities and models from the speckcn2 +package, including EnsembleModel, ComposableLoss, and Normalizer.

+ + + + + + + + +
+ + + + + + + + + +
+ + +

+ assemble_transform(conf) + +

+ + +
+ +

Assembles the transformation to apply to each image.

+ + +

Parameters:

+
    +
  • + conf + (dict) + – +
    +

    Dictionary containing the configuration

    +
    +
    • +
+ + +

Returns:

+
    +
  • +transform ( Compose +) – +
    +

    Transformation to apply to the images

    +
    +
    • +
+ +
+ Source code in src/speckcn2/preprocess.py + 
31
+32
+33
+34
+35
+36
+37
+38
+39
+40
+41
+42
+43
+44
+45
+46
+47
+48
+49
+50
+51
+52
+53
+54
+55
+56
+57
+58
+59
+60
+61
+62
+63
+64
+65
+66
+67
+68
+69
+70
+71
+72
+73
+74
+75
def assemble_transform(conf: dict) -> transforms.Compose:
+    """Assembles the transformation to apply to each image.
+
+    Parameters
+    ----------
+    conf : dict
+        Dictionary containing the configuration
+
+    Returns
+    -------
+    transform : torchvision.transforms.Compose
+        Transformation to apply to the images
+    """
+    list_transforms = []
+
+    if conf['preproc']['centercrop'] > 0:
+        # Take only the center of the image
+        list_transforms.append(
+            transforms.CenterCrop(conf['preproc']['centercrop']))
+
+    if conf['preproc']['polarize']:
+        # make the image larger for better polar conversion
+        list_transforms.append(transforms.Resize(conf['preproc']['polresize']))
+        # Convert to polar coordinates
+        list_transforms.append(PolarCoordinateTransform())
+
+    if conf['preproc']['randomrotate'] and not conf['preproc']['polarize']:
+        # Randomly rotate the image, since it is symmetric (not if it is polarized)
+        list_transforms.append(transforms.RandomRotation(degrees=(-180, 180)))
+
+    if conf['preproc']['resize']:
+        # Optionally, downscale it
+        list_transforms.append(
+            transforms.Resize(
+                (conf['preproc']['resize'], conf['preproc']['resize'])))
+
+    if conf['preproc']['equivariant'] and conf['preproc']['polarize']:
+        # Apply the equivariant transform, which makes sense only in polar coordinates
+        list_transforms.append(ShiftRowsTransform())
+
+
+#    list_transforms.append(SpiderMask())
+    list_transforms.append(ToUnboundTensor())
+
+    return transforms.Compose(list_transforms)
+
+
+
+ +
+ +
+ + +

+ create_average_dataset(dataset, average_size) + +

+ + +
+ +

Creates a dataset of averages from a dataset of single images. The +averages are created by grouping together average_size images.

+ + +

Parameters:

+
    +
  • + dataset + (list) + – +
    +

    List of single images

    +
    +
    • +
  • + average_size + (int) + – +
    +

    The number of images that will be averaged together

    +
    +
    • +
+ + +

Returns:

+
    +
  • +average_dataset ( list +) – +
    +

    List of averages

    +
    +
    • +
+ +
+ Source code in src/speckcn2/preprocess.py + 
394
+395
+396
+397
+398
+399
+400
+401
+402
+403
+404
+405
+406
+407
+408
+409
+410
+411
+412
+413
+414
+415
+416
+417
+418
+419
+420
+421
+422
+423
+424
+425
+426
+427
+428
+429
+430
+431
+432
+433
+434
+435
+436
+437
+438
+439
def create_average_dataset(dataset: list, average_size: int) -> list:
+    """Creates a dataset of averages from a dataset of single images. The
+    averages are created by grouping together average_size images.
+
+    Parameters
+    ----------
+    dataset : list
+        List of single images
+    average_size : int
+        The number of images that will be averaged together
+
+    Returns
+    -------
+    average_dataset : list
+        List of averages
+    """
+    split_averages: dict = {}
+    for item in dataset:
+        key = item[-1]
+        split_averages.setdefault(key, []).append(item)
+
+    average_dataset: list = []
+    for average in split_averages.values():
+        # * In each average, take n_groups groups of average_size datapoints
+        n_groups = len(average) // average_size
+        if n_groups < 1:
+            raise ValueError(f'Average size {average_size} is too large '
+                             f'for groups with size {len(average)}')
+        # Extract the averages randomly
+        sample = random.sample(average, n_groups * average_size)
+        # Split the sample into groups of average_size
+        list_to_avg = [
+            sample[i:i + average_size]
+            for i in range(0, n_groups * average_size, average_size)
+        ]
+        # Average the groups
+        averages = [
+            tuple(
+                sum(element) / average_size
+                for i, element in enumerate(zip(*group)))
+            for group in list_to_avg
+        ]
+        average_dataset.extend(averages)
+
+    random.shuffle(average_dataset)
+    return average_dataset
+
+
+
+ +
+ +
+ + +

+ create_ensemble_dataset(dataset, ensemble_size) + +

+ + +
+ +

Creates a dataset of ensembles from a dataset of single images. The +ensembles are created by grouping together ensemble_size images. These +images will be used to train the model in parallel.

+ + +

Parameters:

+
    +
  • + dataset + (list) + – +
    +

    List of single images

    +
    +
    • +
  • + ensemble_size + (int) + – +
    +

    The number of images that will be processed together as an ensemble

    +
    +
    • +
+ + +

Returns:

+
    +
  • +ensemble_dataset ( list +) – +
    +

    List of ensembles

    +
    +
    • +
+ +
+ Source code in src/speckcn2/preprocess.py + 
464
+465
+466
+467
+468
+469
+470
+471
+472
+473
+474
+475
+476
+477
+478
+479
+480
+481
+482
+483
+484
+485
+486
+487
+488
+489
+490
+491
+492
+493
+494
+495
+496
+497
+498
+499
+500
+501
def create_ensemble_dataset(dataset: list, ensemble_size: int) -> list:
+    """Creates a dataset of ensembles from a dataset of single images. The
+    ensembles are created by grouping together ensemble_size images. These
+    images will be used to train the model in parallel.
+
+    Parameters
+    ----------
+    dataset : list
+        List of single images
+    ensemble_size : int
+        The number of images that will be processed together as an ensemble
+
+    Returns
+    -------
+    ensemble_dataset : list
+        List of ensembles
+    """
+    split_ensembles: dict = {}
+    for item in dataset:
+        key = item[-1]
+        split_ensembles.setdefault(key, []).append(item)
+
+    ensemble_dataset: list = []
+    for ensemble in split_ensembles.values():
+        # * In each ensemble, take n_groups groups of ensemble_size datapoints
+        n_groups = len(ensemble) // ensemble_size
+        if n_groups < 1:
+            raise ValueError(f'Ensemble size {ensemble_size} is too large '
+                             f'for ensembles with size {len(ensemble)}')
+        # Extract the ensembles randomly
+        sample = random.sample(ensemble, n_groups * ensemble_size)
+        # Split the sample into groups of ensemble_size
+        ensemble_dataset.extend(sample[i:i + ensemble_size]
+                                for i in range(0, n_groups *
+                                               ensemble_size, ensemble_size))
+
+    random.shuffle(ensemble_dataset)
+    return ensemble_dataset
+
+
+
+ +
+ +
+ + +

+ get_ensemble_dict(tag_files) + +

+ + +
+ +

Function to associate each Cn2 profile to an ensemble ID for parallel +processing.

+ + +

Parameters:

+
    +
  • + tag_files + (dict) + – +
    +

    Dictionary of image files and their corresponding tag files

    +
    +
    • +
+ + +

Returns:

+
    +
  • +ensemble_dict ( dict +) – +
    +

    Dictionary of image files and their corresponding ensemble IDs

    +
    +
    • +
+ +
+ Source code in src/speckcn2/preprocess.py + 
295
+296
+297
+298
+299
+300
+301
+302
+303
+304
+305
+306
+307
+308
+309
+310
+311
+312
+313
+314
+315
+316
+317
def get_ensemble_dict(tag_files: dict) -> dict:
+    """Function to associate each Cn2 profile to an ensemble ID for parallel
+    processing.
+
+    Parameters
+    ----------
+    tag_files : dict
+        Dictionary of image files and their corresponding tag files
+
+    Returns
+    -------
+    ensemble_dict : dict
+        Dictionary of image files and their corresponding ensemble IDs
+    """
+    ensembles = {}
+    ensemble_counter = 1
+    for value in tag_files.values():
+        # Check if the value is already assigned an ID
+        if value not in ensembles:
+            # Assign a new ID if it's a new value
+            ensembles[value] = ensemble_counter
+            ensemble_counter += 1
+    return {key: ensembles[value] for key, value in tag_files.items()}
+
+
+
+ +
+ +
+ + +

+ get_tag_files(file_list, datadirectory) + +

+ + +
+ +

Function to check the existence of tag files for each image file.

+ + +

Parameters:

+
    +
  • + file_list + (list) + – +
    +

    List of image files

    +
    +
    • +
  • + datadirectory + (str) + – +
    +

    The directory containing the data

    +
    +
    • +
+ + +

Returns:

+
    +
  • +tag_files ( dict +) – +
    +

    Dictionary of image files and their corresponding tag files

    +
    +
    • +
+ +
+ Source code in src/speckcn2/preprocess.py + 
263
+264
+265
+266
+267
+268
+269
+270
+271
+272
+273
+274
+275
+276
+277
+278
+279
+280
+281
+282
+283
+284
+285
+286
+287
+288
+289
+290
+291
+292
def get_tag_files(file_list: list, datadirectory: str) -> dict:
+    """Function to check the existence of tag files for each image file.
+
+    Parameters
+    ----------
+    file_list : list
+        List of image files
+    datadirectory : str
+        The directory containing the data
+
+    Returns
+    -------
+    tag_files : dict
+        Dictionary of image files and their corresponding tag files
+    """
+    tag_files = {}
+    for file_name in file_list:
+        if 'MALES' in file_name:
+            ftagname = file_name.replace('.h5', '_tag.h5')
+        elif 'sample' in file_name:
+            ftagname = file_name.rpartition('-')[0] + '_tag.h5'
+        else:
+            ftagname = file_name.rpartition('_')[0] + '_tag.h5'
+
+        tag_path = os.path.join(datadirectory, ftagname)
+        if os.path.exists(tag_path):
+            tag_files[file_name] = tag_path
+        else:
+            print(f'*** Warning: tag file {ftagname} not found.')
+    return tag_files
+
+
+
+ +
+ +
+ + +

+ imgs_as_single_datapoint(conf, nimg_print=5) + +

+ + +
+ +

Preprocesses the data by loading images and tags from the given +directory, applying a transformation to the images. Each image is treated +as a single data point.

+ + +

Parameters:

+
    +
  • + conf + (dict) + – +
    +

    Dictionary containing the configuration

    +
    +
    • +
  • + nimg_print + (int, default: + 5 +) + – +
    +

    Number of images to print

    +
    +
    • +
+ + +

Returns:

+
    +
  • +all_images ( list +) – +
    +

    List of all images

    +
    +
    • +
  • +all_tags ( list +) – +
    +

    List of all tags

    +
    +
    • +
  • +all_ensemble_ids ( list +) – +
    +

    List of all ensemble ids, representing images from the same Cn2 profile

    +
    +
    • +
+ +
+ Source code in src/speckcn2/preprocess.py + 
142
+143
+144
+145
+146
+147
+148
+149
+150
+151
+152
+153
+154
+155
+156
+157
+158
+159
+160
+161
+162
+163
+164
+165
+166
+167
+168
+169
+170
+171
+172
+173
+174
+175
+176
+177
+178
+179
+180
+181
+182
+183
+184
+185
+186
+187
+188
+189
+190
+191
+192
+193
+194
+195
+196
+197
+198
+199
+200
+201
+202
+203
+204
+205
+206
+207
+208
+209
+210
+211
+212
+213
+214
+215
+216
+217
+218
+219
+220
+221
+222
+223
+224
+225
+226
+227
+228
+229
+230
+231
+232
+233
+234
+235
+236
+237
+238
+239
+240
+241
+242
+243
+244
+245
+246
+247
+248
+249
+250
+251
+252
+253
+254
+255
+256
+257
+258
+259
+260
def imgs_as_single_datapoint(
+    conf: dict,
+    nimg_print: int = 5,
+) -> tuple[list, list, list]:
+    """Preprocesses the data by loading images and tags from the given
+    directory, applying a transformation to the images. Each image is treated
+    as a single data point.
+
+    Parameters
+    ----------
+    conf : dict
+        Dictionary containing the configuration
+    nimg_print: int
+        Number of images to print
+
+    Returns
+    -------
+    all_images : list
+        List of all images
+    all_tags : list
+        List of all tags
+    all_ensemble_ids : list
+        List of all ensemble ids, representing images from the same Cn2 profile
+    """
+    time_start = time.time()
+    datadirectory = conf['speckle']['datadirectory']
+    mname = conf['model']['name']
+    dataname = conf['preproc']['dataname']
+    tagname = dataname.replace('images', 'tags')
+    ensemblename = dataname.replace('images', 'ensemble')
+    nreps = conf['preproc']['speckreps']
+
+    # Dummy transformation to get the original image
+    transform_orig = transforms.Compose([
+        transforms.ToTensor(),
+    ])
+
+    # and get the transformation to apply to each image
+    transform = assemble_transform(conf)
+
+    # Get the list of images
+    file_list = [
+        file_name for file_name in os.listdir(datadirectory)
+        if '.h5' in file_name and 'tag' not in file_name
+    ]
+    np.random.shuffle(file_list)
+    # Optionally, do data augmentation by using each file multiple times
+    # (It makes sense only in combination with random rotations)
+    file_list = file_list * nreps
+
+    all_images, all_tags, all_ensemble_ids = [], [], []
+    show_image = nimg_print > 0
+    if show_image:
+        ensure_directory(f'{datadirectory}/imgs_to_{mname}')
+
+    tag_files = get_tag_files(file_list, datadirectory)
+    ensemble_dict = get_ensemble_dict(tag_files)
+
+    # Load each text file as an image
+    for counter, file_name in enumerate(file_list):
+        # Process only if tags available
+        if file_name in tag_files:
+
+            # Construct the full path to the file
+            file_path = os.path.join(datadirectory, file_name)
+
+            # Open the HDF5 file
+            print(file_path, flush=True)
+            with h5py.File(file_path, 'r') as f:
+                # Load the data from the 'data' dataset
+                pixel_values = np.float32(f['data'][:])
+                # and replace nans with 0
+                np.nan_to_num(pixel_values, copy=False)
+
+            # Create the image
+            image_orig = Image.fromarray(pixel_values, mode='F')
+
+            # Apply the transformation
+            image = transform(image_orig)
+            if show_image:
+                image_orig = transform_orig(image_orig)
+
+            # and add the img to the collection
+            all_images.append(image)
+
+            # Load the tags
+            with h5py.File(tag_files[file_name], 'r') as f:
+                if 'sample' in file_name:
+                    tags = f['J'][:].reshape(8, 1)
+                else:
+                    tags = f['data'][:]
+
+            # Plot the image using maplotlib
+            if counter > nimg_print:
+                show_image = False
+            if show_image:
+                plot_preprocessed_image(image_orig, image, tags, counter,
+                                        datadirectory, mname, file_name)
+
+            # Preprocess the tags
+            np.log10(tags, out=tags)
+            # Add the tag to the collection
+            all_tags.append(tags.squeeze())
+
+            # Get the ensemble ID
+            ensemble_id = ensemble_dict[file_name]
+            # and add it to the collection
+            all_ensemble_ids.append(ensemble_id)
+
+    # Finally, store them before returning
+    torch.save(all_images, os.path.join(datadirectory, dataname))
+    torch.save(all_tags, os.path.join(datadirectory, tagname))
+    torch.save(all_ensemble_ids, os.path.join(datadirectory, ensemblename))
+
+    print('*** Preprocessing complete.', flush=True)
+    print('It took',
+          time.time() - time_start, 'seconds to preprocess the data.')
+
+    return all_images, all_tags, all_ensemble_ids
+
+
+
+ +
+ +
+ + +

+ prepare_data(conf, nimg_print=5) + +

+ + +
+ +

If not already available, preprocesses the data by loading images and +tags from the given directory, applying a transformation to the images.

+ + +

Parameters:

+
    +
  • + conf + (dict) + – +
    +

    Dictionary containing the configuration

    +
    +
    • +
  • + nimg_print + (int, default: + 5 +) + – +
    +

    Number of images to print

    +
    +
    • +
+ + +

Returns:

+
    +
  • +all_images ( list +) – +
    +

    List of all images

    +
    +
    • +
  • +all_tags ( list +) – +
    +

    List of all tags

    +
    +
    • +
  • +all_ensemble_ids ( list +) – +
    +

    List of all ensemble ids, representing images from the same Cn2 profile

    +
    +
    • +
+ +
+ Source code in src/speckcn2/preprocess.py + 
 78
+ 79
+ 80
+ 81
+ 82
+ 83
+ 84
+ 85
+ 86
+ 87
+ 88
+ 89
+ 90
+ 91
+ 92
+ 93
+ 94
+ 95
+ 96
+ 97
+ 98
+ 99
+100
+101
+102
+103
+104
+105
+106
+107
+108
+109
+110
+111
+112
+113
+114
+115
+116
+117
+118
+119
+120
+121
+122
+123
+124
+125
+126
+127
+128
+129
+130
+131
+132
+133
+134
+135
+136
+137
+138
+139
def prepare_data(
+    conf: dict,
+    nimg_print: int = 5,
+) -> tuple[list, list, list]:
+    """If not already available, preprocesses the data by loading images and
+    tags from the given directory, applying a transformation to the images.
+
+    Parameters
+    ----------
+    conf : dict
+        Dictionary containing the configuration
+    nimg_print: int
+        Number of images to print
+
+    Returns
+    -------
+    all_images : list
+        List of all images
+    all_tags : list
+        List of all tags
+    all_ensemble_ids : list
+        List of all ensemble ids, representing images from the same Cn2 profile
+    """
+    datadirectory = conf['speckle']['datadirectory']
+    dataname = conf['preproc']['dataname']
+    tagname = dataname.replace('images', 'tags')
+    ensemblename = dataname.replace('images', 'ensemble')
+
+    # First, check if the data has already been preprocessed
+    if os.path.exists(os.path.join(datadirectory, dataname)):
+        print(f'*** Loading preprocessed data from {dataname}')
+        # If so, load it
+        all_images = torch.load(os.path.join(datadirectory, dataname))
+        all_tags = torch.load(os.path.join(datadirectory, tagname))
+        all_ensemble_ids = torch.load(os.path.join(datadirectory,
+                                                   ensemblename))
+        # print the info about the dataset
+        print(f'*** There are {len(all_images)} images in the dataset.')
+    else:
+        # Check if there is at least one image file in the directory
+        if not any('.h5' in file_name
+                   for file_name in os.listdir(datadirectory)):
+            raise FileNotFoundError(
+                'No image files found in the directory. Please provide the '
+                'correct path to the data directory.')
+        # Otherwise, preprocess the raw data separating the single images
+        all_images, all_tags, all_ensemble_ids = imgs_as_single_datapoint(
+            conf, nimg_print)
+
+    # Get the average value of the pixels, excluding the 0 values
+    non_zero_pixels = 0
+    sum_pixels = 0
+    for image in all_images:
+        non_zero_pixels_in_image = image[image != 0]
+        non_zero_pixels += non_zero_pixels_in_image.numel()
+        sum_pixels += torch.sum(non_zero_pixels_in_image)
+    pixel_average = sum_pixels / non_zero_pixels
+    print('*** Pixel average:', pixel_average)
+    # and store it in the config
+    conf['preproc']['pixel_average'] = pixel_average
+
+    return all_images, all_tags, all_ensemble_ids
+
+
+
+ +
+ +
+ + +

+ print_average_info(dataset, average_size, ttsplit) + +

+ + +
+ +

Prints the information about the average dataset.

+ + +

Parameters:

+
    +
  • + dataset + (list) + – +
    +

    The average dataset

    +
    +
    • +
  • + average_size + (int) + – +
    +

    The number of images in each average

    +
    +
    • +
  • + ttsplit + (int) + – +
    +

    The train-test split

    +
    +
    • +
+ +
+ Source code in src/speckcn2/preprocess.py + 
442
+443
+444
+445
+446
+447
+448
+449
+450
+451
+452
+453
+454
+455
+456
+457
+458
+459
+460
+461
def print_average_info(dataset: list, average_size: int, ttsplit: int):
+    """Prints the information about the average dataset.
+
+    Parameters
+    ----------
+    dataset : list
+        The average dataset
+    average_size : int
+        The number of images in each average
+    ttsplit : int
+        The train-test split
+    """
+    train_size = int(ttsplit * len(dataset))
+    print(
+        f'*** There are {len(dataset)} average groups in the dataset, '
+        f'that I split in {train_size} for training and '
+        f'{len(dataset) - train_size} for testing. Each average is composed by '
+        f'{average_size} images. This corresponds to {train_size * average_size} '
+        f'for training and {(len(dataset) - train_size) * average_size} for testing.'
+    )
+
+
+
+ +
+ +
+ + +

+ print_dataset_info(dataset, ttsplit) + +

+ + +
+ +

Prints the information about the dataset.

+ + +

Parameters:

+
    +
  • + dataset + (list) + – +
    +

    The dataset

    +
    +
    • +
  • + ttsplit + (int) + – +
    +

    The train-test split

    +
    +
    • +
+ +
+ Source code in src/speckcn2/preprocess.py + 
527
+528
+529
+530
+531
+532
+533
+534
+535
+536
+537
+538
+539
+540
def print_dataset_info(dataset: list, ttsplit: int):
+    """Prints the information about the dataset.
+
+    Parameters
+    ----------
+    dataset : list
+        The dataset
+    ttsplit : int
+        The train-test split
+    """
+    train_size = int(ttsplit * len(dataset))
+    print(f'*** There are {len(dataset)} images in the dataset, '
+          f'{train_size} for training and '
+          f'{len(dataset) - train_size} for testing.')
+
+
+
+ +
+ +
+ + +

+ print_ensemble_info(dataset, ensemble_size, ttsplit) + +

+ + +
+ +

Prints the information about the ensemble dataset.

+ + +

Parameters:

+
    +
  • + dataset + (list) + – +
    +

    The ensemble dataset

    +
    +
    • +
  • + ensemble_size + (int) + – +
    +

    The number of images in each ensemble

    +
    +
    • +
  • + ttsplit + (int) + – +
    +

    The train-test split

    +
    +
    • +
+ +
+ Source code in src/speckcn2/preprocess.py + 
504
+505
+506
+507
+508
+509
+510
+511
+512
+513
+514
+515
+516
+517
+518
+519
+520
+521
+522
+523
+524
def print_ensemble_info(dataset: list, ensemble_size: int, ttsplit: int):
+    """Prints the information about the ensemble dataset.
+
+    Parameters
+    ----------
+    dataset : list
+        The ensemble dataset
+    ensemble_size : int
+        The number of images in each ensemble
+    ttsplit : int
+        The train-test split
+    """
+
+    train_size = int(ttsplit * len(dataset))
+    print(
+        f'*** There are {len(dataset)} ensemble groups in the dataset, '
+        f'that I split in {train_size} for training and '
+        f'{len(dataset) - train_size} for testing. Each ensemble is composed by '
+        f'{ensemble_size} images. This corresponds to {train_size * ensemble_size} '
+        f'for training and {(len(dataset) - train_size) * ensemble_size} for testing.'
+    )
+
+
+
+ +
+ +
+ + +

+ split_dataset(dataset, ttsplit) + +

+ + +
+ +

Splits the dataset into training and testing sets.

+ + +

Parameters:

+
    +
  • + dataset + (list) + – +
    +

    The dataset

    +
    +
    • +
  • + ttsplit + (int) + – +
    +

    The train-test split

    +
    +
    • +
+ + +

Returns:

+
    +
  • +train_set ( list +) – +
    +

    The training set

    +
    +
    • +
  • +test_set ( list +) – +
    +

    The testing set

    +
    +
    • +
+ +
+ Source code in src/speckcn2/preprocess.py + 
543
+544
+545
+546
+547
+548
+549
+550
+551
+552
+553
+554
+555
+556
+557
+558
+559
+560
+561
+562
+563
def split_dataset(dataset: list, ttsplit: int) -> tuple[list, list]:
+    """Splits the dataset into training and testing sets.
+
+    Parameters
+    ----------
+    dataset : list
+        The dataset
+    ttsplit : int
+        The train-test split
+
+    Returns
+    -------
+    train_set : list
+        The training set
+    test_set : list
+        The testing set
+    """
+    # First shuffle the dataset
+    random.shuffle(dataset)
+    train_size = int(ttsplit * len(dataset))
+    return dataset[:train_size], dataset[train_size:]
+
+
+
+ +
+ +
+ + +

+ train_test_split(all_images, all_tags, all_ensemble_ids, nz) + +

+ + +
+ +

Splits the data into training and testing sets.

+ + +

Parameters:

+
    +
  • + all_images + (list) + – +
    +

    List of images

    +
    +
    • +
  • + all_tags + (list) + – +
    +

    List of tags

    +
    +
    • +
  • + all_ensemble_ids + (list) + – +
    +

    List of ensemble ids

    +
    +
    • +
  • + nz + (Normalizer) + – +
    +

    The normalizer object to preprocess the data

    +
    +
    • +
+ + +

Returns:

+
    +
  • +train_set ( list +) – +
    +

    Training dataset

    +
    +
    • +
  • +test_set ( list +) – +
    +

    Testing dataset

    +
    +
    • +
+ +
+ Source code in src/speckcn2/preprocess.py + 
320
+321
+322
+323
+324
+325
+326
+327
+328
+329
+330
+331
+332
+333
+334
+335
+336
+337
+338
+339
+340
+341
+342
+343
+344
+345
+346
+347
+348
+349
+350
+351
+352
+353
+354
+355
+356
+357
+358
+359
+360
+361
+362
+363
+364
+365
+366
+367
+368
+369
+370
+371
+372
+373
+374
+375
+376
+377
+378
+379
+380
+381
+382
+383
+384
+385
+386
+387
+388
+389
+390
+391
def train_test_split(
+    all_images: list[torch.tensor],
+    all_tags: list[np.ndarray],
+    all_ensemble_ids: list[int],
+    nz: Normalizer,
+) -> tuple[list, list]:
+    """Splits the data into training and testing sets.
+
+    Parameters
+    ----------
+    all_images : list
+        List of images
+    all_tags : list
+        List of tags
+    all_ensemble_ids : list
+        List of ensemble ids
+    nz: Normalizer
+        The normalizer object to preprocess the data
+
+    Returns
+    -------
+    train_set : list
+        Training dataset
+    test_set : list
+        Testing dataset
+    """
+    # Get the config dict
+    config = nz.conf
+    # extract the model parameters
+    modelname = config['model']['name']
+    datadirectory = config['speckle']['datadirectory']
+    ttsplit = config['hyppar'].get('ttsplit', 0.8)
+    ensemble_size = config['preproc'].get('ensemble', 1)
+    average_size = config['preproc'].get('average', 0)
+
+    # Check if the training and test set are already prepared
+    train_file = f'{datadirectory}/train_set_{modelname}.pickle'
+    test_file = f'{datadirectory}/test_set_{modelname}.pickle'
+
+    if os.path.isfile(train_file) and os.path.isfile(test_file):
+        print('Loading the training and testing set...', flush=True)
+        nz._normalizing_functions(all_images, all_tags, all_ensemble_ids)
+        train_data = pickle.load(open(train_file, 'rb'))
+        test_data = pickle.load(open(test_file, 'rb'))
+        print(f'*** There are {len(train_data)} images in the training set, ')
+        print(f'*** and {len(test_data)} images in the testing set.')
+        return train_data, test_data
+
+    # If the data are not already prepared, first I normalize them using the Normalizer object
+    print('Normalizing the images and tags...', flush=True)
+    dataset = nz.normalize_imgs_and_tags(all_images, all_tags,
+                                         all_ensemble_ids)
+
+    if average_size > 1 and ensemble_size > 1:
+        raise ValueError(
+            'The average_size and ensemble_size cannot be set at the same time.'
+        )
+    elif average_size > 1:
+        dataset = create_average_dataset(dataset, average_size)
+        print_average_info(dataset, average_size, ttsplit)
+    elif ensemble_size > 1:
+        dataset = create_ensemble_dataset(dataset, ensemble_size)
+        print_ensemble_info(dataset, ensemble_size, ttsplit)
+    else:
+        print_dataset_info(dataset, ttsplit)
+
+    train_set, test_set = split_dataset(dataset, ttsplit)
+
+    pickle.dump(train_set, open(train_file, 'wb'))
+    pickle.dump(test_set, open(test_file, 'wb'))
+
+    return train_set, test_set
+
+
+
+ +
+ + + +
+ +
+ +
+ + + + + + + + + + + + + +
+
+ + + +
+ +
+ + + +
+
+
+
+ + + + + + + + + + + + + + \ No newline at end of file diff --git a/api/scnn/index.html b/api/scnn/index.html new file mode 100644 index 0000000..db8b09f --- /dev/null +++ b/api/scnn/index.html @@ -0,0 +1,1049 @@ + + + + + + + + + + + + + + + + + + + + + + + + + SCNN - Speckcn2 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
+ + + + Skip to content + + +
+
+ +
+ + + + + + +
+ + +
+ +
+ + + + + + +
+
+ + + +
+
+
+ + + + + +
+
+
+ + + +
+
+
+ + + +
+
+
+ + + +
+
+ + + + + + + +

SCNN

+ +
+ + + + +
+ +

This module defines a Steerable Convolutional Neural Network (SteerableCNN) +using the escnn library for equivariant neural networks. The primary class, +SteerableCNN, allows for the creation of a convolutional neural network that +handles various symmetries, making it useful for tasks requiring rotational +invariance.

+

The module imports necessary libraries, including torch and escnn, and defines utility +functions such as create_block, compute_new_features, create_pool, and create_final_block. +These functions help construct convolutional blocks, calculate feature map sizes, +create anti-aliased pooling layers, and build fully connected layers.

+

The SteerableCNN class initializes with a configuration dictionary and a symmetry parameter, +setting up parameters like kernel sizes, paddings, strides, and feature fields. +It determines the symmetry group and initializes the input type for the network.

+

The network is built by iterating through specified kernel sizes, creating convolutional +blocks and pooling layers, adding a group pooling layer for invariance, and creating final +fully connected layers using create_final_block.

+

The forward method processes the input tensor through the network, applying each +equivariant block, performing group pooling, and classifying the output using the fully connected layers.

+

Overall, this module provides a flexible framework for building steerable +convolutional neural networks with configurable symmetries and architectures.

+ + + + + + + + +
+ + + + + + + + + +
+ + +

+ create_final_block(config, n_initial, nscreens) + +

+ + +
+ +

Creates a fully connected neural network block based on a predefined +configuration.

+

This function dynamically creates a sequence of PyTorch layers for a fully connected +neural network. The configuration for the layers is read from a global config dictionary +which should contain a 'final_block' key with a list of layer configurations. Each layer +configuration is a dictionary that must include a 'type' key with the name of the layer +class (e.g., 'Linear', 'Dropout', etc.) and can include additional keys for the layer +parameters.

+

The first 'Linear' layer in the configuration has its number of input features set to n_in, +and any 'Linear' layer with 'out_features' set to 'nscreens' has its number of output features +set to nscreens.

+

Args:

+ + +

Parameters:

+
    +
  • + config + (dict) + – +
    +

    The global configuration dictionary containing the layer configurations.

    +
    +
    • +
  • + n_initial + (int) + – +
    +

    The number of input features for the first 'Linear' layer.

    +
    +
    • +
  • + nscreens + (int) + – +
    +

    The number of output features for any 'Linear' layer with 'out_features' set to 'nscreens'.

    +
    +
    • +
+ + +

Returns:

+
    +
  • + torch.nn.Sequential: A sequential container of the configured PyTorch layers. + – +
    + +
    +
    • +
+ +
+ Source code in src/speckcn2/scnn.py + 
 57
+ 58
+ 59
+ 60
+ 61
+ 62
+ 63
+ 64
+ 65
+ 66
+ 67
+ 68
+ 69
+ 70
+ 71
+ 72
+ 73
+ 74
+ 75
+ 76
+ 77
+ 78
+ 79
+ 80
+ 81
+ 82
+ 83
+ 84
+ 85
+ 86
+ 87
+ 88
+ 89
+ 90
+ 91
+ 92
+ 93
+ 94
+ 95
+ 96
+ 97
+ 98
+ 99
+100
+101
+102
+103
+104
+105
+106
+107
+108
def create_final_block(config: dict, n_initial: int,
+                       nscreens: int) -> nn.Sequential:
+    """Creates a fully connected neural network block based on a predefined
+    configuration.
+
+    This function dynamically creates a sequence of PyTorch layers for a fully connected
+    neural network. The configuration for the layers is read from a global `config` dictionary
+    which should contain a 'final_block' key with a list of layer configurations. Each layer
+    configuration is a dictionary that must include a 'type' key with the name of the layer
+    class (e.g., 'Linear', 'Dropout', etc.) and can include additional keys for the layer
+    parameters.
+
+    The first 'Linear' layer in the configuration has its number of input features set to `n_in`,
+    and any 'Linear' layer with 'out_features' set to 'nscreens' has its number of output features
+    set to `nscreens`.
+
+    Args:
+    Parameters
+    ----------
+    config : dict
+        The global configuration dictionary containing the layer configurations.
+    n_initial : int
+        The number of input features for the first 'Linear' layer.
+    nscreens : int
+        The number of output features for any 'Linear' layer with 'out_features' set to 'nscreens'.
+
+    Returns
+    ----------
+    torch.nn.Sequential: A sequential container of the configured PyTorch layers.
+    """
+    layers = []
+
+    for layer_config in config['final_block']:
+        layer_type = layer_config.pop('type')
+
+        if layer_type == 'Linear':
+            # The first linear layer has a constrained number of input features
+            if n_initial != -1:
+                layer_config['in_features'] = n_initial
+                n_initial = -1
+            if layer_config['out_features'] == 'nscreens':
+                layer_config['out_features'] = nscreens
+            # Pass the number of features as args
+            n_in = layer_config.pop('in_features')
+            n_out = layer_config.pop('out_features')
+
+            layers.append(torch.nn.Linear(n_in, n_out, **layer_config))
+        else:
+            layer_class = getattr(torch.nn, layer_type)
+            layers.append(layer_class(**layer_config))
+
+    return torch.nn.Sequential(*layers)
+
+
+
+ +
+ + + +
+ +
+ +
+ + + + + + + + + + + + + +
+
+ + + +
+ +
+ + + +
+
+
+
+ + + + + + + + + + + + + + \ No newline at end of file diff --git a/api/transformations/index.html b/api/transformations/index.html new file mode 100644 index 0000000..d336144 --- /dev/null +++ b/api/transformations/index.html @@ -0,0 +1,1314 @@ + + + + + + + + + + + + + + + + + + + + + + + + + Transformations - Speckcn2 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
+ + + + Skip to content + + +
+
+ +
+ + + + + + +
+ + +
+ +
+ + + + + + +
+
+ + + +
+
+
+ + + + + +
+
+
+ + + +
+
+
+ + + +
+
+
+ + + +
+
+ + + + + + + +

Transformations

+ +
+ + + + +
+ +

This module defines several image transformation classes using PyTorch and +NumPy.

+

The PolarCoordinateTransform class converts a Cartesian image to polar +coordinates, which can be useful for certain types of image analysis. +The ShiftRowsTransform class shifts the rows of an image so that the row with the +smallest sum is positioned at the bottom, which can help in aligning images for further processing. +The ToUnboundTensor class converts an image to a tensor without normalizing it, +preserving the original pixel values. +Lastly, the SpiderMask class applies a circular mask to the image, simulating +the effect of a spider by setting pixels outside the mask to a background value, +which can be useful in certain experimental setups.

+ + + + + + + + +
+ + + + + + + + +
+ + + +

+ PolarCoordinateTransform() + +

+ + +
+

+ Bases: Module

+ + +

Transform a Cartesian image to polar coordinates.

+ + + + + + +
+ Source code in src/speckcn2/transformations.py + 
24
+25
def __init__(self):
+    super(PolarCoordinateTransform, self).__init__()
+
+
+ + + +
+ + + + + + + + + +
+ + +

+ forward(img) + +

+ + +
+ +

forward method of the transform +Args: + img (PIL Image or Tensor): Image to be scaled.

+

Returns: + PIL Image or Tensor: Rescaled image.

+ +
+ Source code in src/speckcn2/transformations.py + 
27
+28
+29
+30
+31
+32
+33
+34
+35
+36
+37
+38
+39
+40
+41
+42
+43
+44
+45
+46
+47
+48
+49
+50
+51
+52
+53
+54
+55
+56
+57
+58
+59
+60
+61
+62
+63
+64
+65
+66
+67
+68
+69
+70
+71
+72
+73
+74
+75
+76
+77
+78
+79
+80
+81
+82
+83
+84
+85
+86
+87
+88
+89
+90
+91
+92
+93
+94
def forward(self, img):
+    """ forward method of the transform
+    Args:
+        img (PIL Image or Tensor): Image to be scaled.
+
+    Returns:
+        PIL Image or Tensor: Rescaled image.
+    """
+
+    img = np.array(img)  # Convert PIL image to NumPy array
+
+    # Assuming img is a grayscale image
+    height, width = img.shape
+    polar_image = np.zeros_like(img)
+
+    # Center of the image
+    center_x, center_y = width // 2, height // 2
+
+    # Maximum possible value of r
+    max_r = np.sqrt(center_x**2 + center_y**2)
+
+    # Create a grid of (x, y) coordinates
+    y, x = np.ogrid[:height, :width]
+
+    # Shift the grid so that the center of the image is at (0, 0)
+    x = x - center_x
+    y = y - center_y
+
+    # Convert Cartesian to Polar coordinates
+    r = np.sqrt(x**2 + y**2)
+    theta = np.arctan2(y, x)
+
+    # Rescale r to [0, height)
+    r = np.round(r * (height - 1) / max_r).astype(int)
+
+    # Rescale theta to [0, width)
+    theta = np.round((theta + 2 * np.pi) % (2 * np.pi) * (width - 1) /
+                     (2 * np.pi)).astype(int)
+
+    # Use a 2D histogram to accumulate all values that map to the same polar coordinate
+    histogram, _, _ = np.histogram2d(theta.flatten(),
+                                     r.flatten(),
+                                     bins=[height, width],
+                                     range=[[0, height], [0, width]],
+                                     weights=img.flatten())
+
+    # Count how many Cartesian coordinates map to each polar coordinate
+    counts, _, _ = np.histogram2d(theta.flatten(),
+                                  r.flatten(),
+                                  bins=[height, width],
+                                  range=[[0, height], [0, width]])
+
+    # Take the average of all values that map to the same polar coordinate
+    polar_image = histogram / counts
+
+    # Handle any divisions by zero
+    polar_image[np.isnan(polar_image)] = 0
+
+    # Crop the large r part that is not used
+    for x in range(width - 1, -1, -1):
+        # If the column contains at least one non-black pixel
+        if np.any(polar_image[:, x] != 0):
+            # Crop at this x position
+            polar_image = polar_image[:, :x]
+            break
+
+    # reconvert to PIL image before returning
+    return Image.fromarray(polar_image)
+
+
+
+ +
+ + + +
+ +
+ +
+ +
+ + + +

+ ShiftRowsTransform() + +

+ + +
+

+ Bases: Module

+ + +

Shift the rows of an image such that the row with the smallest sum is at +the bottom.

+ + + + + + +
+ Source code in src/speckcn2/transformations.py + 
101
+102
def __init__(self):
+    super(ShiftRowsTransform, self).__init__()
+
+
+ + + +
+ + + + + + + + + + + +
+ +
+ +
+ +
+ + + +

+ SpiderMask() + +

+ + +
+

+ Bases: Module

+ + +

Apply a circular mask to the image, representing the effect of the +spider.

+

The pixels outside the spider are set to -0.01, such that their +value is lower than no light in the detector (0).

+ + + + + + +
+ Source code in src/speckcn2/transformations.py + 
139
+140
def __init__(self):
+    super(SpiderMask, self).__init__()
+
+
+ + + +
+ + + + + + + + + + + +
+ +
+ +
+ +
+ + + +

+ ToUnboundTensor() + +

+ + +
+

+ Bases: Module

+ + +

Transform the image into a tensor, but do not normalize it like +torchvision.ToTensor.

