| # For Developers |
| |
| After cloning the repo, run `pre-commit install` so that your commits are automatically formatted for you. Otherwise you'll get yelled at later by our CI. |
| |
| > See https://pre-commit.com/#installation if you don't have `pre-commit` installed. |
| |
| It's a lot of work for us to review and accept Pull Requests. Sometimes things can't merge simply because it's too hard to determine if you change is breaking something else, or because the system you're touching is already being refactored by someone else. Before working on new features, we strongly encourage you to review the project's scope described in the `README.md` file. For large changes, consider writing a design document using [this template](https://goo.gl/YCQttR). |
| |
| ## Architecture |
| |
| Unlike some other Bazel rules, these rules require a separate build step, and |
| can not be used by just pointing your workspace file at this git repo without |
| [some extra work](./examples/from_source/). |
| |
| When you run 'bazel build release', the core parts of these rules will be |
| bundled up into dist/bin/release.tar.gz, and that file can be used in your own |
| workspace by adding something like the following to your workspace file: |
| |
| http_archive( |
| name = "build_bazel_rules_nodejs", |
| urls = [ |
| "file:///tmp/release.tar.gz", |
| ], |
| ) |
| |
| The various submodules in packages/ are not included in release.tar.gz. Most of |
| them contain their own package.json, and they are designed to function like |
| normal npm packages. In the release process, each submodule gets bundled up into |
| an npm package, and uploaded to npm as `@bazel/[name]`. End users then add the |
| desired packages to their own package.json file to use them. |
| |
| If you run bazel build packages/..., you can then see the resulting npm package |
| by looking in, eg, dist/bin/packages/concatjs/npm_package. |
| |
| This separate packaging means that package paths differ between usage inside |
| this repo, and usage by end users. A //packages/something:foo.bzl path needs to be |
| mapped to @bazel/something:foo.bzl, and these modifications are taken care of |
| by a substitutions= argument to pkg_npm() in each package's BUILD.bazel file. |
| |
| ## In-repo tests |
| |
| A number of tests in this repo are designed to function with local repository |
| paths, meaning they can be run directly, and are faster to execute. The `yarn |
| test` command is a shortcut to exclude integration tests, eg |
| |
| yarn test //... |
| |
| or |
| |
| yarn test //packages/... |
| |
| ## Integration tests |
| |
| In order to test that the rules work outside of this repo, this repo uses bazel-in-bazel with the |
| bazel_integration_test rule to set up an environment with the package paths matching what end users |
| will see. The end-to-end tests in e2e, and examples are built this way. |
| |
| The integration tests must be run in series, as they use up too many resources when run in parallel. |
| |
| `bazel test ...` includes all these integration tests. If you wish to exclude some of them, see the output of |
| `yarn test` in the previous section. |
| |
| When running the e2e tests, it is recommended to tune the memory usage of Bazel locally. This can be done with `bazel --host_jvm_args=-Xms256m --host_jvm_args=-Xmx1280m test ... --test_tag_filters=e2e --local_ram_resources=792 --test_arg=--local_ram_resources=13288`. A shortcut for this is `yarn test_e2e //...`. |
| |
| Similarly, for test examples run `bazel --host_jvm_args=-Xms256m --host_jvm_args=-Xmx1280m test ... --test_tag_filters=examples --local_ram_resources=792 --test_arg=--local_ram_resources=13288`. A shortcut for this is `yarn test_examples //...`. |
| |
| To test all targets locally in the main workspace and in all nested workspaces run: |
| |
| ``` |
| yarn test_all |
| ``` |
| |
| To do a full clean run: |
| |
| ``` |
| yarn clean_all |
| ``` |
| |
| ## Debugging |
| |
| See [this page](./docs/debugging.md). |
| |
| ## Patching |
| |
| For small changes, you may find it easier to [patch the standard |
| rules](./docs/changing-rules.md) instead of building your own release products. |
| |
| ## Releasing |
| |
| Start from a clean checkout at master/HEAD. |
| |
| Note: if you are using a new clone, you'll need to configure `git-clang-format` to be able to commit the release: |
| |
| 1. `git config clangFormat.binary node_modules/.bin/clang-format` |
| 1. `git config clangFormat.style file` |
| |
| Googlers: you should npm login using the go/npm-publish service: `$ npm login --registry https://wombat-dressing-room.appspot.com` |
| |
| Check if there are any breaking changes since the last tag - if so, this will be a major. Check if there were new features added since the last tag - if so, this will be a minor. |
| |
| 1. `npm version [major|minor|patch]` (`major` if there are breaking changes, `minor` if there are new features, otherwise `patch`) |
| 1. Manually update the CHANGELOG.md based on the commits since the last release. Look for breaking changes that weren't documented. |
| 1. If publishing from inside Google, set NPM_REGISTRY="--registry https://wombat-dressing-room.appspot.com" in your environment |
| 1. Build npm packages and publish them: `./scripts/publish_release.sh` (for a release candidate, add arguments `publish next`) |
| 1. Run `./scripts/update_nested_lock_files.sh` to update the lock files in all nested workspaces to new release |
| 1. `git commit -a -m 'chore: update lock files for release'` |
| 1. `git push upstream && git push upstream --tags` |
| 1. (Manual for now): go to the [releases] page, edit the release with rough changelog (especially note any breaking changes!) and upload the release artifact from `rules_nodejs-[version].tar.gz`. Also copy the release notes from CHANGELOG.md |
| 1. Announce the release on Bazel slack in `#javascript |
| |
| [releases]: https://github.com/bazelbuild/rules_nodejs/releases |