decaffeinate-coffeescript2

decaffeinate fork of CoffeeScript

This is is a fork of the CoffeeScript implementation with some small patches that are useful to the decaffeinate project. It also contains TypeScript types for working with the CoffeeScript AST.

It is available on npm under the name decaffeinate-coffeescript, but the intention is that it will only be useful for decaffeinate and its dependencies.

Process

The main purpose of this process approach is to make it easy to make patches to CoffeeScript while also not interfering with branch names or version numbers. Any of these steps could probably be automated, but since this fork will likely have very few changes, we'll just follow the process manually for now.

Branch/release strategy

All work is done on a branch directly off of the most recent release, currently 1.12.7. Work is done on a branch called decaffeinate-fork-1.12.7, which is also specified as the GitHub default branch. Ideally, the git history should be clean, and make it clear which changes have been made on top of the official release.

Release names are based on the CoffeeScript release, with a patch number as a suffix. So the first patch is version 1.12.7-patch.1, then 1-12.7-patch.2, etc. Semantic versioning isn't too important here since this package is just for internal use within the decaffeinate project.

Submitting a new patch

Patches are submitted as pull requests to the most recent fork branch. If the patch is a bug fix, we also make a reasonable effort to submit a PR to the official CoffeeScript repo on the master branch.

Once the patch is landed, we make another commit updating package.json with the new version number and publish the result to npm.

Dealing with CoffeeScript updates

Whenever CoffeeScript releases a new version, we will switch to a new branch based on the new release and re-apply the relevant patches using cherry-picks. We'll skip irrelevant changes like the package.json commits. This should hopefully keep the git history understandable and keep this repository focused as a small set of well-understood patches.