+ + + + + + +
+ Source code in src/speckcn2/transformations.py + 
122
+123
def __init__(self):
+    super(ToUnboundTensor, self).__init__()
+
+
+ + + +
+ + + + + + + + + + + +
+ +
+ +
+ + + + +
+ +
+ +
+ + + + + + + + + + + + + +
+
+ + + +
+ +
+ + + +
+
+
+
+ + + + + + + + + + + + + + \ No newline at end of file diff --git a/api/utils/index.html b/api/utils/index.html new file mode 100644 index 0000000..e622ed3 --- /dev/null +++ b/api/utils/index.html @@ -0,0 +1,1379 @@ + + + + + + + + + + + + + + + + + + + + + + + + + Utils - Speckcn2 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
+ + + + Skip to content + + +
+
+ +
+ + + + + + +
+ + +
+ +
+ + + + + + +
+
+ + + +
+
+
+ + + + + +
+
+
+ + + +
+
+
+ + + +
+
+
+ + + +
+
+ + + + + + + +

Utils

+ +
+ + + + +
+ +

This module provides utility functions for image processing and model +optimization.

+

It includes functions to plot original and preprocessed images along +with their tags, ensure the existence of specified directories, set up +optimizers based on configuration files, and create circular masks with +an inner "spider" circle removed. These utilities facilitate various +tasks in image analysis and machine learning model training.

+ + + + + + + + +
+ + + + + + + + + +
+ + +

+ create_circular_mask_with_spider(resolution, bkg_value=0) + +

+ + +
+ +

Creates a circular mask with an inner "spider" circle removed.

+ + +

Parameters:

+
    +
  • + resolution + (int) + – +
    +

    The resolution of the square mask.

    +
    +
    • +
  • + bkg_value + (int, default: + 0 +) + – +
    +

    The background value to set for the masked areas. Defaults to 0.

    +
    +
    • +
+ + +

Returns:

+
    +
  • + torch.Tensor : np.ndarray + – +
    +

    A 2D tensor representing the mask.

    +
    +
    • +
+ +
+ Source code in src/speckcn2/utils.py + 
110
+111
+112
+113
+114
+115
+116
+117
+118
+119
+120
+121
+122
+123
+124
+125
+126
+127
+128
+129
+130
+131
+132
+133
+134
+135
+136
+137
+138
+139
+140
+141
def create_circular_mask_with_spider(resolution: int,
+                                     bkg_value: int = 0) -> torch.Tensor:
+    """Creates a circular mask with an inner "spider" circle removed.
+
+    Parameters
+    ----------
+    resolution : int
+        The resolution of the square mask.
+    bkg_value : int
+        The background value to set for the masked areas. Defaults to 0.
+
+    Returns
+    -------
+    torch.Tensor : np.ndarray
+        A 2D tensor representing the mask.
+    """
+    # Create a circular mask
+    center = (int(resolution / 2), int(resolution / 2))
+    radius = min(center)
+    Y, X = np.ogrid[:resolution, :resolution]
+    mask = (X - center[0])**2 + (Y - center[1])**2 > radius**2
+
+    # Remove the inner circle (spider)
+    spider_radius = int(0.22 * resolution)
+    spider_mask = (X - center[0])**2 + (Y - center[1])**2 < spider_radius**2
+
+    # Apply background value to the mask and spider mask
+    final_mask = np.ones((resolution, resolution), dtype=np.uint8)
+    final_mask[mask] = bkg_value
+    final_mask[spider_mask] = bkg_value
+
+    return torch.Tensor(final_mask)
+
+
+
+ +
+ +
+ + +

+ ensure_directory(data_directory) + +

+ + +
+ +

Ensure that the directory exists.

+ + +

Parameters:

+
    +
  • + data_directory + (str) + – +
    +

    The directory to ensure

    +
    +
    • +
+ +
+ Source code in src/speckcn2/utils.py + 
72
+73
+74
+75
+76
+77
+78
+79
+80
+81
+82
def ensure_directory(data_directory: str) -> None:
+    """Ensure that the directory exists.
+
+    Parameters
+    ----------
+    data_directory : str
+        The directory to ensure
+    """
+
+    if not os.path.isdir(data_directory):
+        os.mkdir(data_directory)
+
+
+
+ +
+ +
+ + +

+ plot_preprocessed_image(image_orig, image, tags, counter, datadirectory, mname, file_name, polar=False) + +

+ + +
+ +

Plots the original and preprocessed image, and the tags.

+ + +

Parameters:

+
    +
  • + image_orig + (tensor) + – +
    +

    The original image

    +
    +
    • +
  • + image + (tensor) + – +
    +

    The preprocessed image

    +
    +
    • +
  • + tags + (tensor) + – +
    +

    The screen tags

    +
    +
    • +
  • + counter + (int) + – +
    +

    The counter of the image

    +
    +
    • +
  • + datadirectory + (str) + – +
    +

    The directory containing the data

    +
    +
    • +
  • + mname + (str) + – +
    +

    The name of the model

    +
    +
    • +
  • + file_name + (str) + – +
    +

    The name of the original image

    +
    +
    • +
  • + polar + (bool, default: + False +) + – +
    +

    If the image is in polar coordinates, by default False

    +
    +
    • +
+ +
+ Source code in src/speckcn2/utils.py + 
20
+21
+22
+23
+24
+25
+26
+27
+28
+29
+30
+31
+32
+33
+34
+35
+36
+37
+38
+39
+40
+41
+42
+43
+44
+45
+46
+47
+48
+49
+50
+51
+52
+53
+54
+55
+56
+57
+58
+59
+60
+61
+62
+63
+64
+65
+66
+67
+68
+69
def plot_preprocessed_image(image_orig: torch.tensor,
+                            image: torch.tensor,
+                            tags: torch.tensor,
+                            counter: int,
+                            datadirectory: str,
+                            mname: str,
+                            file_name: str,
+                            polar: bool = False) -> None:
+    """Plots the original and preprocessed image, and the tags.
+
+    Parameters
+    ----------
+    image_orig : torch.tensor
+        The original image
+    image : torch.tensor
+        The preprocessed image
+    tags : torch.tensor
+        The screen tags
+    counter : int
+        The counter of the image
+    datadirectory : str
+        The directory containing the data
+    mname : str
+        The name of the model
+    file_name : str
+        The name of the original image
+    polar : bool, optional
+        If the image is in polar coordinates, by default False
+    """
+
+    fig, axs = plt.subplots(1, 3, figsize=(15, 5))
+    # Plot the original image
+    axs[0].imshow(image_orig.squeeze(), cmap='bone')
+    axs[0].set_title(f'Training Image {file_name}')
+    # Plot the preprocessd image
+    axs[1].imshow(image.squeeze(), cmap='bone')
+    axs[1].set_title('Processed as')
+    if polar:
+        axs[1].set_xlabel(r'$r$')
+        axs[1].set_ylabel(r'$\theta$')
+
+    # Plot the tags
+    axs[2].plot(tags, 'o')
+    axs[2].set_yscale('log')
+    axs[2].set_title('Screen Tags')
+    axs[2].legend()
+
+    fig.subplots_adjust(wspace=0.3)
+    plt.savefig(f'{datadirectory}/imgs_to_{mname}/{counter}.png')
+    plt.close()
+
+
+
+ +
+ +
+ + +

+ setup_optimizer(config, model) + +

+ + +
+ +

Returns the optimizer specified in the configuration file.

+ + +

Parameters:

+
    +
  • + config + (dict) + – +
    +

    Dictionary containing the configuration

    +
    +
    • +
  • + model + (Module) + – +
    +

    The model to optimize

    +
    +
    • +
+ + +

Returns:

