This is the home of the Scala 2 standard library, compiler, and language spec.
How to contribute
For coordinating broader efforts, we also use the scala/scala-dev tracker.
To contribute here, please open a pull request from your fork of this repository.
Be aware that we can't accept additions to the standard library, only modifications to existing code. Binary compatibility forbids adding new public classes or public methods. Additions are made to scala-library-next instead.
We require that you sign the Scala CLA before we can merge any of your work, to protect Scala's future as open source software.
The general workflow is as follows.
- Find/file an issue in scala/bug (or submit a well-documented PR right away!).
- Fork the scala/scala repo.
- Push your changes to a branch in your forked repo. For coding guidelines, go here.
- Submit a pull request to scala/scala from your forked repo.
For more information on building and developing the core of Scala, read the rest of this README, especially for setting up your machine!
Get in touch!
If you need some help with your PR at any time, please feel free to @-mention anyone from the list below, and we will do our best to help you out:
|username||talk to me about...|
||back end, optimizer, named & default arguments, reporters|
||2.12.x branch, compiler performance, weird compiler bugs, lambdas|
||getting started, build, CI, community build, Jenkins, docs, library, REPL|
||pattern matcher, MiMa, partest|
||collections library, performance|
||interactions with Scala.js|
||higher-kinded types, implicits, variance|
P.S.: If you have some spare time to help out around here, we would be delighted to add your name to this list!
Target the oldest branch you would like your changes to end up in. We periodically merge forward from older release branches (e.g., 2.12.x) to new ones (e.g. 2.13.x).
If your change is difficult to merge forward, you may be asked to also submit a separate PR targeting the newer branch.
If your change is version-specific and shouldn't be merged forward, put
[nomerge] in the PR name.
If your change is a backport from a newer branch and thus doesn't need to be merged forward, put
[backport] in the PR name.
Choosing a branch
Most changes should target 2.13.x. We are increasingly reluctant to target 2.12.x unless there is a special reason (e.g. if an especially bad bug is found, or if there is commercial sponsorship).
The 2.11.x branch is now inactive and no further 2.11.x releases are planned (unless unusual, unforeseeable circumstances arise). You should not target 2.11.x without asking maintainers first.
scala/ +--build.sbt The main sbt build definition +--project/ The rest of the sbt build +--src/ All sources +---/library Scala Standard Library +---/reflect Scala Reflection +---/compiler Scala Compiler +--test/ The Scala test suite +---/files Partest tests +---/junit JUnit tests +---/scalacheck ScalaCheck tests +--spec/ The Scala language specification
scala/ +---/library-aux Scala Auxiliary Library, for bootstrapping and documentation purposes +---/interactive Scala Interactive Compiler, for clients such as an IDE (aka Presentation Compiler) +---/intellij IntelliJ project templates +---/manual Scala's runner scripts "man" (manual) pages +---/partest Scala's internal parallel testing framework +---/partest-javaagent Partest's helper java agent +---/repl Scala REPL core +---/repl-frontend Scala REPL frontend +---/scaladoc Scala's documentation tool +---/scalap Scala's class file decompiler +---/testkit Scala's unit-testing kit +--admin/ Scripts for the CI jobs and releasing +--doc/ Additional licenses and copyrights +--scripts/ Scripts for the CI jobs and releasing +--tools/ Scripts useful for local development +--build/ Build products +--dist/ Build products +--target/ Build products
Get ready to contribute
You need the following tools:
- Java SDK. The baseline version is 8 for both 2.12.x and 2.13.x. It is almost always fine to use a later SDK such as 11 or 15 for local development. CI will verify against the baseline version.
MacOS and Linux work. Windows may work if you use Cygwin. Community help with keeping the build working on Windows and documenting any needed setup is appreciated.
Tools we use
We are grateful for the following OSS licenses:
During ordinary development, a new Scala build is built by the previously released version, known as the "reference compiler" or, slangily, as "STARR" (stable reference release). Building with STARR is sufficient for most kinds of changes.
However, a full build of Scala is bootstrapped. Bootstrapping has two steps: first, build with STARR; then, build again using the freshly built compiler, leaving STARR behind. This guarantees that every Scala version can build itself.
If you change the code generation part of the Scala compiler, your changes will only show up in the bytecode of the library and compiler after a bootstrap. Our CI does a bootstrapped build.
Bootstrapping locally: To perform a bootstrap, run
within an sbt session. This will build and publish the Scala
distribution to your local artifact repository and then switch sbt to
use that version as its new
scalaVersion. You may then revert back
restarrFull will also write the STARR version
buildcharacter.properties so you can switch back to it with
restarr without republishing. This will switch the sbt session to
target-restarr directories instead of
target, which avoids wiping out classfiles and
incremental metadata. IntelliJ will continue to be configured to
compile and run tests using the starr version in
For history on how the current scheme was arrived at, see https://groups.google.com/d/topic/scala-internals/gp5JsM1E0Fo/discussion.
Building with fatal warnings: To make warnings in the project fatal (i.e. turn them into errors), run
set Global / fatalWarnings := true in sbt (replace
Global with the name of a module—such as
reflect—to only make warnings fatal for that module). To disable fatal warnings again, either
reload sbt, or run
set Global / fatalWarnings := false (again, replace
Global with the name of a module if you only enabled fatal warnings for that module). CI always has fatal warnings enabled.
