Skip to content

Latest commit

 

History

History
222 lines (134 loc) · 14.9 KB

CONTRIBUTING.md

File metadata and controls

222 lines (134 loc) · 14.9 KB

Contributing

Welcome to the Vega community. Everyone is welcome to contribute. We value all forms of contributions including code reviews, patches, examples, community participation, tutorial, and blog posts. In this document, we outline the guidelines for contributing to the various aspects of the project.

If you send a pull request from a fork, make sure that GitHub actions run successfully. Make sure to add a GH_PAT secret for a personal access token with public repository permissions.

If you find a bug in the code or a mistake in the documentation or want a new feature, you can help us by creating an issue to our repository, or even submit a pull request (PR).

  • For small fixes, please feel free to submit a pull request. Don't worry about creating an issue first.

  • For major changes, please discuss it with the community via a GitHub issue first. This will help us coordinate our efforts, prevent duplication of work, and help you to craft the change so that it is successfully accepted into the project.

  • Generally, we name a branch using this pattern <your 2-3 letters initial>/<topic>. For example, @kanitw's branch regarding scale type might be called kw/scale-type.

See our issue and pull request templates for more information.

Design Principles

Vega-Lite is a compiler and a declarative language to describing interactive multi view graphics. As a compiler, Vega-Lite compiles to the lower-level Vega specifications. The Vega-Lite project inherits many of the design principles of Vega, especially the declarative design.

  • Provide sensible defaults, but allow customization. Any property that is not specified should have a (somewhat obvious) default value, which users can override. For example, Vega-Lite automatically synthesizes scales and guides (axes and legends).
  • Favor composition over templates. While chart templates (as found in spreadsheet programs) can be convenient, they limit the visualization types that can be created. Instead, Vega-Lite uses a compositional approach, describing a visualization based on the properties of graphical marks. Any new building block should be compatible with the existing building blocks.
  • Support gradual specification. Most people write specifications incrementally, making one atomic change at a time such as changing a single property. Any intermediate step should be valid and produce some visualization.

Development Principles

In the development of Vega-Lite follows, we follow these principles that we have established over the years.

  • Strive to remain backwards compatible. Even if we change a major feature, we aim to support the old syntax. However, the old syntax may not be supported by the JSON schema.
  • Backwards compatibility concerns the input, not the output. Just like Vega, we may change how a declarative specification is interpreted.
  • Generate generic Vega specifications. All decisions about how to compile to Vega have to be made without access to the data that will be used in the chart. This principle enables the compiled Vega charts to work with any dataset that follows the same schema (field names and types).
  • Enable transition to Vega. We aim to generate readable Vega specifications that do not contain unnecessary properties (e.g. Vega's defaults). When specifications are clean, it is easier to handoff specifications to other tools.
  • Fail gracefully. If there is a invalid property in the Vega-Lite specification, show a warning and subsequently ignore it.

Looking for a Task to Contribute

You can find tasks with the "Good first issue" label in the issue tracker 🙏. Please add a comment in issues if you are planning to work on a major task.

Documentation and Website

The website is under site/ and the documentation is under site/docs/. We use GitHub Pages to publish our documentation when we release a new version. To contribute changes to the documentation or website, simply submit a pull request that changes the corresponding markdown files in site/.

Since we only publish the GitHub Pages when we release a new version, it might be slightly outdated compared to main. For development, once you have setup the repository, you can run yarn site to serve the GitHub page locally at http://localhost:4000/vega-lite/.

Note that when you checkout different branches, the compiled JavaScript for the website might be reset. You might have to run yarn build:site to recompile the JavaScript so that interactive examples work.

Documentation Guide

General Guides for Markdown Files:

  • Wrap properties (key) with back ticks.
  • Wrap values with back ticks for numbers and booleans (e.g., 5, true) and wrap string values with both back ticks and double quotes ("red").

Property Table

To generate property tables from the JSON schema (which is in turn generated from the Typescript interfaces, you can include the table.html template. For example, to generate a table that includes rangeStep, scheme, and padding from Scale, you can use

{% include table.html props="rangeStep,scheme,padding" source="Scale" %}

To define a link for types in the table, you can edit _data/link.yml.

For JSDocs comment in the interfaces, please add __Default value:__ line at the end to describe the property's value.

Examples

To include an example specification in the documentation, the specification's .vl.json file must be in examples/specs. Then you can use the following span tag to include the specification (e.g., for point_1d.vl.json):

<span class="vl-example" data-name="point_1d"></span>

Before adding a new example, you might want to search inside examples/ folder to see if there are other redundant examples that you can reuse.

To name the example file:

  • Please begin with mark type and follow by some description for a static single view example. For stacked marks, add stacked_ prefix.
  • For composite views, please begin the file with the operator name (e.g., layer).
  • For interactive example, begin with either interactive_ or selection_.
  • For examples that are only for regression test, begin with test_.

After you push a new branch to GitHub, the CI will automatically run yarn build:examples to recompile all examples and push the changed Vega specs and SVG files in examples/compiled , so that your branch includes these changes. When you add a new example or update the code, you may run yarn build:examples or yarn build:example <examplename> (e.g., yarn build:example bar_1d) to see the change locally. However, do not include these changes in your commit as different systems produce slightly different SVGs (mainly due to floating point differences). To avoid unnecessary SVG diffs, we should just let the CI generate the images. You're still encouraged to run yarn build:examples to make sure that your code does not cause unnecessary changes.

