We welcome contributions of any size and skill level. As an open source project, we believe in giving back to our contributors and are happy to help with guidance on PRs, technical writing, and turning any feature idea into a reality.
Tip
For new contributors: Take a look at https://github.com/firstcontributions/first-contributions for helpful information on contributing
node: "^>=18.17.1"
pnpm: "^9.3.0"
# otherwise, your build will fail
We recommend using Corepack, read PNPM docs.
Astro uses pnpm workspaces, so you should always run pnpm install
from the top-level project directory. Running pnpm install
in the top-level project root will install dependencies for astro
, and every package in the repo.
git clone && cd ...
pnpm install
pnpm run build
In #2254 a .git-blame-ignore-revs
file was added to ignore repo-wide formatting changes. To improve your experience, you should run the following command locally.
git config --local blame.ignoreRevsFile .git-blame-ignore-revs
To automatically handle merge conflicts in pnpm-lock.yaml
, you should run the following commands locally.
pnpm add -g @pnpm/merge-driver
pnpm dlx npm-merge-driver install --driver-name pnpm-merge-driver --driver "pnpm-merge-driver %A %O %B %P" --files pnpm-lock.yaml
To get started, create a codespace for this repository by clicking this 👇
Your new codespace will open in a web-based version of Visual Studio Code. All development dependencies will be preinstalled, and the tests will run automatically ensuring you've got a green base from which to start working.
Note
Dev containers is now an open spec which is supported by GitHub Codespaces and other supporting tools.
# starts a file-watching, live-reloading dev script for active development
pnpm run dev
# build the entire project, one time.
pnpm run build
How can I test my changes while contributing to the repository?
During the development process, you may want to test your changes to ensure they are working as expected. Here are a few ways to do it:
-
Run any of the examples in the
/examples
folder. They are linked to use the local Astro source code, so you can see the effects of your changes.pnpm --filter @example/minimal run dev
-
Write a test and run it. This is useful if you're making a specific fix and want to see if your changes are working as expected.
-
Create a separate project and use your local Astro through
pnpm link
. This is helpful if you're making bigger changes and want to test them in a separate project.
Overall, it's up to personal preference which method to use. For smaller changes, using the examples folder may be sufficient. For larger changes, using a separate project may be more appropriate.
You can debug vite by prefixing any command with DEBUG
like so:
DEBUG=vite:* astro dev # debug everything in Vite
DEBUG=vite:[name] astro dev # debug specific process, e.g. "vite:deps" or "vite:transform"
# run this in the top-level project root to run all tests
pnpm run test
# run only a few tests in the `astro` package, great for working on a single feature
# (example - `pnpm run test:match "cli"` runs tests with "cli" in the name)
pnpm run test:match "$STRING_MATCH"
# run tests on another package
# (example - `pnpm --filter @astrojs/rss run test` runs `packages/astro-rss/test/rss.test.js`)
pnpm --filter $STRING_MATCH run test
Most tests use mocha
as the test runner. We're slowly migrating to use node:test
instead through the custom astro-scripts test
command. For packages that use node:test
, you can run these commands in their directories:
# run all of the package's tests
pnpm run test
# run only a few tests in the package
# (example - `pnpm run test -m "cli"` runs tests with "cli" in the name)
pnpm run test -m "$STRING_MATCH"
# run a single test file, you can use `node --test` directly
node --test ./test/foo.test.js
Sometimes you want to run a single test case (it
or describe
) or a single test file. You can do so by using Node.js.
To run a single test file, for example test/astro-basic.test.js
:
node --test test/astro-basic.test.js
If you wish to run a single test case, you have to postfix it
and describe
functions with .only
:
// test/astro-basic.test.js
- describe("description", () => {
+ describe.only("description", () => {
- it("description", () => {
+ it.only("description", () => {})
})
Then, you have to pass the --test-only
option to the Node.js:
node --test --test-only test/astro-basic.test.js
Warning
- If you have nested
describe
, all of them must postfix with.only
--test-only
and--test
must be placed before declaring the path to the file. Failing to do so will test all files
There might be occasions where some tests fail in certain CI runs due to some timeout issue. If this happens, it will be very difficult to understand which file cause the timeout. That's caused by come quirks of the Node.js test runner combined with our architecture.