+
    +
  • +optimizer ( Module +) – +
    +

    The optimizer with the loaded state

    +
    +
    • +
+ +
+ Source code in src/speckcn2/utils.py + 
 85
+ 86
+ 87
+ 88
+ 89
+ 90
+ 91
+ 92
+ 93
+ 94
+ 95
+ 96
+ 97
+ 98
+ 99
+100
+101
+102
+103
+104
+105
+106
+107
def setup_optimizer(config: dict, model: nn.Module) -> nn.Module:
+    """Returns the optimizer specified in the configuration file.
+
+    Parameters
+    ----------
+    config : dict
+        Dictionary containing the configuration
+    model : torch.nn.Module
+        The model to optimize
+
+    Returns
+    -------
+    optimizer : torch.nn.Module
+        The optimizer with the loaded state
+    """
+
+    optimizer_name = config['hyppar']['optimizer']
+    if optimizer_name == 'Adam':
+        return torch.optim.Adam(model.parameters(), lr=config['hyppar']['lr'])
+    elif optimizer_name == 'SGD':
+        return torch.optim.SGD(model.parameters(), lr=config['hyppar']['lr'])
+    else:
+        raise ValueError(f'Unknown optimizer {optimizer_name}')
+
+
+
+ +
+ + + +
+ +
+ +
+ + + + + + + + + + + + + +
+
+ + + +
+ +
+ + + +
+
+
+
+ + + + + + + + + + + + + + \ No newline at end of file diff --git a/assets/_mkdocstrings.css b/assets/_mkdocstrings.css new file mode 100644 index 0000000..b500381 --- /dev/null +++ b/assets/_mkdocstrings.css @@ -0,0 +1,143 @@ + +/* Avoid breaking parameter names, etc. in table cells. */ +.doc-contents td code { + word-break: normal !important; +} + +/* No line break before first paragraph of descriptions. */ +.doc-md-description, +.doc-md-description>p:first-child { + display: inline; +} + +/* Max width for docstring sections tables. */ +.doc .md-typeset__table, +.doc .md-typeset__table table { + display: table !important; + width: 100%; +} + +.doc .md-typeset__table tr { + display: table-row; +} + +/* Defaults in Spacy table style. */ +.doc-param-default { + float: right; +} + +/* Parameter headings must be inline, not blocks. */ +.doc-heading-parameter { + display: inline; +} + +/* Prefer space on the right, not the left of parameter permalinks. */ +.doc-heading-parameter .headerlink { + margin-left: 0 !important; + margin-right: 0.2rem; +} + +/* Backward-compatibility: docstring section titles in bold. */ +.doc-section-title { + font-weight: bold; +} + +/* Symbols in Navigation and ToC. */ +:root, :host, +[data-md-color-scheme="default"] { + --doc-symbol-parameter-fg-color: #df50af; + --doc-symbol-attribute-fg-color: #953800; + --doc-symbol-function-fg-color: #8250df; + --doc-symbol-method-fg-color: #8250df; + --doc-symbol-class-fg-color: #0550ae; + --doc-symbol-module-fg-color: #5cad0f; + + --doc-symbol-parameter-bg-color: #df50af1a; + --doc-symbol-attribute-bg-color: #9538001a; + --doc-symbol-function-bg-color: #8250df1a; + --doc-symbol-method-bg-color: #8250df1a; + --doc-symbol-class-bg-color: #0550ae1a; + --doc-symbol-module-bg-color: #5cad0f1a; +} + +[data-md-color-scheme="slate"] { + --doc-symbol-parameter-fg-color: #ffa8cc; + --doc-symbol-attribute-fg-color: #ffa657; + --doc-symbol-function-fg-color: #d2a8ff; + --doc-symbol-method-fg-color: #d2a8ff; + --doc-symbol-class-fg-color: #79c0ff; + --doc-symbol-module-fg-color: #baff79; + + --doc-symbol-parameter-bg-color: #ffa8cc1a; + --doc-symbol-attribute-bg-color: #ffa6571a; + --doc-symbol-function-bg-color: #d2a8ff1a; + --doc-symbol-method-bg-color: #d2a8ff1a; + --doc-symbol-class-bg-color: #79c0ff1a; + --doc-symbol-module-bg-color: #baff791a; +} + +code.doc-symbol { + border-radius: .1rem; + font-size: .85em; + padding: 0 .3em; + font-weight: bold; +} + +code.doc-symbol-parameter { + color: var(--doc-symbol-parameter-fg-color); + background-color: var(--doc-symbol-parameter-bg-color); +} + +code.doc-symbol-parameter::after { + content: "param"; +} + +code.doc-symbol-attribute { + color: var(--doc-symbol-attribute-fg-color); + background-color: var(--doc-symbol-attribute-bg-color); +} + +code.doc-symbol-attribute::after { + content: "attr"; +} + +code.doc-symbol-function { + color: var(--doc-symbol-function-fg-color); + background-color: var(--doc-symbol-function-bg-color); +} + +code.doc-symbol-function::after { + content: "func"; +} + +code.doc-symbol-method { + color: var(--doc-symbol-method-fg-color); + background-color: var(--doc-symbol-method-bg-color); +} + +code.doc-symbol-method::after { + content: "meth"; +} + +code.doc-symbol-class { + color: var(--doc-symbol-class-fg-color); + background-color: var(--doc-symbol-class-bg-color); +} + +code.doc-symbol-class::after { + content: "class"; +} + +code.doc-symbol-module { + color: var(--doc-symbol-module-fg-color); + background-color: var(--doc-symbol-module-bg-color); +} + +code.doc-symbol-module::after { + content: "mod"; +} + +.doc-signature .autorefs { + color: inherit; + border-bottom: 1px dotted currentcolor; +} diff --git a/assets/images/favicon.png b/assets/images/favicon.png new file mode 100644 index 0000000..1cf13b9 Binary files /dev/null and b/assets/images/favicon.png differ diff --git a/assets/javascripts/bundle.83f73b43.min.js b/assets/javascripts/bundle.83f73b43.min.js new file mode 100644 index 0000000..43d8b70 --- /dev/null +++ b/assets/javascripts/bundle.83f73b43.min.js @@ -0,0 +1,16 @@ +"use strict";(()=>{var Wi=Object.create;var gr=Object.defineProperty;var Di=Object.getOwnPropertyDescriptor;var Vi=Object.getOwnPropertyNames,Vt=Object.getOwnPropertySymbols,Ni=Object.getPrototypeOf,yr=Object.prototype.hasOwnProperty,ao=Object.prototype.propertyIsEnumerable;var io=(e,t,r)=>t in e?gr(e,t,{enumerable:!0,configurable:!0,writable:!0,value:r}):e[t]=r,$=(e,t)=>{for(var r in t||(t={}))yr.call(t,r)&&io(e,r,t[r]);if(Vt)for(var r of Vt(t))ao.call(t,r)&&io(e,r,t[r]);return e};var so=(e,t)=>{var r={};for(var o in e)yr.call(e,o)&&t.indexOf(o)<0&&(r[o]=e[o]);if(e!=null&&Vt)for(var o of Vt(e))t.indexOf(o)<0&&ao.call(e,o)&&(r[o]=e[o]);return r};var xr=(e,t)=>()=>(t||e((t={exports:{}}).exports,t),t.exports);var zi=(e,t,r,o)=>{if(t&&typeof t=="object"||typeof t=="function")for(let n of Vi(t))!yr.call(e,n)&&n!==r&&gr(e,n,{get:()=>t[n],enumerable:!(o=Di(t,n))||o.enumerable});return e};var Mt=(e,t,r)=>(r=e!=null?Wi(Ni(e)):{},zi(t||!e||!e.__esModule?gr(r,"default",{value:e,enumerable:!0}):r,e));var co=(e,t,r)=>new Promise((o,n)=>{var i=p=>{try{s(r.next(p))}catch(c){n(c)}},a=p=>{try{s(r.throw(p))}catch(c){n(c)}},s=p=>p.done?o(p.value):Promise.resolve(p.value).then(i,a);s((r=r.apply(e,t)).next())});var lo=xr((Er,po)=>{(function(e,t){typeof Er=="object"&&typeof po!="undefined"?t():typeof define=="function"&&define.amd?define(t):t()})(Er,function(){"use strict";function e(r){var o=!0,n=!1,i=null,a={text:!0,search:!0,url:!0,tel:!0,email:!0,password:!0,number:!0,date:!0,month:!0,week:!0,time:!0,datetime:!0,"datetime-local":!0};function s(k){return!!(k&&k!==document&&k.nodeName!=="HTML"&&k.nodeName!=="BODY"&&"classList"in k&&"contains"in k.classList)}function p(k){var ft=k.type,qe=k.tagName;return!!(qe==="INPUT"&&a[ft]&&!k.readOnly||qe==="TEXTAREA"&&!k.readOnly||k.isContentEditable)}function c(k){k.classList.contains("focus-visible")||(k.classList.add("focus-visible"),k.setAttribute("data-focus-visible-added",""))}function l(k){k.hasAttribute("data-focus-visible-added")&&(k.classList.remove("focus-visible"),k.removeAttribute("data-focus-visible-added"))}function f(k){k.metaKey||k.altKey||k.ctrlKey||(s(r.activeElement)&&c(r.activeElement),o=!0)}function u(k){o=!1}function d(k){s(k.target)&&(o||p(k.target))&&c(k.target)}function y(k){s(k.target)&&(k.target.classList.contains("focus-visible")||k.target.hasAttribute("data-focus-visible-added"))&&(n=!0,window.clearTimeout(i),i=window.setTimeout(function(){n=!1},100),l(k.target))}function L(k){document.visibilityState==="hidden"&&(n&&(o=!0),X())}function X(){document.addEventListener("mousemove",J),document.addEventListener("mousedown",J),document.addEventListener("mouseup",J),document.addEventListener("pointermove",J),document.addEventListener("pointerdown",J),document.addEventListener("pointerup",J),document.addEventListener("touchmove",J),document.addEventListener("touchstart",J),document.addEventListener("touchend",J)}function te(){document.removeEventListener("mousemove",J),document.removeEventListener("mousedown",J),document.removeEventListener("mouseup",J),document.removeEventListener("pointermove",J),document.removeEventListener("pointerdown",J),document.removeEventListener("pointerup",J),document.removeEventListener("touchmove",J),document.removeEventListener("touchstart",J),document.removeEventListener("touchend",J)}function J(k){k.target.nodeName&&k.target.nodeName.toLowerCase()==="html"||(o=!1,te())}document.addEventListener("keydown",f,!0),document.addEventListener("mousedown",u,!0),document.addEventListener("pointerdown",u,!0),document.addEventListener("touchstart",u,!0),document.addEventListener("visibilitychange",L,!0),X(),r.addEventListener("focus",d,!0),r.addEventListener("blur",y,!0),r.nodeType===Node.DOCUMENT_FRAGMENT_NODE&&r.host?r.host.setAttribute("data-js-focus-visible",""):r.nodeType===Node.DOCUMENT_NODE&&(document.documentElement.classList.add("js-focus-visible"),document.documentElement.setAttribute("data-js-focus-visible",""))}if(typeof window!="undefined"&&typeof document!="undefined"){window.applyFocusVisiblePolyfill=e;var t;try{t=new CustomEvent("focus-visible-polyfill-ready")}catch(r){t=document.createEvent("CustomEvent"),t.initCustomEvent("focus-visible-polyfill-ready",!1,!1,{})}window.dispatchEvent(t)}typeof document!="undefined"&&e(document)})});var qr=xr((hy,On)=>{"use strict";/*! + * escape-html + * Copyright(c) 2012-2013 TJ Holowaychuk + * Copyright(c) 2015 Andreas Lubbe + * Copyright(c) 2015 Tiancheng "Timothy" Gu + * MIT Licensed + */var $a=/["'&<>]/;On.exports=Pa;function Pa(e){var t=""+e,r=$a.exec(t);if(!r)return t;var o,n="",i=0,a=0;for(i=r.index;i{/*! + * clipboard.js v2.0.11 + * https://clipboardjs.com/ + * + * Licensed MIT © Zeno Rocha + */(function(t,r){typeof It=="object"&&typeof Yr=="object"?Yr.exports=r():typeof define=="function"&&define.amd?define([],r):typeof It=="object"?It.ClipboardJS=r():t.ClipboardJS=r()})(It,function(){return function(){var e={686:function(o,n,i){"use strict";i.d(n,{default:function(){return Ui}});var a=i(279),s=i.n(a),p=i(370),c=i.n(p),l=i(817),f=i.n(l);function u(V){try{return document.execCommand(V)}catch(A){return!1}}var d=function(A){var M=f()(A);return u("cut"),M},y=d;function L(V){var A=document.documentElement.getAttribute("dir")==="rtl",M=document.createElement("textarea");M.style.fontSize="12pt",M.style.border="0",M.style.padding="0",M.style.margin="0",M.style.position="absolute",M.style[A?"right":"left"]="-9999px";var F=window.pageYOffset||document.documentElement.scrollTop;return M.style.top="".concat(F,"px"),M.setAttribute("readonly",""),M.value=V,M}var X=function(A,M){var F=L(A);M.container.appendChild(F);var D=f()(F);return u("copy"),F.remove(),D},te=function(A){var M=arguments.length>1&&arguments[1]!==void 0?arguments[1]:{container:document.body},F="";return typeof A=="string"?F=X(A,M):A instanceof HTMLInputElement&&!["text","search","url","tel","password"].includes(A==null?void 0:A.type)?F=X(A.value,M):(F=f()(A),u("copy")),F},J=te;function k(V){"@babel/helpers - typeof";return typeof Symbol=="function"&&typeof Symbol.iterator=="symbol"?k=function(M){return typeof M}:k=function(M){return M&&typeof Symbol=="function"&&M.constructor===Symbol&&M!==Symbol.prototype?"symbol":typeof M},k(V)}var ft=function(){var A=arguments.length>0&&arguments[0]!==void 0?arguments[0]:{},M=A.action,F=M===void 0?"copy":M,D=A.container,Y=A.target,$e=A.text;if(F!=="copy"&&F!=="cut")throw new Error('Invalid "action" value, use either "copy" or "cut"');if(Y!==void 0)if(Y&&k(Y)==="object"&&Y.nodeType===1){if(F==="copy"&&Y.hasAttribute("disabled"))throw new Error('Invalid "target" attribute. Please use "readonly" instead of "disabled" attribute');if(F==="cut"&&(Y.hasAttribute("readonly")||Y.hasAttribute("disabled")))throw new Error(`Invalid "target" attribute. You can't cut text from elements with "readonly" or "disabled" attributes`)}else throw new Error('Invalid "target" value, use a valid Element');if($e)return J($e,{container:D});if(Y)return F==="cut"?y(Y):J(Y,{container:D})},qe=ft;function Fe(V){"@babel/helpers - typeof";return typeof Symbol=="function"&&typeof Symbol.iterator=="symbol"?Fe=function(M){return typeof M}:Fe=function(M){return M&&typeof Symbol=="function"&&M.constructor===Symbol&&M!==Symbol.prototype?"symbol":typeof M},Fe(V)}function ki(V,A){if(!(V instanceof A))throw new TypeError("Cannot call a class as a function")}function no(V,A){for(var M=0;M0&&arguments[0]!==void 0?arguments[0]:{};this.action=typeof D.action=="function"?D.action:this.defaultAction,this.target=typeof D.target=="function"?D.target:this.defaultTarget,this.text=typeof D.text=="function"?D.text:this.defaultText,this.container=Fe(D.container)==="object"?D.container:document.body}},{key:"listenClick",value:function(D){var Y=this;this.listener=c()(D,"click",function($e){return Y.onClick($e)})}},{key:"onClick",value:function(D){var Y=D.delegateTarget||D.currentTarget,$e=this.action(Y)||"copy",Dt=qe({action:$e,container:this.container,target:this.target(Y),text:this.text(Y)});this.emit(Dt?"success":"error",{action:$e,text:Dt,trigger:Y,clearSelection:function(){Y&&Y.focus(),window.getSelection().removeAllRanges()}})}},{key:"defaultAction",value:function(D){return vr("action",D)}},{key:"defaultTarget",value:function(D){var Y=vr("target",D);if(Y)return document.querySelector(Y)}},{key:"defaultText",value:function(D){return vr("text",D)}},{key:"destroy",value:function(){this.listener.destroy()}}],[{key:"copy",value:function(D){var Y=arguments.length>1&&arguments[1]!==void 0?arguments[1]:{container:document.body};return J(D,Y)}},{key:"cut",value:function(D){return y(D)}},{key:"isSupported",value:function(){var D=arguments.length>0&&arguments[0]!==void 0?arguments[0]:["copy","cut"],Y=typeof D=="string"?[D]:D,$e=!!document.queryCommandSupported;return Y.forEach(function(Dt){$e=$e&&!!document.queryCommandSupported(Dt)}),$e}}]),M}(s()),Ui=Fi},828:function(o){var n=9;if(typeof Element!="undefined"&&!Element.prototype.matches){var i=Element.prototype;i.matches=i.matchesSelector||i.mozMatchesSelector||i.msMatchesSelector||i.oMatchesSelector||i.webkitMatchesSelector}function a(s,p){for(;s&&s.nodeType!==n;){if(typeof s.matches=="function"&&s.matches(p))return s;s=s.parentNode}}o.exports=a},438:function(o,n,i){var a=i(828);function s(l,f,u,d,y){var L=c.apply(this,arguments);return l.addEventListener(u,L,y),{destroy:function(){l.removeEventListener(u,L,y)}}}function p(l,f,u,d,y){return typeof l.addEventListener=="function"?s.apply(null,arguments):typeof u=="function"?s.bind(null,document).apply(null,arguments):(typeof l=="string"&&(l=document.querySelectorAll(l)),Array.prototype.map.call(l,function(L){return s(L,f,u,d,y)}))}function c(l,f,u,d){return function(y){y.delegateTarget=a(y.target,f),y.delegateTarget&&d.call(l,y)}}o.exports=p},879:function(o,n){n.node=function(i){return i!==void 0&&i instanceof HTMLElement&&i.nodeType===1},n.nodeList=function(i){var a=Object.prototype.toString.call(i);return i!==void 0&&(a==="[object NodeList]"||a==="[object HTMLCollection]")&&"length"in i&&(i.length===0||n.node(i[0]))},n.string=function(i){return typeof i=="string"||i instanceof String},n.fn=function(i){var a=Object.prototype.toString.call(i);return a==="[object Function]"}},370:function(o,n,i){var a=i(879),s=i(438);function p(u,d,y){if(!u&&!d&&!y)throw new Error("Missing required arguments");if(!a.string(d))throw new TypeError("Second argument must be a String");if(!a.fn(y))throw new TypeError("Third argument must be a Function");if(a.node(u))return c(u,d,y);if(a.nodeList(u))return l(u,d,y);if(a.string(u))return f(u,d,y);throw new TypeError("First argument must be a String, HTMLElement, HTMLCollection, or NodeList")}function c(u,d,y){return u.addEventListener(d,y),{destroy:function(){u.removeEventListener(d,y)}}}function l(u,d,y){return Array.prototype.forEach.call(u,function(L){L.addEventListener(d,y)}),{destroy:function(){Array.prototype.forEach.call(u,function(L){L.removeEventListener(d,y)})}}}function f(u,d,y){return s(document.body,u,d,y)}o.exports=p},817:function(o){function n(i){var a;if(i.nodeName==="SELECT")i.focus(),a=i.value;else if(i.nodeName==="INPUT"||i.nodeName==="TEXTAREA"){var s=i.hasAttribute("readonly");s||i.setAttribute("readonly",""),i.select(),i.setSelectionRange(0,i.value.length),s||i.removeAttribute("readonly"),a=i.value}else{i.hasAttribute("contenteditable")&&i.focus();var p=window.getSelection(),c=document.createRange();c.selectNodeContents(i),p.removeAllRanges(),p.addRange(c),a=p.toString()}return a}o.exports=n},279:function(o){function n(){}n.prototype={on:function(i,a,s){var p=this.e||(this.e={});return(p[i]||(p[i]=[])).push({fn:a,ctx:s}),this},once:function(i,a,s){var p=this;function c(){p.off(i,c),a.apply(s,arguments)}return c._=a,this.on(i,c,s)},emit:function(i){var a=[].slice.call(arguments,1),s=((this.e||(this.e={}))[i]||[]).slice(),p=0,c=s.length;for(p;p0&&i[i.length-1])&&(c[0]===6||c[0]===2)){r=0;continue}if(c[0]===3&&(!i||c[1]>i[0]&&c[1]=e.length&&(e=void 0),{value:e&&e[o++],done:!e}}};throw new TypeError(t?"Object is not iterable.":"Symbol.iterator is not defined.")}function N(e,t){var r=typeof Symbol=="function"&&e[Symbol.iterator];if(!r)return e;var o=r.call(e),n,i=[],a;try{for(;(t===void 0||t-- >0)&&!(n=o.next()).done;)i.push(n.value)}catch(s){a={error:s}}finally{try{n&&!n.done&&(r=o.return)&&r.call(o)}finally{if(a)throw a.error}}return i}function q(e,t,r){if(r||arguments.length===2)for(var o=0,n=t.length,i;o1||p(d,L)})},y&&(n[d]=y(n[d])))}function p(d,y){try{c(o[d](y))}catch(L){u(i[0][3],L)}}function c(d){d.value instanceof nt?Promise.resolve(d.value.v).then(l,f):u(i[0][2],d)}function l(d){p("next",d)}function f(d){p("throw",d)}function u(d,y){d(y),i.shift(),i.length&&p(i[0][0],i[0][1])}}function uo(e){if(!Symbol.asyncIterator)throw new TypeError("Symbol.asyncIterator is not defined.");var t=e[Symbol.asyncIterator],r;return t?t.call(e):(e=typeof he=="function"?he(e):e[Symbol.iterator](),r={},o("next"),o("throw"),o("return"),r[Symbol.asyncIterator]=function(){return this},r);function o(i){r[i]=e[i]&&function(a){return new Promise(function(s,p){a=e[i](a),n(s,p,a.done,a.value)})}}function n(i,a,s,p){Promise.resolve(p).then(function(c){i({value:c,done:s})},a)}}function H(e){return typeof e=="function"}function ut(e){var t=function(o){Error.call(o),o.stack=new Error().stack},r=e(t);return r.prototype=Object.create(Error.prototype),r.prototype.constructor=r,r}var zt=ut(function(e){return function(r){e(this),this.message=r?r.length+` errors occurred during unsubscription: +`+r.map(function(o,n){return n+1+") "+o.toString()}).join(` + `):"",this.name="UnsubscriptionError",this.errors=r}});function Qe(e,t){if(e){var r=e.indexOf(t);0<=r&&e.splice(r,1)}}var Ue=function(){function e(t){this.initialTeardown=t,this.closed=!1,this._parentage=null,this._finalizers=null}return e.prototype.unsubscribe=function(){var t,r,o,n,i;if(!this.closed){this.closed=!0;var a=this._parentage;if(a)if(this._parentage=null,Array.isArray(a))try{for(var s=he(a),p=s.next();!p.done;p=s.next()){var c=p.value;c.remove(this)}}catch(L){t={error:L}}finally{try{p&&!p.done&&(r=s.return)&&r.call(s)}finally{if(t)throw t.error}}else a.remove(this);var l=this.initialTeardown;if(H(l))try{l()}catch(L){i=L instanceof zt?L.errors:[L]}var f=this._finalizers;if(f){this._finalizers=null;try{for(var u=he(f),d=u.next();!d.done;d=u.next()){var y=d.value;try{ho(y)}catch(L){i=i!=null?i:[],L instanceof zt?i=q(q([],N(i)),N(L.errors)):i.push(L)}}}catch(L){o={error:L}}finally{try{d&&!d.done&&(n=u.return)&&n.call(u)}finally{if(o)throw o.error}}}if(i)throw new zt(i)}},e.prototype.add=function(t){var r;if(t&&t!==this)if(this.closed)ho(t);else{if(t instanceof e){if(t.closed||t._hasParent(this))return;t._addParent(this)}(this._finalizers=(r=this._finalizers)!==null&&r!==void 0?r:[]).push(t)}},e.prototype._hasParent=function(t){var r=this._parentage;return r===t||Array.isArray(r)&&r.includes(t)},e.prototype._addParent=function(t){var r=this._parentage;this._parentage=Array.isArray(r)?(r.push(t),r):r?[r,t]:t},e.prototype._removeParent=function(t){var r=this._parentage;r===t?this._parentage=null:Array.isArray(r)&&Qe(r,t)},e.prototype.remove=function(t){var r=this._finalizers;r&&Qe(r,t),t instanceof e&&t._removeParent(this)},e.EMPTY=function(){var t=new e;return t.closed=!0,t}(),e}();var Tr=Ue.EMPTY;function qt(e){return e instanceof Ue||e&&"closed"in e&&H(e.remove)&&H(e.add)&&H(e.unsubscribe)}function ho(e){H(e)?e():e.unsubscribe()}var Pe={onUnhandledError:null,onStoppedNotification:null,Promise:void 0,useDeprecatedSynchronousErrorHandling:!1,useDeprecatedNextContext:!1};var dt={setTimeout:function(e,t){for(var r=[],o=2;o0},enumerable:!1,configurable:!0}),t.prototype._trySubscribe=function(r){return this._throwIfClosed(),e.prototype._trySubscribe.call(this,r)},t.prototype._subscribe=function(r){return this._throwIfClosed(),this._checkFinalizedStatuses(r),this._innerSubscribe(r)},t.prototype._innerSubscribe=function(r){var o=this,n=this,i=n.hasError,a=n.isStopped,s=n.observers;return i||a?Tr:(this.currentObservers=null,s.push(r),new Ue(function(){o.currentObservers=null,Qe(s,r)}))},t.prototype._checkFinalizedStatuses=function(r){var o=this,n=o.hasError,i=o.thrownError,a=o.isStopped;n?r.error(i):a&&r.complete()},t.prototype.asObservable=function(){var r=new j;return r.source=this,r},t.create=function(r,o){return new To(r,o)},t}(j);var To=function(e){oe(t,e);function t(r,o){var n=e.call(this)||this;return n.destination=r,n.source=o,n}return t.prototype.next=function(r){var o,n;(n=(o=this.destination)===null||o===void 0?void 0:o.next)===null||n===void 0||n.call(o,r)},t.prototype.error=function(r){var o,n;(n=(o=this.destination)===null||o===void 0?void 0:o.error)===null||n===void 0||n.call(o,r)},t.prototype.complete=function(){var r,o;(o=(r=this.destination)===null||r===void 0?void 0:r.complete)===null||o===void 0||o.call(r)},t.prototype._subscribe=function(r){var o,n;return(n=(o=this.source)===null||o===void 0?void 0:o.subscribe(r))!==null&&n!==void 0?n:Tr},t}(g);var _r=function(e){oe(t,e);function t(r){var o=e.call(this)||this;return o._value=r,o}return Object.defineProperty(t.prototype,"value",{get:function(){return this.getValue()},enumerable:!1,configurable:!0}),t.prototype._subscribe=function(r){var o=e.prototype._subscribe.call(this,r);return!o.closed&&r.next(this._value),o},t.prototype.getValue=function(){var r=this,o=r.hasError,n=r.thrownError,i=r._value;if(o)throw n;return this._throwIfClosed(),i},t.prototype.next=function(r){e.prototype.next.call(this,this._value=r)},t}(g);var At={now:function(){return(At.delegate||Date).now()},delegate:void 0};var Ct=function(e){oe(t,e);function t(r,o,n){r===void 0&&(r=1/0),o===void 0&&(o=1/0),n===void 0&&(n=At);var i=e.call(this)||this;return i._bufferSize=r,i._windowTime=o,i._timestampProvider=n,i._buffer=[],i._infiniteTimeWindow=!0,i._infiniteTimeWindow=o===1/0,i._bufferSize=Math.max(1,r),i._windowTime=Math.max(1,o),i}return t.prototype.next=function(r){var o=this,n=o.isStopped,i=o._buffer,a=o._infiniteTimeWindow,s=o._timestampProvider,p=o._windowTime;n||(i.push(r),!a&&i.push(s.now()+p)),this._trimBuffer(),e.prototype.next.call(this,r)},t.prototype._subscribe=function(r){this._throwIfClosed(),this._trimBuffer();for(var o=this._innerSubscribe(r),n=this,i=n._infiniteTimeWindow,a=n._buffer,s=a.slice(),p=0;p0?e.prototype.schedule.call(this,r,o):(this.delay=o,this.state=r,this.scheduler.flush(this),this)},t.prototype.execute=function(r,o){return o>0||this.closed?e.prototype.execute.call(this,r,o):this._execute(r,o)},t.prototype.requestAsyncId=function(r,o,n){return n===void 0&&(n=0),n!=null&&n>0||n==null&&this.delay>0?e.prototype.requestAsyncId.call(this,r,o,n):(r.flush(this),0)},t}(gt);var Lo=function(e){oe(t,e);function t(){return e!==null&&e.apply(this,arguments)||this}return t}(yt);var kr=new Lo(Oo);var Mo=function(e){oe(t,e);function t(r,o){var n=e.call(this,r,o)||this;return n.scheduler=r,n.work=o,n}return t.prototype.requestAsyncId=function(r,o,n){return n===void 0&&(n=0),n!==null&&n>0?e.prototype.requestAsyncId.call(this,r,o,n):(r.actions.push(this),r._scheduled||(r._scheduled=vt.requestAnimationFrame(function(){return r.flush(void 0)})))},t.prototype.recycleAsyncId=function(r,o,n){var i;if(n===void 0&&(n=0),n!=null?n>0:this.delay>0)return e.prototype.recycleAsyncId.call(this,r,o,n);var a=r.actions;o!=null&&((i=a[a.length-1])===null||i===void 0?void 0:i.id)!==o&&(vt.cancelAnimationFrame(o),r._scheduled=void 0)},t}(gt);var _o=function(e){oe(t,e);function t(){return e!==null&&e.apply(this,arguments)||this}return t.prototype.flush=function(r){this._active=!0;var o=this._scheduled;this._scheduled=void 0;var n=this.actions,i;r=r||n.shift();do if(i=r.execute(r.state,r.delay))break;while((r=n[0])&&r.id===o&&n.shift());if(this._active=!1,i){for(;(r=n[0])&&r.id===o&&n.shift();)r.unsubscribe();throw i}},t}(yt);var me=new _o(Mo);var S=new j(function(e){return e.complete()});function Yt(e){return e&&H(e.schedule)}function Hr(e){return e[e.length-1]}function Xe(e){return H(Hr(e))?e.pop():void 0}function ke(e){return Yt(Hr(e))?e.pop():void 0}function Bt(e,t){return typeof Hr(e)=="number"?e.pop():t}var xt=function(e){return e&&typeof e.length=="number"&&typeof e!="function"};function Gt(e){return H(e==null?void 0:e.then)}function Jt(e){return H(e[bt])}function Xt(e){return Symbol.asyncIterator&&H(e==null?void 0:e[Symbol.asyncIterator])}function Zt(e){return new TypeError("You provided "+(e!==null&&typeof e=="object"?"an invalid object":"'"+e+"'")+" where a stream was expected. You can provide an Observable, Promise, ReadableStream, Array, AsyncIterable, or Iterable.")}function Zi(){return typeof Symbol!="function"||!Symbol.iterator?"@@iterator":Symbol.iterator}var er=Zi();function tr(e){return H(e==null?void 0:e[er])}function rr(e){return fo(this,arguments,function(){var r,o,n,i;return Nt(this,function(a){switch(a.label){case 0:r=e.getReader(),a.label=1;case 1:a.trys.push([1,,9,10]),a.label=2;case 2:return[4,nt(r.read())];case 3:return o=a.sent(),n=o.value,i=o.done,i?[4,nt(void 0)]:[3,5];case 4:return[2,a.sent()];case 5:return[4,nt(n)];case 6:return[4,a.sent()];case 7:return a.sent(),[3,2];case 8:return[3,10];case 9:return r.releaseLock(),[7];case 10:return[2]}})})}function or(e){return H(e==null?void 0:e.getReader)}function U(e){if(e instanceof j)return e;if(e!=null){if(Jt(e))return ea(e);if(xt(e))return ta(e);if(Gt(e))return ra(e);if(Xt(e))return Ao(e);if(tr(e))return oa(e);if(or(e))return na(e)}throw Zt(e)}function ea(e){return new j(function(t){var r=e[bt]();if(H(r.subscribe))return r.subscribe(t);throw new TypeError("Provided object does not correctly implement Symbol.observable")})}function ta(e){return new j(function(t){for(var r=0;r=2;return function(o){return o.pipe(e?b(function(n,i){return e(n,i,o)}):le,Te(1),r?De(t):Qo(function(){return new ir}))}}function jr(e){return e<=0?function(){return S}:E(function(t,r){var o=[];t.subscribe(T(r,function(n){o.push(n),e=2,!0))}function pe(e){e===void 0&&(e={});var t=e.connector,r=t===void 0?function(){return new g}:t,o=e.resetOnError,n=o===void 0?!0:o,i=e.resetOnComplete,a=i===void 0?!0:i,s=e.resetOnRefCountZero,p=s===void 0?!0:s;return function(c){var l,f,u,d=0,y=!1,L=!1,X=function(){f==null||f.unsubscribe(),f=void 0},te=function(){X(),l=u=void 0,y=L=!1},J=function(){var k=l;te(),k==null||k.unsubscribe()};return E(function(k,ft){d++,!L&&!y&&X();var qe=u=u!=null?u:r();ft.add(function(){d--,d===0&&!L&&!y&&(f=Ur(J,p))}),qe.subscribe(ft),!l&&d>0&&(l=new at({next:function(Fe){return qe.next(Fe)},error:function(Fe){L=!0,X(),f=Ur(te,n,Fe),qe.error(Fe)},complete:function(){y=!0,X(),f=Ur(te,a),qe.complete()}}),U(k).subscribe(l))})(c)}}function Ur(e,t){for(var r=[],o=2;oe.next(document)),e}function P(e,t=document){return Array.from(t.querySelectorAll(e))}function R(e,t=document){let r=fe(e,t);if(typeof r=="undefined")throw new ReferenceError(`Missing element: expected "${e}" to be present`);return r}function fe(e,t=document){return t.querySelector(e)||void 0}function Ie(){var e,t,r,o;return(o=(r=(t=(e=document.activeElement)==null?void 0:e.shadowRoot)==null?void 0:t.activeElement)!=null?r:document.activeElement)!=null?o:void 0}var wa=O(h(document.body,"focusin"),h(document.body,"focusout")).pipe(_e(1),Q(void 0),m(()=>Ie()||document.body),G(1));function et(e){return wa.pipe(m(t=>e.contains(t)),K())}function $t(e,t){return C(()=>O(h(e,"mouseenter").pipe(m(()=>!0)),h(e,"mouseleave").pipe(m(()=>!1))).pipe(t?Ht(r=>Le(+!r*t)):le,Q(e.matches(":hover"))))}function Jo(e,t){if(typeof t=="string"||typeof t=="number")e.innerHTML+=t.toString();else if(t instanceof Node)e.appendChild(t);else if(Array.isArray(t))for(let r of t)Jo(e,r)}function x(e,t,...r){let o=document.createElement(e);if(t)for(let n of Object.keys(t))typeof t[n]!="undefined"&&(typeof t[n]!="boolean"?o.setAttribute(n,t[n]):o.setAttribute(n,""));for(let n of r)Jo(o,n);return o}function sr(e){if(e>999){let t=+((e-950)%1e3>99);return`${((e+1e-6)/1e3).toFixed(t)}k`}else return e.toString()}function Tt(e){let t=x("script",{src:e});return C(()=>(document.head.appendChild(t),O(h(t,"load"),h(t,"error").pipe(v(()=>$r(()=>new ReferenceError(`Invalid script: ${e}`))))).pipe(m(()=>{}),_(()=>document.head.removeChild(t)),Te(1))))}var Xo=new g,Ta=C(()=>typeof ResizeObserver=="undefined"?Tt("https://unpkg.com/resize-observer-polyfill"):I(void 0)).pipe(m(()=>new ResizeObserver(e=>e.forEach(t=>Xo.next(t)))),v(e=>O(Ye,I(e)).pipe(_(()=>e.disconnect()))),G(1));function ce(e){return{width:e.offsetWidth,height:e.offsetHeight}}function ge(e){let t=e;for(;t.clientWidth===0&&t.parentElement;)t=t.parentElement;return Ta.pipe(w(r=>r.observe(t)),v(r=>Xo.pipe(b(o=>o.target===t),_(()=>r.unobserve(t)))),m(()=>ce(e)),Q(ce(e)))}function St(e){return{width:e.scrollWidth,height:e.scrollHeight}}function cr(e){let t=e.parentElement;for(;t&&(e.scrollWidth<=t.scrollWidth&&e.scrollHeight<=t.scrollHeight);)t=(e=t).parentElement;return t?e:void 0}function Zo(e){let t=[],r=e.parentElement;for(;r;)(e.clientWidth>r.clientWidth||e.clientHeight>r.clientHeight)&&t.push(r),r=(e=r).parentElement;return t.length===0&&t.push(document.documentElement),t}function Ve(e){return{x:e.offsetLeft,y:e.offsetTop}}function en(e){let t=e.getBoundingClientRect();return{x:t.x+window.scrollX,y:t.y+window.scrollY}}function tn(e){return O(h(window,"load"),h(window,"resize")).pipe(Me(0,me),m(()=>Ve(e)),Q(Ve(e)))}function pr(e){return{x:e.scrollLeft,y:e.scrollTop}}function Ne(e){return O(h(e,"scroll"),h(window,"scroll"),h(window,"resize")).pipe(Me(0,me),m(()=>pr(e)),Q(pr(e)))}var rn=new g,Sa=C(()=>I(new IntersectionObserver(e=>{for(let t of e)rn.next(t)},{threshold:0}))).pipe(v(e=>O(Ye,I(e)).pipe(_(()=>e.disconnect()))),G(1));function tt(e){return Sa.pipe(w(t=>t.observe(e)),v(t=>rn.pipe(b(({target:r})=>r===e),_(()=>t.unobserve(e)),m(({isIntersecting:r})=>r))))}function on(e,t=16){return Ne(e).pipe(m(({y:r})=>{let o=ce(e),n=St(e);return r>=n.height-o.height-t}),K())}var lr={drawer:R("[data-md-toggle=drawer]"),search:R("[data-md-toggle=search]")};function nn(e){return lr[e].checked}function Je(e,t){lr[e].checked!==t&&lr[e].click()}function ze(e){let t=lr[e];return h(t,"change").pipe(m(()=>t.checked),Q(t.checked))}function Oa(e,t){switch(e.constructor){case HTMLInputElement:return e.type==="radio"?/^Arrow/.test(t):!0;case HTMLSelectElement:case HTMLTextAreaElement:return!0;default:return e.isContentEditable}}function La(){return O(h(window,"compositionstart").pipe(m(()=>!0)),h(window,"compositionend").pipe(m(()=>!1))).pipe(Q(!1))}function an(){let e=h(window,"keydown").pipe(b(t=>!(t.metaKey||t.ctrlKey)),m(t=>({mode:nn("search")?"search":"global",type:t.key,claim(){t.preventDefault(),t.stopPropagation()}})),b(({mode:t,type:r})=>{if(t==="global"){let o=Ie();if(typeof o!="undefined")return!Oa(o,r)}return!0}),pe());return La().pipe(v(t=>t?S:e))}function ye(){return new URL(location.href)}function lt(e,t=!1){if(B("navigation.instant")&&!t){let r=x("a",{href:e.href});document.body.appendChild(r),r.click(),r.remove()}else location.href=e.href}function sn(){return new g}function cn(){return location.hash.slice(1)}function pn(e){let t=x("a",{href:e});t.addEventListener("click",r=>r.stopPropagation()),t.click()}function Ma(e){return O(h(window,"hashchange"),e).pipe(m(cn),Q(cn()),b(t=>t.length>0),G(1))}function ln(e){return Ma(e).pipe(m(t=>fe(`[id="${t}"]`)),b(t=>typeof t!="undefined"))}function Pt(e){let t=matchMedia(e);return ar(r=>t.addListener(()=>r(t.matches))).pipe(Q(t.matches))}function mn(){let e=matchMedia("print");return O(h(window,"beforeprint").pipe(m(()=>!0)),h(window,"afterprint").pipe(m(()=>!1))).pipe(Q(e.matches))}function Nr(e,t){return e.pipe(v(r=>r?t():S))}function zr(e,t){return new j(r=>{let o=new XMLHttpRequest;return o.open("GET",`${e}`),o.responseType="blob",o.addEventListener("load",()=>{o.status>=200&&o.status<300?(r.next(o.response),r.complete()):r.error(new Error(o.statusText))}),o.addEventListener("error",()=>{r.error(new Error("Network error"))}),o.addEventListener("abort",()=>{r.complete()}),typeof(t==null?void 0:t.progress$)!="undefined"&&(o.addEventListener("progress",n=>{var i;if(n.lengthComputable)t.progress$.next(n.loaded/n.total*100);else{let a=(i=o.getResponseHeader("Content-Length"))!=null?i:0;t.progress$.next(n.loaded/+a*100)}}),t.progress$.next(5)),o.send(),()=>o.abort()})}function je(e,t){return zr(e,t).pipe(v(r=>r.text()),m(r=>JSON.parse(r)),G(1))}function fn(e,t){let r=new DOMParser;return zr(e,t).pipe(v(o=>o.text()),m(o=>r.parseFromString(o,"text/html")),G(1))}function un(e,t){let r=new DOMParser;return zr(e,t).pipe(v(o=>o.text()),m(o=>r.parseFromString(o,"text/xml")),G(1))}function dn(){return{x:Math.max(0,scrollX),y:Math.max(0,scrollY)}}function hn(){return O(h(window,"scroll",{passive:!0}),h(window,"resize",{passive:!0})).pipe(m(dn),Q(dn()))}function bn(){return{width:innerWidth,height:innerHeight}}function vn(){return h(window,"resize",{passive:!0}).pipe(m(bn),Q(bn()))}function gn(){return z([hn(),vn()]).pipe(m(([e,t])=>({offset:e,size:t})),G(1))}function mr(e,{viewport$:t,header$:r}){let o=t.pipe(ee("size")),n=z([o,r]).pipe(m(()=>Ve(e)));return z([r,t,n]).pipe(m(([{height:i},{offset:a,size:s},{x:p,y:c}])=>({offset:{x:a.x-p,y:a.y-c+i},size:s})))}function _a(e){return h(e,"message",t=>t.data)}function Aa(e){let t=new g;return t.subscribe(r=>e.postMessage(r)),t}function yn(e,t=new Worker(e)){let r=_a(t),o=Aa(t),n=new g;n.subscribe(o);let i=o.pipe(Z(),ie(!0));return n.pipe(Z(),Re(r.pipe(W(i))),pe())}var Ca=R("#__config"),Ot=JSON.parse(Ca.textContent);Ot.base=`${new URL(Ot.base,ye())}`;function xe(){return Ot}function B(e){return Ot.features.includes(e)}function Ee(e,t){return typeof t!="undefined"?Ot.translations[e].replace("#",t.toString()):Ot.translations[e]}function Se(e,t=document){return R(`[data-md-component=${e}]`,t)}function ae(e,t=document){return P(`[data-md-component=${e}]`,t)}function ka(e){let t=R(".md-typeset > :first-child",e);return h(t,"click",{once:!0}).pipe(m(()=>R(".md-typeset",e)),m(r=>({hash:__md_hash(r.innerHTML)})))}function xn(e){if(!B("announce.dismiss")||!e.childElementCount)return S;if(!e.hidden){let t=R(".md-typeset",e);__md_hash(t.innerHTML)===__md_get("__announce")&&(e.hidden=!0)}return C(()=>{let t=new g;return t.subscribe(({hash:r})=>{e.hidden=!0,__md_set("__announce",r)}),ka(e).pipe(w(r=>t.next(r)),_(()=>t.complete()),m(r=>$({ref:e},r)))})}function Ha(e,{target$:t}){return t.pipe(m(r=>({hidden:r!==e})))}function En(e,t){let r=new g;return r.subscribe(({hidden:o})=>{e.hidden=o}),Ha(e,t).pipe(w(o=>r.next(o)),_(()=>r.complete()),m(o=>$({ref:e},o)))}function Rt(e,t){return t==="inline"?x("div",{class:"md-tooltip md-tooltip--inline",id:e,role:"tooltip"},x("div",{class:"md-tooltip__inner md-typeset"})):x("div",{class:"md-tooltip",id:e,role:"tooltip"},x("div",{class:"md-tooltip__inner md-typeset"}))}function wn(...e){return x("div",{class:"md-tooltip2",role:"tooltip"},x("div",{class:"md-tooltip2__inner md-typeset"},e))}function Tn(e,t){if(t=t?`${t}_annotation_${e}`:void 0,t){let r=t?`#${t}`:void 0;return x("aside",{class:"md-annotation",tabIndex:0},Rt(t),x("a",{href:r,class:"md-annotation__index",tabIndex:-1},x("span",{"data-md-annotation-id":e})))}else return x("aside",{class:"md-annotation",tabIndex:0},Rt(t),x("span",{class:"md-annotation__index",tabIndex:-1},x("span",{"data-md-annotation-id":e})))}function Sn(e){return x("button",{class:"md-clipboard md-icon",title:Ee("clipboard.copy"),"data-clipboard-target":`#${e} > code`})}var Ln=Mt(qr());function Qr(e,t){let r=t&2,o=t&1,n=Object.keys(e.terms).filter(p=>!e.terms[p]).reduce((p,c)=>[...p,x("del",null,(0,Ln.default)(c))," "],[]).slice(0,-1),i=xe(),a=new URL(e.location,i.base);B("search.highlight")&&a.searchParams.set("h",Object.entries(e.terms).filter(([,p])=>p).reduce((p,[c])=>`${p} ${c}`.trim(),""));let{tags:s}=xe();return x("a",{href:`${a}`,class:"md-search-result__link",tabIndex:-1},x("article",{class:"md-search-result__article md-typeset","data-md-score":e.score.toFixed(2)},r>0&&x("div",{class:"md-search-result__icon md-icon"}),r>0&&x("h1",null,e.title),r<=0&&x("h2",null,e.title),o>0&&e.text.length>0&&e.text,e.tags&&x("nav",{class:"md-tags"},e.tags.map(p=>{let c=s?p in s?`md-tag-icon md-tag--${s[p]}`:"md-tag-icon":"";return x("span",{class:`md-tag ${c}`},p)})),o>0&&n.length>0&&x("p",{class:"md-search-result__terms"},Ee("search.result.term.missing"),": ",...n)))}function Mn(e){let t=e[0].score,r=[...e],o=xe(),n=r.findIndex(l=>!`${new URL(l.location,o.base)}`.includes("#")),[i]=r.splice(n,1),a=r.findIndex(l=>l.scoreQr(l,1)),...p.length?[x("details",{class:"md-search-result__more"},x("summary",{tabIndex:-1},x("div",null,p.length>0&&p.length===1?Ee("search.result.more.one"):Ee("search.result.more.other",p.length))),...p.map(l=>Qr(l,1)))]:[]];return x("li",{class:"md-search-result__item"},c)}function _n(e){return x("ul",{class:"md-source__facts"},Object.entries(e).map(([t,r])=>x("li",{class:`md-source__fact md-source__fact--${t}`},typeof r=="number"?sr(r):r)))}function Kr(e){let t=`tabbed-control tabbed-control--${e}`;return x("div",{class:t,hidden:!0},x("button",{class:"tabbed-button",tabIndex:-1,"aria-hidden":"true"}))}function An(e){return x("div",{class:"md-typeset__scrollwrap"},x("div",{class:"md-typeset__table"},e))}function Ra(e){var o;let t=xe(),r=new URL(`../${e.version}/`,t.base);return x("li",{class:"md-version__item"},x("a",{href:`${r}`,class:"md-version__link"},e.title,((o=t.version)==null?void 0:o.alias)&&e.aliases.length>0&&x("span",{class:"md-version__alias"},e.aliases[0])))}function Cn(e,t){var o;let r=xe();return e=e.filter(n=>{var i;return!((i=n.properties)!=null&&i.hidden)}),x("div",{class:"md-version"},x("button",{class:"md-version__current","aria-label":Ee("select.version")},t.title,((o=r.version)==null?void 0:o.alias)&&t.aliases.length>0&&x("span",{class:"md-version__alias"},t.aliases[0])),x("ul",{class:"md-version__list"},e.map(Ra)))}var Ia=0;function ja(e){let t=z([et(e),$t(e)]).pipe(m(([o,n])=>o||n),K()),r=C(()=>Zo(e)).pipe(ne(Ne),pt(1),He(t),m(()=>en(e)));return t.pipe(Ae(o=>o),v(()=>z([t,r])),m(([o,n])=>({active:o,offset:n})),pe())}function Fa(e,t){let{content$:r,viewport$:o}=t,n=`__tooltip2_${Ia++}`;return C(()=>{let i=new g,a=new _r(!1);i.pipe(Z(),ie(!1)).subscribe(a);let s=a.pipe(Ht(c=>Le(+!c*250,kr)),K(),v(c=>c?r:S),w(c=>c.id=n),pe());z([i.pipe(m(({active:c})=>c)),s.pipe(v(c=>$t(c,250)),Q(!1))]).pipe(m(c=>c.some(l=>l))).subscribe(a);let p=a.pipe(b(c=>c),re(s,o),m(([c,l,{size:f}])=>{let u=e.getBoundingClientRect(),d=u.width/2;if(l.role==="tooltip")return{x:d,y:8+u.height};if(u.y>=f.height/2){let{height:y}=ce(l);return{x:d,y:-16-y}}else return{x:d,y:16+u.height}}));return z([s,i,p]).subscribe(([c,{offset:l},f])=>{c.style.setProperty("--md-tooltip-host-x",`${l.x}px`),c.style.setProperty("--md-tooltip-host-y",`${l.y}px`),c.style.setProperty("--md-tooltip-x",`${f.x}px`),c.style.setProperty("--md-tooltip-y",`${f.y}px`),c.classList.toggle("md-tooltip2--top",f.y<0),c.classList.toggle("md-tooltip2--bottom",f.y>=0)}),a.pipe(b(c=>c),re(s,(c,l)=>l),b(c=>c.role==="tooltip")).subscribe(c=>{let l=ce(R(":scope > *",c));c.style.setProperty("--md-tooltip-width",`${l.width}px`),c.style.setProperty("--md-tooltip-tail","0px")}),a.pipe(K(),ve(me),re(s)).subscribe(([c,l])=>{l.classList.toggle("md-tooltip2--active",c)}),z([a.pipe(b(c=>c)),s]).subscribe(([c,l])=>{l.role==="dialog"?(e.setAttribute("aria-controls",n),e.setAttribute("aria-haspopup","dialog")):e.setAttribute("aria-describedby",n)}),a.pipe(b(c=>!c)).subscribe(()=>{e.removeAttribute("aria-controls"),e.removeAttribute("aria-describedby"),e.removeAttribute("aria-haspopup")}),ja(e).pipe(w(c=>i.next(c)),_(()=>i.complete()),m(c=>$({ref:e},c)))})}function mt(e,{viewport$:t},r=document.body){return Fa(e,{content$:new j(o=>{let n=e.title,i=wn(n);return o.next(i),e.removeAttribute("title"),r.append(i),()=>{i.remove(),e.setAttribute("title",n)}}),viewport$:t})}function Ua(e,t){let r=C(()=>z([tn(e),Ne(t)])).pipe(m(([{x:o,y:n},i])=>{let{width:a,height:s}=ce(e);return{x:o-i.x+a/2,y:n-i.y+s/2}}));return et(e).pipe(v(o=>r.pipe(m(n=>({active:o,offset:n})),Te(+!o||1/0))))}function kn(e,t,{target$:r}){let[o,n]=Array.from(e.children);return C(()=>{let i=new g,a=i.pipe(Z(),ie(!0));return i.subscribe({next({offset:s}){e.style.setProperty("--md-tooltip-x",`${s.x}px`),e.style.setProperty("--md-tooltip-y",`${s.y}px`)},complete(){e.style.removeProperty("--md-tooltip-x"),e.style.removeProperty("--md-tooltip-y")}}),tt(e).pipe(W(a)).subscribe(s=>{e.toggleAttribute("data-md-visible",s)}),O(i.pipe(b(({active:s})=>s)),i.pipe(_e(250),b(({active:s})=>!s))).subscribe({next({active:s}){s?e.prepend(o):o.remove()},complete(){e.prepend(o)}}),i.pipe(Me(16,me)).subscribe(({active:s})=>{o.classList.toggle("md-tooltip--active",s)}),i.pipe(pt(125,me),b(()=>!!e.offsetParent),m(()=>e.offsetParent.getBoundingClientRect()),m(({x:s})=>s)).subscribe({next(s){s?e.style.setProperty("--md-tooltip-0",`${-s}px`):e.style.removeProperty("--md-tooltip-0")},complete(){e.style.removeProperty("--md-tooltip-0")}}),h(n,"click").pipe(W(a),b(s=>!