Skip to content

Latest commit

 

History

History
275 lines (187 loc) · 14.3 KB

File metadata and controls

275 lines (187 loc) · 14.3 KB

Thank you for your interest in contributing to Turborepo!

General dependencies

You will need to have these dependencies installed on your machine to work on this repository:

Optional dependencies

  • Bun is required to build @turbo/gen (the turbo gen code generator). The @turbo/gen package is compiled into a standalone binary using bun build --compile.
  • For running tests locally, jq and zstd are also required.
    • macOS: brew install jq zstd
    • Linux: sudo apt update && sudo apt install jq zstd
    • Windows: choco install jq zstandard
  • On Linux, ensure LLD (LLVM Linker) is installed, as it's not installed by default on many Linux distributions (e.g. apt install lld).

Structure of the repository

In general, there are two major areas in the repository:

  • The crates directory with the Rust code for the turbo binary
  • The packages directory with JavaScript packages
  • The examples directory with examples of how to use Turborepo with other tools and frameworks
  • The docs directory with the documentation for Turborepo

Building Turborepo

  1. Run pnpm install at the root of the repository
  2. Run cargo build

TLS Implementation

Turborepo uses reqwest to make requests to the Remote Cache.

reqwest supports two TLS implementations: rustls and native-tls. rustls is a pure Rust implementation of TLS, while native-tls is a wrapper around OpenSSL. You may select which implementation you want with the native-tls and rustls-tls features.

By default, the rustls-tls feature is selected so that cargo build works out of the box. If you wish to select native-tls, you may do so by running cargo build --no-default-features --features native-tls.

OpenTelemetry Observability

Turborepo includes OpenTelemetry (OTel) support for exporting metrics in default builds.

If you need to build without OTel support, use:

cargo build -p turbo --no-default-features --features rustls-tls

experimentalObservability is still gated by futureFlags.experimentalObservability in turbo.json.

Running tests

Important

You will need to have jq and zstd installed on your system in order to run tests. See General dependencies for instructions on how to install these tools.

First, install Turborepo globally with your package manager of choice. For instance, with npm, npm install -g turbo. This will install the turbo binary in your system's PATH, making it globally available.

Now, from the root directory, you can run:

  • Unit tests
  cargo test
  • A module's unit tests
cargo test -p <module>
  • Integration tests

    cargo test -p turbo
  • A single integration test file

    cargo test -p turbo --test force_test

Manually testing turbo

After building turbo, you can manually test turbo for the behaviors you're affecting with your changes. We recommend setting an alias to the built binary so you can call it with your alias.

alias devturbo='~/projects/turbo/target/debug/turbo'
devturbo run build --skip-infer

Important

The --skip-infer flag is required so that turbo doesn't try to use a locally installed binary of turbo. Forgetting to use this flag will cause devturbo to defer to the binary installed into the repository.

A non-exhaustive list of things to check on:

  • Features related to your changes
  • Test with and without daemon (daemon is deprecated for turbo run but still used by turbo watch)
  • Installation scenarios
    • Global only. turbo is installed as global binary without a local turbo in repository.
    • Local only. turbo is installed as local binary without global turbo in PATH. turbo is invoked via a root package script.
    • Global and local. turbo is installed as global binary, and local turbo in repository. Global turbo delegates to local turbo

Repositories to test with

There are many open-source Turborepos out in the community that you can test with. A few are listed below:

Debugging tips

Logging & Output

All output in Turborepo flows through turborepo-log. Use the following decision tree when adding new output:

What you're writing Use this Why
Internal debug/trace info tracing::{debug,trace,info,warn}! Developer diagnostics, filtered by TURBO_LOG_VERBOSITY. Not user-facing.
User-facing warning/error/status turborepo_log::{warn,error,info}(Source::turbo(Subsystem::X), ...) Renders in both terminal and TUI via sinks. Structured, sanitized.
Task child process output TaskHandle::task_output(channel, bytes) Routes through GroupingLayer for grouped/passthrough buffering, then to sinks for rendering. TerminalSink adds per-line task prefixes.
Task-scoped error/warning task_handle.emit(LogEvent::new(Level::Error, Source::task(id), msg)) Appears inline in the task's output stream (grouped with child process bytes).
Cache status message task_handle.task_output(OutputChannel::Stdout, message.as_bytes()) Plain text, rendered with task prefix by TerminalSink. Also call TaskSender::status() for TUI lifecycle.

Adding a new subsystem

To emit logs from a new area of the codebase, add a variant to the Subsystem enum in crates/turborepo-log/src/event.rs. The enum is #[non_exhaustive] so this is non-breaking. Each variant maps to a lowercase string via the Display impl.

