Kittl/Vectorizing

Utility to vectorize raster images 🚀

Local development

First time run

Open this repository in the dev container:

1. Install dev container extension in vscode
2. At the top of the repo click the Open in dev container badge, or, in vscode press cmd+shift+p and search for the command: Dev Containers: Open Folder in Container
3. The dev container has already installed all necessary tools: conda, black, flake8, pre-commit, AWS CLI, Act

The first time the execution can take few minutes, as it is pulling the dev container docker image, and installing Vectorizing system and python dependencies inside of it.

Development workflow

1. There is a debug configuration named Vectorizing setup, press F5 to run. It will start the flask application in development mode, listening on localhost:8080. If neeed, this can be changed in the debug configuration file.

2. If you want to execute any python command that uses Vectorizing packages, you need to first activate the conda environment in every new terminal you open, by running: conda activate dev

3. If you want to add or remove dependencies python dependencies, add / remove the corresponding package from either requirements/prod.in or requirements/dev.in. Then run, from the root of the repo:

   bash scripts/compile_requirements.sh
   bash scripts/compile_envs.sh

   This will compile dependencies and environments, ensuring a consistent development workflow and deployment.

4. If you want to add or remove system dependencies, update the script: scripts/install_system_dependencies.sh. This is used both in CI and at dev container creation, to keep them consistent.

Linting and formatting

To perform linting and formatting, run from the root of the repo:

pre-commit run --all-files

The first execution might take a bit longer, as it will set up the virtual environment where the linter and the formatter will run.

Server

The server has a single endpoint that receives POST requests. The request format is the following:

{url: string,// Image URLsolver: number,// Solver. 0 -> Binary, 1 -> Colorcolor_count: number,// Number of colors (if applicable)raw: boolean// If true, plain return plain SVG markup}

A typical response would be

{success: boolean,// Whether the request was successfulobjectId: string,// The object id in the S3 bucketinfo: {// Bounds of the computed vectorsbounds: {left: number,top: number,bottom: number,right: number,width: number,height: number},image_width: number,// Image widthimage_height: number// Image height}

Or, if raw = true was supplied, just plain SVG markup

Testing

Tests work on rasterized versions of vectorized markup. i.e

1. Images are vectorized
2. SVG markup is then rasterized

They should primarily focus on making sure that

• Random requests with different input data don't crash the server
• Changes in results don't go unnoticed

To run tests, you can run

python -m pytest vectorizing/tests/test.py

Required setup

If needed, update the S3_TEST_BUCKET environment variable in your .env file. You should have read + write access to it.

Adding new tests

To add a new test:

• Place the image you want to be tested inside vectorizing/tests/images

• Add an entry to the TEST object in vectorizing/tests/config.py in the following form:

  # New object entry'[name_of_image]': [
  # Here, a list of all tests to be ran
  {
  'id': '[test_case_id]'# Test case id, used to generate file names'solver': 0, # Solver to use
  },
  {
  'id': '[test_case_id_2]''solver': 1,
  'color_count': # Amount of colors
  }
  ]

• Run the tests

• Verify the output in vectorizing/testing/results. A new image of the test result should have been placed there. If it looks correct, keep it. It will be used as a baseline for subsequent test runs.

Failing tests

• When any of the test cases fail, an entry will be placed in vectorizing/tests/diff_output highlighting the parts of the test result that are too far apart from the baseline entries.