To understand which file causes the issue, you can modify the test
script inside the package.json
by adding the --parallel
option:
{
- "test": "astro-scripts test \"test/**/*.test.js\"",
+ "test": "astro-scripts test --parallel \"test/**/*.test.js\"",
}
Save the change and push it to your PR. This change will make the test CI slower, but it will allow to see which files causes the timeout. Once you fixed the issue revert the change and push it.
Certain features, like HMR and client hydration, need end-to-end tests to verify functionality in the dev server. Playwright is used to test against the dev server.
# run this in the top-level project root to run all E2E tests
pnpm run test:e2e
# run only a few tests, great for working on a single feature
# (example - `pnpm run test:e2e:match "Tailwind CSS" runs `tailwindcss.test.js`)
pnpm run test:e2e:match "$STRING_MATCH"
When should you add E2E tests?
Any tests for astro build
output should use the main mocha
tests rather than E2E - these tests will run faster than having Playwright start the astro preview
server.
If a test needs to validate what happens on the page after it's loading in the browser, that's a perfect use for E2E dev server tests, i.e. to verify that hot-module reloading works in astro dev
or that components were client hydrated and are interactive.
When creating new tests, it's best to reference other existing test files and replicate the same setup. Some other tips include:
- When re-using a fixture multiple times with different configurations, you should also configure unique
outDir
,build.client
, andbuild.server
values so the build output runtime isn't cached and shared by ESM between test runs.
# auto-format the entire project
# (optional - a GitHub Action formats every commit after a PR is merged)
pnpm run format
# lint the project
# (optional - our linter creates helpful warnings, but not errors.)
pnpm run lint
When making a pull request, be sure to add a changeset when something has changed with Astro. Non-packages (examples/*
) do not need changesets.
pnpm exec changeset
We have benchmarks to keep performance under control. They are located in the benchmarks
directory, and it exposes a CLI you can use to run them.
You can run all available benchmarks sequentially by running (from the project root):
pnpm run benchmark
To run a specific benchmark only, you can add the name of the benchmark after the command:
pnpm run benchmark memory
Use pnpm run benchmark --help
to see all available options.
To run these benchmarks in a PR on GitHub instead of using the CLI, you can comment !bench
. The benchmarks will run on both the PR branch and the main
branch, and the results will be posted as a new comment.
To run only a specific benchmark on CI, add its name after the command in your comment, for example, !bench memory
.
This paragraph provides some guidance to the maintainers of the monorepo. The guidelines explained here aren't necessarily followed by other repositories of the same GitHub organisation.
graph TD;
start{Followed issue\ntemplate?}
start --NO--> close1[Close and ask to\nfollow template]
start --YES--> dupe{Is duplicate?}
dupe --YES--> close2[Close and point\nto duplicate]
dupe --NO--> repro{Has proper\nreproduction?}
repro --NO--> close3[Label: 'needs reproduction'\nbot will auto close if no update\nhas been made in 3 days]
repro --YES--> real{Is actually a bug?}
real --NO--> maybefeat{Is it a feature request?}
maybefeat -- YES --> roadmap[Close the issue.\n Point user to the roadmap.]
maybefeat -- NO --> intended{Is the intended\nbehaviour?}
intended --YES--> explain[Explain and close\npoint to docs if needed]
intended --NO--> open[Add label 'needs discussion'\nRemove 'needs triage' label]
real --YES--> real2["1. Remove 'needs triage' label\n2. Add related feature label if\napplicable (e.g. 'feat: ssr')\n3. Add priority and meta labels (see below)"]
real2 --> tolabel[Use the framework below to decide the priority of the issue,\nand choose the correct label]
The Astro project has five levels of priority to issues, where p5
is the highest in priority, and p1
is the lowest in priority.