(s.metaKey||s.ctrlKey))).subscribe(s=>{s.stopPropagation(),s.preventDefault()}),h(n,"mousedown").pipe(W(a),re(i)).subscribe(([s,{active:p}])=>{var c;if(s.button!==0||s.metaKey||s.ctrlKey)s.preventDefault();else if(p){s.preventDefault();let l=e.parentElement.closest(".md-annotation");l instanceof HTMLElement?l.focus():(c=Ie())==null||c.blur()}}),r.pipe(W(a),b(s=>s===o),Ge(125)).subscribe(()=>e.focus()),Ua(e,t).pipe(w(s=>i.next(s)),_(()=>i.complete()),m(s=>$({ref:e},s)))})}function Wa(e){return e.tagName==="CODE"?P(".c, .c1, .cm",e):[e]}function Da(e){let t=[];for(let r of Wa(e)){let o=[],n=document.createNodeIterator(r,NodeFilter.SHOW_TEXT);for(let i=n.nextNode();i;i=n.nextNode())o.push(i);for(let i of o){let a;for(;a=/(\(\d+\))(!)?/.exec(i.textContent);){let[,s,p]=a;if(typeof p=="undefined"){let c=i.splitText(a.index);i=c.splitText(s.length),t.push(c)}else{i.textContent=s,t.push(i);break}}}}return t}function Hn(e,t){t.append(...Array.from(e.childNodes))}function fr(e,t,{target$:r,print$:o}){let n=t.closest("[id]"),i=n==null?void 0:n.id,a=new Map;for(let s of Da(t)){let[,p]=s.textContent.match(/\((\d+)\)/);fe(`:scope > li:nth-child(${p})`,e)&&(a.set(p,Tn(p,i)),s.replaceWith(a.get(p)))}return a.size===0?S:C(()=>{let s=new g,p=s.pipe(Z(),ie(!0)),c=[];for(let[l,f]of a)c.push([R(".md-typeset",f),R(`:scope > li:nth-child(${l})`,e)]);return o.pipe(W(p)).subscribe(l=>{e.hidden=!l,e.classList.toggle("md-annotation-list",l);for(let[f,u]of c)l?Hn(f,u):Hn(u,f)}),O(...[...a].map(([,l])=>kn(l,t,{target$:r}))).pipe(_(()=>s.complete()),pe())})}function $n(e){if(e.nextElementSibling){let t=e.nextElementSibling;if(t.tagName==="OL")return t;if(t.tagName==="P"&&!t.children.length)return $n(t)}}function Pn(e,t){return C(()=>{let r=$n(e);return typeof r!="undefined"?fr(r,e,t):S})}var Rn=Mt(Br());var Va=0;function In(e){if(e.nextElementSibling){let t=e.nextElementSibling;if(t.tagName==="OL")return t;if(t.tagName==="P"&&!t.children.length)return In(t)}}function Na(e){return ge(e).pipe(m(({width:t})=>({scrollable:St(e).width>t})),ee("scrollable"))}function jn(e,t){let{matches:r}=matchMedia("(hover)"),o=C(()=>{let n=new g,i=n.pipe(jr(1));n.subscribe(({scrollable:c})=>{c&&r?e.setAttribute("tabindex","0"):e.removeAttribute("tabindex")});let a=[];if(Rn.default.isSupported()&&(e.closest(".copy")||B("content.code.copy")&&!e.closest(".no-copy"))){let c=e.closest("pre");c.id=`__code_${Va++}`;let l=Sn(c.id);c.insertBefore(l,e),B("content.tooltips")&&a.push(mt(l,{viewport$}))}let s=e.closest(".highlight");if(s instanceof HTMLElement){let c=In(s);if(typeof c!="undefined"&&(s.classList.contains("annotate")||B("content.code.annotate"))){let l=fr(c,e,t);a.push(ge(s).pipe(W(i),m(({width:f,height:u})=>f&&u),K(),v(f=>f?l:S)))}}return P(":scope > span[id]",e).length&&e.classList.add("md-code__content"),Na(e).pipe(w(c=>n.next(c)),_(()=>n.complete()),m(c=>$({ref:e},c)),Re(...a))});return B("content.lazy")?tt(e).pipe(b(n=>n),Te(1),v(()=>o)):o}function za(e,{target$:t,print$:r}){let o=!0;return O(t.pipe(m(n=>n.closest("details:not([open])")),b(n=>e===n),m(()=>({action:"open",reveal:!0}))),r.pipe(b(n=>n||!o),w(()=>o=e.open),m(n=>({action:n?"open":"close"}))))}function Fn(e,t){return C(()=>{let r=new g;return r.subscribe(({action:o,reveal:n})=>{e.toggleAttribute("open",o==="open"),n&&e.scrollIntoView()}),za(e,t).pipe(w(o=>r.next(o)),_(()=>r.complete()),m(o=>$({ref:e},o)))})}var Un=".node circle,.node ellipse,.node path,.node polygon,.node rect{fill:var(--md-mermaid-node-bg-color);stroke:var(--md-mermaid-node-fg-color)}marker{fill:var(--md-mermaid-edge-color)!important}.edgeLabel .label rect{fill:#0000}.label{color:var(--md-mermaid-label-fg-color);font-family:var(--md-mermaid-font-family)}.label foreignObject{line-height:normal;overflow:visible}.label div .edgeLabel{color:var(--md-mermaid-label-fg-color)}.edgeLabel,.edgeLabel p,.label div .edgeLabel{background-color:var(--md-mermaid-label-bg-color)}.edgeLabel,.edgeLabel p{fill:var(--md-mermaid-label-bg-color);color:var(--md-mermaid-edge-color)}.edgePath .path,.flowchart-link{stroke:var(--md-mermaid-edge-color);stroke-width:.05rem}.edgePath .arrowheadPath{fill:var(--md-mermaid-edge-color);stroke:none}.cluster rect{fill:var(--md-default-fg-color--lightest);stroke:var(--md-default-fg-color--lighter)}.cluster span{color:var(--md-mermaid-label-fg-color);font-family:var(--md-mermaid-font-family)}g #flowchart-circleEnd,g #flowchart-circleStart,g #flowchart-crossEnd,g #flowchart-crossStart,g #flowchart-pointEnd,g #flowchart-pointStart{stroke:none}g.classGroup line,g.classGroup rect{fill:var(--md-mermaid-node-bg-color);stroke:var(--md-mermaid-node-fg-color)}g.classGroup text{fill:var(--md-mermaid-label-fg-color);font-family:var(--md-mermaid-font-family)}.classLabel .box{fill:var(--md-mermaid-label-bg-color);background-color:var(--md-mermaid-label-bg-color);opacity:1}.classLabel .label{fill:var(--md-mermaid-label-fg-color);font-family:var(--md-mermaid-font-family)}.node .divider{stroke:var(--md-mermaid-node-fg-color)}.relation{stroke:var(--md-mermaid-edge-color)}.cardinality{fill:var(--md-mermaid-label-fg-color);font-family:var(--md-mermaid-font-family)}.cardinality text{fill:inherit!important}defs #classDiagram-compositionEnd,defs #classDiagram-compositionStart,defs #classDiagram-dependencyEnd,defs #classDiagram-dependencyStart,defs #classDiagram-extensionEnd,defs #classDiagram-extensionStart{fill:var(--md-mermaid-edge-color)!important;stroke:var(--md-mermaid-edge-color)!important}defs #classDiagram-aggregationEnd,defs #classDiagram-aggregationStart{fill:var(--md-mermaid-label-bg-color)!important;stroke:var(--md-mermaid-edge-color)!important}g.stateGroup rect{fill:var(--md-mermaid-node-bg-color);stroke:var(--md-mermaid-node-fg-color)}g.stateGroup .state-title{fill:var(--md-mermaid-label-fg-color)!important;font-family:var(--md-mermaid-font-family)}g.stateGroup .composit{fill:var(--md-mermaid-label-bg-color)}.nodeLabel,.nodeLabel p{color:var(--md-mermaid-label-fg-color);font-family:var(--md-mermaid-font-family)}a .nodeLabel{text-decoration:underline}.node circle.state-end,.node circle.state-start,.start-state{fill:var(--md-mermaid-edge-color);stroke:none}.end-state-inner,.end-state-outer{fill:var(--md-mermaid-edge-color)}.end-state-inner,.node circle.state-end{stroke:var(--md-mermaid-label-bg-color)}.transition{stroke:var(--md-mermaid-edge-color)}[id^=state-fork] rect,[id^=state-join] rect{fill:var(--md-mermaid-edge-color)!important;stroke:none!important}.statediagram-cluster.statediagram-cluster .inner{fill:var(--md-default-bg-color)}.statediagram-cluster rect{fill:var(--md-mermaid-node-bg-color);stroke:var(--md-mermaid-node-fg-color)}.statediagram-state rect.divider{fill:var(--md-default-fg-color--lightest);stroke:var(--md-default-fg-color--lighter)}defs #statediagram-barbEnd{stroke:var(--md-mermaid-edge-color)}.attributeBoxEven,.attributeBoxOdd{fill:var(--md-mermaid-node-bg-color);stroke:var(--md-mermaid-node-fg-color)}.entityBox{fill:var(--md-mermaid-label-bg-color);stroke:var(--md-mermaid-node-fg-color)}.entityLabel{fill:var(--md-mermaid-label-fg-color);font-family:var(--md-mermaid-font-family)}.relationshipLabelBox{fill:var(--md-mermaid-label-bg-color);fill-opacity:1;background-color:var(--md-mermaid-label-bg-color);opacity:1}.relationshipLabel{fill:var(--md-mermaid-label-fg-color)}.relationshipLine{stroke:var(--md-mermaid-edge-color)}defs #ONE_OR_MORE_END *,defs #ONE_OR_MORE_START *,defs #ONLY_ONE_END *,defs #ONLY_ONE_START *,defs #ZERO_OR_MORE_END *,defs #ZERO_OR_MORE_START *,defs #ZERO_OR_ONE_END *,defs #ZERO_OR_ONE_START *{stroke:var(--md-mermaid-edge-color)!important}defs #ZERO_OR_MORE_END circle,defs #ZERO_OR_MORE_START circle{fill:var(--md-mermaid-label-bg-color)}.actor{fill:var(--md-mermaid-sequence-actor-bg-color);stroke:var(--md-mermaid-sequence-actor-border-color)}text.actor>tspan{fill:var(--md-mermaid-sequence-actor-fg-color);font-family:var(--md-mermaid-font-family)}line{stroke:var(--md-mermaid-sequence-actor-line-color)}.actor-man circle,.actor-man line{fill:var(--md-mermaid-sequence-actorman-bg-color);stroke:var(--md-mermaid-sequence-actorman-line-color)}.messageLine0,.messageLine1{stroke:var(--md-mermaid-sequence-message-line-color)}.note{fill:var(--md-mermaid-sequence-note-bg-color);stroke:var(--md-mermaid-sequence-note-border-color)}.loopText,.loopText>tspan,.messageText,.noteText>tspan{stroke:none;font-family:var(--md-mermaid-font-family)!important}.messageText{fill:var(--md-mermaid-sequence-message-fg-color)}.loopText,.loopText>tspan{fill:var(--md-mermaid-sequence-loop-fg-color)}.noteText>tspan{fill:var(--md-mermaid-sequence-note-fg-color)}#arrowhead path{fill:var(--md-mermaid-sequence-message-line-color);stroke:none}.loopLine{fill:var(--md-mermaid-sequence-loop-bg-color);stroke:var(--md-mermaid-sequence-loop-border-color)}.labelBox{fill:var(--md-mermaid-sequence-label-bg-color);stroke:none}.labelText,.labelText>span{fill:var(--md-mermaid-sequence-label-fg-color);font-family:var(--md-mermaid-font-family)}.sequenceNumber{fill:var(--md-mermaid-sequence-number-fg-color)}rect.rect{fill:var(--md-mermaid-sequence-box-bg-color);stroke:none}rect.rect+text.text{fill:var(--md-mermaid-sequence-box-fg-color)}defs #sequencenumber{fill:var(--md-mermaid-sequence-number-bg-color)!important}";var Gr,Qa=0;function Ka(){return typeof mermaid=="undefined"||mermaid instanceof Element?Tt("https://unpkg.com/mermaid@11/dist/mermaid.min.js"):I(void 0)}function Wn(e){return e.classList.remove("mermaid"),Gr||(Gr=Ka().pipe(w(()=>mermaid.initialize({startOnLoad:!1,themeCSS:Un,sequence:{actorFontSize:"16px",messageFontSize:"16px",noteFontSize:"16px"}})),m(()=>{}),G(1))),Gr.subscribe(()=>co(this,null,function*(){e.classList.add("mermaid");let t=`__mermaid_${Qa++}`,r=x("div",{class:"mermaid"}),o=e.textContent,{svg:n,fn:i}=yield mermaid.render(t,o),a=r.attachShadow({mode:"closed"});a.innerHTML=n,e.replaceWith(r),i==null||i(a)})),Gr.pipe(m(()=>({ref:e})))}var Dn=x("table");function Vn(e){return e.replaceWith(Dn),Dn.replaceWith(An(e)),I({ref:e})}function Ya(e){let t=e.find(r=>r.checked)||e[0];return O(...e.map(r=>h(r,"change").pipe(m(()=>R(`label[for="${r.id}"]`))))).pipe(Q(R(`label[for="${t.id}"]`)),m(r=>({active:r})))}function Nn(e,{viewport$:t,target$:r}){let o=R(".tabbed-labels",e),n=P(":scope > input",e),i=Kr("prev");e.append(i);let a=Kr("next");return e.append(a),C(()=>{let s=new g,p=s.pipe(Z(),ie(!0));z([s,ge(e),tt(e)]).pipe(W(p),Me(1,me)).subscribe({next([{active:c},l]){let f=Ve(c),{width:u}=ce(c);e.style.setProperty("--md-indicator-x",`${f.x}px`),e.style.setProperty("--md-indicator-width",`${u}px`);let d=pr(o);(f.xd.x+l.width)&&o.scrollTo({left:Math.max(0,f.x-16),behavior:"smooth"})},complete(){e.style.removeProperty("--md-indicator-x"),e.style.removeProperty("--md-indicator-width")}}),z([Ne(o),ge(o)]).pipe(W(p)).subscribe(([c,l])=>{let f=St(o);i.hidden=c.x<16,a.hidden=c.x>f.width-l.width-16}),O(h(i,"click").pipe(m(()=>-1)),h(a,"click").pipe(m(()=>1))).pipe(W(p)).subscribe(c=>{let{width:l}=ce(o);o.scrollBy({left:l*c,behavior:"smooth"})}),r.pipe(W(p),b(c=>n.includes(c))).subscribe(c=>c.click()),o.classList.add("tabbed-labels--linked");for(let c of n){let l=R(`label[for="${c.id}"]`);l.replaceChildren(x("a",{href:`#${l.htmlFor}`,tabIndex:-1},...Array.from(l.childNodes))),h(l.firstElementChild,"click").pipe(W(p),b(f=>!(f.metaKey||f.ctrlKey)),w(f=>{f.preventDefault(),f.stopPropagation()})).subscribe(()=>{history.replaceState({},"",`#${l.htmlFor}`),l.click()})}return B("content.tabs.link")&&s.pipe(Ce(1),re(t)).subscribe(([{active:c},{offset:l}])=>{let f=c.innerText.trim();if(c.hasAttribute("data-md-switching"))c.removeAttribute("data-md-switching");else{let u=e.offsetTop-l.y;for(let y of P("[data-tabs]"))for(let L of P(":scope > input",y)){let X=R(`label[for="${L.id}"]`);if(X!==c&&X.innerText.trim()===f){X.setAttribute("data-md-switching",""),L.click();break}}window.scrollTo({top:e.offsetTop-u});let d=__md_get("__tabs")||[];__md_set("__tabs",[...new Set([f,...d])])}}),s.pipe(W(p)).subscribe(()=>{for(let c of P("audio, video",e))c.pause()}),Ya(n).pipe(w(c=>s.next(c)),_(()=>s.complete()),m(c=>$({ref:e},c)))}).pipe(Ke(se))}function zn(e,{viewport$:t,target$:r,print$:o}){return O(...P(".annotate:not(.highlight)",e).map(n=>Pn(n,{target$:r,print$:o})),...P("pre:not(.mermaid) > code",e).map(n=>jn(n,{target$:r,print$:o})),...P("pre.mermaid",e).map(n=>Wn(n)),...P("table:not([class])",e).map(n=>Vn(n)),...P("details",e).map(n=>Fn(n,{target$:r,print$:o})),...P("[data-tabs]",e).map(n=>Nn(n,{viewport$:t,target$:r})),...P("[title]",e).filter(()=>B("content.tooltips")).map(n=>mt(n,{viewport$:t})))}function Ba(e,{alert$:t}){return t.pipe(v(r=>O(I(!0),I(!1).pipe(Ge(2e3))).pipe(m(o=>({message:r,active:o})))))}function qn(e,t){let r=R(".md-typeset",e);return C(()=>{let o=new g;return o.subscribe(({message:n,active:i})=>{e.classList.toggle("md-dialog--active",i),r.textContent=n}),Ba(e,t).pipe(w(n=>o.next(n)),_(()=>o.complete()),m(n=>$({ref:e},n)))})}var Ga=0;function Ja(e,t){document.body.append(e);let{width:r}=ce(e);e.style.setProperty("--md-tooltip-width",`${r}px`),e.remove();let o=cr(t),n=typeof o!="undefined"?Ne(o):I({x:0,y:0}),i=O(et(t),$t(t)).pipe(K());return z([i,n]).pipe(m(([a,s])=>{let{x:p,y:c}=Ve(t),l=ce(t),f=t.closest("table");return f&&t.parentElement&&(p+=f.offsetLeft+t.parentElement.offsetLeft,c+=f.offsetTop+t.parentElement.offsetTop),{active:a,offset:{x:p-s.x+l.width/2-r/2,y:c-s.y+l.height+8}}}))}function Qn(e){let t=e.title;if(!t.length)return S;let r=`__tooltip_${Ga++}`,o=Rt(r,"inline"),n=R(".md-typeset",o);return n.innerHTML=t,C(()=>{let i=new g;return i.subscribe({next({offset:a}){o.style.setProperty("--md-tooltip-x",`${a.x}px`),o.style.setProperty("--md-tooltip-y",`${a.y}px`)},complete(){o.style.removeProperty("--md-tooltip-x"),o.style.removeProperty("--md-tooltip-y")}}),O(i.pipe(b(({active:a})=>a)),i.pipe(_e(250),b(({active:a})=>!a))).subscribe({next({active:a}){a?(e.insertAdjacentElement("afterend",o),e.setAttribute("aria-describedby",r),e.removeAttribute("title")):(o.remove(),e.removeAttribute("aria-describedby"),e.setAttribute("title",t))},complete(){o.remove(),e.removeAttribute("aria-describedby"),e.setAttribute("title",t)}}),i.pipe(Me(16,me)).subscribe(({active:a})=>{o.classList.toggle("md-tooltip--active",a)}),i.pipe(pt(125,me),b(()=>!!e.offsetParent),m(()=>e.offsetParent.getBoundingClientRect()),m(({x:a})=>a)).subscribe({next(a){a?o.style.setProperty("--md-tooltip-0",`${-a}px`):o.style.removeProperty("--md-tooltip-0")},complete(){o.style.removeProperty("--md-tooltip-0")}}),Ja(o,e).pipe(w(a=>i.next(a)),_(()=>i.complete()),m(a=>$({ref:e},a)))}).pipe(Ke(se))}function Xa({viewport$:e}){if(!B("header.autohide"))return I(!1);let t=e.pipe(m(({offset:{y:n}})=>n),Be(2,1),m(([n,i])=>[nMath.abs(i-n.y)>100),m(([,[n]])=>n),K()),o=ze("search");return z([e,o]).pipe(m(([{offset:n},i])=>n.y>400&&!i),K(),v(n=>n?r:I(!1)),Q(!1))}function Kn(e,t){return C(()=>z([ge(e),Xa(t)])).pipe(m(([{height:r},o])=>({height:r,hidden:o})),K((r,o)=>r.height===o.height&&r.hidden===o.hidden),G(1))}function Yn(e,{header$:t,main$:r}){return C(()=>{let o=new g,n=o.pipe(Z(),ie(!0));o.pipe(ee("active"),He(t)).subscribe(([{active:a},{hidden:s}])=>{e.classList.toggle("md-header--shadow",a&&!s),e.hidden=s});let i=ue(P("[title]",e)).pipe(b(()=>B("content.tooltips")),ne(a=>Qn(a)));return r.subscribe(o),t.pipe(W(n),m(a=>$({ref:e},a)),Re(i.pipe(W(n))))})}function Za(e,{viewport$:t,header$:r}){return mr(e,{viewport$:t,header$:r}).pipe(m(({offset:{y:o}})=>{let{height:n}=ce(e);return{active:o>=n}}),ee("active"))}function Bn(e,t){return C(()=>{let r=new g;r.subscribe({next({active:n}){e.classList.toggle("md-header__title--active",n)},complete(){e.classList.remove("md-header__title--active")}});let o=fe(".md-content h1");return typeof o=="undefined"?S:Za(o,t).pipe(w(n=>r.next(n)),_(()=>r.complete()),m(n=>$({ref:e},n)))})}function Gn(e,{viewport$:t,header$:r}){let o=r.pipe(m(({height:i})=>i),K()),n=o.pipe(v(()=>ge(e).pipe(m(({height:i})=>({top:e.offsetTop,bottom:e.offsetTop+i})),ee("bottom"))));return z([o,n,t]).pipe(m(([i,{top:a,bottom:s},{offset:{y:p},size:{height:c}}])=>(c=Math.max(0,c-Math.max(0,a-p,i)-Math.max(0,c+p-s)),{offset:a-i,height:c,active:a-i<=p})),K((i,a)=>i.offset===a.offset&&i.height===a.height&&i.active===a.active))}function es(e){let t=__md_get("__palette")||{index:e.findIndex(o=>matchMedia(o.getAttribute("data-md-color-media")).matches)},r=Math.max(0,Math.min(t.index,e.length-1));return I(...e).pipe(ne(o=>h(o,"change").pipe(m(()=>o))),Q(e[r]),m(o=>({index:e.indexOf(o),color:{media:o.getAttribute("data-md-color-media"),scheme:o.getAttribute("data-md-color-scheme"),primary:o.getAttribute("data-md-color-primary"),accent:o.getAttribute("data-md-color-accent")}})),G(1))}function Jn(e){let t=P("input",e),r=x("meta",{name:"theme-color"});document.head.appendChild(r);let o=x("meta",{name:"color-scheme"});document.head.appendChild(o);let n=Pt("(prefers-color-scheme: light)");return C(()=>{let i=new g;return i.subscribe(a=>{if(document.body.setAttribute("data-md-color-switching",""),a.color.media==="(prefers-color-scheme)"){let s=matchMedia("(prefers-color-scheme: light)"),p=document.querySelector(s.matches?"[data-md-color-media='(prefers-color-scheme: light)']":"[data-md-color-media='(prefers-color-scheme: dark)']");a.color.scheme=p.getAttribute("data-md-color-scheme"),a.color.primary=p.getAttribute("data-md-color-primary"),a.color.accent=p.getAttribute("data-md-color-accent")}for(let[s,p]of Object.entries(a.color))document.body.setAttribute(`data-md-color-${s}`,p);for(let s=0;sa.key==="Enter"),re(i,(a,s)=>s)).subscribe(({index:a})=>{a=(a+1)%t.length,t[a].click(),t[a].focus()}),i.pipe(m(()=>{let a=Se("header"),s=window.getComputedStyle(a);return o.content=s.colorScheme,s.backgroundColor.match(/\d+/g).map(p=>(+p).toString(16).padStart(2,"0")).join("")})).subscribe(a=>r.content=`#${a}`),i.pipe(ve(se)).subscribe(()=>{document.body.removeAttribute("data-md-color-switching")}),es(t).pipe(W(n.pipe(Ce(1))),ct(),w(a=>i.next(a)),_(()=>i.complete()),m(a=>$({ref:e},a)))})}function Xn(e,{progress$:t}){return C(()=>{let r=new g;return r.subscribe(({value:o})=>{e.style.setProperty("--md-progress-value",`${o}`)}),t.pipe(w(o=>r.next({value:o})),_(()=>r.complete()),m(o=>({ref:e,value:o})))})}var Jr=Mt(Br());function ts(e){e.setAttribute("data-md-copying","");let t=e.closest("[data-copy]"),r=t?t.getAttribute("data-copy"):e.innerText;return e.removeAttribute("data-md-copying"),r.trimEnd()}function Zn({alert$:e}){Jr.default.isSupported()&&new j(t=>{new Jr.default("[data-clipboard-target], [data-clipboard-text]",{text:r=>r.getAttribute("data-clipboard-text")||ts(R(r.getAttribute("data-clipboard-target")))}).on("success",r=>t.next(r))}).pipe(w(t=>{t.trigger.focus()}),m(()=>Ee("clipboard.copied"))).subscribe(e)}function ei(e,t){return e.protocol=t.protocol,e.hostname=t.hostname,e}function rs(e,t){let r=new Map;for(let o of P("url",e)){let n=R("loc",o),i=[ei(new URL(n.textContent),t)];r.set(`${i[0]}`,i);for(let a of P("[rel=alternate]",o)){let s=a.getAttribute("href");s!=null&&i.push(ei(new URL(s),t))}}return r}function ur(e){return un(new URL("sitemap.xml",e)).pipe(m(t=>rs(t,new URL(e))),de(()=>I(new Map)))}function os(e,t){if(!(e.target instanceof Element))return S;let r=e.target.closest("a");if(r===null)return S;if(r.target||e.metaKey||e.ctrlKey)return S;let o=new URL(r.href);return o.search=o.hash="",t.has(`${o}`)?(e.preventDefault(),I(new URL(r.href))):S}function ti(e){let t=new Map;for(let r of P(":scope > *",e.head))t.set(r.outerHTML,r);return t}function ri(e){for(let t of P("[href], [src]",e))for(let r of["href","src"]){let o=t.getAttribute(r);if(o&&!/^(?:[a-z]+:)?\/\//i.test(o)){t[r]=t[r];break}}return I(e)}function ns(e){for(let o of["[data-md-component=announce]","[data-md-component=container]","[data-md-component=header-topic]","[data-md-component=outdated]","[data-md-component=logo]","[data-md-component=skip]",...B("navigation.tabs.sticky")?["[data-md-component=tabs]"]:[]]){let n=fe(o),i=fe(o,e);typeof n!="undefined"&&typeof i!="undefined"&&n.replaceWith(i)}let t=ti(document);for(let[o,n]of ti(e))t.has(o)?t.delete(o):document.head.appendChild(n);for(let o of t.values()){let n=o.getAttribute("name");n!=="theme-color"&&n!=="color-scheme"&&o.remove()}let r=Se("container");return We(P("script",r)).pipe(v(o=>{let n=e.createElement("script");if(o.src){for(let i of o.getAttributeNames())n.setAttribute(i,o.getAttribute(i));return o.replaceWith(n),new j(i=>{n.onload=()=>i.complete()})}else return n.textContent=o.textContent,o.replaceWith(n),S}),Z(),ie(document))}function oi({location$:e,viewport$:t,progress$:r}){let o=xe();if(location.protocol==="file:")return S;let n=ur(o.base);I(document).subscribe(ri);let i=h(document.body,"click").pipe(He(n),v(([p,c])=>os(p,c)),pe()),a=h(window,"popstate").pipe(m(ye),pe());i.pipe(re(t)).subscribe(([p,{offset:c}])=>{history.replaceState(c,""),history.pushState(null,"",p)}),O(i,a).subscribe(e);let s=e.pipe(ee("pathname"),v(p=>fn(p,{progress$:r}).pipe(de(()=>(lt(p,!0),S)))),v(ri),v(ns),pe());return O(s.pipe(re(e,(p,c)=>c)),s.pipe(v(()=>e),ee("pathname"),v(()=>e),ee("hash")),e.pipe(K((p,c)=>p.pathname===c.pathname&&p.hash===c.hash),v(()=>i),w(()=>history.back()))).subscribe(p=>{var c,l;history.state!==null||!p.hash?window.scrollTo(0,(l=(c=history.state)==null?void 0:c.y)!=null?l:0):(history.scrollRestoration="auto",pn(p.hash),history.scrollRestoration="manual")}),e.subscribe(()=>{history.scrollRestoration="manual"}),h(window,"beforeunload").subscribe(()=>{history.scrollRestoration="auto"}),t.pipe(ee("offset"),_e(100)).subscribe(({offset:p})=>{history.replaceState(p,"")}),s}var ni=Mt(qr());function ii(e){let t=e.separator.split("|").map(n=>n.replace(/(\(\?[!=<][^)]+\))/g,"").length===0?"\uFFFD":n).join("|"),r=new RegExp(t,"img"),o=(n,i,a)=>`${i}${a}`;return n=>{n=n.replace(/[\s*+\-:~^]+/g," ").trim();let i=new RegExp(`(^|${e.separator}|)(${n.replace(/[|\\{}()[\]^$+*?.-]/g,"\\$&").replace(r,"|")})`,"img");return a=>(0,ni.default)(a).replace(i,o).replace(/<\/mark>(\s+)]*>/img,"$1")}}function jt(e){return e.type===1}function dr(e){return e.type===3}function ai(e,t){let r=yn(e);return O(I(location.protocol!=="file:"),ze("search")).pipe(Ae(o=>o),v(()=>t)).subscribe(({config:o,docs:n})=>r.next({type:0,data:{config:o,docs:n,options:{suggest:B("search.suggest")}}})),r}function si(e){var l;let{selectedVersionSitemap:t,selectedVersionBaseURL:r,currentLocation:o,currentBaseURL:n}=e,i=(l=Xr(n))==null?void 0:l.pathname;if(i===void 0)return;let a=ss(o.pathname,i);if(a===void 0)return;let s=ps(t.keys());if(!t.has(s))return;let p=Xr(a,s);if(!p||!t.has(p.href))return;let c=Xr(a,r);if(c)return c.hash=o.hash,c.search=o.search,c}function Xr(e,t){try{return new URL(e,t)}catch(r){return}}function ss(e,t){if(e.startsWith(t))return e.slice(t.length)}function cs(e,t){let r=Math.min(e.length,t.length),o;for(o=0;oS)),o=r.pipe(m(n=>{let[,i]=t.base.match(/([^/]+)\/?$/);return n.find(({version:a,aliases:s})=>a===i||s.includes(i))||n[0]}));r.pipe(m(n=>new Map(n.map(i=>[`${new URL(`../${i.version}/`,t.base)}`,i]))),v(n=>h(document.body,"click").pipe(b(i=>!i.metaKey&&!i.ctrlKey),re(o),v(([i,a])=>{if(i.target instanceof Element){let s=i.target.closest("a");if(s&&!s.target&&n.has(s.href)){let p=s.href;return!i.target.closest(".md-version")&&n.get(p)===a?S:(i.preventDefault(),I(new URL(p)))}}return S}),v(i=>ur(i).pipe(m(a=>{var s;return(s=si({selectedVersionSitemap:a,selectedVersionBaseURL:i,currentLocation:ye(),currentBaseURL:t.base}))!=null?s:i})))))).subscribe(n=>lt(n,!0)),z([r,o]).subscribe(([n,i])=>{R(".md-header__topic").appendChild(Cn(n,i))}),e.pipe(v(()=>o)).subscribe(n=>{var a;let i=__md_get("__outdated",sessionStorage);if(i===null){i=!0;let s=((a=t.version)==null?void 0:a.default)||"latest";Array.isArray(s)||(s=[s]);e:for(let p of s)for(let c of n.aliases.concat(n.version))if(new RegExp(p,"i").test(c)){i=!1;break e}__md_set("__outdated",i,sessionStorage)}if(i)for(let s of ae("outdated"))s.hidden=!1})}function ls(e,{worker$:t}){let{searchParams:r}=ye();r.has("q")&&(Je("search",!0),e.value=r.get("q"),e.focus(),ze("search").pipe(Ae(i=>!i)).subscribe(()=>{let i=ye();i.searchParams.delete("q"),history.replaceState({},"",`${i}`)}));let o=et(e),n=O(t.pipe(Ae(jt)),h(e,"keyup"),o).pipe(m(()=>e.value),K());return z([n,o]).pipe(m(([i,a])=>({value:i,focus:a})),G(1))}function pi(e,{worker$:t}){let r=new g,o=r.pipe(Z(),ie(!0));z([t.pipe(Ae(jt)),r],(i,a)=>a).pipe(ee("value")).subscribe(({value:i})=>t.next({type:2,data:i})),r.pipe(ee("focus")).subscribe(({focus:i})=>{i&&Je("search",i)}),h(e.form,"reset").pipe(W(o)).subscribe(()=>e.focus());let n=R("header [for=__search]");return h(n,"click").subscribe(()=>e.focus()),ls(e,{worker$:t}).pipe(w(i=>r.next(i)),_(()=>r.complete()),m(i=>$({ref:e},i)),G(1))}function li(e,{worker$:t,query$:r}){let o=new g,n=on(e.parentElement).pipe(b(Boolean)),i=e.parentElement,a=R(":scope > :first-child",e),s=R(":scope > :last-child",e);ze("search").subscribe(l=>s.setAttribute("role",l?"list":"presentation")),o.pipe(re(r),Wr(t.pipe(Ae(jt)))).subscribe(([{items:l},{value:f}])=>{switch(l.length){case 0:a.textContent=f.length?Ee("search.result.none"):Ee("search.result.placeholder");break;case 1:a.textContent=Ee("search.result.one");break;default:let u=sr(l.length);a.textContent=Ee("search.result.other",u)}});let p=o.pipe(w(()=>s.innerHTML=""),v(({items:l})=>O(I(...l.slice(0,10)),I(...l.slice(10)).pipe(Be(4),Vr(n),v(([f])=>f)))),m(Mn),pe());return p.subscribe(l=>s.appendChild(l)),p.pipe(ne(l=>{let f=fe("details",l);return typeof f=="undefined"?S:h(f,"toggle").pipe(W(o),m(()=>f))})).subscribe(l=>{l.open===!1&&l.offsetTop<=i.scrollTop&&i.scrollTo({top:l.offsetTop})}),t.pipe(b(dr),m(({data:l})=>l)).pipe(w(l=>o.next(l)),_(()=>o.complete()),m(l=>$({ref:e},l)))}function ms(e,{query$:t}){return t.pipe(m(({value:r})=>{let o=ye();return o.hash="",r=r.replace(/\s+/g,"+").replace(/&/g,"%26").replace(/=/g,"%3D"),o.search=`q=${r}`,{url:o}}))}function mi(e,t){let r=new g,o=r.pipe(Z(),ie(!0));return r.subscribe(({url:n})=>{e.setAttribute("data-clipboard-text",e.href),e.href=`${n}`}),h(e,"click").pipe(W(o)).subscribe(n=>n.preventDefault()),ms(e,t).pipe(w(n=>r.next(n)),_(()=>r.complete()),m(n=>$({ref:e},n)))}function fi(e,{worker$:t,keyboard$:r}){let o=new g,n=Se("search-query"),i=O(h(n,"keydown"),h(n,"focus")).pipe(ve(se),m(()=>n.value),K());return o.pipe(He(i),m(([{suggest:s},p])=>{let c=p.split(/([\s-]+)/);if(s!=null&&s.length&&c[c.length-1]){let l=s[s.length-1];l.startsWith(c[c.length-1])&&(c[c.length-1]=l)}else c.length=0;return c})).subscribe(s=>e.innerHTML=s.join("").replace(/\s/g," ")),r.pipe(b(({mode:s})=>s==="search")).subscribe(s=>{switch(s.type){case"ArrowRight":e.innerText.length&&n.selectionStart===n.value.length&&(n.value=e.innerText);break}}),t.pipe(b(dr),m(({data:s})=>s)).pipe(w(s=>o.next(s)),_(()=>o.complete()),m(()=>({ref:e})))}function ui(e,{index$:t,keyboard$:r}){let o=xe();try{let n=ai(o.search,t),i=Se("search-query",e),a=Se("search-result",e);h(e,"click").pipe(b(({target:p})=>p instanceof Element&&!!p.closest("a"))).subscribe(()=>Je("search",!1)),r.pipe(b(({mode:p})=>p==="search")).subscribe(p=>{let c=Ie();switch(p.type){case"Enter":if(c===i){let l=new Map;for(let f of P(":first-child [href]",a)){let u=f.firstElementChild;l.set(f,parseFloat(u.getAttribute("data-md-score")))}if(l.size){let[[f]]=[...l].sort(([,u],[,d])=>d-u);f.click()}p.claim()}break;case"Escape":case"Tab":Je("search",!1),i.blur();break;case"ArrowUp":case"ArrowDown":if(typeof c=="undefined")i.focus();else{let l=[i,...P(":not(details) > [href], summary, details[open] [href]",a)],f=Math.max(0,(Math.max(0,l.indexOf(c))+l.length+(p.type==="ArrowUp"?-1:1))%l.length);l[f].focus()}p.claim();break;default:i!==Ie()&&i.focus()}}),r.pipe(b(({mode:p})=>p==="global")).subscribe(p=>{switch(p.type){case"f":case"s":case"/":i.focus(),i.select(),p.claim();break}});let s=pi(i,{worker$:n});return O(s,li(a,{worker$:n,query$:s})).pipe(Re(...ae("search-share",e).map(p=>mi(p,{query$:s})),...ae("search-suggest",e).map(p=>fi(p,{worker$:n,keyboard$:r}))))}catch(n){return e.hidden=!0,Ye}}function di(e,{index$:t,location$:r}){return z([t,r.pipe(Q(ye()),b(o=>!!o.searchParams.get("h")))]).pipe(m(([o,n])=>ii(o.config)(n.searchParams.get("h"))),m(o=>{var a;let n=new Map,i=document.createNodeIterator(e,NodeFilter.SHOW_TEXT);for(let s=i.nextNode();s;s=i.nextNode())if((a=s.parentElement)!=null&&a.offsetHeight){let p=s.textContent,c=o(p);c.length>p.length&&n.set(s,c)}for(let[s,p]of n){let{childNodes:c}=x("span",null,p);s.replaceWith(...Array.from(c))}return{ref:e,nodes:n}}))}function fs(e,{viewport$:t,main$:r}){let o=e.closest(".md-grid"),n=o.offsetTop-o.parentElement.offsetTop;return z([r,t]).pipe(m(([{offset:i,height:a},{offset:{y:s}}])=>(a=a+Math.min(n,Math.max(0,s-i))-n,{height:a,locked:s>=i+n})),K((i,a)=>i.height===a.height&&i.locked===a.locked))}function Zr(e,o){var n=o,{header$:t}=n,r=so(n,["header$"]);let i=R(".md-sidebar__scrollwrap",e),{y:a}=Ve(i);return C(()=>{let s=new g,p=s.pipe(Z(),ie(!0)),c=s.pipe(Me(0,me));return c.pipe(re(t)).subscribe({next([{height:l},{height:f}]){i.style.height=`${l-2*a}px`,e.style.top=`${f}px`},complete(){i.style.height="",e.style.top=""}}),c.pipe(Ae()).subscribe(()=>{for(let l of P(".md-nav__link--active[href]",e)){if(!l.clientHeight)continue;let f=l.closest(".md-sidebar__scrollwrap");if(typeof f!="undefined"){let u=l.offsetTop-f.offsetTop,{height:d}=ce(f);f.scrollTo({top:u-d/2})}}}),ue(P("label[tabindex]",e)).pipe(ne(l=>h(l,"click").pipe(ve(se),m(()=>l),W(p)))).subscribe(l=>{let f=R(`[id="${l.htmlFor}"]`);R(`[aria-labelledby="${l.id}"]`).setAttribute("aria-expanded",`${f.checked}`)}),fs(e,r).pipe(w(l=>s.next(l)),_(()=>s.complete()),m(l=>$({ref:e},l)))})}function hi(e,t){if(typeof t!="undefined"){let r=`https://api.github.com/repos/${e}/${t}`;return st(je(`${r}/releases/latest`).pipe(de(()=>S),m(o=>({version:o.tag_name})),De({})),je(r).pipe(de(()=>S),m(o=>({stars:o.stargazers_count,forks:o.forks_count})),De({}))).pipe(m(([o,n])=>$($({},o),n)))}else{let r=`https://api.github.com/users/${e}`;return je(r).pipe(m(o=>({repositories:o.public_repos})),De({}))}}function bi(e,t){let r=`https://${e}/api/v4/projects/${encodeURIComponent(t)}`;return st(je(`${r}/releases/permalink/latest`).pipe(de(()=>S),m(({tag_name:o})=>({version:o})),De({})),je(r).pipe(de(()=>S),m(({star_count:o,forks_count:n})=>({stars:o,forks:n})),De({}))).pipe(m(([o,n])=>$($({},o),n)))}function vi(e){let t=e.match(/^.+github\.com\/([^/]+)\/?([^/]+)?/i);if(t){let[,r,o]=t;return hi(r,o)}if(t=e.match(/^.+?([^/]*gitlab[^/]+)\/(.+?)\/?$/i),t){let[,r,o]=t;return bi(r,o)}return S}var us;function ds(e){return us||(us=C(()=>{let t=__md_get("__source",sessionStorage);if(t)return I(t);if(ae("consent").length){let o=__md_get("__consent");if(!(o&&o.github))return S}return vi(e.href).pipe(w(o=>__md_set("__source",o,sessionStorage)))}).pipe(de(()=>S),b(t=>Object.keys(t).length>0),m(t=>({facts:t})),G(1)))}function gi(e){let t=R(":scope > :last-child",e);return C(()=>{let r=new g;return r.subscribe(({facts:o})=>{t.appendChild(_n(o)),t.classList.add("md-source__repository--active")}),ds(e).pipe(w(o=>r.next(o)),_(()=>r.complete()),m(o=>$({ref:e},o)))})}function hs(e,{viewport$:t,header$:r}){return ge(document.body).pipe(v(()=>mr(e,{header$:r,viewport$:t})),m(({offset:{y:o}})=>({hidden:o>=10})),ee("hidden"))}function yi(e,t){return C(()=>{let r=new g;return r.subscribe({next({hidden:o}){e.hidden=o},complete(){e.hidden=!1}}),(B("navigation.tabs.sticky")?I({hidden:!1}):hs(e,t)).pipe(w(o=>r.next(o)),_(()=>r.complete()),m(o=>$({ref:e},o)))})}function bs(e,{viewport$:t,header$:r}){let o=new Map,n=P(".md-nav__link",e);for(let s of n){let p=decodeURIComponent(s.hash.substring(1)),c=fe(`[id="${p}"]`);typeof c!="undefined"&&o.set(s,c)}let i=r.pipe(ee("height"),m(({height:s})=>{let p=Se("main"),c=R(":scope > :first-child",p);return s+.8*(c.offsetTop-p.offsetTop)}),pe());return ge(document.body).pipe(ee("height"),v(s=>C(()=>{let p=[];return I([...o].reduce((c,[l,f])=>{for(;p.length&&o.get(p[p.length-1]).tagName>=f.tagName;)p.pop();let u=f.offsetTop;for(;!u&&f.parentElement;)f=f.parentElement,u=f.offsetTop;let d=f.offsetParent;for(;d;d=d.offsetParent)u+=d.offsetTop;return c.set([...p=[...p,l]].reverse(),u)},new Map))}).pipe(m(p=>new Map([...p].sort(([,c],[,l])=>c-l))),He(i),v(([p,c])=>t.pipe(Fr(([l,f],{offset:{y:u},size:d})=>{let y=u+d.height>=Math.floor(s.height);for(;f.length;){let[,L]=f[0];if(L-c=u&&!y)f=[l.pop(),...f];else break}return[l,f]},[[],[...p]]),K((l,f)=>l[0]===f[0]&&l[1]===f[1])))))).pipe(m(([s,p])=>({prev:s.map(([c])=>c),next:p.map(([c])=>c)})),Q({prev:[],next:[]}),Be(2,1),m(([s,p])=>s.prev.length{let i=new g,a=i.pipe(Z(),ie(!0));if(i.subscribe(({prev:s,next:p})=>{for(let[c]of p)c.classList.remove("md-nav__link--passed"),c.classList.remove("md-nav__link--active");for(let[c,[l]]of s.entries())l.classList.add("md-nav__link--passed"),l.classList.toggle("md-nav__link--active",c===s.length-1)}),B("toc.follow")){let s=O(t.pipe(_e(1),m(()=>{})),t.pipe(_e(250),m(()=>"smooth")));i.pipe(b(({prev:p})=>p.length>0),He(o.pipe(ve(se))),re(s)).subscribe(([[{prev:p}],c])=>{let[l]=p[p.length-1];if(l.offsetHeight){let f=cr(l);if(typeof f!="undefined"){let u=l.offsetTop-f.offsetTop,{height:d}=ce(f);f.scrollTo({top:u-d/2,behavior:c})}}})}return B("navigation.tracking")&&t.pipe(W(a),ee("offset"),_e(250),Ce(1),W(n.pipe(Ce(1))),ct({delay:250}),re(i)).subscribe(([,{prev:s}])=>{let p=ye(),c=s[s.length-1];if(c&&c.length){let[l]=c,{hash:f}=new URL(l.href);p.hash!==f&&(p.hash=f,history.replaceState({},"",`${p}`))}else p.hash="",history.replaceState({},"",`${p}`)}),bs(e,{viewport$:t,header$:r}).pipe(w(s=>i.next(s)),_(()=>i.complete()),m(s=>$({ref:e},s)))})}function vs(e,{viewport$:t,main$:r,target$:o}){let n=t.pipe(m(({offset:{y:a}})=>a),Be(2,1),m(([a,s])=>a>s&&s>0),K()),i=r.pipe(m(({active:a})=>a));return z([i,n]).pipe(m(([a,s])=>!(a&&s)),K(),W(o.pipe(Ce(1))),ie(!0),ct({delay:250}),m(a=>({hidden:a})))}function Ei(e,{viewport$:t,header$:r,main$:o,target$:n}){let i=new g,a=i.pipe(Z(),ie(!0));return i.subscribe({next({hidden:s}){e.hidden=s,s?(e.setAttribute("tabindex","-1"),e.blur()):e.removeAttribute("tabindex")},complete(){e.style.top="",e.hidden=!0,e.removeAttribute("tabindex")}}),r.pipe(W(a),ee("height")).subscribe(({height:s})=>{e.style.top=`${s+16}px`}),h(e,"click").subscribe(s=>{s.preventDefault(),window.scrollTo({top:0})}),vs(e,{viewport$:t,main$:o,target$:n}).pipe(w(s=>i.next(s)),_(()=>i.complete()),m(s=>$({ref:e},s)))}function wi({document$:e,viewport$:t}){e.pipe(v(()=>P(".md-ellipsis")),ne(r=>tt(r).pipe(W(e.pipe(Ce(1))),b(o=>o),m(()=>r),Te(1))),b(r=>r.offsetWidth{let o=r.innerText,n=r.closest("a")||r;return n.title=o,B("content.tooltips")?mt(n,{viewport$:t}).pipe(W(e.pipe(Ce(1))),_(()=>n.removeAttribute("title"))):S})).subscribe(),B("content.tooltips")&&e.pipe(v(()=>P(".md-status")),ne(r=>mt(r,{viewport$:t}))).subscribe()}function Ti({document$:e,tablet$:t}){e.pipe(v(()=>P(".md-toggle--indeterminate")),w(r=>{r.indeterminate=!0,r.checked=!1}),ne(r=>h(r,"change").pipe(Dr(()=>r.classList.contains("md-toggle--indeterminate")),m(()=>r))),re(t)).subscribe(([r,o])=>{r.classList.remove("md-toggle--indeterminate"),o&&(r.checked=!1)})}function gs(){return/(iPad|iPhone|iPod)/.test(navigator.userAgent)}function Si({document$:e}){e.pipe(v(()=>P("[data-md-scrollfix]")),w(t=>t.removeAttribute("data-md-scrollfix")),b(gs),ne(t=>h(t,"touchstart").pipe(m(()=>t)))).subscribe(t=>{let r=t.scrollTop;r===0?t.scrollTop=1:r+t.offsetHeight===t.scrollHeight&&(t.scrollTop=r-1)})}function Oi({viewport$:e,tablet$:t}){z([ze("search"),t]).pipe(m(([r,o])=>r&&!o),v(r=>I(r).pipe(Ge(r?400:100))),re(e)).subscribe(([r,{offset:{y:o}}])=>{if(r)document.body.setAttribute("data-md-scrolllock",""),document.body.style.top=`-${o}px`;else{let n=-1*parseInt(document.body.style.top,10);document.body.removeAttribute("data-md-scrolllock"),document.body.style.top="",n&&window.scrollTo(0,n)}})}Object.entries||(Object.entries=function(e){let t=[];for(let r of Object.keys(e))t.push([r,e[r]]);return t});Object.values||(Object.values=function(e){let t=[];for(let r of Object.keys(e))t.push(e[r]);return t});typeof Element!="undefined"&&(Element.prototype.scrollTo||(Element.prototype.scrollTo=function(e,t){typeof e=="object"?(this.scrollLeft=e.left,this.scrollTop=e.top):(this.scrollLeft=e,this.scrollTop=t)}),Element.prototype.replaceWith||(Element.prototype.replaceWith=function(...e){let t=this.parentNode;if(t){e.length===0&&t.removeChild(this);for(let r=e.length-1;r>=0;r--){let o=e[r];typeof o=="string"?o=document.createTextNode(o):o.parentNode&&o.parentNode.removeChild(o),r?t.insertBefore(this.previousSibling,o):t.replaceChild(o,this)}}}));function ys(){return location.protocol==="file:"?Tt(`${new URL("search/search_index.js",eo.base)}`).pipe(m(()=>__index),G(1)):je(new URL("search/search_index.json",eo.base))}document.documentElement.classList.remove("no-js");document.documentElement.classList.add("js");var ot=Go(),Ut=sn(),Lt=ln(Ut),to=an(),Oe=gn(),hr=Pt("(min-width: 960px)"),Mi=Pt("(min-width: 1220px)"),_i=mn(),eo=xe(),Ai=document.forms.namedItem("search")?ys():Ye,ro=new g;Zn({alert$:ro});var oo=new g;B("navigation.instant")&&oi({location$:Ut,viewport$:Oe,progress$:oo}).subscribe(ot);var Li;((Li=eo.version)==null?void 0:Li.provider)==="mike"&&ci({document$:ot});O(Ut,Lt).pipe(Ge(125)).subscribe(()=>{Je("drawer",!1),Je("search",!1)});to.pipe(b(({mode:e})=>e==="global")).subscribe(e=>{switch(e.type){case"p":case",":let t=fe("link[rel=prev]");typeof t!="undefined"&<(t);break;case"n":case".":let r=fe("link[rel=next]");typeof r!