Using the sbt build
Once you've started an
sbt session you can run one of the core commands:
compilecompiles all sub-projects (library, reflect, compiler, scaladoc, etc)
scalacrun the REPL / compiler directly from sbt (accept options / arguments)
enableOptimizerreloads the build with the Scala optimizer enabled. Our releases are built this way. Enable this when working on compiler performance improvements. When the optimizer is enabled the build will be slower and incremental builds can be incorrect.
enableOptimizerand configures a version number based on the current Git SHA. Often used as part of bootstrapping:
sbt setupPublishCore publishLocal && sbt -Dstarr.version=<VERSION> testAll
dist/mkBingenerates runner scripts (
scalac, etc) in
dist/mkPackcreates a build in the Scala distribution format in
junit/testruns the JUnit tests;
junit/testOnly *Fooruns a subset
scalacheck/testruns scalacheck tests, use
testOnlyto run a subset
partestruns partest tests (accepts options, try
publishLocalpublishes a distribution locally (can be used as
scalaVersionin other sbt projects)
set baseVersionSuffix := "bin-abcd123-SNAPSHOT"where
abcd123is the git hash of the revision being published. You can also use something custom like
"bin-mypatch". This changes the version number from
2.13.2-SNAPSHOTto something more stable (
- Note that the
-binstring marks the version binary compatible. Using it in sbt will cause the
2.13. If the version is not binary compatible, we recommend using
set publishArtifact in (Compile, packageDoc) in ThisBuild := falseto skip generating / publishing API docs (speeds up the process).
If a command results in an error message like
a module is not authorized to depend on itself, it may be that a global sbt plugin is causing
a cyclical dependency. Try disabling global sbt plugins (perhaps by
temporarily commenting them out in
We recommend keeping local test files in the
sandbox directory which is listed in
.gitignore of the Scala repo.
Note that sbt's incremental compilation is often too coarse for the Scala compiler codebase and re-compiles too many files, resulting in long build times (check sbt#1104 for progress on that front). In the meantime you can:
- Use IntelliJ IDEA for incremental compiles (see IDE Setup below) - its incremental compiler is a bit less conservative, but usually correct.
We suggest using IntelliJ IDEA (see src/intellij/README.md).
Metals may also work, but we don't yet have instructions or sample configuration for that. A pull request in this area would be exceedingly welcome. In the meantime, we are collecting guidance at scala/scala-dev#668.
In order to use IntelliJ's incremental compiler:
dist/mkBinin sbt to get a build and the runner scripts in
- run "Build" - "Make Project" in IntelliJ
Now you can edit and build in IntelliJ and use the scripts (compiler, REPL) to
directly test your changes. You can also run the
commands in sbt. Enable "Ant mode" (explained above) to prevent sbt's incremental
compiler from re-compiling (too many) files before each
Our guidelines for contributing are explained in CONTRIBUTING.md. It contains useful information on our coding standards, testing, documentation, how we use git and GitHub and how to get your code reviewed.
You may also want to check out the following resources:
- The "Scala Hacker Guide" covers some of the same ground as this README, but in greater detail and in a more tutorial style, using a running example.
- Scala documentation site
Once you submit a PR your commits will be automatically tested by the Scala CI.
Our CI setup is always evolving. See scala/scala-dev#751 for more details on how things currently work and how we expect they might change.
If you see a spurious failure on Jenkins, you can post
/rebuild as a PR comment.
The scabot README lists all available commands.
If you'd like to test your patch before having everything polished for review,
you can have Travis CI build your branch (make sure you have a fork and have Travis CI
enabled for branch builds on it first, and then push your branch). Also
feel free to submit a draft PR. In case your draft branch contains
a large number of commits (that you didn't clean up / squash yet for review),
[ci: last-only] to the PR title. That way only the last commit
will be tested, saving some energy and CI-resources. Note that inactive draft PRs
will be closed eventually, which does not mean the change is being rejected.
CI performs a compiler bootstrap. The first task,
a build of your commit to the temporary repository
Note that this build is not yet bootstrapped, its bytecode is built using the
current STARR. The version number is
is the commit hash. For binary incompatible builds, the version number is
You can use Scala builds in the validation repository locally by adding a resolver
and specifying the corresponding
$ sbt > set resolvers += "pr" at "https://scala-ci.typesafe.com/artifactory/scala-pr-validation-snapshots/" > set scalaVersion := "2.12.2-bin-abcd123-SNAPSHOT" > console
The Scala CI builds nightly download releases and publishes them to https://scala-ci.typesafe.com/artifactory/scala-integration/ .
Using a nightly build in sbt is explained in this Stack Overflow answer
Although we casually refer to these as "nightly" builds, they aren't actually built nightly, but "mergely". That is to say, a build is published for every merged PR.
Scala CI internals
The build bot that watches PRs, triggers testing builds and applies the "reviewed" label after an LGTM comment is in the scala/scabot repo.
The Scala community build is an important method for testing Scala releases. A community build can be launched for any Scala commit, even before the commit's PR has been merged. That commit is then used to build a large number of open-source projects from source and run their test suites.
To request a community build run on your PR, just ask in a comment on the PR and a Scala team member (probably @SethTisue) will take care of it. (details)
Community builds run on the Scala Jenkins instance. The jobs are
..-integrate-community-build. See the