Architecture

Task Executor / Commands
        |
   GroupingLayer          <- per-task buffering (passthrough vs grouped)
        |
     Logger               <- broadcasts to all sinks
        |
   +---------+---------+
   |         |         |
Terminal   TuiSink   FileSink
 Sink
  • TerminalSink: Owns line prefixing (ColorSelector, per-task line buffers), CI group markers, and color rendering.
  • TuiSink: Routes task output to the TUI's task panes, normalizes \n to \r\n for the VT100 parser.
  • FileSink / CollectorSink: Structured JSON capture.
  • GroupingLayer: Manages per-task buffering. In passthrough mode (stream), events flow immediately. In grouped mode, events buffer per-task and flush atomically on task completion.

Links in error messages

Many of Turborepo's error messages include links to information or documentation to help end users.

The base URL for the links can be set to a value of your choosing by providing a TURBO_SITE environment variable at compilation time.

TURBO_SITE="http://localhost:3000" cargo build

Verbose logging

Verbose logging can be enabled by using the -v, -vv, or -vvv flag on your turbo command, depending on the level of logging you're looking for:

turbo build --vvv

Crash logs

In the event of a crash, Rust's crash logs will be written to your temp directory. When turbo crashes, the location of the crash log will be printed to the console.

Terminal UI debugging

The architecture of the Terminal UI makes for a tricky debugging experience. Because the UI writes to the console through stdout in a specific way, using println!() statements won't work as expected.

Instead, use eprintln!() to print to stdout and output stdout to a file:

# devturbo is an alias to the debug binary of `turbo` in this case
devturbo run build --ui=tui --skip-infer 2&> ~/tmp/logs.txt

Important

The --skip-infer flag is required so that turbo doesn't try to use a locally installed binary of turbo. Forgetting to use this flag will cause devturbo to defer to the binary installed into the repository rather than the one you're developing.

Publishing turbo to the npm registry

See the publishing guide.

Contributing to examples

Contributing to examples helps the Turborepo community by showcasing how to use Turborepo in real-world scenarios with other tools and frameworks. They can be found in the examples directory of this repository.

Important

As Turborepo usage has grown, the community has contributed more and more examples to the repository. While this is exciting for us on the core team, we're unable to maintain the full surface area of every example, given the constant updates across the breadth of tooling that Turborepo works with in the examples.

Because of this, a handful of the examples are explicitly marked as maintained by the core team. For the rest, we work with the community to keep them as up to date and correct as possible. If you find a problem with a community-supported template, we ask that you do not open a GitHub Issue for it. Instead, please open a pull request with the needed fixes.

The basic example is the default used by create-turbo.

For simplicity, each example is treated as a standalone "repository", separate from the rest of the repository, with its own dependencies, lockfile, turbo version, etc. You are able to run code and make code updates in an example without needing to install the dependencies of the rest of the repository.

Note

You may find that opening your code editor directly in the example's directory that you're working on can give you a better sense of how the example will feel to community members who download the example.

Contributing to an existing example

To contribute to an existing example, create your code updates and submit a pull request to the repository. No special steps are required to contribute to an example.

Philosophy for new examples

Turborepo works with any framework, tool, or even language. Because of this, the community often expresses interest in creating new examples to showcase Turborepo working with other tooling.

However, we aim to have as few examples in the repository while still showcasing Turborepo's flexibility. By having fewer examples, the core team has a better chance to maintain the collection of examples, keeping them at a higher quality. The ecosystem evolves quickly, and keeping every example up-to-date for every tool requires a wealth of attention. Our goal is to balance the needs of the core team and the community together to keep the Turboverse in a healthy state.

Due to these considerations, we ask that you first open a Discussion before working on a new example for the repository. It's best to make sure ahead of time that the example you'd like to propose will be accepted. Once you have received approval, you can work on and create a pull request for your example.

Designing a new example

Each example should have a specific focus when compared to the basic example. The goal is for an example to show how to use a singular, distinct technology's usage in a Turborepo.

You're encouraged to start with the basic example and add your specific tool of interest to it. Each example should have as few modifications to the basic example as possible required to showcase the tool or framework.

Key characteristics of a great example include:

  • One technology added to the basic example
  • An updated README at the root of the example directory. Make sure to include any steps required to run the example
  • All tasks in turbo.json in the example run successfully without any code changes needed
  • Works with every package manager listed in our Support Policy

Once you've created your example (with prior approval, as discussed above), you can submit a pull request to the repository.

Testing examples

To test out the experience of your example with create-turbo, run create-turbo with the --example flag pointed to a URL to your example's source code:

npx create-turbo@latest --example https://github.com/your-org/your-repo/tree/your-branch/...

This will allow you to use the example as a user would.