="undefined"&<(r);break;case"Enter":let o=Ie();o instanceof HTMLLabelElement&&o.click()}});wi({viewport$:Oe,document$:ot});Ti({document$:ot,tablet$:hr});Si({document$:ot});Oi({viewport$:Oe,tablet$:hr});var rt=Kn(Se("header"),{viewport$:Oe}),Ft=ot.pipe(m(()=>Se("main")),v(e=>Gn(e,{viewport$:Oe,header$:rt})),G(1)),xs=O(...ae("consent").map(e=>En(e,{target$:Lt})),...ae("dialog").map(e=>qn(e,{alert$:ro})),...ae("palette").map(e=>Jn(e)),...ae("progress").map(e=>Xn(e,{progress$:oo})),...ae("search").map(e=>ui(e,{index$:Ai,keyboard$:to})),...ae("source").map(e=>gi(e))),Es=C(()=>O(...ae("announce").map(e=>xn(e)),...ae("content").map(e=>zn(e,{viewport$:Oe,target$:Lt,print$:_i})),...ae("content").map(e=>B("search.highlight")?di(e,{index$:Ai,location$:Ut}):S),...ae("header").map(e=>Yn(e,{viewport$:Oe,header$:rt,main$:Ft})),...ae("header-title").map(e=>Bn(e,{viewport$:Oe,header$:rt})),...ae("sidebar").map(e=>e.getAttribute("data-md-type")==="navigation"?Nr(Mi,()=>Zr(e,{viewport$:Oe,header$:rt,main$:Ft})):Nr(hr,()=>Zr(e,{viewport$:Oe,header$:rt,main$:Ft}))),...ae("tabs").map(e=>yi(e,{viewport$:Oe,header$:rt})),...ae("toc").map(e=>xi(e,{viewport$:Oe,header$:rt,main$:Ft,target$:Lt})),...ae("top").map(e=>Ei(e,{viewport$:Oe,header$:rt,main$:Ft,target$:Lt})))),Ci=ot.pipe(v(()=>Es),Re(xs),G(1));Ci.subscribe();window.document$=ot;window.location$=Ut;window.target$=Lt;window.keyboard$=to;window.viewport$=Oe;window.tablet$=hr;window.screen$=Mi;window.print$=_i;window.alert$=ro;window.progress$=oo;window.component$=Ci;})(); +//# sourceMappingURL=bundle.83f73b43.min.js.map + diff --git a/assets/javascripts/bundle.83f73b43.min.js.map b/assets/javascripts/bundle.83f73b43.min.js.map new file mode 100644 index 0000000..fe920b7 --- /dev/null +++ b/assets/javascripts/bundle.83f73b43.min.js.map @@ -0,0 +1,7 @@ +{ + "version": 3, + "sources": ["node_modules/focus-visible/dist/focus-visible.js", "node_modules/escape-html/index.js", "node_modules/clipboard/dist/clipboard.js", "src/templates/assets/javascripts/bundle.ts", "node_modules/tslib/tslib.es6.mjs", "node_modules/rxjs/src/internal/util/isFunction.ts", "node_modules/rxjs/src/internal/util/createErrorClass.ts", "node_modules/rxjs/src/internal/util/UnsubscriptionError.ts", "node_modules/rxjs/src/internal/util/arrRemove.ts", "node_modules/rxjs/src/internal/Subscription.ts", "node_modules/rxjs/src/internal/config.ts", "node_modules/rxjs/src/internal/scheduler/timeoutProvider.ts", "node_modules/rxjs/src/internal/util/reportUnhandledError.ts", "node_modules/rxjs/src/internal/util/noop.ts", "node_modules/rxjs/src/internal/NotificationFactories.ts", "node_modules/rxjs/src/internal/util/errorContext.ts", "node_modules/rxjs/src/internal/Subscriber.ts", "node_modules/rxjs/src/internal/symbol/observable.ts", "node_modules/rxjs/src/internal/util/identity.ts", "node_modules/rxjs/src/internal/util/pipe.ts", "node_modules/rxjs/src/internal/Observable.ts", "node_modules/rxjs/src/internal/util/lift.ts", "node_modules/rxjs/src/internal/operators/OperatorSubscriber.ts", "node_modules/rxjs/src/internal/scheduler/animationFrameProvider.ts", "node_modules/rxjs/src/internal/util/ObjectUnsubscribedError.ts", "node_modules/rxjs/src/internal/Subject.ts", "node_modules/rxjs/src/internal/BehaviorSubject.ts", "node_modules/rxjs/src/internal/scheduler/dateTimestampProvider.ts", "node_modules/rxjs/src/internal/ReplaySubject.ts", "node_modules/rxjs/src/internal/scheduler/Action.ts", "node_modules/rxjs/src/internal/scheduler/intervalProvider.ts", "node_modules/rxjs/src/internal/scheduler/AsyncAction.ts", "node_modules/rxjs/src/internal/Scheduler.ts", "node_modules/rxjs/src/internal/scheduler/AsyncScheduler.ts", "node_modules/rxjs/src/internal/scheduler/async.ts", "node_modules/rxjs/src/internal/scheduler/QueueAction.ts", "node_modules/rxjs/src/internal/scheduler/QueueScheduler.ts", "node_modules/rxjs/src/internal/scheduler/queue.ts", "node_modules/rxjs/src/internal/scheduler/AnimationFrameAction.ts", "node_modules/rxjs/src/internal/scheduler/AnimationFrameScheduler.ts", "node_modules/rxjs/src/internal/scheduler/animationFrame.ts", "node_modules/rxjs/src/internal/observable/empty.ts", "node_modules/rxjs/src/internal/util/isScheduler.ts", "node_modules/rxjs/src/internal/util/args.ts", "node_modules/rxjs/src/internal/util/isArrayLike.ts", "node_modules/rxjs/src/internal/util/isPromise.ts", "node_modules/rxjs/src/internal/util/isInteropObservable.ts", "node_modules/rxjs/src/internal/util/isAsyncIterable.ts", "node_modules/rxjs/src/internal/util/throwUnobservableError.ts", "node_modules/rxjs/src/internal/symbol/iterator.ts", "node_modules/rxjs/src/internal/util/isIterable.ts", "node_modules/rxjs/src/internal/util/isReadableStreamLike.ts", "node_modules/rxjs/src/internal/observable/innerFrom.ts", "node_modules/rxjs/src/internal/util/executeSchedule.ts", "node_modules/rxjs/src/internal/operators/observeOn.ts", "node_modules/rxjs/src/internal/operators/subscribeOn.ts", "node_modules/rxjs/src/internal/scheduled/scheduleObservable.ts", "node_modules/rxjs/src/internal/scheduled/schedulePromise.ts", "node_modules/rxjs/src/internal/scheduled/scheduleArray.ts", "node_modules/rxjs/src/internal/scheduled/scheduleIterable.ts", "node_modules/rxjs/src/internal/scheduled/scheduleAsyncIterable.ts", "node_modules/rxjs/src/internal/scheduled/scheduleReadableStreamLike.ts", "node_modules/rxjs/src/internal/scheduled/scheduled.ts", "node_modules/rxjs/src/internal/observable/from.ts", "node_modules/rxjs/src/internal/observable/of.ts", "node_modules/rxjs/src/internal/observable/throwError.ts", "node_modules/rxjs/src/internal/util/EmptyError.ts", "node_modules/rxjs/src/internal/util/isDate.ts", "node_modules/rxjs/src/internal/operators/map.ts", "node_modules/rxjs/src/internal/util/mapOneOrManyArgs.ts", "node_modules/rxjs/src/internal/util/argsArgArrayOrObject.ts", "node_modules/rxjs/src/internal/util/createObject.ts", "node_modules/rxjs/src/internal/observable/combineLatest.ts", "node_modules/rxjs/src/internal/operators/mergeInternals.ts", "node_modules/rxjs/src/internal/operators/mergeMap.ts", "node_modules/rxjs/src/internal/operators/mergeAll.ts", "node_modules/rxjs/src/internal/operators/concatAll.ts", "node_modules/rxjs/src/internal/observable/concat.ts", "node_modules/rxjs/src/internal/observable/defer.ts", "node_modules/rxjs/src/internal/observable/fromEvent.ts", "node_modules/rxjs/src/internal/observable/fromEventPattern.ts", "node_modules/rxjs/src/internal/observable/timer.ts", "node_modules/rxjs/src/internal/observable/merge.ts", "node_modules/rxjs/src/internal/observable/never.ts", "node_modules/rxjs/src/internal/util/argsOrArgArray.ts", "node_modules/rxjs/src/internal/operators/filter.ts", "node_modules/rxjs/src/internal/observable/zip.ts", "node_modules/rxjs/src/internal/operators/audit.ts", "node_modules/rxjs/src/internal/operators/auditTime.ts", "node_modules/rxjs/src/internal/operators/bufferCount.ts", "node_modules/rxjs/src/internal/operators/catchError.ts", "node_modules/rxjs/src/internal/operators/scanInternals.ts", "node_modules/rxjs/src/internal/operators/combineLatest.ts", "node_modules/rxjs/src/internal/operators/combineLatestWith.ts", "node_modules/rxjs/src/internal/operators/debounce.ts", "node_modules/rxjs/src/internal/operators/debounceTime.ts", "node_modules/rxjs/src/internal/operators/defaultIfEmpty.ts", "node_modules/rxjs/src/internal/operators/take.ts", "node_modules/rxjs/src/internal/operators/ignoreElements.ts", "node_modules/rxjs/src/internal/operators/mapTo.ts", "node_modules/rxjs/src/internal/operators/delayWhen.ts", "node_modules/rxjs/src/internal/operators/delay.ts", "node_modules/rxjs/src/internal/operators/distinctUntilChanged.ts", "node_modules/rxjs/src/internal/operators/distinctUntilKeyChanged.ts", "node_modules/rxjs/src/internal/operators/throwIfEmpty.ts", "node_modules/rxjs/src/internal/operators/endWith.ts", "node_modules/rxjs/src/internal/operators/finalize.ts", "node_modules/rxjs/src/internal/operators/first.ts", "node_modules/rxjs/src/internal/operators/takeLast.ts", "node_modules/rxjs/src/internal/operators/merge.ts", "node_modules/rxjs/src/internal/operators/mergeWith.ts", "node_modules/rxjs/src/internal/operators/repeat.ts", "node_modules/rxjs/src/internal/operators/scan.ts", "node_modules/rxjs/src/internal/operators/share.ts", "node_modules/rxjs/src/internal/operators/shareReplay.ts", "node_modules/rxjs/src/internal/operators/skip.ts", "node_modules/rxjs/src/internal/operators/skipUntil.ts", "node_modules/rxjs/src/internal/operators/startWith.ts", "node_modules/rxjs/src/internal/operators/switchMap.ts", "node_modules/rxjs/src/internal/operators/takeUntil.ts", "node_modules/rxjs/src/internal/operators/takeWhile.ts", "node_modules/rxjs/src/internal/operators/tap.ts", "node_modules/rxjs/src/internal/operators/throttle.ts", "node_modules/rxjs/src/internal/operators/throttleTime.ts", "node_modules/rxjs/src/internal/operators/withLatestFrom.ts", "node_modules/rxjs/src/internal/operators/zip.ts", "node_modules/rxjs/src/internal/operators/zipWith.ts", "src/templates/assets/javascripts/browser/document/index.ts", "src/templates/assets/javascripts/browser/element/_/index.ts", "src/templates/assets/javascripts/browser/element/focus/index.ts", "src/templates/assets/javascripts/browser/element/hover/index.ts", "src/templates/assets/javascripts/utilities/h/index.ts", "src/templates/assets/javascripts/utilities/round/index.ts", "src/templates/assets/javascripts/browser/script/index.ts", "src/templates/assets/javascripts/browser/element/size/_/index.ts", "src/templates/assets/javascripts/browser/element/size/content/index.ts", "src/templates/assets/javascripts/browser/element/offset/_/index.ts", "src/templates/assets/javascripts/browser/element/offset/content/index.ts", "src/templates/assets/javascripts/browser/element/visibility/index.ts", "src/templates/assets/javascripts/browser/toggle/index.ts", "src/templates/assets/javascripts/browser/keyboard/index.ts", "src/templates/assets/javascripts/browser/location/_/index.ts", "src/templates/assets/javascripts/browser/location/hash/index.ts", "src/templates/assets/javascripts/browser/media/index.ts", "src/templates/assets/javascripts/browser/request/index.ts", "src/templates/assets/javascripts/browser/viewport/offset/index.ts", "src/templates/assets/javascripts/browser/viewport/size/index.ts", "src/templates/assets/javascripts/browser/viewport/_/index.ts", "src/templates/assets/javascripts/browser/viewport/at/index.ts", "src/templates/assets/javascripts/browser/worker/index.ts", "src/templates/assets/javascripts/_/index.ts", "src/templates/assets/javascripts/components/_/index.ts", "src/templates/assets/javascripts/components/announce/index.ts", "src/templates/assets/javascripts/components/consent/index.ts", "src/templates/assets/javascripts/templates/tooltip/index.tsx", "src/templates/assets/javascripts/templates/annotation/index.tsx", "src/templates/assets/javascripts/templates/clipboard/index.tsx", "src/templates/assets/javascripts/templates/search/index.tsx", "src/templates/assets/javascripts/templates/source/index.tsx", "src/templates/assets/javascripts/templates/tabbed/index.tsx", "src/templates/assets/javascripts/templates/table/index.tsx", "src/templates/assets/javascripts/templates/version/index.tsx", "src/templates/assets/javascripts/components/tooltip2/index.ts", "src/templates/assets/javascripts/components/content/annotation/_/index.ts", "src/templates/assets/javascripts/components/content/annotation/list/index.ts", "src/templates/assets/javascripts/components/content/annotation/block/index.ts", "src/templates/assets/javascripts/components/content/code/_/index.ts", "src/templates/assets/javascripts/components/content/details/index.ts", "src/templates/assets/javascripts/components/content/mermaid/index.css", "src/templates/assets/javascripts/components/content/mermaid/index.ts", "src/templates/assets/javascripts/components/content/table/index.ts", "src/templates/assets/javascripts/components/content/tabs/index.ts", "src/templates/assets/javascripts/components/content/_/index.ts", "src/templates/assets/javascripts/components/dialog/index.ts", "src/templates/assets/javascripts/components/tooltip/index.ts", "src/templates/assets/javascripts/components/header/_/index.ts", "src/templates/assets/javascripts/components/header/title/index.ts", "src/templates/assets/javascripts/components/main/index.ts", "src/templates/assets/javascripts/components/palette/index.ts", "src/templates/assets/javascripts/components/progress/index.ts", "src/templates/assets/javascripts/integrations/clipboard/index.ts", "src/templates/assets/javascripts/integrations/sitemap/index.ts", "src/templates/assets/javascripts/integrations/instant/index.ts", "src/templates/assets/javascripts/integrations/search/highlighter/index.ts", "src/templates/assets/javascripts/integrations/search/worker/message/index.ts", "src/templates/assets/javascripts/integrations/search/worker/_/index.ts", "src/templates/assets/javascripts/integrations/version/findurl/index.ts", "src/templates/assets/javascripts/integrations/version/index.ts", "src/templates/assets/javascripts/components/search/query/index.ts", "src/templates/assets/javascripts/components/search/result/index.ts", "src/templates/assets/javascripts/components/search/share/index.ts", "src/templates/assets/javascripts/components/search/suggest/index.ts", "src/templates/assets/javascripts/components/search/_/index.ts", "src/templates/assets/javascripts/components/search/highlight/index.ts", "src/templates/assets/javascripts/components/sidebar/index.ts", "src/templates/assets/javascripts/components/source/facts/github/index.ts", "src/templates/assets/javascripts/components/source/facts/gitlab/index.ts", "src/templates/assets/javascripts/components/source/facts/_/index.ts", "src/templates/assets/javascripts/components/source/_/index.ts", "src/templates/assets/javascripts/components/tabs/index.ts", "src/templates/assets/javascripts/components/toc/index.ts", "src/templates/assets/javascripts/components/top/index.ts", "src/templates/assets/javascripts/patches/ellipsis/index.ts", "src/templates/assets/javascripts/patches/indeterminate/index.ts", "src/templates/assets/javascripts/patches/scrollfix/index.ts", "src/templates/assets/javascripts/patches/scrolllock/index.ts", "src/templates/assets/javascripts/polyfills/index.ts"], + "sourcesContent": ["(function (global, factory) {\n typeof exports === 'object' && typeof module !== 'undefined' ? factory() :\n typeof define === 'function' && define.amd ? define(factory) :\n (factory());\n}(this, (function () { 'use strict';\n\n /**\n * Applies the :focus-visible polyfill at the given scope.\n * A scope in this case is either the top-level Document or a Shadow Root.\n *\n * @param {(Document|ShadowRoot)} scope\n * @see https://github.com/WICG/focus-visible\n */\n function applyFocusVisiblePolyfill(scope) {\n var hadKeyboardEvent = true;\n var hadFocusVisibleRecently = false;\n var hadFocusVisibleRecentlyTimeout = null;\n\n var inputTypesAllowlist = {\n text: true,\n search: true,\n url: true,\n tel: true,\n email: true,\n password: true,\n number: true,\n date: true,\n month: true,\n week: true,\n time: true,\n datetime: true,\n 'datetime-local': true\n };\n\n /**\n * Helper function for legacy browsers and iframes which sometimes focus\n * elements like document, body, and non-interactive SVG.\n * @param {Element} el\n */\n function isValidFocusTarget(el) {\n if (\n el &&\n el !== document &&\n el.nodeName !== 'HTML' &&\n el.nodeName !== 'BODY' &&\n 'classList' in el &&\n 'contains' in el.classList\n ) {\n return true;\n }\n return false;\n }\n\n /**\n * Computes whether the given element should automatically trigger the\n * `focus-visible` class being added, i.e. whether it should always match\n * `:focus-visible` when focused.\n * @param {Element} el\n * @return {boolean}\n */\n function focusTriggersKeyboardModality(el) {\n var type = el.type;\n var tagName = el.tagName;\n\n if (tagName === 'INPUT' && inputTypesAllowlist[type] && !el.readOnly) {\n return true;\n }\n\n if (tagName === 'TEXTAREA' && !el.readOnly) {\n return true;\n }\n\n if (el.isContentEditable) {\n return true;\n }\n\n return false;\n }\n\n /**\n * Add the `focus-visible` class to the given element if it was not added by\n * the author.\n * @param {Element} el\n */\n function addFocusVisibleClass(el) {\n if (el.classList.contains('focus-visible')) {\n return;\n }\n el.classList.add('focus-visible');\n el.setAttribute('data-focus-visible-added', '');\n }\n\n /**\n * Remove the `focus-visible` class from the given element if it was not\n * originally added by the author.\n * @param {Element} el\n */\n function removeFocusVisibleClass(el) {\n if (!el.hasAttribute('data-focus-visible-added')) {\n return;\n }\n el.classList.remove('focus-visible');\n el.removeAttribute('data-focus-visible-added');\n }\n\n /**\n * If the most recent user interaction was via the keyboard;\n * and the key press did not include a meta, alt/option, or control key;\n * then the modality is keyboard. Otherwise, the modality is not keyboard.\n * Apply `focus-visible` to any current active element and keep track\n * of our keyboard modality state with `hadKeyboardEvent`.\n * @param {KeyboardEvent} e\n */\n function onKeyDown(e) {\n if (e.metaKey || e.altKey || e.ctrlKey) {\n return;\n }\n\n if (isValidFocusTarget(scope.activeElement)) {\n addFocusVisibleClass(scope.activeElement);\n }\n\n hadKeyboardEvent = true;\n }\n\n /**\n * If at any point a user clicks with a pointing device, ensure that we change\n * the modality away from keyboard.\n * This avoids the situation where a user presses a key on an already focused\n * element, and then clicks on a different element, focusing it with a\n * pointing device, while we still think we're in keyboard modality.\n * @param {Event} e\n */\n function onPointerDown(e) {\n hadKeyboardEvent = false;\n }\n\n /**\n * On `focus`, add the `focus-visible` class to the target if:\n * - the target received focus as a result of keyboard navigation, or\n * - the event target is an element that will likely require interaction\n * via the keyboard (e.g. a text box)\n * @param {Event} e\n */\n function onFocus(e) {\n // Prevent IE from focusing the document or HTML element.\n if (!isValidFocusTarget(e.target)) {\n return;\n }\n\n if (hadKeyboardEvent || focusTriggersKeyboardModality(e.target)) {\n addFocusVisibleClass(e.target);\n }\n }\n\n /**\n * On `blur`, remove the `focus-visible` class from the target.\n * @param {Event} e\n */\n function onBlur(e) {\n if (!isValidFocusTarget(e.target)) {\n return;\n }\n\n if (\n e.target.classList.contains('focus-visible') ||\n e.target.hasAttribute('data-focus-visible-added')\n ) {\n // To detect a tab/window switch, we look for a blur event followed\n // rapidly by a visibility change.\n // If we don't see a visibility change within 100ms, it's probably a\n // regular focus change.\n hadFocusVisibleRecently = true;\n window.clearTimeout(hadFocusVisibleRecentlyTimeout);\n hadFocusVisibleRecentlyTimeout = window.setTimeout(function() {\n hadFocusVisibleRecently = false;\n }, 100);\n removeFocusVisibleClass(e.target);\n }\n }\n\n /**\n * If the user changes tabs, keep track of whether or not the previously\n * focused element had .focus-visible.\n * @param {Event} e\n */\n function onVisibilityChange(e) {\n if (document.visibilityState === 'hidden') {\n // If the tab becomes active again, the browser will handle calling focus\n // on the element (Safari actually calls it twice).\n // If this tab change caused a blur on an element with focus-visible,\n // re-apply the class when the user switches back to the tab.\n if (hadFocusVisibleRecently) {\n hadKeyboardEvent = true;\n }\n addInitialPointerMoveListeners();\n }\n }\n\n /**\n * Add a group of listeners to detect usage of any pointing devices.\n * These listeners will be added when the polyfill first loads, and anytime\n * the window is blurred, so that they are active when the window regains\n * focus.\n */\n function addInitialPointerMoveListeners() {\n document.addEventListener('mousemove', onInitialPointerMove);\n document.addEventListener('mousedown', onInitialPointerMove);\n document.addEventListener('mouseup', onInitialPointerMove);\n document.addEventListener('pointermove', onInitialPointerMove);\n document.addEventListener('pointerdown', onInitialPointerMove);\n document.addEventListener('pointerup', onInitialPointerMove);\n document.addEventListener('touchmove', onInitialPointerMove);\n document.addEventListener('touchstart', onInitialPointerMove);\n document.addEventListener('touchend', onInitialPointerMove);\n }\n\n function removeInitialPointerMoveListeners() {\n document.removeEventListener('mousemove', onInitialPointerMove);\n document.removeEventListener('mousedown', onInitialPointerMove);\n document.removeEventListener('mouseup', onInitialPointerMove);\n document.removeEventListener('pointermove', onInitialPointerMove);\n document.removeEventListener('pointerdown', onInitialPointerMove);\n document.removeEventListener('pointerup', onInitialPointerMove);\n document.removeEventListener('touchmove', onInitialPointerMove);\n document.removeEventListener('touchstart', onInitialPointerMove);\n document.removeEventListener('touchend', onInitialPointerMove);\n }\n\n /**\n * When the polfyill first loads, assume the user is in keyboard modality.\n * If any event is received from a pointing device (e.g. mouse, pointer,\n * touch), turn off keyboard modality.\n * This accounts for situations where focus enters the page from the URL bar.\n * @param {Event} e\n */\n function onInitialPointerMove(e) {\n // Work around a Safari quirk that fires a mousemove on whenever the\n // window blurs, even if you're tabbing out of the page. \u00AF\\_(\u30C4)_/\u00AF\n if (e.target.nodeName && e.target.nodeName.toLowerCase() === 'html') {\n return;\n }\n\n hadKeyboardEvent = false;\n removeInitialPointerMoveListeners();\n }\n\n // For some kinds of state, we are interested in changes at the global scope\n // only. For example, global pointer input, global key presses and global\n // visibility change should affect the state at every scope:\n document.addEventListener('keydown', onKeyDown, true);\n document.addEventListener('mousedown', onPointerDown, true);\n document.addEventListener('pointerdown', onPointerDown, true);\n document.addEventListener('touchstart', onPointerDown, true);\n document.addEventListener('visibilitychange', onVisibilityChange, true);\n\n addInitialPointerMoveListeners();\n\n // For focus and blur, we specifically care about state changes in the local\n // scope. This is because focus / blur events that originate from within a\n // shadow root are not re-dispatched from the host element if it was already\n // the active element in its own scope:\n scope.addEventListener('focus', onFocus, true);\n scope.addEventListener('blur', onBlur, true);\n\n // We detect that a node is a ShadowRoot by ensuring that it is a\n // DocumentFragment and also has a host property. This check covers native\n // implementation and polyfill implementation transparently. If we only cared\n // about the native implementation, we could just check if the scope was\n // an instance of a ShadowRoot.\n if (scope.nodeType === Node.DOCUMENT_FRAGMENT_NODE && scope.host) {\n // Since a ShadowRoot is a special kind of DocumentFragment, it does not\n // have a root element to add a class to. So, we add this attribute to the\n // host element instead:\n scope.host.setAttribute('data-js-focus-visible', '');\n } else if (scope.nodeType === Node.DOCUMENT_NODE) {\n document.documentElement.classList.add('js-focus-visible');\n document.documentElement.setAttribute('data-js-focus-visible', '');\n }\n }\n\n // It is important to wrap all references to global window and document in\n // these checks to support server-side rendering use cases\n // @see https://github.com/WICG/focus-visible/issues/199\n if (typeof window !== 'undefined' && typeof document !== 'undefined') {\n // Make the polyfill helper globally available. This can be used as a signal\n // to interested libraries that wish to coordinate with the polyfill for e.g.,\n // applying the polyfill to a shadow root:\n window.applyFocusVisiblePolyfill = applyFocusVisiblePolyfill;\n\n // Notify interested libraries of the polyfill's presence, in case the\n // polyfill was loaded lazily:\n var event;\n\n try {\n event = new CustomEvent('focus-visible-polyfill-ready');\n } catch (error) {\n // IE11 does not support using CustomEvent as a constructor directly:\n event = document.createEvent('CustomEvent');\n event.initCustomEvent('focus-visible-polyfill-ready', false, false, {});\n }\n\n window.dispatchEvent(event);\n }\n\n if (typeof document !== 'undefined') {\n // Apply the polyfill to the global document, so that no JavaScript\n // coordination is required to use the polyfill in the top-level document:\n applyFocusVisiblePolyfill(document);\n }\n\n})));\n", "/*!\n * escape-html\n * Copyright(c) 2012-2013 TJ Holowaychuk\n * Copyright(c) 2015 Andreas Lubbe\n * Copyright(c) 2015 Tiancheng \"Timothy\" Gu\n * MIT Licensed\n */\n\n'use strict';\n\n/**\n * Module variables.\n * @private\n */\n\nvar matchHtmlRegExp = /[\"'&<>]/;\n\n/**\n * Module exports.\n * @public\n */\n\nmodule.exports = escapeHtml;\n\n/**\n * Escape special characters in the given string of html.\n *\n * @param {string} string The string to escape for inserting into HTML\n * @return {string}\n * @public\n */\n\nfunction escapeHtml(string) {\n var str = '' + string;\n var match = matchHtmlRegExp.exec(str);\n\n if (!match) {\n return str;\n }\n\n var escape;\n var html = '';\n var index = 0;\n var lastIndex = 0;\n\n for (index = match.index; index < str.length; index++) {\n switch (str.charCodeAt(index)) {\n case 34: // \"\n escape = '"';\n break;\n case 38: // &\n escape = '&';\n break;\n case 39: // '\n escape = ''';\n break;\n case 60: // <\n escape = '<';\n break;\n case 62: // >\n escape = '>';\n break;\n default:\n continue;\n }\n\n if (lastIndex !== index) {\n html += str.substring(lastIndex, index);\n }\n\n lastIndex = index + 1;\n html += escape;\n }\n\n return lastIndex !== index\n ? html + str.substring(lastIndex, index)\n : html;\n}\n", "/*!\n * clipboard.js v2.0.11\n * https://clipboardjs.com/\n *\n * Licensed MIT \u00A9 Zeno Rocha\n */\n(function webpackUniversalModuleDefinition(root, factory) {\n\tif(typeof exports === 'object' && typeof module === 'object')\n\t\tmodule.exports = factory();\n\telse if(typeof define === 'function' && define.amd)\n\t\tdefine([], factory);\n\telse if(typeof exports === 'object')\n\t\texports[\"ClipboardJS\"] = factory();\n\telse\n\t\troot[\"ClipboardJS\"] = factory();\n})(this, function() {\nreturn /******/ (function() { // webpackBootstrap\n/******/ \tvar __webpack_modules__ = ({\n\n/***/ 686:\n/***/ (function(__unused_webpack_module, __webpack_exports__, __webpack_require__) {\n\n\"use strict\";\n\n// EXPORTS\n__webpack_require__.d(__webpack_exports__, {\n \"default\": function() { return /* binding */ clipboard; }\n});\n\n// EXTERNAL MODULE: ./node_modules/tiny-emitter/index.js\nvar tiny_emitter = __webpack_require__(279);\nvar tiny_emitter_default = /*#__PURE__*/__webpack_require__.n(tiny_emitter);\n// EXTERNAL MODULE: ./node_modules/good-listener/src/listen.js\nvar listen = __webpack_require__(370);\nvar listen_default = /*#__PURE__*/__webpack_require__.n(listen);\n// EXTERNAL MODULE: ./node_modules/select/src/select.js\nvar src_select = __webpack_require__(817);\nvar select_default = /*#__PURE__*/__webpack_require__.n(src_select);\n;// CONCATENATED MODULE: ./src/common/command.js\n/**\n * Executes a given operation type.\n * @param {String} type\n * @return {Boolean}\n */\nfunction command(type) {\n try {\n return document.execCommand(type);\n } catch (err) {\n return false;\n }\n}\n;// CONCATENATED MODULE: ./src/actions/cut.js\n\n\n/**\n * Cut action wrapper.\n * @param {String|HTMLElement} target\n * @return {String}\n */\n\nvar ClipboardActionCut = function ClipboardActionCut(target) {\n var selectedText = select_default()(target);\n command('cut');\n return selectedText;\n};\n\n/* harmony default export */ var actions_cut = (ClipboardActionCut);\n;// CONCATENATED MODULE: ./src/common/create-fake-element.js\n/**\n * Creates a fake textarea element with a value.\n * @param {String} value\n * @return {HTMLElement}\n */\nfunction createFakeElement(value) {\n var isRTL = document.documentElement.getAttribute('dir') === 'rtl';\n var fakeElement = document.createElement('textarea'); // Prevent zooming on iOS\n\n fakeElement.style.fontSize = '12pt'; // Reset box model\n\n fakeElement.style.border = '0';\n fakeElement.style.padding = '0';\n fakeElement.style.margin = '0'; // Move element out of screen horizontally\n\n fakeElement.style.position = 'absolute';\n fakeElement.style[isRTL ? 'right' : 'left'] = '-9999px'; // Move element to the same position vertically\n\n var yPosition = window.pageYOffset || document.documentElement.scrollTop;\n fakeElement.style.top = \"\".concat(yPosition, \"px\");\n fakeElement.setAttribute('readonly', '');\n fakeElement.value = value;\n return fakeElement;\n}\n;// CONCATENATED MODULE: ./src/actions/copy.js\n\n\n\n/**\n * Create fake copy action wrapper using a fake element.\n * @param {String} target\n * @param {Object} options\n * @return {String}\n */\n\nvar fakeCopyAction = function fakeCopyAction(value, options) {\n var fakeElement = createFakeElement(value);\n options.container.appendChild(fakeElement);\n var selectedText = select_default()(fakeElement);\n command('copy');\n fakeElement.remove();\n return selectedText;\n};\n/**\n * Copy action wrapper.\n * @param {String|HTMLElement} target\n * @param {Object} options\n * @return {String}\n */\n\n\nvar ClipboardActionCopy = function ClipboardActionCopy(target) {\n var options = arguments.length > 1 && arguments[1] !== undefined ? arguments[1] : {\n container: document.body\n };\n var selectedText = '';\n\n if (typeof target === 'string') {\n selectedText = fakeCopyAction(target, options);\n } else if (target instanceof HTMLInputElement && !['text', 'search', 'url', 'tel', 'password'].includes(target === null || target === void 0 ? void 0 : target.type)) {\n // If input type doesn't support `setSelectionRange`. Simulate it. https://developer.mozilla.org/en-US/docs/Web/API/HTMLInputElement/setSelectionRange\n selectedText = fakeCopyAction(target.value, options);\n } else {\n selectedText = select_default()(target);\n command('copy');\n }\n\n return selectedText;\n};\n\n/* harmony default export */ var actions_copy = (ClipboardActionCopy);\n;// CONCATENATED MODULE: ./src/actions/default.js\nfunction _typeof(obj) { \"@babel/helpers - typeof\"; if (typeof Symbol === \"function\" && typeof Symbol.iterator === \"symbol\") { _typeof = function _typeof(obj) { return typeof obj; }; } else { _typeof = function _typeof(obj) { return obj && typeof Symbol === \"function\" && obj.constructor === Symbol && obj !== Symbol.prototype ? \"symbol\" : typeof obj; }; } return _typeof(obj); }\n\n\n\n/**\n * Inner function which performs selection from either `text` or `target`\n * properties and then executes copy or cut operations.\n * @param {Object} options\n */\n\nvar ClipboardActionDefault = function ClipboardActionDefault() {\n var options = arguments.length > 0 && arguments[0] !== undefined ? arguments[0] : {};\n // Defines base properties passed from constructor.\n var _options$action = options.action,\n action = _options$action === void 0 ? 'copy' : _options$action,\n container = options.container,\n target = options.target,\n text = options.text; // Sets the `action` to be performed which can be either 'copy' or 'cut'.\n\n if (action !== 'copy' && action !== 'cut') {\n throw new Error('Invalid \"action\" value, use either \"copy\" or \"cut\"');\n } // Sets the `target` property using an element that will be have its content copied.\n\n\n if (target !== undefined) {\n if (target && _typeof(target) === 'object' && target.nodeType === 1) {\n if (action === 'copy' && target.hasAttribute('disabled')) {\n throw new Error('Invalid \"target\" attribute. Please use \"readonly\" instead of \"disabled\" attribute');\n }\n\n if (action === 'cut' && (target.hasAttribute('readonly') || target.hasAttribute('disabled'))) {\n throw new Error('Invalid \"target\" attribute. You can\\'t cut text from elements with \"readonly\" or \"disabled\" attributes');\n }\n } else {\n throw new Error('Invalid \"target\" value, use a valid Element');\n }\n } // Define selection strategy based on `text` property.\n\n\n if (text) {\n return actions_copy(text, {\n container: container\n });\n } // Defines which selection strategy based on `target` property.\n\n\n if (target) {\n return action === 'cut' ? actions_cut(target) : actions_copy(target, {\n container: container\n });\n }\n};\n\n/* harmony default export */ var actions_default = (ClipboardActionDefault);\n;// CONCATENATED MODULE: ./src/clipboard.js\nfunction clipboard_typeof(obj) { \"@babel/helpers - typeof\"; if (typeof Symbol === \"function\" && typeof Symbol.iterator === \"symbol\") { clipboard_typeof = function _typeof(obj) { return typeof obj; }; } else { clipboard_typeof = function _typeof(obj) { return obj && typeof Symbol === \"function\" && obj.constructor === Symbol && obj !== Symbol.prototype ? \"symbol\" : typeof obj; }; } return clipboard_typeof(obj); }\n\nfunction _classCallCheck(instance, Constructor) { if (!(instance instanceof Constructor)) { throw new TypeError(\"Cannot call a class as a function\"); } }\n\nfunction _defineProperties(target, props) { for (var i = 0; i < props.length; i++) { var descriptor = props[i]; descriptor.enumerable = descriptor.enumerable || false; descriptor.configurable = true; if (\"value\" in descriptor) descriptor.writable = true; Object.defineProperty(target, descriptor.key, descriptor); } }\n\nfunction _createClass(Constructor, protoProps, staticProps) { if (protoProps) _defineProperties(Constructor.prototype, protoProps); if (staticProps) _defineProperties(Constructor, staticProps); return Constructor; }\n\nfunction _inherits(subClass, superClass) { if (typeof superClass !== \"function\" && superClass !== null) { throw new TypeError(\"Super expression must either be null or a function\"); } subClass.prototype = Object.create(superClass && superClass.prototype, { constructor: { value: subClass, writable: true, configurable: true } }); if (superClass) _setPrototypeOf(subClass, superClass); }\n\nfunction _setPrototypeOf(o, p) { _setPrototypeOf = Object.setPrototypeOf || function _setPrototypeOf(o, p) { o.__proto__ = p; return o; }; return _setPrototypeOf(o, p); }\n\nfunction _createSuper(Derived) { var hasNativeReflectConstruct = _isNativeReflectConstruct(); return function _createSuperInternal() { var Super = _getPrototypeOf(Derived), result; if (hasNativeReflectConstruct) { var NewTarget = _getPrototypeOf(this).constructor; result = Reflect.construct(Super, arguments, NewTarget); } else { result = Super.apply(this, arguments); } return _possibleConstructorReturn(this, result); }; }\n\nfunction _possibleConstructorReturn(self, call) { if (call && (clipboard_typeof(call) === \"object\" || typeof call === \"function\")) { return call; } return _assertThisInitialized(self); }\n\nfunction _assertThisInitialized(self) { if (self === void 0) { throw new ReferenceError(\"this hasn't been initialised - super() hasn't been called\"); } return self; }\n\nfunction _isNativeReflectConstruct() { if (typeof Reflect === \"undefined\" || !Reflect.construct) return false; if (Reflect.construct.sham) return false; if (typeof Proxy === \"function\") return true; try { Date.prototype.toString.call(Reflect.construct(Date, [], function () {})); return true; } catch (e) { return false; } }\n\nfunction _getPrototypeOf(o) { _getPrototypeOf = Object.setPrototypeOf ? Object.getPrototypeOf : function _getPrototypeOf(o) { return o.__proto__ || Object.getPrototypeOf(o); }; return _getPrototypeOf(o); }\n\n\n\n\n\n\n/**\n * Helper function to retrieve attribute value.\n * @param {String} suffix\n * @param {Element} element\n */\n\nfunction getAttributeValue(suffix, element) {\n var attribute = \"data-clipboard-\".concat(suffix);\n\n if (!element.hasAttribute(attribute)) {\n return;\n }\n\n return element.getAttribute(attribute);\n}\n/**\n * Base class which takes one or more elements, adds event listeners to them,\n * and instantiates a new `ClipboardAction` on each click.\n */\n\n\nvar Clipboard = /*#__PURE__*/function (_Emitter) {\n _inherits(Clipboard, _Emitter);\n\n var _super = _createSuper(Clipboard);\n\n /**\n * @param {String|HTMLElement|HTMLCollection|NodeList} trigger\n * @param {Object} options\n */\n function Clipboard(trigger, options) {\n var _this;\n\n _classCallCheck(this, Clipboard);\n\n _this = _super.call(this);\n\n _this.resolveOptions(options);\n\n _this.listenClick(trigger);\n\n return _this;\n }\n /**\n * Defines if attributes would be resolved using internal setter functions\n * or custom functions that were passed in the constructor.\n * @param {Object} options\n */\n\n\n _createClass(Clipboard, [{\n key: \"resolveOptions\",\n value: function resolveOptions() {\n var options = arguments.length > 0 && arguments[0] !== undefined ? arguments[0] : {};\n this.action = typeof options.action === 'function' ? options.action : this.defaultAction;\n this.target = typeof options.target === 'function' ? options.target : this.defaultTarget;\n this.text = typeof options.text === 'function' ? options.text : this.defaultText;\n this.container = clipboard_typeof(options.container) === 'object' ? options.container : document.body;\n }\n /**\n * Adds a click event listener to the passed trigger.\n * @param {String|HTMLElement|HTMLCollection|NodeList} trigger\n */\n\n }, {\n key: \"listenClick\",\n value: function listenClick(trigger) {\n var _this2 = this;\n\n this.listener = listen_default()(trigger, 'click', function (e) {\n return _this2.onClick(e);\n });\n }\n /**\n * Defines a new `ClipboardAction` on each click event.\n * @param {Event} e\n */\n\n }, {\n key: \"onClick\",\n value: function onClick(e) {\n var trigger = e.delegateTarget || e.currentTarget;\n var action = this.action(trigger) || 'copy';\n var text = actions_default({\n action: action,\n container: this.container,\n target: this.target(trigger),\n text: this.text(trigger)\n }); // Fires an event based on the copy operation result.\n\n this.emit(text ? 'success' : 'error', {\n action: action,\n text: text,\n trigger: trigger,\n clearSelection: function clearSelection() {\n if (trigger) {\n trigger.focus();\n }\n\n window.getSelection().removeAllRanges();\n }\n });\n }\n /**\n * Default `action` lookup function.\n * @param {Element} trigger\n */\n\n }, {\n key: \"defaultAction\",\n value: function defaultAction(trigger) {\n return getAttributeValue('action', trigger);\n }\n /**\n * Default `target` lookup function.\n * @param {Element} trigger\n */\n\n }, {\n key: \"defaultTarget\",\n value: function defaultTarget(trigger) {\n var selector = getAttributeValue('target', trigger);\n\n if (selector) {\n return document.querySelector(selector);\n }\n }\n /**\n * Allow fire programmatically a copy action\n * @param {String|HTMLElement} target\n * @param {Object} options\n * @returns Text copied.\n */\n\n }, {\n key: \"defaultText\",\n\n /**\n * Default `text` lookup function.\n * @param {Element} trigger\n */\n value: function defaultText(trigger) {\n return getAttributeValue('text', trigger);\n }\n /**\n * Destroy lifecycle.\n */\n\n }, {\n key: \"destroy\",\n value: function destroy() {\n this.listener.destroy();\n }\n }], [{\n key: \"copy\",\n value: function copy(target) {\n var options = arguments.length > 1 && arguments[1] !== undefined ? arguments[1] : {\n container: document.body\n };\n return actions_copy(target, options);\n }\n /**\n * Allow fire programmatically a cut action\n * @param {String|HTMLElement} target\n * @returns Text cutted.\n */\n\n }, {\n key: \"cut\",\n value: function cut(target) {\n return actions_cut(target);\n }\n /**\n * Returns the support of the given action, or all actions if no action is\n * given.\n * @param {String} [action]\n */\n\n }, {\n key: \"isSupported\",\n value: function isSupported() {\n var action = arguments.length > 0 && arguments[0] !== undefined ? arguments[0] : ['copy', 'cut'];\n var actions = typeof action === 'string' ? [action] : action;\n var support = !!document.queryCommandSupported;\n actions.forEach(function (action) {\n support = support && !!document.queryCommandSupported(action);\n });\n return support;\n }\n }]);\n\n return Clipboard;\n}((tiny_emitter_default()));\n\n/* harmony default export */ var clipboard = (Clipboard);\n\n/***/ }),\n\n/***/ 828:\n/***/ (function(module) {\n\nvar DOCUMENT_NODE_TYPE = 9;\n\n/**\n * A polyfill for Element.matches()\n */\nif (typeof Element !== 'undefined' && !Element.prototype.matches) {\n var proto = Element.prototype;\n\n proto.matches = proto.matchesSelector ||\n proto.mozMatchesSelector ||\n proto.msMatchesSelector ||\n proto.oMatchesSelector ||\n proto.webkitMatchesSelector;\n}\n\n/**\n * Finds the closest parent that matches a selector.\n *\n * @param {Element} element\n * @param {String} selector\n * @return {Function}\n */\nfunction closest (element, selector) {\n while (element && element.nodeType !== DOCUMENT_NODE_TYPE) {\n if (typeof element.matches === 'function' &&\n element.matches(selector)) {\n return element;\n }\n element = element.parentNode;\n }\n}\n\nmodule.exports = closest;\n\n\n/***/ }),\n\n/***/ 438:\n/***/ (function(module, __unused_webpack_exports, __webpack_require__) {\n\nvar closest = __webpack_require__(828);\n\n/**\n * Delegates event to a selector.\n *\n * @param {Element} element\n * @param {String} selector\n * @param {String} type\n * @param {Function} callback\n * @param {Boolean} useCapture\n * @return {Object}\n */\nfunction _delegate(element, selector, type, callback, useCapture) {\n var listenerFn = listener.apply(this, arguments);\n\n element.addEventListener(type, listenerFn, useCapture);\n\n return {\n destroy: function() {\n element.removeEventListener(type, listenerFn, useCapture);\n }\n }\n}\n\n/**\n * Delegates event to a selector.\n *\n * @param {Element|String|Array} [elements]\n * @param {String} selector\n * @param {String} type\n * @param {Function} callback\n * @param {Boolean} useCapture\n * @return {Object}\n */\nfunction delegate(elements, selector, type, callback, useCapture) {\n // Handle the regular Element usage\n if (typeof elements.addEventListener === 'function') {\n return _delegate.apply(null, arguments);\n }\n\n // Handle Element-less usage, it defaults to global delegation\n if (typeof type === 'function') {\n // Use `document` as the first parameter, then apply arguments\n // This is a short way to .unshift `arguments` without running into deoptimizations\n return _delegate.bind(null, document).apply(null, arguments);\n }\n\n // Handle Selector-based usage\n if (typeof elements === 'string') {\n elements = document.querySelectorAll(elements);\n }\n\n // Handle Array-like based usage\n return Array.prototype.map.call(elements, function (element) {\n return _delegate(element, selector, type, callback, useCapture);\n });\n}\n\n/**\n * Finds closest match and invokes callback.\n *\n * @param {Element} element\n * @param {String} selector\n * @param {String} type\n * @param {Function} callback\n * @return {Function}\n */\nfunction listener(element, selector, type, callback) {\n return function(e) {\n e.delegateTarget = closest(e.target, selector);\n\n if (e.delegateTarget) {\n callback.call(element, e);\n }\n }\n}\n\nmodule.exports = delegate;\n\n\n/***/ }),\n\n/***/ 879:\n/***/ (function(__unused_webpack_module, exports) {\n\n/**\n * Check if argument is a HTML element.\n *\n * @param {Object} value\n * @return {Boolean}\n */\nexports.node = function(value) {\n return value !== undefined\n && value instanceof HTMLElement\n && value.nodeType === 1;\n};\n\n/**\n * Check if argument is a list of HTML elements.\n *\n * @param {Object} value\n * @return {Boolean}\n */\nexports.nodeList = function(value) {\n var type = Object.prototype.toString.call(value);\n\n return value !== undefined\n && (type === '[object NodeList]' || type === '[object HTMLCollection]')\n && ('length' in value)\n && (value.length === 0 || exports.node(value[0]));\n};\n\n/**\n * Check if argument is a string.\n *\n * @param {Object} value\n * @return {Boolean}\n */\nexports.string = function(value) {\n return typeof value === 'string'\n || value instanceof String;\n};\n\n/**\n * Check if argument is a function.\n *\n * @param {Object} value\n * @return {Boolean}\n */\nexports.fn = function(value) {\n var type = Object.prototype.toString.call(value);\n\n return type === '[object Function]';\n};\n\n\n/***/ }),\n\n/***/ 370:\n/***/ (function(module, __unused_webpack_exports, __webpack_require__) {\n\nvar is = __webpack_require__(879);\nvar delegate = __webpack_require__(438);\n\n/**\n * Validates all params and calls the right\n * listener function based on its target type.\n *\n * @param {String|HTMLElement|HTMLCollection|NodeList} target\n * @param {String} type\n * @param {Function} callback\n * @return {Object}\n */\nfunction listen(target, type, callback) {\n if (!target && !type && !callback) {\n throw new Error('Missing required arguments');\n }\n\n if (!is.string(type)) {\n throw new TypeError('Second argument must be a String');\n }\n\n if (!is.fn(callback)) {\n throw new TypeError('Third argument must be a Function');\n }\n\n if (is.node(target)) {\n return listenNode(target, type, callback);\n }\n else if (is.nodeList(target)) {\n return listenNodeList(target, type, callback);\n }\n else if (is.string(target)) {\n return listenSelector(target, type, callback);\n }\n else {\n throw new TypeError('First argument must be a String, HTMLElement, HTMLCollection, or NodeList');\n }\n}\n\n/**\n * Adds an event listener to a HTML element\n * and returns a remove listener function.\n *\n * @param {HTMLElement} node\n * @param {String} type\n * @param {Function} callback\n * @return {Object}\n */\nfunction listenNode(node, type, callback) {\n node.addEventListener(type, callback);\n\n return {\n destroy: function() {\n node.removeEventListener(type, callback);\n }\n }\n}\n\n/**\n * Add an event listener to a list of HTML elements\n * and returns a remove listener function.