Skip to content

Take screenshots of WMDE fundraising banners in different browsers and resolutions

Notifications You must be signed in to change notification settings

wmde/banner-screenshots

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

WMDE Banner Screenshots

This is a tool for taking screenshots of WMDE fundraising banners on wikipedia.org in different browsers and resolutions.

The system consists of four parts:

  1. A message queue (RabbitMQ)
  2. A worker daemon, screenshot_worker.ts, that processes screenshot messages from the queue and takes screenshots
  3. A worker daemon metadata_worker.ts, that creates and updates metadata for every batch of screenshots.
  4. A CLI tool that creates a test matrix (a batch of screenshots) from the campaign configuration and pushes them into the queue.

Please note that the instructions in this README are for local development only! To create screenshots, use docker-compose environment and the instructions in PRODUCTION.md.

Installing the dependencies

npm install

Starting the message queue

make start-rabbitmq

This will start RabbitMQ inside a docker container (using the official RabbitMQ docker image), listening on port 5672.

If you get an error that port 5672 is blocked, make sure that no other docker containers (e.g. the docker-compose.yml file in this repo) are using the port. Use the command docker ps to see which containers use which port.

Preparing the configuration file

Create the file .env and put in the following content:

QUEUE_URL="amqp://localhost"
TB_SECRET="<secret>"
TB_KEY="<key>"

Replace the placeholders <secret> and <key> with the real API data from your account at Testingbot.

Running the workers

npm run metadata-worker
npm run screenshot-worker

Debugging the screenshot messages

Instead of the full-fledged screenshot worker you can also run

npm run simple-worker

The "simple_worker" echoes the data it receives and does not take any screenshots.

Running the screenshot tool

Creating screenshots for a GitHub branch

The following command will pull the specified campaign branch/tag from the Banner Repository, figure out which channel (desktop, mobile, ipad, etc) the campaign belongs to and queue screenshot requests for this channel:

npx ts-node queue_screenshots.ts <CAMPAIGN_NAME>

Creating screenshots for a local campaign file

npx ts-node queue_screenshots_from_file.ts -c path/to/campaign_info.toml <CHANNEL_NAME>

Downloading the campaign file for a branch

Run the command

make BRANCH_NAME=<CAMPAIGN_NAME> fetch-campaign-info

Replace the placeholder <CAMPAIGN_NAME> with the branch name you want to fetch from. It defaults to main.

Using your local copy of the banner repository

The queue_screenshots_from_file.ts script will look first in the main directory of this repository for a campaign_info.toml. You can use this behavior by creating a symbolic link from the campaign_info.toml of your local copy of the wmde/fundraising-banners/ repository to the main directory of the screenshot repository. You can then omit the -c path/to/campaign_info.toml part of the command:

npx ts-node queue_screenshots_from_file.ts <CHANNEL_NAME>

Running the unit tests

npm run test

Use the following command to run individual tests

npx mocha test/specs/name_of_your_test.js 

Configuration file format

This is the same file used in the wmde/fundraising-banners repository for configuring campaigns. It also contains test matrix configurations.

The TOML file has the following (abbreviated) format

[campaign]
campaign_tracking = 211215-ba-11
preview_url = "https://example.com/banner/{{PLACEHOLDER}}"
...


[campaign.banners.ctrl]
pagename = "B21_WMDE_Test_11_ctrl"
...


[campaign.test_matrix]

platform = ["edge", "firefox_win10" ]
resolution = ["1280x960", "1024x768"]

[campaign] is the key you pass to the screenshot tool. The TOML file can contain several campaigns, each one for a specific channel (a combination of device type and language). [campaign] in the example above is a placeholder, the campaign name can consist of all alphanumeric characters, e.g. desktop or ipad_en.

Each campaign will also have at least one banner, configured in the [campaign.banners.BANNER_NAME] section. Usually, BANNER_NAME is ctrl or var. Each banner has a unique pagename which designates its page name in CentralNotice. To preview a banner (and to render the screenshot), you replace the {{PLACEHOLDER}} with the banner name in the preview_url of the campaign. The screenshot tool will run the test matrix for each banner.

The [campaign.test_matrix] key can contain one of two test dimensions: device and orientation (for mobile devices) or platform and resolution (for desktop browsers).

To see the available values for platform and device, look at the CapabilityFactory class that defines what these values mean in the context of the testing service (testingbot.com or saucelabs.com).

All values are arrays with at least one element. The screenshot tool will use each matrix key with each other matrix key and the banners.
Example: platform=["edge", "safari", "chrome_linux"], resolutions=["1280x960", "1024x768"] and 2 banners, ctrl and var, will create 12 (324) screenshots.

Different test functions for different banners

Add a different test function to the src/test_functions/ directory, im- and export it in src/test_functions/index.js and specify its name with the -t flag of queue_screenshots.ts or queue_screenshots_from_file.ts.

Flushing the message queue

To throw away all messages currently in the queue, run the following commands:

docker stop banner-screenshots-rabbitmq
make start-rabbitmq

The RabbitMQ Docker container uses container-internal storage for its messages and when you stop the container, it will delete all its data.

Architecture

Architecture - Component Diagram

About

Take screenshots of WMDE fundraising banners in different browsers and resolutions

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published