-
p5
: the bug impacts the majority of Astro projects, it doesn't have a workaround and makes Astro unusable/unstable.Some examples:
- the dev server crashes;
- the build breaks and doesn't complete;
- huge regressions in terms of performance;
Bugs violate the documentation/intended behaviour of the feature, although sometimes the documentation might not cover possible edge cases.
Usually we don't assign this priority to packages that aren't
astro
, but that can change. -
p4
: the bug impacts many Astro projects, it doesn't have a workaround but Astro is still stable/usable. -
p3
: any bug that doesn't fall in thep4
orp5
category. If the documentation doesn't cover the case reported by the user, it's useful to initiate a discussion via the"needs discussion"
label. Seek opinions from OP and other maintainers. -
p2
: all the bugs that have workarounds. -
p1
: very minor bug, that impacts a small amount of users. Sometimes it's an edge case and it's easy to fix. Very useful if you want to assign the fix to a first-time contributor.
Important
The priority of a bug isn't set on stone. It can change based on different factors.
Assigning labels isn't always easy and many times the distinction between the different levels of priority is blurry, hence try to follow these guidelines:
- When assigning a
p2
, always add a comment that explains the workaround. If a workaround isn't provided, ping the person that assigned the label and ask them to provide one. - Astro has many features, but there are some that have a larger impact than others: development server, build command, HMR (TBD, we don't have a page that explains expectations of HMR in Astro), evident regressions in performance.
- In case the number of reactions of an issue grows, the number of users affected grows, or a discussion uncovers some insights that weren't clear before, it's OK to change the priority of the issue. The maintainer should provide an explanation when assigning a different label. As with any other contribution, triaging is voluntary and best-efforts. We welcome and appreciate all the help you're happy to give (including reading this!) and nothing more. If you are not confident about an issue, you are welcome to leave an issue untriaged for someone who would have more context, or to bring it to their attention.
Server-side rendering (SSR) can be complicated. The Astro package (packages/astro
) is structured in a way to help think about the different systems.
components/
: Built-in components to use in your project (e.g.import Code from 'astro/components/Code.astro'
)src/
: Astro source@types/
: TypeScript types. These are centralized to cut down on circular dependenciescli/
: Code that powers theastro
CLI commandcore/
: Code that executes in the top-level scope (in Node). Within, you’ll find code that powers theastro build
andastro dev
commands, as well as top-level SSR code.runtime/
: Code that executes in different scopes (i.e. not in a pure Node context). You’ll have to think about code differently here.client/
: Code that executes in the browser. Astro’s partial hydration code lives here, and only browser-compatible code can be used.server/
: Code that executes inside Vite’s SSR. Though this is a Node environment inside, this will be executed independently ofcore/
and may have to be structured differently.
vite-plugin-*/
: Any Vite plugins that Astro needs to run. For the most part, these also execute within Vite similar tosrc/runtime/server/
, but it’s also helpful to think about them as independent modules. Note: at the moment these are internal while they’re in development
There are 3 contexts in which code executes:
- Node.js: this code lives in
src/core/
. - Inside Vite: this code lives in
src/runtime/server/
. - In the browser: this code lives in
src/runtime/client/
.
Understanding in which environment code runs, and at which stage in the process, can help clarify thinking about what Astro is doing. It also helps with debugging, for instance, if you’re working within src/core/
, you know that your code isn’t executing within Vite, so you don’t have to debug Vite’s setup. But you will have to debug vite inside runtime/server/
.
Active Astro development happens on the main
branch. main
always reflects the latest code.
Note
During certain periods, we put main
into a prerelease state. Read more about Releasing Astro.
The stable release of Astro can always be found on the latest
branch. latest
is automatically updated every time we publish a stable (not prerelease) version of Astro.
By default, create-astro
and astro.new point to this branch.
Note
Only core maintainers (L3+) can release new versions of Astro.