\n *\n * @param {NodeList|HTMLCollection} nodeList\n * @param {String} type\n * @param {Function} callback\n * @return {Object}\n */\nfunction listenNodeList(nodeList, type, callback) {\n Array.prototype.forEach.call(nodeList, function(node) {\n node.addEventListener(type, callback);\n });\n\n return {\n destroy: function() {\n Array.prototype.forEach.call(nodeList, function(node) {\n node.removeEventListener(type, callback);\n });\n }\n }\n}\n\n/**\n * Add an event listener to a selector\n * and returns a remove listener function.\n *\n * @param {String} selector\n * @param {String} type\n * @param {Function} callback\n * @return {Object}\n */\nfunction listenSelector(selector, type, callback) {\n return delegate(document.body, selector, type, callback);\n}\n\nmodule.exports = listen;\n\n\n/***/ }),\n\n/***/ 817:\n/***/ (function(module) {\n\nfunction select(element) {\n var selectedText;\n\n if (element.nodeName === 'SELECT') {\n element.focus();\n\n selectedText = element.value;\n }\n else if (element.nodeName === 'INPUT' || element.nodeName === 'TEXTAREA') {\n var isReadOnly = element.hasAttribute('readonly');\n\n if (!isReadOnly) {\n element.setAttribute('readonly', '');\n }\n\n element.select();\n element.setSelectionRange(0, element.value.length);\n\n if (!isReadOnly) {\n element.removeAttribute('readonly');\n }\n\n selectedText = element.value;\n }\n else {\n if (element.hasAttribute('contenteditable')) {\n element.focus();\n }\n\n var selection = window.getSelection();\n var range = document.createRange();\n\n range.selectNodeContents(element);\n selection.removeAllRanges();\n selection.addRange(range);\n\n selectedText = selection.toString();\n }\n\n return selectedText;\n}\n\nmodule.exports = select;\n\n\n/***/ }),\n\n/***/ 279:\n/***/ (function(module) {\n\nfunction E () {\n // Keep this empty so it's easier to inherit from\n // (via https://github.com/lipsmack from https://github.com/scottcorgan/tiny-emitter/issues/3)\n}\n\nE.prototype = {\n on: function (name, callback, ctx) {\n var e = this.e || (this.e = {});\n\n (e[name] || (e[name] = [])).push({\n fn: callback,\n ctx: ctx\n });\n\n return this;\n },\n\n once: function (name, callback, ctx) {\n var self = this;\n function listener () {\n self.off(name, listener);\n callback.apply(ctx, arguments);\n };\n\n listener._ = callback\n return this.on(name, listener, ctx);\n },\n\n emit: function (name) {\n var data = [].slice.call(arguments, 1);\n var evtArr = ((this.e || (this.e = {}))[name] || []).slice();\n var i = 0;\n var len = evtArr.length;\n\n for (i; i < len; i++) {\n evtArr[i].fn.apply(evtArr[i].ctx, data);\n }\n\n return this;\n },\n\n off: function (name, callback) {\n var e = this.e || (this.e = {});\n var evts = e[name];\n var liveEvents = [];\n\n if (evts && callback) {\n for (var i = 0, len = evts.length; i < len; i++) {\n if (evts[i].fn !== callback && evts[i].fn._ !== callback)\n liveEvents.push(evts[i]);\n }\n }\n\n // Remove event from queue to prevent memory leak\n // Suggested by https://github.com/lazd\n // Ref: https://github.com/scottcorgan/tiny-emitter/commit/c6ebfaa9bc973b33d110a84a307742b7cf94c953#commitcomment-5024910\n\n (liveEvents.length)\n ? e[name] = liveEvents\n : delete e[name];\n\n return this;\n }\n};\n\nmodule.exports = E;\nmodule.exports.TinyEmitter = E;\n\n\n/***/ })\n\n/******/ \t});\n/************************************************************************/\n/******/ \t// The module cache\n/******/ \tvar __webpack_module_cache__ = {};\n/******/ \t\n/******/ \t// The require function\n/******/ \tfunction __webpack_require__(moduleId) {\n/******/ \t\t// Check if module is in cache\n/******/ \t\tif(__webpack_module_cache__[moduleId]) {\n/******/ \t\t\treturn __webpack_module_cache__[moduleId].exports;\n/******/ \t\t}\n/******/ \t\t// Create a new module (and put it into the cache)\n/******/ \t\tvar module = __webpack_module_cache__[moduleId] = {\n/******/ \t\t\t// no module.id needed\n/******/ \t\t\t// no module.loaded needed\n/******/ \t\t\texports: {}\n/******/ \t\t};\n/******/ \t\n/******/ \t\t// Execute the module function\n/******/ \t\t__webpack_modules__[moduleId](module, module.exports, __webpack_require__);\n/******/ \t\n/******/ \t\t// Return the exports of the module\n/******/ \t\treturn module.exports;\n/******/ \t}\n/******/ \t\n/************************************************************************/\n/******/ \t/* webpack/runtime/compat get default export */\n/******/ \t!function() {\n/******/ \t\t// getDefaultExport function for compatibility with non-harmony modules\n/******/ \t\t__webpack_require__.n = function(module) {\n/******/ \t\t\tvar getter = module && module.__esModule ?\n/******/ \t\t\t\tfunction() { return module['default']; } :\n/******/ \t\t\t\tfunction() { return module; };\n/******/ \t\t\t__webpack_require__.d(getter, { a: getter });\n/******/ \t\t\treturn getter;\n/******/ \t\t};\n/******/ \t}();\n/******/ \t\n/******/ \t/* webpack/runtime/define property getters */\n/******/ \t!function() {\n/******/ \t\t// define getter functions for harmony exports\n/******/ \t\t__webpack_require__.d = function(exports, definition) {\n/******/ \t\t\tfor(var key in definition) {\n/******/ \t\t\t\tif(__webpack_require__.o(definition, key) && !__webpack_require__.o(exports, key)) {\n/******/ \t\t\t\t\tObject.defineProperty(exports, key, { enumerable: true, get: definition[key] });\n/******/ \t\t\t\t}\n/******/ \t\t\t}\n/******/ \t\t};\n/******/ \t}();\n/******/ \t\n/******/ \t/* webpack/runtime/hasOwnProperty shorthand */\n/******/ \t!function() {\n/******/ \t\t__webpack_require__.o = function(obj, prop) { return Object.prototype.hasOwnProperty.call(obj, prop); }\n/******/ \t}();\n/******/ \t\n/************************************************************************/\n/******/ \t// module exports must be returned from runtime so entry inlining is disabled\n/******/ \t// startup\n/******/ \t// Load entry module and return exports\n/******/ \treturn __webpack_require__(686);\n/******/ })()\n.default;\n});", "/*\n * Copyright (c) 2016-2024 Martin Donath \n *\n * Permission is hereby granted, free of charge, to any person obtaining a copy\n * of this software and associated documentation files (the \"Software\"), to\n * deal in the Software without restriction, including without limitation the\n * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or\n * sell copies of the Software, and to permit persons to whom the Software is\n * furnished to do so, subject to the following conditions:\n *\n * The above copyright notice and this permission notice shall be included in\n * all copies or substantial portions of the Software.\n *\n * THE SOFTWARE IS PROVIDED \"AS IS\", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR\n * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,\n * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL THE\n * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER\n * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING\n * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS\n * IN THE SOFTWARE.\n */\n\nimport \"focus-visible\"\n\nimport {\n EMPTY,\n NEVER,\n Observable,\n Subject,\n defer,\n delay,\n filter,\n map,\n merge,\n mergeWith,\n shareReplay,\n switchMap\n} from \"rxjs\"\n\nimport { configuration, feature } from \"./_\"\nimport {\n at,\n getActiveElement,\n getOptionalElement,\n requestJSON,\n setLocation,\n setToggle,\n watchDocument,\n watchKeyboard,\n watchLocation,\n watchLocationTarget,\n watchMedia,\n watchPrint,\n watchScript,\n watchViewport\n} from \"./browser\"\nimport {\n getComponentElement,\n getComponentElements,\n mountAnnounce,\n mountBackToTop,\n mountConsent,\n mountContent,\n mountDialog,\n mountHeader,\n mountHeaderTitle,\n mountPalette,\n mountProgress,\n mountSearch,\n mountSearchHiglight,\n mountSidebar,\n mountSource,\n mountTableOfContents,\n mountTabs,\n watchHeader,\n watchMain\n} from \"./components\"\nimport {\n SearchIndex,\n setupClipboardJS,\n setupInstantNavigation,\n setupVersionSelector\n} from \"./integrations\"\nimport {\n patchEllipsis,\n patchIndeterminate,\n patchScrollfix,\n patchScrolllock\n} from \"./patches\"\nimport \"./polyfills\"\n\n/* ----------------------------------------------------------------------------\n * Functions - @todo refactor\n * ------------------------------------------------------------------------- */\n\n/**\n * Fetch search index\n *\n * @returns Search index observable\n */\nfunction fetchSearchIndex(): Observable {\n if (location.protocol === \"file:\") {\n return watchScript(\n `${new URL(\"search/search_index.js\", config.base)}`\n )\n .pipe(\n // @ts-ignore - @todo fix typings\n map(() => __index),\n shareReplay(1)\n )\n } else {\n return requestJSON(\n new URL(\"search/search_index.json\", config.base)\n )\n }\n}\n\n/* ----------------------------------------------------------------------------\n * Application\n * ------------------------------------------------------------------------- */\n\n/* Yay, JavaScript is available */\ndocument.documentElement.classList.remove(\"no-js\")\ndocument.documentElement.classList.add(\"js\")\n\n/* Set up navigation observables and subjects */\nconst document$ = watchDocument()\nconst location$ = watchLocation()\nconst target$ = watchLocationTarget(location$)\nconst keyboard$ = watchKeyboard()\n\n/* Set up media observables */\nconst viewport$ = watchViewport()\nconst tablet$ = watchMedia(\"(min-width: 960px)\")\nconst screen$ = watchMedia(\"(min-width: 1220px)\")\nconst print$ = watchPrint()\n\n/* Retrieve search index, if search is enabled */\nconst config = configuration()\nconst index$ = document.forms.namedItem(\"search\")\n ? fetchSearchIndex()\n : NEVER\n\n/* Set up Clipboard.js integration */\nconst alert$ = new Subject()\nsetupClipboardJS({ alert$ })\n\n/* Set up progress indicator */\nconst progress$ = new Subject()\n\n/* Set up instant navigation, if enabled */\nif (feature(\"navigation.instant\"))\n setupInstantNavigation({ location$, viewport$, progress$ })\n .subscribe(document$)\n\n/* Set up version selector */\nif (config.version?.provider === \"mike\")\n setupVersionSelector({ document$ })\n\n/* Always close drawer and search on navigation */\nmerge(location$, target$)\n .pipe(\n delay(125)\n )\n .subscribe(() => {\n setToggle(\"drawer\", false)\n setToggle(\"search\", false)\n })\n\n/* Set up global keyboard handlers */\nkeyboard$\n .pipe(\n filter(({ mode }) => mode === \"global\")\n )\n .subscribe(key => {\n switch (key.type) {\n\n /* Go to previous page */\n case \"p\":\n case \",\":\n const prev = getOptionalElement(\"link[rel=prev]\")\n if (typeof prev !== \"undefined\")\n setLocation(prev)\n break\n\n /* Go to next page */\n case \"n\":\n case \".\":\n const next = getOptionalElement(\"link[rel=next]\")\n if (typeof next !== \"undefined\")\n setLocation(next)\n break\n\n /* Expand navigation, see https://bit.ly/3ZjG5io */\n case \"Enter\":\n const active = getActiveElement()\n if (active instanceof HTMLLabelElement)\n active.click()\n }\n })\n\n/* Set up patches */\npatchEllipsis({ viewport$, document$ })\npatchIndeterminate({ document$, tablet$ })\npatchScrollfix({ document$ })\npatchScrolllock({ viewport$, tablet$ })\n\n/* Set up header and main area observable */\nconst header$ = watchHeader(getComponentElement(\"header\"), { viewport$ })\nconst main$ = document$\n .pipe(\n map(() => getComponentElement(\"main\")),\n switchMap(el => watchMain(el, { viewport$, header$ })),\n shareReplay(1)\n )\n\n/* Set up control component observables */\nconst control$ = merge(\n\n /* Consent */\n ...getComponentElements(\"consent\")\n .map(el => mountConsent(el, { target$ })),\n\n /* Dialog */\n ...getComponentElements(\"dialog\")\n .map(el => mountDialog(el, { alert$ })),\n\n /* Color palette */\n ...getComponentElements(\"palette\")\n .map(el => mountPalette(el)),\n\n /* Progress bar */\n ...getComponentElements(\"progress\")\n .map(el => mountProgress(el, { progress$ })),\n\n /* Search */\n ...getComponentElements(\"search\")\n .map(el => mountSearch(el, { index$, keyboard$ })),\n\n /* Repository information */\n ...getComponentElements(\"source\")\n .map(el => mountSource(el))\n)\n\n/* Set up content component observables */\nconst content$ = defer(() => merge(\n\n /* Announcement bar */\n ...getComponentElements(\"announce\")\n .map(el => mountAnnounce(el)),\n\n /* Content */\n ...getComponentElements(\"content\")\n .map(el => mountContent(el, { viewport$, target$, print$ })),\n\n /* Search highlighting */\n ...getComponentElements(\"content\")\n .map(el => feature(\"search.highlight\")\n ? mountSearchHiglight(el, { index$, location$ })\n : EMPTY\n ),\n\n /* Header */\n ...getComponentElements(\"header\")\n .map(el => mountHeader(el, { viewport$, header$, main$ })),\n\n /* Header title */\n ...getComponentElements(\"header-title\")\n .map(el => mountHeaderTitle(el, { viewport$, header$ })),\n\n /* Sidebar */\n ...getComponentElements(\"sidebar\")\n .map(el => el.getAttribute(\"data-md-type\") === \"navigation\"\n ? at(screen$, () => mountSidebar(el, { viewport$, header$, main$ }))\n : at(tablet$, () => mountSidebar(el, { viewport$, header$, main$ }))\n ),\n\n /* Navigation tabs */\n ...getComponentElements(\"tabs\")\n .map(el => mountTabs(el, { viewport$, header$ })),\n\n /* Table of contents */\n ...getComponentElements(\"toc\")\n .map(el => mountTableOfContents(el, {\n viewport$, header$, main$, target$\n })),\n\n /* Back-to-top button */\n ...getComponentElements(\"top\")\n .map(el => mountBackToTop(el, { viewport$, header$, main$, target$ }))\n))\n\n/* Set up component observables */\nconst component$ = document$\n .pipe(\n switchMap(() => content$),\n mergeWith(control$),\n shareReplay(1)\n )\n\n/* Subscribe to all components */\ncomponent$.subscribe()\n\n/* ----------------------------------------------------------------------------\n * Exports\n * ------------------------------------------------------------------------- */\n\nwindow.document$ = document$ /* Document observable */\nwindow.location$ = location$ /* Location subject */\nwindow.target$ = target$ /* Location target observable */\nwindow.keyboard$ = keyboard$ /* Keyboard observable */\nwindow.viewport$ = viewport$ /* Viewport observable */\nwindow.tablet$ = tablet$ /* Media tablet observable */\nwindow.screen$ = screen$ /* Media screen observable */\nwindow.print$ = print$ /* Media print observable */\nwindow.alert$ = alert$ /* Alert subject */\nwindow.progress$ = progress$ /* Progress indicator subject */\nwindow.component$ = component$ /* Component observable */\n", "/******************************************************************************\nCopyright (c) Microsoft Corporation.\n\nPermission to use, copy, modify, and/or distribute this software for any\npurpose with or without fee is hereby granted.\n\nTHE SOFTWARE IS PROVIDED \"AS IS\" AND THE AUTHOR DISCLAIMS ALL WARRANTIES WITH\nREGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY\nAND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY SPECIAL, DIRECT,\nINDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM\nLOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR\nOTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR\nPERFORMANCE OF THIS SOFTWARE.\n***************************************************************************** */\n/* global Reflect, Promise, SuppressedError, Symbol, Iterator */\n\nvar extendStatics = function(d, b) {\n extendStatics = Object.setPrototypeOf ||\n ({ __proto__: [] } instanceof Array && function (d, b) { d.__proto__ = b; }) ||\n function (d, b) { for (var p in b) if (Object.prototype.hasOwnProperty.call(b, p)) d[p] = b[p]; };\n return extendStatics(d, b);\n};\n\nexport function __extends(d, b) {\n if (typeof b !== \"function\" && b !== null)\n throw new TypeError(\"Class extends value \" + String(b) + \" is not a constructor or null\");\n extendStatics(d, b);\n function __() { this.constructor = d; }\n d.prototype = b === null ? Object.create(b) : (__.prototype = b.prototype, new __());\n}\n\nexport var __assign = function() {\n __assign = Object.assign || function __assign(t) {\n for (var s, i = 1, n = arguments.length; i < n; i++) {\n s = arguments[i];\n for (var p in s) if (Object.prototype.hasOwnProperty.call(s, p)) t[p] = s[p];\n }\n return t;\n }\n return __assign.apply(this, arguments);\n}\n\nexport function __rest(s, e) {\n var t = {};\n for (var p in s) if (Object.prototype.hasOwnProperty.call(s, p) && e.indexOf(p) < 0)\n t[p] = s[p];\n if (s != null && typeof Object.getOwnPropertySymbols === \"function\")\n for (var i = 0, p = Object.getOwnPropertySymbols(s); i < p.length; i++) {\n if (e.indexOf(p[i]) < 0 && Object.prototype.propertyIsEnumerable.call(s, p[i]))\n t[p[i]] = s[p[i]];\n }\n return t;\n}\n\nexport function __decorate(decorators, target, key, desc) {\n var c = arguments.length, r = c < 3 ? target : desc === null ? desc = Object.getOwnPropertyDescriptor(target, key) : desc, d;\n if (typeof Reflect === \"object\" && typeof Reflect.decorate === \"function\") r = Reflect.decorate(decorators, target, key, desc);\n else for (var i = decorators.length - 1; i >= 0; i--) if (d = decorators[i]) r = (c < 3 ? d(r) : c > 3 ? d(target, key, r) : d(target, key)) || r;\n return c > 3 && r && Object.defineProperty(target, key, r), r;\n}\n\nexport function __param(paramIndex, decorator) {\n return function (target, key) { decorator(target, key, paramIndex); }\n}\n\nexport function __esDecorate(ctor, descriptorIn, decorators, contextIn, initializers, extraInitializers) {\n function accept(f) { if (f !== void 0 && typeof f !== \"function\") throw new TypeError(\"Function expected\"); return f; }\n var kind = contextIn.kind, key = kind === \"getter\" ? \"get\" : kind === \"setter\" ? \"set\" : \"value\";\n var target = !descriptorIn && ctor ? contextIn[\"static\"] ? ctor : ctor.prototype : null;\n var descriptor = descriptorIn || (target ? Object.getOwnPropertyDescriptor(target, contextIn.name) : {});\n var _, done = false;\n for (var i = decorators.length - 1; i >= 0; i--) {\n var context = {};\n for (var p in contextIn) context[p] = p === \"access\" ? {} : contextIn[p];\n for (var p in contextIn.access) context.access[p] = contextIn.access[p];\n context.addInitializer = function (f) { if (done) throw new TypeError(\"Cannot add initializers after decoration has completed\"); extraInitializers.push(accept(f || null)); };\n var result = (0, decorators[i])(kind === \"accessor\" ? { get: descriptor.get, set: descriptor.set } : descriptor[key], context);\n if (kind === \"accessor\") {\n if (result === void 0) continue;\n if (result === null || typeof result !== \"object\") throw new TypeError(\"Object expected\");\n if (_ = accept(result.get)) descriptor.get = _;\n if (_ = accept(result.set)) descriptor.set = _;\n if (_ = accept(result.init)) initializers.unshift(_);\n }\n else if (_ = accept(result)) {\n if (kind === \"field\") initializers.unshift(_);\n else descriptor[key] = _;\n }\n }\n if (target) Object.defineProperty(target, contextIn.name, descriptor);\n done = true;\n};\n\nexport function __runInitializers(thisArg, initializers, value) {\n var useValue = arguments.length > 2;\n for (var i = 0; i < initializers.length; i++) {\n value = useValue ? initializers[i].call(thisArg, value) : initializers[i].call(thisArg);\n }\n return useValue ? value : void 0;\n};\n\nexport function __propKey(x) {\n return typeof x === \"symbol\" ? x : \"\".concat(x);\n};\n\nexport function __setFunctionName(f, name, prefix) {\n if (typeof name === \"symbol\") name = name.description ? \"[\".concat(name.description, \"]\") : \"\";\n return Object.defineProperty(f, \"name\", { configurable: true, value: prefix ? \"\".concat(prefix, \" \", name) : name });\n};\n\nexport function __metadata(metadataKey, metadataValue) {\n if (typeof Reflect === \"object\" && typeof Reflect.metadata === \"function\") return Reflect.metadata(metadataKey, metadataValue);\n}\n\nexport function __awaiter(thisArg, _arguments, P, generator) {\n function adopt(value) { return value instanceof P ? value : new P(function (resolve) { resolve(value); }); }\n return new (P || (P = Promise))(function (resolve, reject) {\n function fulfilled(value) { try { step(generator.next(value)); } catch (e) { reject(e); } }\n function rejected(value) { try { step(generator[\"throw\"](value)); } catch (e) { reject(e); } }\n function step(result) { result.done ? resolve(result.value) : adopt(result.value).then(fulfilled, rejected); }\n step((generator = generator.apply(thisArg, _arguments || [])).next());\n });\n}\n\nexport function __generator(thisArg, body) {\n var _ = { label: 0, sent: function() { if (t[0] & 1) throw t[1]; return t[1]; }, trys: [], ops: [] }, f, y, t, g = Object.create((typeof Iterator === \"function\" ? Iterator : Object).prototype);\n return g.next = verb(0), g[\"throw\"] = verb(1), g[\"return\"] = verb(2), typeof Symbol === \"function\" && (g[Symbol.iterator] = function() { return this; }), g;\n function verb(n) { return function (v) { return step([n, v]); }; }\n function step(op) {\n if (f) throw new TypeError(\"Generator is already executing.\");\n while (g && (g = 0, op[0] && (_ = 0)), _) try {\n if (f = 1, y && (t = op[0] & 2 ? y[\"return\"] : op[0] ? y[\"throw\"] || ((t = y[\"return\"]) && t.call(y), 0) : y.next) && !(t = t.call(y, op[1])).done) return t;\n if (y = 0, t) op = [op[0] & 2, t.value];\n switch (op[0]) {\n case 0: case 1: t = op; break;\n case 4: _.label++; return { value: op[1], done: false };\n case 5: _.label++; y = op[1]; op = [0]; continue;\n case 7: op = _.ops.pop(); _.trys.pop(); continue;\n default:\n if (!(t = _.trys, t = t.length > 0 && t[t.length - 1]) && (op[0] === 6 || op[0] === 2)) { _ = 0; continue; }\n if (op[0] === 3 && (!t || (op[1] > t[0] && op[1] < t[3]))) { _.label = op[1]; break; }\n if (op[0] === 6 && _.label < t[1]) { _.label = t[1]; t = op; break; }\n if (t && _.label < t[2]) { _.label = t[2]; _.ops.push(op); break; }\n if (t[2]) _.ops.pop();\n _.trys.pop(); continue;\n }\n op = body.call(thisArg, _);\n } catch (e) { op = [6, e]; y = 0; } finally { f = t = 0; }\n if (op[0] & 5) throw op[1]; return { value: op[0] ? op[1] : void 0, done: true };\n }\n}\n\nexport var __createBinding = Object.create ? (function(o, m, k, k2) {\n if (k2 === undefined) k2 = k;\n var desc = Object.getOwnPropertyDescriptor(m, k);\n if (!desc || (\"get\" in desc ? !m.__esModule : desc.writable || desc.configurable)) {\n desc = { enumerable: true, get: function() { return m[k]; } };\n }\n Object.defineProperty(o, k2, desc);\n}) : (function(o, m, k, k2) {\n if (k2 === undefined) k2 = k;\n o[k2] = m[k];\n});\n\nexport function __exportStar(m, o) {\n for (var p in m) if (p !== \"default\" && !Object.prototype.hasOwnProperty.call(o, p)) __createBinding(o, m, p);\n}\n\nexport function __values(o) {\n var s = typeof Symbol === \"function\" && Symbol.iterator, m = s && o[s], i = 0;\n if (m) return m.call(o);\n if (o && typeof o.length === \"number\") return {\n next: function () {\n if (o && i >= o.length) o = void 0;\n return { value: o && o[i++], done: !o };\n }\n };\n throw new TypeError(s ? \"Object is not iterable.\" : \"Symbol.iterator is not defined.\");\n}\n\nexport function __read(o, n) {\n var m = typeof Symbol === \"function\" && o[Symbol.iterator];\n if (!m) return o;\n var i = m.call(o), r, ar = [], e;\n try {\n while ((n === void 0 || n-- > 0) && !(r = i.next()).done) ar.push(r.value);\n }\n catch (error) { e = { error: error }; }\n finally {\n try {\n if (r && !r.done && (m = i[\"return\"])) m.call(i);\n }\n finally { if (e) throw e.error; }\n }\n return ar;\n}\n\n/** @deprecated */\nexport function __spread() {\n for (var ar = [], i = 0; i < arguments.length; i++)\n ar = ar.concat(__read(arguments[i]));\n return ar;\n}\n\n/** @deprecated */\nexport function __spreadArrays() {\n for (var s = 0, i = 0, il = arguments.length; i < il; i++) s += arguments[i].length;\n for (var r = Array(s), k = 0, i = 0; i < il; i++)\n for (var a = arguments[i], j = 0, jl = a.length; j < jl; j++, k++)\n r[k] = a[j];\n return r;\n}\n\nexport function __spreadArray(to, from, pack) {\n if (pack || arguments.length === 2) for (var i = 0, l = from.length, ar; i < l; i++) {\n if (ar || !(i in from)) {\n if (!ar) ar = Array.prototype.slice.call(from, 0, i);\n ar[i] = from[i];\n }\n }\n return to.concat(ar || Array.prototype.slice.call(from));\n}\n\nexport function __await(v) {\n return this instanceof __await ? (this.v = v, this) : new __await(v);\n}\n\nexport function __asyncGenerator(thisArg, _arguments, generator) {\n if (!Symbol.asyncIterator) throw new TypeError(\"Symbol.asyncIterator is not defined.\");\n var g = generator.apply(thisArg, _arguments || []), i, q = [];\n return i = Object.create((typeof AsyncIterator === \"function\" ? AsyncIterator : Object).prototype), verb(\"next\"), verb(\"throw\"), verb(\"return\", awaitReturn), i[Symbol.asyncIterator] = function () { return this; }, i;\n function awaitReturn(f) { return function (v) { return Promise.resolve(v).then(f, reject); }; }\n function verb(n, f) { if (g[n]) { i[n] = function (v) { return new Promise(function (a, b) { q.push([n, v, a, b]) > 1 || resume(n, v); }); }; if (f) i[n] = f(i[n]); } }\n function resume(n, v) { try { step(g[n](v)); } catch (e) { settle(q[0][3], e); } }\n function step(r) { r.value instanceof __await ? Promise.resolve(r.value.v).then(fulfill, reject) : settle(q[0][2], r); }\n function fulfill(value) { resume(\"next\", value); }\n function reject(value) { resume(\"throw\", value); }\n function settle(f, v) { if (f(v), q.shift(), q.length) resume(q[0][0], q[0][1]); }\n}\n\nexport function __asyncDelegator(o) {\n var i, p;\n return i = {}, verb(\"next\"), verb(\"throw\", function (e) { throw e; }), verb(\"return\"), i[Symbol.iterator] = function () { return this; }, i;\n function verb(n, f) { i[n] = o[n] ? function (v) { return (p = !p) ? { value: __await(o[n](v)), done: false } : f ? f(v) : v; } : f; }\n}\n\nexport function __asyncValues(o) {\n if (!Symbol.asyncIterator) throw new TypeError(\"Symbol.asyncIterator is not defined.\");\n var m = o[Symbol.asyncIterator], i;\n return m ? m.call(o) : (o = typeof __values === \"function\" ? __values(o) : o[Symbol.iterator](), i = {}, verb(\"next\"), verb(\"throw\"), verb(\"return\"), i[Symbol.asyncIterator] = function () { return this; }, i);\n function verb(n) { i[n] = o[n] && function (v) { return new Promise(function (resolve, reject) { v = o[n](v), settle(resolve, reject, v.done, v.value); }); }; }\n function settle(resolve, reject, d, v) { Promise.resolve(v).then(function(v) { resolve({ value: v, done: d }); }, reject); }\n}\n\nexport function __makeTemplateObject(cooked, raw) {\n if (Object.defineProperty) { Object.defineProperty(cooked, \"raw\", { value: raw }); } else { cooked.raw = raw; }\n return cooked;\n};\n\nvar __setModuleDefault = Object.create ? (function(o, v) {\n Object.defineProperty(o, \"default\", { enumerable: true, value: v });\n}) : function(o, v) {\n o[\"default\"] = v;\n};\n\nexport function __importStar(mod) {\n if (mod && mod.__esModule) return mod;\n var result = {};\n if (mod != null) for (var k in mod) if (k !== \"default\" && Object.prototype.hasOwnProperty.call(mod, k)) __createBinding(result, mod, k);\n __setModuleDefault(result, mod);\n return result;\n}\n\nexport function __importDefault(mod) {\n return (mod && mod.__esModule) ? mod : { default: mod };\n}\n\nexport function __classPrivateFieldGet(receiver, state, kind, f) {\n if (kind === \"a\" && !f) throw new TypeError(\"Private accessor was defined without a getter\");\n if (typeof state === \"function\" ? receiver !== state || !f : !state.has(receiver)) throw new TypeError(\"Cannot read private member from an object whose class did not declare it\");\n return kind === \"m\" ? f : kind === \"a\" ? f.call(receiver) : f ? f.value : state.get(receiver);\n}\n\nexport function __classPrivateFieldSet(receiver, state, value, kind, f) {\n if (kind === \"m\") throw new TypeError(\"Private method is not writable\");\n if (kind === \"a\" && !f) throw new TypeError(\"Private accessor was defined without a setter\");\n if (typeof state === \"function\" ? receiver !== state || !f : !state.has(receiver)) throw new TypeError(\"Cannot write private member to an object whose class did not declare it\");\n return (kind === \"a\" ? f.call(receiver, value) : f ? f.value = value : state.set(receiver, value)), value;\n}\n\nexport function __classPrivateFieldIn(state, receiver) {\n if (receiver === null || (typeof receiver !== \"object\" && typeof receiver !== \"function\")) throw new TypeError(\"Cannot use 'in' operator on non-object\");\n return typeof state === \"function\" ? receiver === state : state.has(receiver);\n}\n\nexport function __addDisposableResource(env, value, async) {\n if (value !== null && value !== void 0) {\n if (typeof value !== \"object\" && typeof value !== \"function\") throw new TypeError(\"Object expected.\");\n var dispose, inner;\n if (async) {\n if (!Symbol.asyncDispose) throw new TypeError(\"Symbol.asyncDispose is not defined.\");\n dispose = value[Symbol.asyncDispose];\n }\n if (dispose === void 0) {\n if (!Symbol.dispose) throw new TypeError(\"Symbol.dispose is not defined.\");\n dispose = value[Symbol.dispose];\n if (async) inner = dispose;\n }\n if (typeof dispose !== \"function\") throw new TypeError(\"Object not disposable.\");\n if (inner) dispose = function() { try { inner.call(this); } catch (e) { return Promise.reject(e); } };\n env.stack.push({ value: value, dispose: dispose, async: async });\n }\n else if (async) {\n env.stack.push({ async: true });\n }\n return value;\n}\n\nvar _SuppressedError = typeof SuppressedError === \"function\" ? SuppressedError : function (error, suppressed, message) {\n var e = new Error(message);\n return e.name = \"SuppressedError\", e.error = error, e.suppressed = suppressed, e;\n};\n\nexport function __disposeResources(env) {\n function fail(e) {\n env.error = env.hasError ? new _SuppressedError(e, env.error, \"An error was suppressed during disposal.\") : e;\n env.hasError = true;\n }\n var r, s = 0;\n function next() {\n while (r = env.stack.pop()) {\n try {\n if (!r.async && s === 1) return s = 0, env.stack.push(r), Promise.resolve().then(next);\n if (r.dispose) {\n var result = r.dispose.call(r.value);\n if (r.async) return s |= 2, Promise.resolve(result).then(next, function(e) { fail(e); return next(); });\n }\n else s |= 1;\n }\n catch (e) {\n fail(e);\n }\n }\n if (s === 1) return env.hasError ? Promise.reject(env.error) : Promise.resolve();\n if (env.hasError) throw env.error;\n }\n return next();\n}\n\nexport default {\n __extends,\n __assign,\n __rest,\n __decorate,\n __param,\n __metadata,\n __awaiter,\n __generator,\n __createBinding,\n __exportStar,\n __values,\n __read,\n __spread,\n __spreadArrays,\n __spreadArray,\n __await,\n __asyncGenerator,\n __asyncDelegator,\n __asyncValues,\n __makeTemplateObject,\n __importStar,\n __importDefault,\n __classPrivateFieldGet,\n __classPrivateFieldSet,\n __classPrivateFieldIn,\n __addDisposableResource,\n __disposeResources,\n};\n", "/**\n * Returns true if the object is a function.\n * @param value The value to check\n */\nexport function isFunction(value: any): value is (...args: any[]) => any {\n return typeof value === 'function';\n}\n", "/**\n * Used to create Error subclasses until the community moves away from ES5.\n *\n * This is because compiling from TypeScript down to ES5 has issues with subclassing Errors\n * as well as other built-in types: https://github.com/Microsoft/TypeScript/issues/12123\n *\n * @param createImpl A factory function to create the actual constructor implementation. The returned\n * function should be a named function that calls `_super` internally.\n */\nexport function createErrorClass(createImpl: (_super: any) => any): T {\n const _super = (instance: any) => {\n Error.call(instance);\n instance.stack = new Error().stack;\n };\n\n const ctorFunc = createImpl(_super);\n ctorFunc.prototype = Object.create(Error.prototype);\n ctorFunc.prototype.constructor = ctorFunc;\n return ctorFunc;\n}\n", "import { createErrorClass } from './createErrorClass';\n\nexport interface UnsubscriptionError extends Error {\n readonly errors: any[];\n}\n\nexport interface UnsubscriptionErrorCtor {\n /**\n * @deprecated Internal implementation detail. Do not construct error instances.\n * Cannot be tagged as internal: https://github.com/ReactiveX/rxjs/issues/6269\n */\n new (errors: any[]): UnsubscriptionError;\n}\n\n/**\n * An error thrown when one or more errors have occurred during the\n * `unsubscribe` of a {@link Subscription}.\n */\nexport const UnsubscriptionError: UnsubscriptionErrorCtor = createErrorClass(\n (_super) =>\n function UnsubscriptionErrorImpl(this: any, errors: (Error | string)[]) {\n _super(this);\n this.message = errors\n ? `${errors.length} errors occurred during unsubscription:\n${errors.map((err, i) => `${i + 1}) ${err.toString()}`).join('\\n ')}`\n : '';\n this.name = 'UnsubscriptionError';\n this.errors = errors;\n }\n);\n", "/**\n * Removes an item from an array, mutating it.\n * @param arr The array to remove the item from\n * @param item The item to remove\n */\nexport function arrRemove(arr: T[] | undefined | null, item: T) {\n if (arr) {\n const index = arr.indexOf(item);\n 0 <= index && arr.splice(index, 1);\n }\n}\n", "import { isFunction } from './util/isFunction';\nimport { UnsubscriptionError } from './util/UnsubscriptionError';\nimport { SubscriptionLike, TeardownLogic, Unsubscribable } from './types';\nimport { arrRemove } from './util/arrRemove';\n\n/**\n * Represents a disposable resource, such as the execution of an Observable. A\n * Subscription has one important method, `unsubscribe`, that takes no argument\n * and just disposes the resource held by the subscription.\n *\n * Additionally, subscriptions may be grouped together through the `add()`\n * method, which will attach a child Subscription to the current Subscription.\n * When a Subscription is unsubscribed, all its children (and its grandchildren)\n * will be unsubscribed as well.\n *\n * @class Subscription\n */\nexport class Subscription implements SubscriptionLike {\n /** @nocollapse */\n public static EMPTY = (() => {\n const empty = new Subscription();\n empty.closed = true;\n return empty;\n })();\n\n /**\n * A flag to indicate whether this Subscription has already been unsubscribed.\n */\n public closed = false;\n\n private _parentage: Subscription[] | Subscription | null = null;\n\n /**\n * The list of registered finalizers to execute upon unsubscription. Adding and removing from this\n * list occurs in the {@link #add} and {@link #remove} methods.\n */\n private _finalizers: Exclude[] | null = null;\n\n /**\n * @param initialTeardown A function executed first as part of the finalization\n * process that is kicked off when {@link #unsubscribe} is called.\n */\n constructor(private initialTeardown?: () => void) {}\n\n /**\n * Disposes the resources held by the subscription. May, for instance, cancel\n * an ongoing Observable execution or cancel any other type of work that\n * started when the Subscription was created.\n * @return {void}\n */\n unsubscribe(): void {\n let errors: any[] | undefined;\n\n if (!this.closed) {\n this.closed = true;\n\n // Remove this from it's parents.\n const { _parentage } = this;\n if (_parentage) {\n this._parentage = null;\n if (Array.isArray(_parentage)) {\n for (const parent of _parentage) {\n parent.remove(this);\n }\n } else {\n _parentage.remove(this);\n }\n }\n\n const { initialTeardown: initialFinalizer } = this;\n if (isFunction(initialFinalizer)) {\n try {\n initialFinalizer();\n } catch (e) {\n errors = e instanceof UnsubscriptionError ? e.errors : [e];\n }\n }\n\n const { _finalizers } = this;\n if (_finalizers) {\n this._finalizers = null;\n for (const finalizer of _finalizers) {\n try {\n execFinalizer(finalizer);\n } catch (err) {\n errors = errors ?? [];\n if (err instanceof UnsubscriptionError) {\n errors = [...errors, ...err.errors];\n } else {\n errors.push(err);\n }\n }\n }\n }\n\n if (errors) {\n throw new UnsubscriptionError(errors);\n }\n }\n }\n\n /**\n * Adds a finalizer to this subscription, so that finalization will be unsubscribed/called\n * when this subscription is unsubscribed. If this subscription is already {@link #closed},\n * because it has already been unsubscribed, then whatever finalizer is passed to it\n * will automatically be executed (unless the finalizer itself is also a closed subscription).\n *\n * Closed Subscriptions cannot be added as finalizers to any subscription. Adding a closed\n * subscription to a any subscription will result in no operation. (A noop).\n *\n * Adding a subscription to itself, or adding `null` or `undefined` will not perform any\n * operation at all. (A noop).\n *\n * `Subscription` instances that are added to this instance will automatically remove themselves\n * if they are unsubscribed. Functions and {@link Unsubscribable} objects that you wish to remove\n * will need to be removed manually with {@link #remove}\n *\n * @param teardown The finalization logic to add to this subscription.\n */\n add(teardown: TeardownLogic): void {\n // Only add the finalizer if it's not undefined\n // and don't add a subscription to itself.\n if (teardown && teardown !== this) {\n if (this.closed) {\n // If this subscription is already closed,\n // execute whatever finalizer is handed to it automatically.\n execFinalizer(teardown);\n } else {\n if (teardown instanceof Subscription) {\n // We don't add closed subscriptions, and we don't add the same subscription\n // twice. Subscription unsubscribe is idempotent.\n if (teardown.closed || teardown._hasParent(this)) {\n return;\n }\n teardown._addParent(this);\n }\n (this._finalizers = this._finalizers ?? []).push(teardown);\n }\n }\n }\n\n /**\n * Checks to see if a this subscription already has a particular parent.\n * This will signal that this subscription has already been added to the parent in question.\n * @param parent the parent to check for\n */\n private _hasParent(parent: Subscription) {\n const { _parentage } = this;\n return _parentage === parent || (Array.isArray(_parentage) && _parentage.includes(parent));\n }\n\n /**\n * Adds a parent to this subscription so it can be removed from the parent if it\n * unsubscribes on it's own.\n *\n * NOTE: THIS ASSUMES THAT {@link _hasParent} HAS ALREADY BEEN CHECKED.\n * @param parent The parent subscription to add\n */\n private _addParent(parent: Subscription) {\n const { _parentage } = this;\n this._parentage = Array.isArray(_parentage) ? (_parentage.push(parent), _parentage) : _parentage ? [_parentage, parent] : parent;\n }\n\n /**\n * Called on a child when it is removed via {@link #remove}.\n * @param parent The parent to remove\n */\n private _removeParent(parent: Subscription) {\n const { _parentage } = this;\n if (_parentage === parent) {\n this._parentage = null;\n } else if (Array.isArray(_parentage)) {\n arrRemove(_parentage, parent);\n }\n }\n\n /**\n * Removes a finalizer from this subscription that was previously added with the {@link #add} method.\n *\n * Note that `Subscription` instances, when unsubscribed, will automatically remove themselves\n * from every other `Subscription` they have been added to. This means that using the `remove` method\n * is not a common thing and should be used thoughtfully.\n *\n * If you add the same finalizer instance of a function or an unsubscribable object to a `Subscription` instance\n * more than once, you will need to call `remove` the same number of times to remove all instances.\n *\n * All finalizer instances are removed to free up memory upon unsubscription.\n *\n * @param teardown The finalizer to remove from this subscription\n */\n remove(teardown: Exclude): void {\n const { _finalizers } = this;\n _finalizers && arrRemove(_finalizers, teardown);\n\n if (teardown instanceof Subscription) {\n teardown._removeParent(this);\n }\n }\n}\n\nexport const EMPTY_SUBSCRIPTION = Subscription.EMPTY;\n\nexport function isSubscription(value: any): value is Subscription {\n return (\n value instanceof Subscription ||\n (value && 'closed' in value && isFunction(value.remove) && isFunction(value.add) && isFunction(value.unsubscribe))\n );\n}\n\nfunction execFinalizer(finalizer: Unsubscribable | (() => void)) {\n if (isFunction(finalizer)) {\n finalizer();\n } else {\n finalizer.unsubscribe();\n }\n}\n", "import { Subscriber } from './Subscriber';\nimport { ObservableNotification } from './types';\n\n/**\n * The {@link GlobalConfig} object for RxJS. It is used to configure things\n * like how to react on unhandled errors.\n */\nexport const config: GlobalConfig = {\n onUnhandledError: null,\n onStoppedNotification: null,\n Promise: undefined,\n useDeprecatedSynchronousErrorHandling: false,\n useDeprecatedNextContext: false,\n};\n\n/**\n * The global configuration object for RxJS, used to configure things\n * like how to react on unhandled errors. Accessible via {@link config}\n * object.\n */\nexport interface GlobalConfig {\n /**\n * A registration point for unhandled errors from RxJS. These are errors that\n * cannot were not handled by consuming code in the usual subscription path. For\n * example, if you have this configured, and you subscribe to an observable without\n * providing an error handler, errors from that subscription will end up here. This\n * will _always_ be called asynchronously on another job in the runtime. This is because\n * we do not want errors thrown in this user-configured handler to interfere with the\n * behavior of the library.\n */\n onUnhandledError: ((err: any) => void) | null;\n\n /**\n * A registration point for notifications that cannot be sent to subscribers because they\n * have completed, errored or have been explicitly unsubscribed. By default, next, complete\n * and error notifications sent to stopped subscribers are noops. However, sometimes callers\n * might want a different behavior. For example, with sources that attempt to report errors\n * to stopped subscribers, a caller can configure RxJS to throw an unhandled error instead.\n * This will _always_ be called asynchronously on another job in the runtime. This is because\n * we do not want errors thrown in this user-configured handler to interfere with the\n * behavior of the library.