Merge pull request #150 from navis-org/fastcore

Integrate fastcore functions

.github/workflows/test-package.yml

@@ -39,7 +39,7 @@ jobs:
         pip install k3d
         pip install pyarrow
     - name: Install navis
-      run: pip install -e .[dev,vispy-pyqt5,pathos,cloudvolume]
+      run: pip install -e .[dev,vispy-pyqt5,pathos,cloudvolume,fastcore]
     - run: pip install python-igraph
       if: ${{ matrix.igraph == 'igraph' }}
     - name: Report dependency versions

README.md

@@ -19,6 +19,7 @@ NAVis is on [ReadTheDocs](http://navis.readthedocs.io/ "NAVis ReadTheDocs").
 * **render**: use Blender 3D for high quality [visualizations](https://youtu.be/wl3sFG7WQJc)
 * **R** neuron libraries: interfaces with [nat](https://github.com/jefferis/nat), [rcatmaid](https://github.com/jefferis/rcatmaid), [elmr](https://github.com/jefferis/elmr) and more
 * **import-export**: read/write SWCs, neuroglancer's ["*precomputed*"](https://github.com/google/neuroglancer/tree/master/src/datasource/precomputed) format, NMX/NML, NRRD, mesh-files and more
+* **fast**: uses functions compiled in Rust under-the-hood (see [fastcore](https://github.com/schlegelp/fastcore-rs))
 * **scalable**: out-of-the-box support for multiprocessing
 * **extensible**: build your own package on top of navis - see [pymaid](https://pymaid.readthedocs.io/en/latest/) for example
 

conftest.py

@@ -4,15 +4,15 @@ def pytest_ignore_collect(path, config):
     more specific hooks.
     """
     path = str(path)
-    if 'interfaces' in path:
-        return True
-    if '/docs' in path:
-        return True
-    if '/stubs' in path:
-        return True
-    if '/examples' in path:
-        return True
-    if '/dist/' in path:
-        return True
-    if '/binder' in path:
-        return True
+    for pattern in (
+        "interfaces",
+        "/docs",
+        "/stubs",
+        "/examples",
+        "/dist/",
+        "/binder",
+        "/site",
+        "/scripts",
+    ):
+        if pattern in path:
+            return True

docs/source/install.rst

@@ -3,7 +3,7 @@
 Install
 =======
 
-NAVis currently requires Python 3.8 or later. Below instructions assume that
+NAVis currently requires Python 3.9 or later. Below instructions assume that
 you have already installed Python and its package manager ``pip``.
 
 .. topic:: By the way
@@ -67,6 +67,19 @@ These extras can be installed directly, or along with navis with
 The user-facing extras, the dependencies they install,
 and how to install those dependencies directly, are below.
 
+.. _fastcore:
+
+``fastcore``: `navis-fastcore <https://github.com/schlegelp/fastcore-rs>`_
+  ``navis-fastcore`` re-implements a bunch of low-level functions in Rust
+  and wraps them in Python. ``navis`` will use ``fastcore`` under-the-hood
+  if it is available. This is a highly recommended extra as it will
+  speed up certain operations such as geodesic distances, Strahler Index
+  or pruning by several orders of magnitude.
+
+  ::
+
+    pip3 install navis-fastcore
+
 
 .. _pykd:
 

docs/source/tutorials/morph_analysis.ipynb

@@ -737,7 +737,7 @@
     "import networkx as nx\n",
     "\n",
     "# First we need to find the path between the soma and the terminal node\n",
-    "path = nx.shortest_path(n.graph.to_undirected(), n.soma[0], end)\n",
+    "path = nx.shortest_path(n.graph.to_undirected(), n.soma, end)\n",
     "\n",
     "# Get coordinates for the path\n",
     "path_co = n.nodes.set_index('node_id').loc[path, ['x', 'y', 'z']].copy()\n",
@@ -754,7 +754,7 @@
     "\n",
     "# Add Euclidian path\n",
     "end_loc = n.nodes.set_index('node_id').loc[end, ['x', 'y', 'z']]\n",
-    "soma_loc = n.nodes.set_index('node_id').loc[n.soma[0], ['x', 'y', 'z']]\n",
+    "soma_loc = n.nodes.set_index('node_id').loc[n.soma, ['x', 'y', 'z']]\n",
     "ax.plot([soma_loc.x, end_loc.x], [-soma_loc.z, -end_loc.z], c='g', ls='--')\n",
     "\n",
     "d_eucl = np.sqrt(np.sum((end_loc - soma_loc)**2)) * n.units\n",

navis/connectivity/adjacency.py

@@ -152,7 +152,7 @@ def to_adjacency(self, include_other=True) -> pd.DataFrame:
         data = np.zeros((len(index), len(index)), np.uint64)
         df = pd.DataFrame(data, index, index)
         for _, src, tgt, _, _ in self.edges(include_other):
-            df[tgt][src] += 1
+            df.loc[src, tgt] += 1
 
         return df
 

navis/core/dotprop.py

@@ -15,7 +15,6 @@
 import numbers
 import pint
 import types
-import uuid
 import warnings
 
 import numpy as np

navis/core/skeleton.py

@@ -217,13 +217,13 @@ def __truediv__(self, other, copy=True):
 
             # If a number, consider this an offset for coordinates
             n = self.copy() if copy else self
-            n.nodes.loc[:, ['x', 'y', 'z', 'radius']] /= other
+            n.nodes[['x', 'y', 'z', 'radius']] /= other
 
             # At this point we can ditch any 4th unit
             if utils.is_iterable(other):
                 other = other[:3]
             if n.has_connectors:
-                n.connectors.loc[:, ['x', 'y', 'z']] /= other
+                n.connectors[['x', 'y', 'z']] /= other
 
             if hasattr(n, 'soma_radius'):
                 if isinstance(n.soma_radius, numbers.Number):
@@ -254,13 +254,13 @@ def __mul__(self, other, copy=True):
 
             # If a number, consider this an offset for coordinates
             n = self.copy() if copy else self
-            n.nodes.loc[:, ['x', 'y', 'z', 'radius']] *= other
+            n.nodes[['x', 'y', 'z', 'radius']] *= other
 
             # At this point we can ditch any 4th unit
             if utils.is_iterable(other):
                 other = other[:3]
             if n.has_connectors:
-                n.connectors.loc[:, ['x', 'y', 'z']] *= other
+                n.connectors[['x', 'y', 'z']] *= other
 
             if hasattr(n, 'soma_radius'):
                 if isinstance(n.soma_radius, numbers.Number):
@@ -606,17 +606,8 @@ def n_leafs(self) -> Optional[int]:
     @temp_property
     def cable_length(self) -> Union[int, float]:
         """Cable length."""
-        if hasattr(self, '_cable_length'):
-            return self._cable_length
-
-        # The by far fastest way to get the cable length is to work on the node table
-        # Using the igraph representation is about the same speed - if it is already calculated!
-        # However, one problem with the graph representation is that with large neuronlists
-        # it adds a lot to the memory footprint.
-        not_root = (self.nodes.parent_id >= 0).values
-        xyz = self.nodes[['x', 'y', 'z']].values[not_root]
-        xyz_parent = self.nodes.set_index('node_id').loc[self.nodes.parent_id.values[not_root], ['x', 'y', 'z']].values
-        self._cable_length = np.sum(np.linalg.norm(xyz - xyz_parent, axis=1))
+        if not hasattr(self, '_cable_length'):
+            self._cable_length = morpho.cable_length(self)
         return self._cable_length
 
     @property