The repo is set up with automatic releases, using the changeset GitHub action & bot.
To release a new version of Astro, find the Version Packages
PR, read it over, and merge it.
Our release tool changeset
has a feature for releasing "snapshot" releases from a PR or custom branch. These are npm package publishes that live temporarily, so that you can give users a way to test a PR before merging. This can be a great way to get early user feedback while still in the PR review process.
To run changeset version
locally, you'll need to create a GitHub personal access token and set it as a GITHUB_TOKEN
environment variable.
To release a snapshot, run the following locally:
# Notes:
# - YYY should be a keyword to identify this release. Ex: `--snapshot routing` & `--tag next--routing`
# - Use npm/npx instead of pnpm, since npm handles registry login, authentication and publishing.
# - Adding GITHUB_TOKEN in the command adds that token to your bash history. Set a short expiration!
# 1: Tag the new release versions
GITHUB_TOKEN=XXX npx changeset version --snapshot YYY
# 2: Review the diff, and make sure that you're not releasing more than you need to.
git checkout -- examples/
# 3: Release
npm run release --tag next--YYY
# 4: If you're satisfied, you can now throw out all local changes
git reset --hard
By default, every package with a changeset will be released. If you only want to target a smaller subset of packages for release, you can consider clearing out the .changesets
directory to replace all existing changesets with a single changeset of only the packages that you want to release. Just be sure not to commit or push this to main
, since it will destroy existing changesets that you will still want to eventually release.
Full documentation: https://github.com/changesets/changesets/blob/main/docs/snapshot-releases.md
Sometimes, the repo will enter into "prerelease mode". In prerelease mode, our normal release process will publish npm versions under the next
dist-tag, instead of the default latest
tag. We do this from time-to-time to test large features before sharing them with the larger Astro audience.
While in prerelease mode, follow the normal release process to release astro@next
instead of astro@latest
. To release astro@latest
instead, see Releasing astro@latest
while in prerelease mode.
Full documentation: https://github.com/changesets/changesets/blob/main/docs/prereleases.md
If you have gotten permission from the core contributors, you can enter into prerelease mode by following the following steps:
- Run:
pnpm exec changeset pre enter next
in the project root - Create a new PR from the changes created by this command
- Review, approve, and more the PR to enter prerelease mode.
- If successful, The "[ci] release" PR (if one exists) will now say "[ci] release (next)".
Exiting prerelease mode should happen once an experimental release is ready to go from npm install astro@next
to npm install astro
. Only a core contributor run these steps. These steps should be run before
- Run:
pnpm exec changeset pre exit
in the project root - Create a new PR from the changes created by this command.
- Review, approve, and more the PR to enter prerelease mode.
- If successful, The "[ci] release (next)" PR (if one exists) will now say "[ci] release".
When in prerelease mode, the automatic PR release process will no longer release astro@latest
, and will instead release astro@next
. That means that releasing to latest
becomes a manual process. To release latest manually while in prerelease mode:
- In the code snippets below, replace
0.X
with your version (ex:0.18
,release/0.18
, etc.). - Create a new
release/0.X
branch, if none exists. - Point
release/0.X
to the latest commit for thev0.X
version. git cherry-pick
commits frommain
, as needed.- Make sure that all changesets for the new release are included. You can create some manually (via
pnpm exec changeset
) if needed. - Run
pnpm exec changeset version
to create your new release. - Run
pnpm exec release
to publish your new release. - Run
git push && git push --tags
to push your new release to GitHub. - Run
git push release/0.X:latest
to push your release branch tolatest
. - Go to https://github.com/withastro/astro/releases/new and create a new release. Copy the new changelog entry from https://github.com/withastro/astro/blob/latest/packages/astro/CHANGELOG.md.
- Post in Discord #announcements channel, if needed!
Help us make docs.astro.build as accurate and easy-to-use as possible. Contributing to documentation can be a great way to get involved with open source development without having to code.
Head over to the withastro/docs
repo to get involved!