\n */\n onStoppedNotification: ((notification: ObservableNotification, subscriber: Subscriber) => void) | null;\n\n /**\n * The promise constructor used by default for {@link Observable#toPromise toPromise} and {@link Observable#forEach forEach}\n * methods.\n *\n * @deprecated As of version 8, RxJS will no longer support this sort of injection of a\n * Promise constructor. If you need a Promise implementation other than native promises,\n * please polyfill/patch Promise as you see appropriate. Will be removed in v8.\n */\n Promise?: PromiseConstructorLike;\n\n /**\n * If true, turns on synchronous error rethrowing, which is a deprecated behavior\n * in v6 and higher. This behavior enables bad patterns like wrapping a subscribe\n * call in a try/catch block. It also enables producer interference, a nasty bug\n * where a multicast can be broken for all observers by a downstream consumer with\n * an unhandled error. DO NOT USE THIS FLAG UNLESS IT'S NEEDED TO BUY TIME\n * FOR MIGRATION REASONS.\n *\n * @deprecated As of version 8, RxJS will no longer support synchronous throwing\n * of unhandled errors. All errors will be thrown on a separate call stack to prevent bad\n * behaviors described above. Will be removed in v8.\n */\n useDeprecatedSynchronousErrorHandling: boolean;\n\n /**\n * If true, enables an as-of-yet undocumented feature from v5: The ability to access\n * `unsubscribe()` via `this` context in `next` functions created in observers passed\n * to `subscribe`.\n *\n * This is being removed because the performance was severely problematic, and it could also cause\n * issues when types other than POJOs are passed to subscribe as subscribers, as they will likely have\n * their `this` context overwritten.\n *\n * @deprecated As of version 8, RxJS will no longer support altering the\n * context of next functions provided as part of an observer to Subscribe. Instead,\n * you will have access to a subscription or a signal or token that will allow you to do things like\n * unsubscribe and test closed status. Will be removed in v8.\n */\n useDeprecatedNextContext: boolean;\n}\n", "import type { TimerHandle } from './timerHandle';\ntype SetTimeoutFunction = (handler: () => void, timeout?: number, ...args: any[]) => TimerHandle;\ntype ClearTimeoutFunction = (handle: TimerHandle) => void;\n\ninterface TimeoutProvider {\n setTimeout: SetTimeoutFunction;\n clearTimeout: ClearTimeoutFunction;\n delegate:\n | {\n setTimeout: SetTimeoutFunction;\n clearTimeout: ClearTimeoutFunction;\n }\n | undefined;\n}\n\nexport const timeoutProvider: TimeoutProvider = {\n // When accessing the delegate, use the variable rather than `this` so that\n // the functions can be called without being bound to the provider.\n setTimeout(handler: () => void, timeout?: number, ...args) {\n const { delegate } = timeoutProvider;\n if (delegate?.setTimeout) {\n return delegate.setTimeout(handler, timeout, ...args);\n }\n return setTimeout(handler, timeout, ...args);\n },\n clearTimeout(handle) {\n const { delegate } = timeoutProvider;\n return (delegate?.clearTimeout || clearTimeout)(handle as any);\n },\n delegate: undefined,\n};\n", "import { config } from '../config';\nimport { timeoutProvider } from '../scheduler/timeoutProvider';\n\n/**\n * Handles an error on another job either with the user-configured {@link onUnhandledError},\n * or by throwing it on that new job so it can be picked up by `window.onerror`, `process.on('error')`, etc.\n *\n * This should be called whenever there is an error that is out-of-band with the subscription\n * or when an error hits a terminal boundary of the subscription and no error handler was provided.\n *\n * @param err the error to report\n */\nexport function reportUnhandledError(err: any) {\n timeoutProvider.setTimeout(() => {\n const { onUnhandledError } = config;\n if (onUnhandledError) {\n // Execute the user-configured error handler.\n onUnhandledError(err);\n } else {\n // Throw so it is picked up by the runtime's uncaught error mechanism.\n throw err;\n }\n });\n}\n", "/* tslint:disable:no-empty */\nexport function noop() { }\n", "import { CompleteNotification, NextNotification, ErrorNotification } from './types';\n\n/**\n * A completion object optimized for memory use and created to be the\n * same \"shape\" as other notifications in v8.\n * @internal\n */\nexport const COMPLETE_NOTIFICATION = (() => createNotification('C', undefined, undefined) as CompleteNotification)();\n\n/**\n * Internal use only. Creates an optimized error notification that is the same \"shape\"\n * as other notifications.\n * @internal\n */\nexport function errorNotification(error: any): ErrorNotification {\n return createNotification('E', undefined, error) as any;\n}\n\n/**\n * Internal use only. Creates an optimized next notification that is the same \"shape\"\n * as other notifications.\n * @internal\n */\nexport function nextNotification(value: T) {\n return createNotification('N', value, undefined) as NextNotification;\n}\n\n/**\n * Ensures that all notifications created internally have the same \"shape\" in v8.\n *\n * TODO: This is only exported to support a crazy legacy test in `groupBy`.\n * @internal\n */\nexport function createNotification(kind: 'N' | 'E' | 'C', value: any, error: any) {\n return {\n kind,\n value,\n error,\n };\n}\n", "import { config } from '../config';\n\nlet context: { errorThrown: boolean; error: any } | null = null;\n\n/**\n * Handles dealing with errors for super-gross mode. Creates a context, in which\n * any synchronously thrown errors will be passed to {@link captureError}. Which\n * will record the error such that it will be rethrown after the call back is complete.\n * TODO: Remove in v8\n * @param cb An immediately executed function.\n */\nexport function errorContext(cb: () => void) {\n if (config.useDeprecatedSynchronousErrorHandling) {\n const isRoot = !context;\n if (isRoot) {\n context = { errorThrown: false, error: null };\n }\n cb();\n if (isRoot) {\n const { errorThrown, error } = context!;\n context = null;\n if (errorThrown) {\n throw error;\n }\n }\n } else {\n // This is the general non-deprecated path for everyone that\n // isn't crazy enough to use super-gross mode (useDeprecatedSynchronousErrorHandling)\n cb();\n }\n}\n\n/**\n * Captures errors only in super-gross mode.\n * @param err the error to capture\n */\nexport function captureError(err: any) {\n if (config.useDeprecatedSynchronousErrorHandling && context) {\n context.errorThrown = true;\n context.error = err;\n }\n}\n", "import { isFunction } from './util/isFunction';\nimport { Observer, ObservableNotification } from './types';\nimport { isSubscription, Subscription } from './Subscription';\nimport { config } from './config';\nimport { reportUnhandledError } from './util/reportUnhandledError';\nimport { noop } from './util/noop';\nimport { nextNotification, errorNotification, COMPLETE_NOTIFICATION } from './NotificationFactories';\nimport { timeoutProvider } from './scheduler/timeoutProvider';\nimport { captureError } from './util/errorContext';\n\n/**\n * Implements the {@link Observer} interface and extends the\n * {@link Subscription} class. While the {@link Observer} is the public API for\n * consuming the values of an {@link Observable}, all Observers get converted to\n * a Subscriber, in order to provide Subscription-like capabilities such as\n * `unsubscribe`. Subscriber is a common type in RxJS, and crucial for\n * implementing operators, but it is rarely used as a public API.\n *\n * @class Subscriber\n */\nexport class Subscriber extends Subscription implements Observer {\n /**\n * A static factory for a Subscriber, given a (potentially partial) definition\n * of an Observer.\n * @param next The `next` callback of an Observer.\n * @param error The `error` callback of an\n * Observer.\n * @param complete The `complete` callback of an\n * Observer.\n * @return A Subscriber wrapping the (partially defined)\n * Observer represented by the given arguments.\n * @nocollapse\n * @deprecated Do not use. Will be removed in v8. There is no replacement for this\n * method, and there is no reason to be creating instances of `Subscriber` directly.\n * If you have a specific use case, please file an issue.\n */\n static create(next?: (x?: T) => void, error?: (e?: any) => void, complete?: () => void): Subscriber {\n return new SafeSubscriber(next, error, complete);\n }\n\n /** @deprecated Internal implementation detail, do not use directly. Will be made internal in v8. */\n protected isStopped: boolean = false;\n /** @deprecated Internal implementation detail, do not use directly. Will be made internal in v8. */\n protected destination: Subscriber | Observer; // this `any` is the escape hatch to erase extra type param (e.g. R)\n\n /**\n * @deprecated Internal implementation detail, do not use directly. Will be made internal in v8.\n * There is no reason to directly create an instance of Subscriber. This type is exported for typings reasons.\n */\n constructor(destination?: Subscriber | Observer) {\n super();\n if (destination) {\n this.destination = destination;\n // Automatically chain subscriptions together here.\n // if destination is a Subscription, then it is a Subscriber.\n if (isSubscription(destination)) {\n destination.add(this);\n }\n } else {\n this.destination = EMPTY_OBSERVER;\n }\n }\n\n /**\n * The {@link Observer} callback to receive notifications of type `next` from\n * the Observable, with a value. The Observable may call this method 0 or more\n * times.\n * @param {T} [value] The `next` value.\n * @return {void}\n */\n next(value?: T): void {\n if (this.isStopped) {\n handleStoppedNotification(nextNotification(value), this);\n } else {\n this._next(value!);\n }\n }\n\n /**\n * The {@link Observer} callback to receive notifications of type `error` from\n * the Observable, with an attached `Error`. Notifies the Observer that\n * the Observable has experienced an error condition.\n * @param {any} [err] The `error` exception.\n * @return {void}\n */\n error(err?: any): void {\n if (this.isStopped) {\n handleStoppedNotification(errorNotification(err), this);\n } else {\n this.isStopped = true;\n this._error(err);\n }\n }\n\n /**\n * The {@link Observer} callback to receive a valueless notification of type\n * `complete` from the Observable. Notifies the Observer that the Observable\n * has finished sending push-based notifications.\n * @return {void}\n */\n complete(): void {\n if (this.isStopped) {\n handleStoppedNotification(COMPLETE_NOTIFICATION, this);\n } else {\n this.isStopped = true;\n this._complete();\n }\n }\n\n unsubscribe(): void {\n if (!this.closed) {\n this.isStopped = true;\n super.unsubscribe();\n this.destination = null!;\n }\n }\n\n protected _next(value: T): void {\n this.destination.next(value);\n }\n\n protected _error(err: any): void {\n try {\n this.destination.error(err);\n } finally {\n this.unsubscribe();\n }\n }\n\n protected _complete(): void {\n try {\n this.destination.complete();\n } finally {\n this.unsubscribe();\n }\n }\n}\n\n/**\n * This bind is captured here because we want to be able to have\n * compatibility with monoid libraries that tend to use a method named\n * `bind`. In particular, a library called Monio requires this.\n */\nconst _bind = Function.prototype.bind;\n\nfunction bind any>(fn: Fn, thisArg: any): Fn {\n return _bind.call(fn, thisArg);\n}\n\n/**\n * Internal optimization only, DO NOT EXPOSE.\n * @internal\n */\nclass ConsumerObserver implements Observer {\n constructor(private partialObserver: Partial>) {}\n\n next(value: T): void {\n const { partialObserver } = this;\n if (partialObserver.next) {\n try {\n partialObserver.next(value);\n } catch (error) {\n handleUnhandledError(error);\n }\n }\n }\n\n error(err: any): void {\n const { partialObserver } = this;\n if (partialObserver.error) {\n try {\n partialObserver.error(err);\n } catch (error) {\n handleUnhandledError(error);\n }\n } else {\n handleUnhandledError(err);\n }\n }\n\n complete(): void {\n const { partialObserver } = this;\n if (partialObserver.complete) {\n try {\n partialObserver.complete();\n } catch (error) {\n handleUnhandledError(error);\n }\n }\n }\n}\n\nexport class SafeSubscriber extends Subscriber {\n constructor(\n observerOrNext?: Partial> | ((value: T) => void) | null,\n error?: ((e?: any) => void) | null,\n complete?: (() => void) | null\n ) {\n super();\n\n let partialObserver: Partial>;\n if (isFunction(observerOrNext) || !observerOrNext) {\n // The first argument is a function, not an observer. The next\n // two arguments *could* be observers, or they could be empty.\n partialObserver = {\n next: (observerOrNext ?? undefined) as (((value: T) => void) | undefined),\n error: error ?? undefined,\n complete: complete ?? undefined,\n };\n } else {\n // The first argument is a partial observer.\n let context: any;\n if (this && config.useDeprecatedNextContext) {\n // This is a deprecated path that made `this.unsubscribe()` available in\n // next handler functions passed to subscribe. This only exists behind a flag\n // now, as it is *very* slow.\n context = Object.create(observerOrNext);\n context.unsubscribe = () => this.unsubscribe();\n partialObserver = {\n next: observerOrNext.next && bind(observerOrNext.next, context),\n error: observerOrNext.error && bind(observerOrNext.error, context),\n complete: observerOrNext.complete && bind(observerOrNext.complete, context),\n };\n } else {\n // The \"normal\" path. Just use the partial observer directly.\n partialObserver = observerOrNext;\n }\n }\n\n // Wrap the partial observer to ensure it's a full observer, and\n // make sure proper error handling is accounted for.\n this.destination = new ConsumerObserver(partialObserver);\n }\n}\n\nfunction handleUnhandledError(error: any) {\n if (config.useDeprecatedSynchronousErrorHandling) {\n captureError(error);\n } else {\n // Ideal path, we report this as an unhandled error,\n // which is thrown on a new call stack.\n reportUnhandledError(error);\n }\n}\n\n/**\n * An error handler used when no error handler was supplied\n * to the SafeSubscriber -- meaning no error handler was supplied\n * do the `subscribe` call on our observable.\n * @param err The error to handle\n */\nfunction defaultErrorHandler(err: any) {\n throw err;\n}\n\n/**\n * A handler for notifications that cannot be sent to a stopped subscriber.\n * @param notification The notification being sent\n * @param subscriber The stopped subscriber\n */\nfunction handleStoppedNotification(notification: ObservableNotification, subscriber: Subscriber) {\n const { onStoppedNotification } = config;\n onStoppedNotification && timeoutProvider.setTimeout(() => onStoppedNotification(notification, subscriber));\n}\n\n/**\n * The observer used as a stub for subscriptions where the user did not\n * pass any arguments to `subscribe`. Comes with the default error handling\n * behavior.\n */\nexport const EMPTY_OBSERVER: Readonly> & { closed: true } = {\n closed: true,\n next: noop,\n error: defaultErrorHandler,\n complete: noop,\n};\n", "/**\n * Symbol.observable or a string \"@@observable\". Used for interop\n *\n * @deprecated We will no longer be exporting this symbol in upcoming versions of RxJS.\n * Instead polyfill and use Symbol.observable directly *or* use https://www.npmjs.com/package/symbol-observable\n */\nexport const observable: string | symbol = (() => (typeof Symbol === 'function' && Symbol.observable) || '@@observable')();\n", "/**\n * This function takes one parameter and just returns it. Simply put,\n * this is like `(x: T): T => x`.\n *\n * ## Examples\n *\n * This is useful in some cases when using things like `mergeMap`\n *\n * ```ts\n * import { interval, take, map, range, mergeMap, identity } from 'rxjs';\n *\n * const source$ = interval(1000).pipe(take(5));\n *\n * const result$ = source$.pipe(\n * map(i => range(i)),\n * mergeMap(identity) // same as mergeMap(x => x)\n * );\n *\n * result$.subscribe({\n * next: console.log\n * });\n * ```\n *\n * Or when you want to selectively apply an operator\n *\n * ```ts\n * import { interval, take, identity } from 'rxjs';\n *\n * const shouldLimit = () => Math.random() < 0.5;\n *\n * const source$ = interval(1000);\n *\n * const result$ = source$.pipe(shouldLimit() ? take(5) : identity);\n *\n * result$.subscribe({\n * next: console.log\n * });\n * ```\n *\n * @param x Any value that is returned by this function\n * @returns The value passed as the first parameter to this function\n */\nexport function identity(x: T): T {\n return x;\n}\n", "import { identity } from './identity';\nimport { UnaryFunction } from '../types';\n\nexport function pipe(): typeof identity;\nexport function pipe(fn1: UnaryFunction): UnaryFunction;\nexport function pipe(fn1: UnaryFunction, fn2: UnaryFunction): UnaryFunction;\nexport function pipe(fn1: UnaryFunction, fn2: UnaryFunction, fn3: UnaryFunction): UnaryFunction;\nexport function pipe(\n fn1: UnaryFunction,\n fn2: UnaryFunction,\n fn3: UnaryFunction,\n fn4: UnaryFunction\n): UnaryFunction;\nexport function pipe(\n fn1: UnaryFunction,\n fn2: UnaryFunction,\n fn3: UnaryFunction,\n fn4: UnaryFunction,\n fn5: UnaryFunction\n): UnaryFunction;\nexport function pipe(\n fn1: UnaryFunction,\n fn2: UnaryFunction,\n fn3: UnaryFunction,\n fn4: UnaryFunction,\n fn5: UnaryFunction,\n fn6: UnaryFunction\n): UnaryFunction;\nexport function pipe(\n fn1: UnaryFunction,\n fn2: UnaryFunction,\n fn3: UnaryFunction,\n fn4: UnaryFunction,\n fn5: UnaryFunction,\n fn6: UnaryFunction,\n fn7: UnaryFunction\n): UnaryFunction;\nexport function pipe(\n fn1: UnaryFunction,\n fn2: UnaryFunction,\n fn3: UnaryFunction,\n fn4: UnaryFunction,\n fn5: UnaryFunction,\n fn6: UnaryFunction,\n fn7: UnaryFunction,\n fn8: UnaryFunction\n): UnaryFunction;\nexport function pipe(\n fn1: UnaryFunction,\n fn2: UnaryFunction,\n fn3: UnaryFunction,\n fn4: UnaryFunction,\n fn5: UnaryFunction,\n fn6: UnaryFunction,\n fn7: UnaryFunction,\n fn8: UnaryFunction,\n fn9: UnaryFunction\n): UnaryFunction;\nexport function pipe(\n fn1: UnaryFunction,\n fn2: UnaryFunction,\n fn3: UnaryFunction,\n fn4: UnaryFunction,\n fn5: UnaryFunction,\n fn6: UnaryFunction,\n fn7: UnaryFunction,\n fn8: UnaryFunction,\n fn9: UnaryFunction,\n ...fns: UnaryFunction[]\n): UnaryFunction;\n\n/**\n * pipe() can be called on one or more functions, each of which can take one argument (\"UnaryFunction\")\n * and uses it to return a value.\n * It returns a function that takes one argument, passes it to the first UnaryFunction, and then\n * passes the result to the next one, passes that result to the next one, and so on. \n */\nexport function pipe(...fns: Array>): UnaryFunction {\n return pipeFromArray(fns);\n}\n\n/** @internal */\nexport function pipeFromArray(fns: Array>): UnaryFunction {\n if (fns.length === 0) {\n return identity as UnaryFunction;\n }\n\n if (fns.length === 1) {\n return fns[0];\n }\n\n return function piped(input: T): R {\n return fns.reduce((prev: any, fn: UnaryFunction) => fn(prev), input as any);\n };\n}\n", "import { Operator } from './Operator';\nimport { SafeSubscriber, Subscriber } from './Subscriber';\nimport { isSubscription, Subscription } from './Subscription';\nimport { TeardownLogic, OperatorFunction, Subscribable, Observer } from './types';\nimport { observable as Symbol_observable } from './symbol/observable';\nimport { pipeFromArray } from './util/pipe';\nimport { config } from './config';\nimport { isFunction } from './util/isFunction';\nimport { errorContext } from './util/errorContext';\n\n/**\n * A representation of any set of values over any amount of time. This is the most basic building block\n * of RxJS.\n *\n * @class Observable\n */\nexport class Observable implements Subscribable {\n /**\n * @deprecated Internal implementation detail, do not use directly. Will be made internal in v8.\n */\n source: Observable | undefined;\n\n /**\n * @deprecated Internal implementation detail, do not use directly. Will be made internal in v8.\n */\n operator: Operator | undefined;\n\n /**\n * @constructor\n * @param {Function} subscribe the function that is called when the Observable is\n * initially subscribed to. This function is given a Subscriber, to which new values\n * can be `next`ed, or an `error` method can be called to raise an error, or\n * `complete` can be called to notify of a successful completion.\n */\n constructor(subscribe?: (this: Observable, subscriber: Subscriber) => TeardownLogic) {\n if (subscribe) {\n this._subscribe = subscribe;\n }\n }\n\n // HACK: Since TypeScript inherits static properties too, we have to\n // fight against TypeScript here so Subject can have a different static create signature\n /**\n * Creates a new Observable by calling the Observable constructor\n * @owner Observable\n * @method create\n * @param {Function} subscribe? the subscriber function to be passed to the Observable constructor\n * @return {Observable} a new observable\n * @nocollapse\n * @deprecated Use `new Observable()` instead. Will be removed in v8.\n */\n static create: (...args: any[]) => any = (subscribe?: (subscriber: Subscriber) => TeardownLogic) => {\n return new Observable(subscribe);\n };\n\n /**\n * Creates a new Observable, with this Observable instance as the source, and the passed\n * operator defined as the new observable's operator.\n * @method lift\n * @param operator the operator defining the operation to take on the observable\n * @return a new observable with the Operator applied\n * @deprecated Internal implementation detail, do not use directly. Will be made internal in v8.\n * If you have implemented an operator using `lift`, it is recommended that you create an\n * operator by simply returning `new Observable()` directly. See \"Creating new operators from\n * scratch\" section here: https://rxjs.dev/guide/operators\n */\n lift(operator?: Operator): Observable {\n const observable = new Observable();\n observable.source = this;\n observable.operator = operator;\n return observable;\n }\n\n subscribe(observerOrNext?: Partial> | ((value: T) => void)): Subscription;\n /** @deprecated Instead of passing separate callback arguments, use an observer argument. Signatures taking separate callback arguments will be removed in v8. Details: https://rxjs.dev/deprecations/subscribe-arguments */\n subscribe(next?: ((value: T) => void) | null, error?: ((error: any) => void) | null, complete?: (() => void) | null): Subscription;\n /**\n * Invokes an execution of an Observable and registers Observer handlers for notifications it will emit.\n *\n * Use it when you have all these Observables, but still nothing is happening.\n *\n * `subscribe` is not a regular operator, but a method that calls Observable's internal `subscribe` function. It\n * might be for example a function that you passed to Observable's constructor, but most of the time it is\n * a library implementation, which defines what will be emitted by an Observable, and when it be will emitted. This means\n * that calling `subscribe` is actually the moment when Observable starts its work, not when it is created, as it is often\n * the thought.\n *\n * Apart from starting the execution of an Observable, this method allows you to listen for values\n * that an Observable emits, as well as for when it completes or errors. You can achieve this in two\n * of the following ways.\n *\n * The first way is creating an object that implements {@link Observer} interface. It should have methods\n * defined by that interface, but note that it should be just a regular JavaScript object, which you can create\n * yourself in any way you want (ES6 class, classic function constructor, object literal etc.). In particular, do\n * not attempt to use any RxJS implementation details to create Observers - you don't need them. Remember also\n * that your object does not have to implement all methods. If you find yourself creating a method that doesn't\n * do anything, you can simply omit it. Note however, if the `error` method is not provided and an error happens,\n * it will be thrown asynchronously. Errors thrown asynchronously cannot be caught using `try`/`catch`. Instead,\n * use the {@link onUnhandledError} configuration option or use a runtime handler (like `window.onerror` or\n * `process.on('error)`) to be notified of unhandled errors. Because of this, it's recommended that you provide\n * an `error` method to avoid missing thrown errors.\n *\n * The second way is to give up on Observer object altogether and simply provide callback functions in place of its methods.\n * This means you can provide three functions as arguments to `subscribe`, where the first function is equivalent\n * of a `next` method, the second of an `error` method and the third of a `complete` method. Just as in case of an Observer,\n * if you do not need to listen for something, you can omit a function by passing `undefined` or `null`,\n * since `subscribe` recognizes these functions by where they were placed in function call. When it comes\n * to the `error` function, as with an Observer, if not provided, errors emitted by an Observable will be thrown asynchronously.\n *\n * You can, however, subscribe with no parameters at all. This may be the case where you're not interested in terminal events\n * and you also handled emissions internally by using operators (e.g. using `tap`).\n *\n * Whichever style of calling `subscribe` you use, in both cases it returns a Subscription object.\n * This object allows you to call `unsubscribe` on it, which in turn will stop the work that an Observable does and will clean\n * up all resources that an Observable used. Note that cancelling a subscription will not call `complete` callback\n * provided to `subscribe` function, which is reserved for a regular completion signal that comes from an Observable.\n *\n * Remember that callbacks provided to `subscribe` are not guaranteed to be called asynchronously.\n * It is an Observable itself that decides when these functions will be called. For example {@link of}\n * by default emits all its values synchronously. Always check documentation for how given Observable\n * will behave when subscribed and if its default behavior can be modified with a `scheduler`.\n *\n * #### Examples\n *\n * Subscribe with an {@link guide/observer Observer}\n *\n * ```ts\n * import { of } from 'rxjs';\n *\n * const sumObserver = {\n * sum: 0,\n * next(value) {\n * console.log('Adding: ' + value);\n * this.sum = this.sum + value;\n * },\n * error() {\n * // We actually could just remove this method,\n * // since we do not really care about errors right now.\n * },\n * complete() {\n * console.log('Sum equals: ' + this.sum);\n * }\n * };\n *\n * of(1, 2, 3) // Synchronously emits 1, 2, 3 and then completes.\n * .subscribe(sumObserver);\n *\n * // Logs:\n * // 'Adding: 1'\n * // 'Adding: 2'\n * // 'Adding: 3'\n * // 'Sum equals: 6'\n * ```\n *\n * Subscribe with functions ({@link deprecations/subscribe-arguments deprecated})\n *\n * ```ts\n * import { of } from 'rxjs'\n *\n * let sum = 0;\n *\n * of(1, 2, 3).subscribe(\n * value => {\n * console.log('Adding: ' + value);\n * sum = sum + value;\n * },\n * undefined,\n * () => console.log('Sum equals: ' + sum)\n * );\n *\n * // Logs:\n * // 'Adding: 1'\n * // 'Adding: 2'\n * // 'Adding: 3'\n * // 'Sum equals: 6'\n * ```\n *\n * Cancel a subscription\n *\n * ```ts\n * import { interval } from 'rxjs';\n *\n * const subscription = interval(1000).subscribe({\n * next(num) {\n * console.log(num)\n * },\n * complete() {\n * // Will not be called, even when cancelling subscription.\n * console.log('completed!');\n * }\n * });\n *\n * setTimeout(() => {\n * subscription.unsubscribe();\n * console.log('unsubscribed!');\n * }, 2500);\n *\n * // Logs:\n * // 0 after 1s\n * // 1 after 2s\n * // 'unsubscribed!' after 2.5s\n * ```\n *\n * @param {Observer|Function} observerOrNext (optional) Either an observer with methods to be called,\n * or the first of three possible handlers, which is the handler for each value emitted from the subscribed\n * Observable.\n * @param {Function} error (optional) A handler for a terminal event resulting from an error. If no error handler is provided,\n * the error will be thrown asynchronously as unhandled.\n * @param {Function} complete (optional) A handler for a terminal event resulting from successful completion.\n * @return {Subscription} a subscription reference to the registered handlers\n * @method subscribe\n */\n subscribe(\n observerOrNext?: Partial> | ((value: T) => void) | null,\n error?: ((error: any) => void) | null,\n complete?: (() => void) | null\n ): Subscription {\n const subscriber = isSubscriber(observerOrNext) ? observerOrNext : new SafeSubscriber(observerOrNext, error, complete);\n\n errorContext(() => {\n const { operator, source } = this;\n subscriber.add(\n operator\n ? // We're dealing with a subscription in the\n // operator chain to one of our lifted operators.\n operator.call(subscriber, source)\n : source\n ? // If `source` has a value, but `operator` does not, something that\n // had intimate knowledge of our API, like our `Subject`, must have\n // set it. We're going to just call `_subscribe` directly.\n this._subscribe(subscriber)\n : // In all other cases, we're likely wrapping a user-provided initializer\n // function, so we need to catch errors and handle them appropriately.\n this._trySubscribe(subscriber)\n );\n });\n\n return subscriber;\n }\n\n /** @internal */\n protected _trySubscribe(sink: Subscriber): TeardownLogic {\n try {\n return this._subscribe(sink);\n } catch (err) {\n // We don't need to return anything in this case,\n // because it's just going to try to `add()` to a subscription\n // above.\n sink.error(err);\n }\n }\n\n /**\n * Used as a NON-CANCELLABLE means of subscribing to an observable, for use with\n * APIs that expect promises, like `async/await`. You cannot unsubscribe from this.\n *\n * **WARNING**: Only use this with observables you *know* will complete. If the source\n * observable does not complete, you will end up with a promise that is hung up, and\n * potentially all of the state of an async function hanging out in memory. To avoid\n * this situation, look into adding something like {@link timeout}, {@link take},\n * {@link takeWhile}, or {@link takeUntil} amongst others.\n *\n * #### Example\n *\n * ```ts\n * import { interval, take } from 'rxjs';\n *\n * const source$ = interval(1000).pipe(take(4));\n *\n * async function getTotal() {\n * let total = 0;\n *\n * await source$.forEach(value => {\n * total += value;\n * console.log('observable -> ' + value);\n * });\n *\n * return total;\n * }\n *\n * getTotal().then(\n * total => console.log('Total: ' + total)\n * );\n *\n * // Expected:\n * // 'observable -> 0'\n * // 'observable -> 1'\n * // 'observable -> 2'\n * // 'observable -> 3'\n * // 'Total: 6'\n * ```\n *\n * @param next a handler for each value emitted by the observable\n * @return a promise that either resolves on observable completion or\n * rejects with the handled error\n */\n forEach(next: (value: T) => void): Promise;\n\n /**\n * @param next a handler for each value emitted by the observable\n * @param promiseCtor a constructor function used to instantiate the Promise\n * @return a promise that either resolves on observable completion or\n * rejects with the handled error\n * @deprecated Passing a Promise constructor will no longer be available\n * in upcoming versions of RxJS. This is because it adds weight to the library, for very\n * little benefit. If you need this functionality, it is recommended that you either\n * polyfill Promise, or you create an adapter to convert the returned native promise\n * to whatever promise implementation you wanted. Will be removed in v8.\n */\n forEach(next: (value: T) => void, promiseCtor: PromiseConstructorLike): Promise;\n\n forEach(next: (value: T) => void, promiseCtor?: PromiseConstructorLike): Promise {\n promiseCtor = getPromiseCtor(promiseCtor);\n\n return new promiseCtor((resolve, reject) => {\n const subscriber = new SafeSubscriber({\n next: (value) => {\n try {\n next(value);\n } catch (err) {\n reject(err);\n subscriber.unsubscribe();\n }\n },\n error: reject,\n complete: resolve,\n });\n this.subscribe(subscriber);\n }) as Promise;\n }\n\n /** @internal */\n protected _subscribe(subscriber: Subscriber): TeardownLogic {\n return this.source?.subscribe(subscriber);\n }\n\n /**\n * An interop point defined by the es7-observable spec https://github.com/zenparsing/es-observable\n * @method Symbol.observable\n * @return {Observable} this instance of the observable\n */\n [Symbol_observable]() {\n return this;\n }\n\n /* tslint:disable:max-line-length */\n pipe(): Observable;\n pipe(op1: OperatorFunction): Observable;\n pipe(op1: OperatorFunction, op2: OperatorFunction): Observable;\n pipe(op1: OperatorFunction, op2: OperatorFunction, op3: OperatorFunction): Observable;\n pipe(\n op1: OperatorFunction,\n op2: OperatorFunction,\n op3: OperatorFunction,\n op4: OperatorFunction\n ): Observable;\n pipe(\n op1: OperatorFunction,\n op2: OperatorFunction,\n op3: OperatorFunction,\n op4: OperatorFunction,\n op5: OperatorFunction\n ): Observable;\n pipe(\n op1: OperatorFunction,\n op2: OperatorFunction,\n op3: OperatorFunction,\n op4: OperatorFunction,\n op5: OperatorFunction,\n op6: OperatorFunction\n ): Observable;\n pipe(\n op1: OperatorFunction,\n op2: OperatorFunction,\n op3: OperatorFunction,\n op4: OperatorFunction,\n op5: OperatorFunction,\n op6: OperatorFunction,\n op7: OperatorFunction\n ): Observable;\n pipe(\n op1: OperatorFunction,\n op2: OperatorFunction,\n op3: OperatorFunction,\n op4: OperatorFunction,\n op5: OperatorFunction,\n op6: OperatorFunction,\n op7: OperatorFunction,\n op8: OperatorFunction\n ): Observable;\n pipe(\n op1: OperatorFunction,\n op2: OperatorFunction,\n op3: OperatorFunction,\n op4: OperatorFunction,\n op5: OperatorFunction,\n op6: OperatorFunction,\n op7: OperatorFunction,\n op8: OperatorFunction,\n op9: OperatorFunction\n ): Observable;\n pipe(\n op1: OperatorFunction,\n op2: OperatorFunction,\n op3: OperatorFunction,\n op4: OperatorFunction,\n op5: OperatorFunction,\n op6: OperatorFunction,\n op7: OperatorFunction,\n op8: OperatorFunction,\n op9: OperatorFunction,\n ...operations: OperatorFunction[]\n ): Observable;\n /* tslint:enable:max-line-length */\n\n /**\n * Used to stitch together functional operators into a chain.\n * @method pipe\n * @return {Observable} the Observable result of all of the operators having\n * been called in the order they were passed in.\n *\n * ## Example\n *\n * ```ts\n * import { interval, filter, map, scan } from 'rxjs';\n *\n * interval(1000)\n * .pipe(\n * filter(x => x % 2 === 0),\n * map(x => x + x),\n * scan((acc, x) => acc + x)\n * )\n * .subscribe(x => console.log(x));\n * ```\n */\n pipe(...operations: OperatorFunction[]): Observable {\n return pipeFromArray(operations)(this);\n }\n\n /* tslint:disable:max-line-length */\n /** @deprecated Replaced with {@link firstValueFrom} and {@link lastValueFrom}. Will be removed in v8. Details: https://rxjs.dev/deprecations/to-promise */\n toPromise(): Promise;\n /** @deprecated Replaced with {@link firstValueFrom} and {@link lastValueFrom}. Will be removed in v8. Details: https://rxjs.dev/deprecations/to-promise */\n toPromise(PromiseCtor: typeof Promise): Promise;\n /** @deprecated Replaced with {@link firstValueFrom} and {@link lastValueFrom}. Will be removed in v8. Details: https://rxjs.dev/deprecations/to-promise */\n toPromise(PromiseCtor: PromiseConstructorLike): Promise;\n /* tslint:enable:max-line-length */\n\n /**\n * Subscribe to this Observable and get a Promise resolving on\n * `complete` with the last emission (if any).\n *\n * **WARNING**: Only use this with observables you *know* will complete. If the source\n * observable does not complete, you will end up with a promise that is hung up, and\n * potentially all of the state of an async function hanging out in memory. To avoid\n * this situation, look into adding something like {@link timeout}, {@link take},\n * {@link takeWhile}, or {@link takeUntil} amongst others.\n *\n * @method toPromise\n * @param [promiseCtor] a constructor function used to instantiate\n * the Promise\n * @return A Promise that resolves with the last value emit, or\n * rejects on an error. If there were no emissions, Promise\n * resolves with undefined.\n * @deprecated Replaced with {@link firstValueFrom} and {@link lastValueFrom}. Will be removed in v8. Details: https://rxjs.dev/deprecations/to-promise\n */\n toPromise(promiseCtor?: PromiseConstructorLike): Promise {\n promiseCtor = getPromiseCtor(promiseCtor);\n\n return new promiseCtor((resolve, reject) => {\n let value: T | undefined;\n this.subscribe(\n (x: T) => (value = x),\n (err: any) => reject(err),\n () => resolve(value)\n );\n }) as Promise;\n }\n}\n\n/**\n * Decides between a passed promise constructor from consuming code,\n * A default configured promise constructor, and the native promise\n * constructor and returns it. If nothing can be found, it will throw\n * an error.\n * @param promiseCtor The optional promise constructor to passed by consuming code\n */\nfunction getPromiseCtor(promiseCtor: PromiseConstructorLike | undefined) {\n return promiseCtor ?? config.Promise ?? Promise;\n}\n\nfunction isObserver(value: any): value is Observer {\n return value && isFunction(value.next) && isFunction(value.error) && isFunction(value.complete);\n}\n\nfunction isSubscriber(value: any): value is Subscriber {\n return (value && value instanceof Subscriber) || (isObserver(value) && isSubscription(value));\n}\n", "import { Observable } from '../Observable';\nimport { Subscriber } from '../Subscriber';\nimport { OperatorFunction } from '../types';\nimport { isFunction } from './isFunction';\n\n/**\n * Used to determine if an object is an Observable with a lift function.\n */\nexport function hasLift(source: any): source is { lift: InstanceType['lift'] } {\n return isFunction(source?.lift);\n}\n\n/**\n * Creates an `OperatorFunction`. Used to define operators throughout the library in a concise way.\n * @param init The logic to connect the liftedSource to the subscriber at the moment of subscription.\n */\nexport function operate(\n init: (liftedSource: Observable, subscriber: Subscriber) => (() => void) | void\n): OperatorFunction {\n return (source: Observable) => {\n if (hasLift(source)) {\n return source.lift(function (this: Subscriber, liftedSource: Observable) {\n try {\n return init(liftedSource, this);\n } catch (err) {\n this.error(err);\n }\n });\n }\n throw new TypeError('Unable to lift unknown Observable type');\n };\n}\n", "import { Subscriber } from '../Subscriber';\n\n/**\n * Creates an instance of an `OperatorSubscriber`.\n * @param destination The downstream subscriber.\n * @param onNext Handles next values, only called if this subscriber is not stopped or closed. Any\n * error that occurs in this function is caught and sent to the `error` method of this subscriber.\n * @param onError Handles errors from the subscription, any errors that occur in this handler are caught\n * and send to the `destination` error handler.\n * @param onComplete Handles completion notification from the subscription. Any errors that occur in\n * this handler are sent to the `destination` error handler.\n * @param onFinalize Additional teardown logic here. This will only be called on teardown if the\n * subscriber itself is not already closed. This is called after all other teardown logic is executed.\n */\nexport function createOperatorSubscriber(\n destination: Subscriber,\n onNext?: (value: T) => void,\n onComplete?: () => void,\n onError?: (err: any) => void,\n onFinalize?: () => void\n): Subscriber {\n return new OperatorSubscriber(destination, onNext, onComplete, onError, onFinalize);\n}\n\n/**\n * A generic helper for allowing operators to be created with a Subscriber and\n * use closures to capture necessary state from the operator function itself.\n */\nexport class OperatorSubscriber extends Subscriber {\n /**\n * Creates an instance of an `OperatorSubscriber`.\n * @param destination The downstream subscriber.\n * @param onNext Handles next values, only called if this subscriber is not stopped or closed. Any\n * error that occurs in this function is caught and sent to the `error` method of this subscriber.\n * @param onError Handles errors from the subscription, any errors that occur in this handler are caught\n * and send to the `destination` error handler.\n * @param onComplete Handles completion notification from the subscription. Any errors that occur in\n * this handler are sent to the `destination` error handler.\n * @param onFinalize Additional finalization logic here. This will only be called on finalization if the\n * subscriber itself is not already closed. This is called after all other finalization logic is executed.\n * @param shouldUnsubscribe An optional check to see if an unsubscribe call should truly unsubscribe.\n * NOTE: This currently **ONLY** exists to support the strange behavior of {@link groupBy}, where unsubscription\n * to the resulting observable does not actually disconnect from the source if there are active subscriptions\n * to any grouped observable. (DO NOT EXPOSE OR USE EXTERNALLY!!!)\n */\n constructor(\n destination: Subscriber,\n onNext?: (value: T) => void,\n onComplete?: () => void,\n onError?: (err: any) => void,\n private onFinalize?: () => void,\n private shouldUnsubscribe?: () => boolean\n ) {\n // It's important - for performance reasons - that all of this class's\n // members are initialized and that they are always initialized in the same\n // order. This will ensure that all OperatorSubscriber instances have the\n // same hidden class in V8. This, in turn, will help keep the number of\n // hidden classes involved in property accesses within the base class as\n // low as possible. If the number of hidden classes involved exceeds four,\n // the property accesses will become megamorphic and performance penalties\n // will be incurred - i.e. inline caches won't be used.