Notes:

  1. yarn build:examples only re-compile SVGs if the output Vega file changes (so it runs way faster). If you want to enforce re-compilation of all SVGs, use yarn build:examples-full.
  2. To make Travis run yarn build:examples-full, include [SVG] in your commit message of the last commit in your branch.
  3. To run yarn build:examples, you need to install gnu parallel. (For Mac, you can simply do brew install parallel.)

Development Guide

Repository Setup

  1. Make sure you have node.js. For mac users, we recommend using homebrew and simply run:
brew install node
  1. Clone this repository and cd into your local clone of the repository, and install all the npm dependencies. We use yarn to have reproducible dependencies:
git clone https://github.com/vega/vega-lite.git
cd vega-lite
yarn

Now you should be able to build and test the code.

  1. To serve the website and documentation, you will need ruby, bundler and Jekyll.

For ruby, Mac users can use homebrew to add it:

brew install ruby

For bundler:

gem install bundler

Or to install the same version that was used to create the bundle:

gem install bundler -v "$(grep -A 1 "BUNDLED WITH" Gemfile.lock | tail -n 1)"

For Jekyll and its dependencies, because we already have the Gemfile in the repo, you can simply run:

pushd site && bundle install && popd

Directory Structure

  • bin/ – Scripts for using Vega-Lite with command line.

  • site/ – Vega-Lite website including documentation.

    • _data/ – Jekyll data.
    • _includes/ – Jekyll includes.
    • _layouts/ – Jekyll layout files.
    • data/ – Example data.
    • examples/ – Example images, specifications, and pages for the website.
    • static/ – Static files for the website.
  • examples/ – Example Vega-Lite specifications.

    • specs Vega-Lite examples.
    • compiled The generated Vega specifications and SVG files of the Vega-Lite examples.
  • scripts/ - Scripts for NPM commands.

  • src/ - Main source code directory.

    • All interfaces for Vega-Lite syntax should be declared at the top-level of the src/ folder.
      • src/index.ts is the root file for Vega-Lite that exports the global vegaLite object.
      • Other files under src/ reflect the namespace structure. All methods for vegaLite.xxx will be in either src/xxx.ts or src/xxx/xxx.ts. For example, vegaLite.channel.* methods are in src/channel.ts while vegaLite.compile is in src/compile/compile.ts.
  • test/ - Code for unit testing. test's structure reflects src's directory structure. For example, test/compile/ tests files inside src/compile/.

  • test-runtime/ - Code for runtime tests.

  • typings/ - TypeScript typing declaration for dependencies.

Understanding How Vega-Lite Works

  • The main compiler code is in src/compile/compile.ts. To try to understand how Vega-Lite works, first start by reading the compile method in the file and try to understand different phases in the compilation process. You can browse the code online with Sourcegraph or GitHub1s.

Commands

This section lists commands that are commonly used during development. See package.json for other commands.

Build

You can run yarn build to compile Vega-Lite and regenerate vega-lite-schema.json.

Basic Lint & Test & Test Coverage

yarn test run linting and all unit-tests respectively. yarn format automatically fixes linting issues if possible. yarn test:inspect to inspect tests

yarn test:cover includes test coverage and generates a report inside coverage/index.html. You can see if specific lines are covered in the unit test by running open coverage/index.html and browsing through the report.

Watch tasks

During development, it can be convenient to rebuild automatically or to run tests in the background. You can use:

  • yarn watch:test to start a watcher task that lints and runs tests when any .ts file changes.

  • yarn watch to start a watcher task that re-compiles Vega-Lite when .ts files related to VL change.

Website

yarn site. See details in Documentation and Website.

Publishing

To make a release, run npm run release. After the release notes are generated (from commits), please take a look to clean up so they are readable to the community. Triggered by the GitHub release, the CI will automatically deploy the website and update the schema repo. The website deployment will require an approval from the maintainers on GitHub.

Suggested Programming Environment.

We use the Visual Studio Code editor.

  • VSCode has nice built-in Typescript support!
  • We already include project settings to hide compiled files (*.js, *.js.map). This should work automatically if you open the vega-lite folder with VSCode.
  • Make sure to install ESLint, Prettier extensions.
  • The vscode-jest-runner extension is also very helpful for debugging tests.

Manually Testing with Vega-Editor

To manually test your changes locally, you should have a local instance of Vega Editor and link Vega-Lite to the editor (See Vega Editor's README for instructions).

To update the Vega-Lite code in the editor, you need to compile TypeScript to JavaScript. The easiest way is to run yarn watch in the Vega-Lite directory. This command will automatically recompile the code whenever you make changes.

Pull Requests and Continuous Integration (CI)

All pull requests will be tested on GitHub Actions. If your PR does not pass the checks, your PR will not be approved. The CI will run yarn test, generate Vega specs and SVG files from your updated code, compare them with the existing compiled outputs in examples/compiled/, and check code coverage of your code. If you don't want your PR reviewed until checks pass, mark the pull request as draft. Once you're ready for review, convert the pull request to mark it as ready for review.

Code Coverage

When checking for code coverage, we require that your PR tests cover at least the same percentage of code that was being covered before. To check the code coverage, you can see the link in the job log of your CI test, from the GitHub page of your PR, or on https://codecov.io/gh/vega/vega-lite/commits. It'll be usually in the form of https://codecov.io/gh/vega/vega-lite/commit/your-full-head-commit-number. Under the Files and Diff tab, you can check your code coverage differences and total. In Files, you can check which lines in your files are being tested (marked in green) and which are not (marked in red). We appreciate PRs that improve our overall code coverage!