\n //\n // The reasons for ensuring all instances have the same hidden class are\n // further discussed in this blog post from Benedikt Meurer:\n // https://benediktmeurer.de/2018/03/23/impact-of-polymorphism-on-component-based-frameworks-like-react/\n super(destination);\n this._next = onNext\n ? function (this: OperatorSubscriber, value: T) {\n try {\n onNext(value);\n } catch (err) {\n destination.error(err);\n }\n }\n : super._next;\n this._error = onError\n ? function (this: OperatorSubscriber, err: any) {\n try {\n onError(err);\n } catch (err) {\n // Send any errors that occur down stream.\n destination.error(err);\n } finally {\n // Ensure finalization.\n this.unsubscribe();\n }\n }\n : super._error;\n this._complete = onComplete\n ? function (this: OperatorSubscriber) {\n try {\n onComplete();\n } catch (err) {\n // Send any errors that occur down stream.\n destination.error(err);\n } finally {\n // Ensure finalization.\n this.unsubscribe();\n }\n }\n : super._complete;\n }\n\n unsubscribe() {\n if (!this.shouldUnsubscribe || this.shouldUnsubscribe()) {\n const { closed } = this;\n super.unsubscribe();\n // Execute additional teardown if we have any and we didn't already do so.\n !closed && this.onFinalize?.();\n }\n }\n}\n", "import { Subscription } from '../Subscription';\n\ninterface AnimationFrameProvider {\n schedule(callback: FrameRequestCallback): Subscription;\n requestAnimationFrame: typeof requestAnimationFrame;\n cancelAnimationFrame: typeof cancelAnimationFrame;\n delegate:\n | {\n requestAnimationFrame: typeof requestAnimationFrame;\n cancelAnimationFrame: typeof cancelAnimationFrame;\n }\n | undefined;\n}\n\nexport const animationFrameProvider: AnimationFrameProvider = {\n // When accessing the delegate, use the variable rather than `this` so that\n // the functions can be called without being bound to the provider.\n schedule(callback) {\n let request = requestAnimationFrame;\n let cancel: typeof cancelAnimationFrame | undefined = cancelAnimationFrame;\n const { delegate } = animationFrameProvider;\n if (delegate) {\n request = delegate.requestAnimationFrame;\n cancel = delegate.cancelAnimationFrame;\n }\n const handle = request((timestamp) => {\n // Clear the cancel function. The request has been fulfilled, so\n // attempting to cancel the request upon unsubscription would be\n // pointless.\n cancel = undefined;\n callback(timestamp);\n });\n return new Subscription(() => cancel?.(handle));\n },\n requestAnimationFrame(...args) {\n const { delegate } = animationFrameProvider;\n return (delegate?.requestAnimationFrame || requestAnimationFrame)(...args);\n },\n cancelAnimationFrame(...args) {\n const { delegate } = animationFrameProvider;\n return (delegate?.cancelAnimationFrame || cancelAnimationFrame)(...args);\n },\n delegate: undefined,\n};\n", "import { createErrorClass } from './createErrorClass';\n\nexport interface ObjectUnsubscribedError extends Error {}\n\nexport interface ObjectUnsubscribedErrorCtor {\n /**\n * @deprecated Internal implementation detail. Do not construct error instances.\n * Cannot be tagged as internal: https://github.com/ReactiveX/rxjs/issues/6269\n */\n new (): ObjectUnsubscribedError;\n}\n\n/**\n * An error thrown when an action is invalid because the object has been\n * unsubscribed.\n *\n * @see {@link Subject}\n * @see {@link BehaviorSubject}\n *\n * @class ObjectUnsubscribedError\n */\nexport const ObjectUnsubscribedError: ObjectUnsubscribedErrorCtor = createErrorClass(\n (_super) =>\n function ObjectUnsubscribedErrorImpl(this: any) {\n _super(this);\n this.name = 'ObjectUnsubscribedError';\n this.message = 'object unsubscribed';\n }\n);\n", "import { Operator } from './Operator';\nimport { Observable } from './Observable';\nimport { Subscriber } from './Subscriber';\nimport { Subscription, EMPTY_SUBSCRIPTION } from './Subscription';\nimport { Observer, SubscriptionLike, TeardownLogic } from './types';\nimport { ObjectUnsubscribedError } from './util/ObjectUnsubscribedError';\nimport { arrRemove } from './util/arrRemove';\nimport { errorContext } from './util/errorContext';\n\n/**\n * A Subject is a special type of Observable that allows values to be\n * multicasted to many Observers. Subjects are like EventEmitters.\n *\n * Every Subject is an Observable and an Observer. You can subscribe to a\n * Subject, and you can call next to feed values as well as error and complete.\n */\nexport class Subject extends Observable implements SubscriptionLike {\n closed = false;\n\n private currentObservers: Observer[] | null = null;\n\n /** @deprecated Internal implementation detail, do not use directly. Will be made internal in v8. */\n observers: Observer[] = [];\n /** @deprecated Internal implementation detail, do not use directly. Will be made internal in v8. */\n isStopped = false;\n /** @deprecated Internal implementation detail, do not use directly. Will be made internal in v8. */\n hasError = false;\n /** @deprecated Internal implementation detail, do not use directly. Will be made internal in v8. */\n thrownError: any = null;\n\n /**\n * Creates a \"subject\" by basically gluing an observer to an observable.\n *\n * @nocollapse\n * @deprecated Recommended you do not use. Will be removed at some point in the future. Plans for replacement still under discussion.\n */\n static create: (...args: any[]) => any = (destination: Observer, source: Observable): AnonymousSubject => {\n return new AnonymousSubject(destination, source);\n };\n\n constructor() {\n // NOTE: This must be here to obscure Observable's constructor.\n super();\n }\n\n /** @deprecated Internal implementation detail, do not use directly. Will be made internal in v8. */\n lift(operator: Operator): Observable {\n const subject = new AnonymousSubject(this, this);\n subject.operator = operator as any;\n return subject as any;\n }\n\n /** @internal */\n protected _throwIfClosed() {\n if (this.closed) {\n throw new ObjectUnsubscribedError();\n }\n }\n\n next(value: T) {\n errorContext(() => {\n this._throwIfClosed();\n if (!this.isStopped) {\n if (!this.currentObservers) {\n this.currentObservers = Array.from(this.observers);\n }\n for (const observer of this.currentObservers) {\n observer.next(value);\n }\n }\n });\n }\n\n error(err: any) {\n errorContext(() => {\n this._throwIfClosed();\n if (!this.isStopped) {\n this.hasError = this.isStopped = true;\n this.thrownError = err;\n const { observers } = this;\n while (observers.length) {\n observers.shift()!.error(err);\n }\n }\n });\n }\n\n complete() {\n errorContext(() => {\n this._throwIfClosed();\n if (!this.isStopped) {\n this.isStopped = true;\n const { observers } = this;\n while (observers.length) {\n observers.shift()!.complete();\n }\n }\n });\n }\n\n unsubscribe() {\n this.isStopped = this.closed = true;\n this.observers = this.currentObservers = null!;\n }\n\n get observed() {\n return this.observers?.length > 0;\n }\n\n /** @internal */\n protected _trySubscribe(subscriber: Subscriber): TeardownLogic {\n this._throwIfClosed();\n return super._trySubscribe(subscriber);\n }\n\n /** @internal */\n protected _subscribe(subscriber: Subscriber): Subscription {\n this._throwIfClosed();\n this._checkFinalizedStatuses(subscriber);\n return this._innerSubscribe(subscriber);\n }\n\n /** @internal */\n protected _innerSubscribe(subscriber: Subscriber) {\n const { hasError, isStopped, observers } = this;\n if (hasError || isStopped) {\n return EMPTY_SUBSCRIPTION;\n }\n this.currentObservers = null;\n observers.push(subscriber);\n return new Subscription(() => {\n this.currentObservers = null;\n arrRemove(observers, subscriber);\n });\n }\n\n /** @internal */\n protected _checkFinalizedStatuses(subscriber: Subscriber) {\n const { hasError, thrownError, isStopped } = this;\n if (hasError) {\n subscriber.error(thrownError);\n } else if (isStopped) {\n subscriber.complete();\n }\n }\n\n /**\n * Creates a new Observable with this Subject as the source. You can do this\n * to create custom Observer-side logic of the Subject and conceal it from\n * code that uses the Observable.\n * @return {Observable} Observable that the Subject casts to\n */\n asObservable(): Observable {\n const observable: any = new Observable();\n observable.source = this;\n return observable;\n }\n}\n\n/**\n * @class AnonymousSubject\n */\nexport class AnonymousSubject extends Subject {\n constructor(\n /** @deprecated Internal implementation detail, do not use directly. Will be made internal in v8. */\n public destination?: Observer,\n source?: Observable\n ) {\n super();\n this.source = source;\n }\n\n next(value: T) {\n this.destination?.next?.(value);\n }\n\n error(err: any) {\n this.destination?.error?.(err);\n }\n\n complete() {\n this.destination?.complete?.();\n }\n\n /** @internal */\n protected _subscribe(subscriber: Subscriber): Subscription {\n return this.source?.subscribe(subscriber) ?? EMPTY_SUBSCRIPTION;\n }\n}\n", "import { Subject } from './Subject';\nimport { Subscriber } from './Subscriber';\nimport { Subscription } from './Subscription';\n\n/**\n * A variant of Subject that requires an initial value and emits its current\n * value whenever it is subscribed to.\n *\n * @class BehaviorSubject\n */\nexport class BehaviorSubject extends Subject {\n constructor(private _value: T) {\n super();\n }\n\n get value(): T {\n return this.getValue();\n }\n\n /** @internal */\n protected _subscribe(subscriber: Subscriber): Subscription {\n const subscription = super._subscribe(subscriber);\n !subscription.closed && subscriber.next(this._value);\n return subscription;\n }\n\n getValue(): T {\n const { hasError, thrownError, _value } = this;\n if (hasError) {\n throw thrownError;\n }\n this._throwIfClosed();\n return _value;\n }\n\n next(value: T): void {\n super.next((this._value = value));\n }\n}\n", "import { TimestampProvider } from '../types';\n\ninterface DateTimestampProvider extends TimestampProvider {\n delegate: TimestampProvider | undefined;\n}\n\nexport const dateTimestampProvider: DateTimestampProvider = {\n now() {\n // Use the variable rather than `this` so that the function can be called\n // without being bound to the provider.\n return (dateTimestampProvider.delegate || Date).now();\n },\n delegate: undefined,\n};\n", "import { Subject } from './Subject';\nimport { TimestampProvider } from './types';\nimport { Subscriber } from './Subscriber';\nimport { Subscription } from './Subscription';\nimport { dateTimestampProvider } from './scheduler/dateTimestampProvider';\n\n/**\n * A variant of {@link Subject} that \"replays\" old values to new subscribers by emitting them when they first subscribe.\n *\n * `ReplaySubject` has an internal buffer that will store a specified number of values that it has observed. Like `Subject`,\n * `ReplaySubject` \"observes\" values by having them passed to its `next` method. When it observes a value, it will store that\n * value for a time determined by the configuration of the `ReplaySubject`, as passed to its constructor.\n *\n * When a new subscriber subscribes to the `ReplaySubject` instance, it will synchronously emit all values in its buffer in\n * a First-In-First-Out (FIFO) manner. The `ReplaySubject` will also complete, if it has observed completion; and it will\n * error if it has observed an error.\n *\n * There are two main configuration items to be concerned with:\n *\n * 1. `bufferSize` - This will determine how many items are stored in the buffer, defaults to infinite.\n * 2. `windowTime` - The amount of time to hold a value in the buffer before removing it from the buffer.\n *\n * Both configurations may exist simultaneously. So if you would like to buffer a maximum of 3 values, as long as the values\n * are less than 2 seconds old, you could do so with a `new ReplaySubject(3, 2000)`.\n *\n * ### Differences with BehaviorSubject\n *\n * `BehaviorSubject` is similar to `new ReplaySubject(1)`, with a couple of exceptions:\n *\n * 1. `BehaviorSubject` comes \"primed\" with a single value upon construction.\n * 2. `ReplaySubject` will replay values, even after observing an error, where `BehaviorSubject` will not.\n *\n * @see {@link Subject}\n * @see {@link BehaviorSubject}\n * @see {@link shareReplay}\n */\nexport class ReplaySubject extends Subject {\n private _buffer: (T | number)[] = [];\n private _infiniteTimeWindow = true;\n\n /**\n * @param bufferSize The size of the buffer to replay on subscription\n * @param windowTime The amount of time the buffered items will stay buffered\n * @param timestampProvider An object with a `now()` method that provides the current timestamp. This is used to\n * calculate the amount of time something has been buffered.\n */\n constructor(\n private _bufferSize = Infinity,\n private _windowTime = Infinity,\n private _timestampProvider: TimestampProvider = dateTimestampProvider\n ) {\n super();\n this._infiniteTimeWindow = _windowTime === Infinity;\n this._bufferSize = Math.max(1, _bufferSize);\n this._windowTime = Math.max(1, _windowTime);\n }\n\n next(value: T): void {\n const { isStopped, _buffer, _infiniteTimeWindow, _timestampProvider, _windowTime } = this;\n if (!isStopped) {\n _buffer.push(value);\n !_infiniteTimeWindow && _buffer.push(_timestampProvider.now() + _windowTime);\n }\n this._trimBuffer();\n super.next(value);\n }\n\n /** @internal */\n protected _subscribe(subscriber: Subscriber): Subscription {\n this._throwIfClosed();\n this._trimBuffer();\n\n const subscription = this._innerSubscribe(subscriber);\n\n const { _infiniteTimeWindow, _buffer } = this;\n // We use a copy here, so reentrant code does not mutate our array while we're\n // emitting it to a new subscriber.\n const copy = _buffer.slice();\n for (let i = 0; i < copy.length && !subscriber.closed; i += _infiniteTimeWindow ? 1 : 2) {\n subscriber.next(copy[i] as T);\n }\n\n this._checkFinalizedStatuses(subscriber);\n\n return subscription;\n }\n\n private _trimBuffer() {\n const { _bufferSize, _timestampProvider, _buffer, _infiniteTimeWindow } = this;\n // If we don't have an infinite buffer size, and we're over the length,\n // use splice to truncate the old buffer values off. Note that we have to\n // double the size for instances where we're not using an infinite time window\n // because we're storing the values and the timestamps in the same array.\n const adjustedBufferSize = (_infiniteTimeWindow ? 1 : 2) * _bufferSize;\n _bufferSize < Infinity && adjustedBufferSize < _buffer.length && _buffer.splice(0, _buffer.length - adjustedBufferSize);\n\n // Now, if we're not in an infinite time window, remove all values where the time is\n // older than what is allowed.\n if (!_infiniteTimeWindow) {\n const now = _timestampProvider.now();\n let last = 0;\n // Search the array for the first timestamp that isn't expired and\n // truncate the buffer up to that point.\n for (let i = 1; i < _buffer.length && (_buffer[i] as number) <= now; i += 2) {\n last = i;\n }\n last && _buffer.splice(0, last + 1);\n }\n }\n}\n", "import { Scheduler } from '../Scheduler';\nimport { Subscription } from '../Subscription';\nimport { SchedulerAction } from '../types';\n\n/**\n * A unit of work to be executed in a `scheduler`. An action is typically\n * created from within a {@link SchedulerLike} and an RxJS user does not need to concern\n * themselves about creating and manipulating an Action.\n *\n * ```ts\n * class Action extends Subscription {\n * new (scheduler: Scheduler, work: (state?: T) => void);\n * schedule(state?: T, delay: number = 0): Subscription;\n * }\n * ```\n *\n * @class Action\n */\nexport class Action extends Subscription {\n constructor(scheduler: Scheduler, work: (this: SchedulerAction, state?: T) => void) {\n super();\n }\n /**\n * Schedules this action on its parent {@link SchedulerLike} for execution. May be passed\n * some context object, `state`. May happen at some point in the future,\n * according to the `delay` parameter, if specified.\n * @param {T} [state] Some contextual data that the `work` function uses when\n * called by the Scheduler.\n * @param {number} [delay] Time to wait before executing the work, where the\n * time unit is implicit and defined by the Scheduler.\n * @return {void}\n */\n public schedule(state?: T, delay: number = 0): Subscription {\n return this;\n }\n}\n", "import type { TimerHandle } from './timerHandle';\ntype SetIntervalFunction = (handler: () => void, timeout?: number, ...args: any[]) => TimerHandle;\ntype ClearIntervalFunction = (handle: TimerHandle) => void;\n\ninterface IntervalProvider {\n setInterval: SetIntervalFunction;\n clearInterval: ClearIntervalFunction;\n delegate:\n | {\n setInterval: SetIntervalFunction;\n clearInterval: ClearIntervalFunction;\n }\n | undefined;\n}\n\nexport const intervalProvider: IntervalProvider = {\n // When accessing the delegate, use the variable rather than `this` so that\n // the functions can be called without being bound to the provider.\n setInterval(handler: () => void, timeout?: number, ...args) {\n const { delegate } = intervalProvider;\n if (delegate?.setInterval) {\n return delegate.setInterval(handler, timeout, ...args);\n }\n return setInterval(handler, timeout, ...args);\n },\n clearInterval(handle) {\n const { delegate } = intervalProvider;\n return (delegate?.clearInterval || clearInterval)(handle as any);\n },\n delegate: undefined,\n};\n", "import { Action } from './Action';\nimport { SchedulerAction } from '../types';\nimport { Subscription } from '../Subscription';\nimport { AsyncScheduler } from './AsyncScheduler';\nimport { intervalProvider } from './intervalProvider';\nimport { arrRemove } from '../util/arrRemove';\nimport { TimerHandle } from './timerHandle';\n\nexport class AsyncAction extends Action {\n public id: TimerHandle | undefined;\n public state?: T;\n // @ts-ignore: Property has no initializer and is not definitely assigned\n public delay: number;\n protected pending: boolean = false;\n\n constructor(protected scheduler: AsyncScheduler, protected work: (this: SchedulerAction, state?: T) => void) {\n super(scheduler, work);\n }\n\n public schedule(state?: T, delay: number = 0): Subscription {\n if (this.closed) {\n return this;\n }\n\n // Always replace the current state with the new state.\n this.state = state;\n\n const id = this.id;\n const scheduler = this.scheduler;\n\n //\n // Important implementation note:\n //\n // Actions only execute once by default, unless rescheduled from within the\n // scheduled callback. This allows us to implement single and repeat\n // actions via the same code path, without adding API surface area, as well\n // as mimic traditional recursion but across asynchronous boundaries.\n //\n // However, JS runtimes and timers distinguish between intervals achieved by\n // serial `setTimeout` calls vs. a single `setInterval` call. An interval of\n // serial `setTimeout` calls can be individually delayed, which delays\n // scheduling the next `setTimeout`, and so on. `setInterval` attempts to\n // guarantee the interval callback will be invoked more precisely to the\n // interval period, regardless of load.\n //\n // Therefore, we use `setInterval` to schedule single and repeat actions.\n // If the action reschedules itself with the same delay, the interval is not\n // canceled. If the action doesn't reschedule, or reschedules with a\n // different delay, the interval will be canceled after scheduled callback\n // execution.\n //\n if (id != null) {\n this.id = this.recycleAsyncId(scheduler, id, delay);\n }\n\n // Set the pending flag indicating that this action has been scheduled, or\n // has recursively rescheduled itself.\n this.pending = true;\n\n this.delay = delay;\n // If this action has already an async Id, don't request a new one.\n this.id = this.id ?? this.requestAsyncId(scheduler, this.id, delay);\n\n return this;\n }\n\n protected requestAsyncId(scheduler: AsyncScheduler, _id?: TimerHandle, delay: number = 0): TimerHandle {\n return intervalProvider.setInterval(scheduler.flush.bind(scheduler, this), delay);\n }\n\n protected recycleAsyncId(_scheduler: AsyncScheduler, id?: TimerHandle, delay: number | null = 0): TimerHandle | undefined {\n // If this action is rescheduled with the same delay time, don't clear the interval id.\n if (delay != null && this.delay === delay && this.pending === false) {\n return id;\n }\n // Otherwise, if the action's delay time is different from the current delay,\n // or the action has been rescheduled before it's executed, clear the interval id\n if (id != null) {\n intervalProvider.clearInterval(id);\n }\n\n return undefined;\n }\n\n /**\n * Immediately executes this action and the `work` it contains.\n * @return {any}\n */\n public execute(state: T, delay: number): any {\n if (this.closed) {\n return new Error('executing a cancelled action');\n }\n\n this.pending = false;\n const error = this._execute(state, delay);\n if (error) {\n return error;\n } else if (this.pending === false && this.id != null) {\n // Dequeue if the action didn't reschedule itself. Don't call\n // unsubscribe(), because the action could reschedule later.\n // For example:\n // ```\n // scheduler.schedule(function doWork(counter) {\n // /* ... I'm a busy worker bee ... */\n // var originalAction = this;\n // /* wait 100ms before rescheduling the action */\n // setTimeout(function () {\n // originalAction.schedule(counter + 1);\n // }, 100);\n // }, 1000);\n // ```\n this.id = this.recycleAsyncId(this.scheduler, this.id, null);\n }\n }\n\n protected _execute(state: T, _delay: number): any {\n let errored: boolean = false;\n let errorValue: any;\n try {\n this.work(state);\n } catch (e) {\n errored = true;\n // HACK: Since code elsewhere is relying on the \"truthiness\" of the\n // return here, we can't have it return \"\" or 0 or false.\n // TODO: Clean this up when we refactor schedulers mid-version-8 or so.\n errorValue = e ? e : new Error('Scheduled action threw falsy error');\n }\n if (errored) {\n this.unsubscribe();\n return errorValue;\n }\n }\n\n unsubscribe() {\n if (!this.closed) {\n const { id, scheduler } = this;\n const { actions } = scheduler;\n\n this.work = this.state = this.scheduler = null!;\n this.pending = false;\n\n arrRemove(actions, this);\n if (id != null) {\n this.id = this.recycleAsyncId(scheduler, id, null);\n }\n\n this.delay = null!;\n super.unsubscribe();\n }\n }\n}\n", "import { Action } from './scheduler/Action';\nimport { Subscription } from './Subscription';\nimport { SchedulerLike, SchedulerAction } from './types';\nimport { dateTimestampProvider } from './scheduler/dateTimestampProvider';\n\n/**\n * An execution context and a data structure to order tasks and schedule their\n * execution. Provides a notion of (potentially virtual) time, through the\n * `now()` getter method.\n *\n * Each unit of work in a Scheduler is called an `Action`.\n *\n * ```ts\n * class Scheduler {\n * now(): number;\n * schedule(work, delay?, state?): Subscription;\n * }\n * ```\n *\n * @class Scheduler\n * @deprecated Scheduler is an internal implementation detail of RxJS, and\n * should not be used directly. Rather, create your own class and implement\n * {@link SchedulerLike}. Will be made internal in v8.\n */\nexport class Scheduler implements SchedulerLike {\n public static now: () => number = dateTimestampProvider.now;\n\n constructor(private schedulerActionCtor: typeof Action, now: () => number = Scheduler.now) {\n this.now = now;\n }\n\n /**\n * A getter method that returns a number representing the current time\n * (at the time this function was called) according to the scheduler's own\n * internal clock.\n * @return {number} A number that represents the current time. May or may not\n * have a relation to wall-clock time. May or may not refer to a time unit\n * (e.g. milliseconds).\n */\n public now: () => number;\n\n /**\n * Schedules a function, `work`, for execution. May happen at some point in\n * the future, according to the `delay` parameter, if specified. May be passed\n * some context object, `state`, which will be passed to the `work` function.\n *\n * The given arguments will be processed an stored as an Action object in a\n * queue of actions.\n *\n * @param {function(state: ?T): ?Subscription} work A function representing a\n * task, or some unit of work to be executed by the Scheduler.\n * @param {number} [delay] Time to wait before executing the work, where the\n * time unit is implicit and defined by the Scheduler itself.\n * @param {T} [state] Some contextual data that the `work` function uses when\n * called by the Scheduler.\n * @return {Subscription} A subscription in order to be able to unsubscribe\n * the scheduled work.\n */\n public schedule(work: (this: SchedulerAction, state?: T) => void, delay: number = 0, state?: T): Subscription {\n return new this.schedulerActionCtor(this, work).schedule(state, delay);\n }\n}\n", "import { Scheduler } from '../Scheduler';\nimport { Action } from './Action';\nimport { AsyncAction } from './AsyncAction';\nimport { TimerHandle } from './timerHandle';\n\nexport class AsyncScheduler extends Scheduler {\n public actions: Array> = [];\n /**\n * A flag to indicate whether the Scheduler is currently executing a batch of\n * queued actions.\n * @type {boolean}\n * @internal\n */\n public _active: boolean = false;\n /**\n * An internal ID used to track the latest asynchronous task such as those\n * coming from `setTimeout`, `setInterval`, `requestAnimationFrame`, and\n * others.\n * @type {any}\n * @internal\n */\n public _scheduled: TimerHandle | undefined;\n\n constructor(SchedulerAction: typeof Action, now: () => number = Scheduler.now) {\n super(SchedulerAction, now);\n }\n\n public flush(action: AsyncAction): void {\n const { actions } = this;\n\n if (this._active) {\n actions.push(action);\n return;\n }\n\n let error: any;\n this._active = true;\n\n do {\n if ((error = action.execute(action.state, action.delay))) {\n break;\n }\n } while ((action = actions.shift()!)); // exhaust the scheduler queue\n\n this._active = false;\n\n if (error) {\n while ((action = actions.shift()!)) {\n action.unsubscribe();\n }\n throw error;\n }\n }\n}\n", "import { AsyncAction } from './AsyncAction';\nimport { AsyncScheduler } from './AsyncScheduler';\n\n/**\n *\n * Async Scheduler\n *\n * Schedule task as if you used setTimeout(task, duration)\n *\n * `async` scheduler schedules tasks asynchronously, by putting them on the JavaScript\n * event loop queue. It is best used to delay tasks in time or to schedule tasks repeating\n * in intervals.\n *\n * If you just want to \"defer\" task, that is to perform it right after currently\n * executing synchronous code ends (commonly achieved by `setTimeout(deferredTask, 0)`),\n * better choice will be the {@link asapScheduler} scheduler.\n *\n * ## Examples\n * Use async scheduler to delay task\n * ```ts\n * import { asyncScheduler } from 'rxjs';\n *\n * const task = () => console.log('it works!');\n *\n * asyncScheduler.schedule(task, 2000);\n *\n * // After 2 seconds logs:\n * // \"it works!\"\n * ```\n *\n * Use async scheduler to repeat task in intervals\n * ```ts\n * import { asyncScheduler } from 'rxjs';\n *\n * function task(state) {\n * console.log(state);\n * this.schedule(state + 1, 1000); // `this` references currently executing Action,\n * // which we reschedule with new state and delay\n * }\n *\n * asyncScheduler.schedule(task, 3000, 0);\n *\n * // Logs:\n * // 0 after 3s\n * // 1 after 4s\n * // 2 after 5s\n * // 3 after 6s\n * ```\n */\n\nexport const asyncScheduler = new AsyncScheduler(AsyncAction);\n\n/**\n * @deprecated Renamed to {@link asyncScheduler}. Will be removed in v8.\n */\nexport const async = asyncScheduler;\n", "import { AsyncAction } from './AsyncAction';\nimport { Subscription } from '../Subscription';\nimport { QueueScheduler } from './QueueScheduler';\nimport { SchedulerAction } from '../types';\nimport { TimerHandle } from './timerHandle';\n\nexport class QueueAction extends AsyncAction {\n constructor(protected scheduler: QueueScheduler, protected work: (this: SchedulerAction, state?: T) => void) {\n super(scheduler, work);\n }\n\n public schedule(state?: T, delay: number = 0): Subscription {\n if (delay > 0) {\n return super.schedule(state, delay);\n }\n this.delay = delay;\n this.state = state;\n this.scheduler.flush(this);\n return this;\n }\n\n public execute(state: T, delay: number): any {\n return delay > 0 || this.closed ? super.execute(state, delay) : this._execute(state, delay);\n }\n\n protected requestAsyncId(scheduler: QueueScheduler, id?: TimerHandle, delay: number = 0): TimerHandle {\n // If delay exists and is greater than 0, or if the delay is null (the\n // action wasn't rescheduled) but was originally scheduled as an async\n // action, then recycle as an async action.\n\n if ((delay != null && delay > 0) || (delay == null && this.delay > 0)) {\n return super.requestAsyncId(scheduler, id, delay);\n }\n\n // Otherwise flush the scheduler starting with this action.\n scheduler.flush(this);\n\n // HACK: In the past, this was returning `void`. However, `void` isn't a valid\n // `TimerHandle`, and generally the return value here isn't really used. So the\n // compromise is to return `0` which is both \"falsy\" and a valid `TimerHandle`,\n // as opposed to refactoring every other instanceo of `requestAsyncId`.\n return 0;\n }\n}\n", "import { AsyncScheduler } from './AsyncScheduler';\n\nexport class QueueScheduler extends AsyncScheduler {\n}\n", "import { QueueAction } from './QueueAction';\nimport { QueueScheduler } from './QueueScheduler';\n\n/**\n *\n * Queue Scheduler\n *\n * Put every next task on a queue, instead of executing it immediately\n *\n * `queue` scheduler, when used with delay, behaves the same as {@link asyncScheduler} scheduler.\n *\n * When used without delay, it schedules given task synchronously - executes it right when\n * it is scheduled. However when called recursively, that is when inside the scheduled task,\n * another task is scheduled with queue scheduler, instead of executing immediately as well,\n * that task will be put on a queue and wait for current one to finish.\n *\n * This means that when you execute task with `queue` scheduler, you are sure it will end\n * before any other task scheduled with that scheduler will start.\n *\n * ## Examples\n * Schedule recursively first, then do something\n * ```ts\n * import { queueScheduler } from 'rxjs';\n *\n * queueScheduler.schedule(() => {\n * queueScheduler.schedule(() => console.log('second')); // will not happen now, but will be put on a queue\n *\n * console.log('first');\n * });\n *\n * // Logs:\n * // \"first\"\n * // \"second\"\n * ```\n *\n * Reschedule itself recursively\n * ```ts\n * import { queueScheduler } from 'rxjs';\n *\n * queueScheduler.schedule(function(state) {\n * if (state !== 0) {\n * console.log('before', state);\n * this.schedule(state - 1); // `this` references currently executing Action,\n * // which we reschedule with new state\n * console.log('after', state);\n * }\n * }, 0, 3);\n *\n * // In scheduler that runs recursively, you would expect:\n * // \"before\", 3\n * // \"before\", 2\n * // \"before\", 1\n * // \"after\", 1\n * // \"after\", 2\n * // \"after\", 3\n *\n * // But with queue it logs:\n * // \"before\", 3\n * // \"after\", 3\n * // \"before\", 2\n * // \"after\", 2\n * // \"before\", 1\n * // \"after\", 1\n * ```\n */\n\nexport const queueScheduler = new QueueScheduler(QueueAction);\n\n/**\n * @deprecated Renamed to {@link queueScheduler}. Will be removed in v8.\n */\nexport const queue = queueScheduler;\n", "import { AsyncAction } from './AsyncAction';\nimport { AnimationFrameScheduler } from './AnimationFrameScheduler';\nimport { SchedulerAction } from '../types';\nimport { animationFrameProvider } from './animationFrameProvider';\nimport { TimerHandle } from './timerHandle';\n\nexport class AnimationFrameAction extends AsyncAction {\n constructor(protected scheduler: AnimationFrameScheduler, protected work: (this: SchedulerAction, state?: T) => void) {\n super(scheduler, work);\n }\n\n protected requestAsyncId(scheduler: AnimationFrameScheduler, id?: TimerHandle, delay: number = 0): TimerHandle {\n // If delay is greater than 0, request as an async action.\n if (delay !== null && delay > 0) {\n return super.requestAsyncId(scheduler, id, delay);\n }\n // Push the action to the end of the scheduler queue.\n scheduler.actions.push(this);\n // If an animation frame has already been requested, don't request another\n // one. If an animation frame hasn't been requested yet, request one. Return\n // the current animation frame request id.\n return scheduler._scheduled || (scheduler._scheduled = animationFrameProvider.requestAnimationFrame(() => scheduler.flush(undefined)));\n }\n\n protected recycleAsyncId(scheduler: AnimationFrameScheduler, id?: TimerHandle, delay: number = 0): TimerHandle | undefined {\n // If delay exists and is greater than 0, or if the delay is null (the\n // action wasn't rescheduled) but was originally scheduled as an async\n // action, then recycle as an async action.\n if (delay != null ? delay > 0 : this.delay > 0) {\n return super.recycleAsyncId(scheduler, id, delay);\n }\n // If the scheduler queue has no remaining actions with the same async id,\n // cancel the requested animation frame and set the scheduled flag to\n // undefined so the next AnimationFrameAction will request its own.\n const { actions } = scheduler;\n if (id != null && actions[actions.length - 1]?.id !== id) {\n animationFrameProvider.cancelAnimationFrame(id as number);\n scheduler._scheduled = undefined;\n }\n // Return undefined so the action knows to request a new async id if it's rescheduled.\n return undefined;\n }\n}\n", "import { AsyncAction } from './AsyncAction';\nimport { AsyncScheduler } from './AsyncScheduler';\n\nexport class AnimationFrameScheduler extends AsyncScheduler {\n public flush(action?: AsyncAction): void {\n this._active = true;\n // The async id that effects a call to flush is stored in _scheduled.\n // Before executing an action, it's necessary to check the action's async\n // id to determine whether it's supposed to be executed in the current\n // flush.\n // Previous implementations of this method used a count to determine this,\n // but that was unsound, as actions that are unsubscribed - i.e. cancelled -\n // are removed from the actions array and that can shift actions that are\n // scheduled to be executed in a subsequent flush into positions at which\n // they are executed within the current flush.\n const flushId = this._scheduled;\n this._scheduled = undefined;\n\n const { actions } = this;\n let error: any;\n action = action || actions.shift()!;\n\n do {\n if ((error = action.execute(action.state, action.delay))) {\n break;\n }\n } while ((action = actions[0]) && action.id === flushId && actions.shift());\n\n this._active = false;\n\n if (error) {\n while ((action = actions[0]) && action.id === flushId && actions.shift()) {\n action.unsubscribe();\n }\n throw error;\n }\n }\n}\n", "import { AnimationFrameAction } from './AnimationFrameAction';\nimport { AnimationFrameScheduler } from './AnimationFrameScheduler';\n\n/**\n *\n * Animation Frame Scheduler\n *\n * Perform task when `window.requestAnimationFrame` would fire\n *\n * When `animationFrame` scheduler is used with delay, it will fall back to {@link asyncScheduler} scheduler\n * behaviour.\n *\n * Without delay, `animationFrame` scheduler can be used to create smooth browser animations.\n * It makes sure scheduled task will happen just before next browser content repaint,\n * thus performing animations as efficiently as possible.\n *\n * ## Example\n * Schedule div height animation\n * ```ts\n * // html:
\n * import { animationFrameScheduler } from 'rxjs';\n *\n * const div = document.querySelector('div');\n *\n * animationFrameScheduler.schedule(function(height) {\n * div.style.height = height + \"px\";\n *\n * this.schedule(height + 1); // `this` references currently executing Action,\n * // which we reschedule with new state\n * }, 0, 0);\n *\n * // You will see a div element growing in height\n * ```\n */\n\nexport const animationFrameScheduler = new AnimationFrameScheduler(AnimationFrameAction);\n\n/**\n * @deprecated Renamed to {@link animationFrameScheduler}. Will be removed in v8.\n */\nexport const animationFrame = animationFrameScheduler;\n", "import { Observable } from '../Observable';\nimport { SchedulerLike } from '../types';\n\n/**\n * A simple Observable that emits no items to the Observer and immediately\n * emits a complete notification.\n *\n * Just emits 'complete', and nothing else.\n *\n * ![](empty.png)\n *\n * A simple Observable that only emits the complete notification. It can be used\n * for composing with other Observables, such as in a {@link mergeMap}.\n *\n * ## Examples\n *\n * Log complete notification\n *\n * ```ts\n * import { EMPTY } from 'rxjs';\n *\n * EMPTY.subscribe({\n * next: () => console.log('Next'),\n * complete: () => console.log('Complete!')\n * });\n *\n * // Outputs\n * // Complete!\n * ```\n *\n * Emit the number 7, then complete\n *\n * ```ts\n * import { EMPTY, startWith } from 'rxjs';\n *\n * const result = EMPTY.pipe(startWith(7));\n * result.subscribe(x => console.log(x));\n *\n * // Outputs\n * // 7\n * ```\n *\n * Map and flatten only odd numbers to the sequence `'a'`, `'b'`, `'c'`\n *\n * ```ts\n * import { interval, mergeMap, of, EMPTY } from 'rxjs';\n *\n * const interval$ = interval(1000);\n * const result = interval$.pipe(\n * mergeMap(x => x % 2 === 1 ? of('a', 'b', 'c') : EMPTY),\n * );\n * result.subscribe(x => console.log(x));\n *\n * // Results in the following to the console:\n * // x is equal to the count on the interval, e.g. (0, 1, 2, 3, ...)\n * // x will occur every 1000ms\n * // if x % 2 is equal to 1, print a, b, c (each on its own)\n * // if x % 2 is not equal to 1, nothing will be output\n * ```\n *\n * @see {@link Observable}\n * @see {@link NEVER}\n * @see {@link of}\n * @see {@link throwError}\n */\nexport const EMPTY = new Observable((subscriber) => subscriber.complete());\n\n/**\n * @param scheduler A {@link SchedulerLike} to use for scheduling\n * the emission of the complete notification.\n * @deprecated Replaced with the {@link EMPTY} constant or {@link scheduled} (e.g. `scheduled([], scheduler)`). Will be removed in v8.\n */\nexport function empty(scheduler?: SchedulerLike) {\n return scheduler ? emptyScheduled(scheduler) : EMPTY;\n}\n\nfunction emptyScheduled(scheduler: SchedulerLike) {\n return new Observable((subscriber) => scheduler.schedule(() => subscriber.complete()));\n}\n", "import { SchedulerLike } from '../types';\nimport { isFunction } from './isFunction';\n\nexport function isScheduler(value: any): value is SchedulerLike {\n return value && isFunction(value.schedule);\n}\n", "import { SchedulerLike } from '../types';\nimport { isFunction } from './isFunction';\nimport { isScheduler } from './isScheduler';\n\nfunction last(arr: T[]): T | undefined {\n return arr[arr.length - 1];\n}\n\nexport function popResultSelector(args: any[]): ((...args: unknown[]) => unknown) | undefined {\n return isFunction(last(args)) ? args.pop() : undefined;\n}\n\nexport function popScheduler(args: any[]): SchedulerLike | undefined {\n return isScheduler(last(args)) ? args.pop() : undefined;\n}\n\nexport function popNumber(args: any[], defaultValue: number): number {\n return typeof last(args) === 'number' ? args.pop()! : defaultValue;\n}\n", "export const isArrayLike = ((x: any): x is ArrayLike => x && typeof x.length === 'number' && typeof x !== 'function');", "import { isFunction } from \"./isFunction\";\n\n/**\n * Tests to see if the object is \"thennable\".\n * @param value the object to test\n */\nexport function isPromise(value: any): value is PromiseLike {\n return isFunction(value?.then);\n}\n", "import { InteropObservable } from '../types';\nimport { observable as Symbol_observable } from '../symbol/observable';\nimport { isFunction } from './isFunction';\n\n/** Identifies an input as being Observable (but not necessary an Rx Observable) */\nexport function isInteropObservable(input: any): input is InteropObservable {\n return isFunction(input[Symbol_observable]);\n}\n", "import { isFunction } from './isFunction';\n\nexport function isAsyncIterable(obj: any): obj is AsyncIterable {\n return Symbol.asyncIterator && isFunction(obj?.[Symbol.asyncIterator]);\n}\n", "/**\n * Creates the TypeError to throw if an invalid object is passed to `from` or `scheduled`.\n * @param input The object that was passed.\n */\nexport function createInvalidObservableTypeError(input: any) {\n // TODO: We should create error codes that can be looked up, so this can be less verbose.\n return new TypeError(\n `You provided ${\n input !== null && typeof input === 'object' ? 'an invalid object' : `'${input}'`\n } where a stream was expected. You can provide an Observable, Promise, ReadableStream, Array, AsyncIterable, or Iterable.`\n );\n}\n", "export function getSymbolIterator(): symbol {\n if (typeof Symbol !== 'function' || !Symbol.iterator) {\n return '@@iterator' as any;\n }\n\n return Symbol.iterator;\n}\n\nexport const iterator = getSymbolIterator();\n", "import { iterator as Symbol_iterator } from '../symbol/iterator';\nimport { isFunction } from './isFunction';\n\n/** Identifies an input as being an Iterable */\nexport function isIterable(input: any): input is Iterable {\n return isFunction(input?.[Symbol_iterator]);\n}\n", "import { ReadableStreamLike } from '../types';\nimport { isFunction } from './isFunction';\n\nexport async function* readableStreamLikeToAsyncGenerator(readableStream: ReadableStreamLike): AsyncGenerator {\n const reader = readableStream.getReader();\n try {\n while (true) {\n const { value, done } = await reader.read();\n if (done) {\n return;\n }\n yield value!;\n }\n } finally {\n reader.releaseLock();\n }\n}\n\nexport function isReadableStreamLike(obj: any): obj is ReadableStreamLike {\n // We don't want to use instanceof checks because they would return\n // false for instances from another Realm, like an