Edit

CLI tutorial

The Bloop CLI allows you to run Bloop commands from the comfort of the terminal. It is nothing but another build client for Bloop.

The Bloop CLI is often handy and allows you to run as many build commands as you want with a different workflow as most host build tools you use to build your project code. It makes building, testing and running Scala code feel snappy and productive.

Using the Bloop CLI is optional. If you want to learn more about how to use it, you're in the right place. The following page teaches you common CLI commands and patterns run on a simple build.

Requirements

  1. A build with some projects in it (the guide assumes foo and foo-test)
  2. You have followed the Installation Guide, which means:
    • The build server is running in the background.
    • The bloop CLI tool is installed in your machine.
    • You have exported your build to Bloop.

Compile

Let's compile first the test project from scratch.

→ bloop compile foo-test
Compiling foo (1 Scala source)
Non-compiled module 'compiler-bridge_2.12' for Scala 2.12.7. Compiling...
  Compilation completed in 12.165s.
Compiled foo (13176ms)
Compiling foo-test (1 Scala source)
Compiled foo-test (578ms)

The first time you compile, Bloop resolves and compiles the compilation bridges for your Scala version if they are not found in the bloop cache. Then, it follows up with the incremental compilation.

Compile downstream projects

The compilation of a project always requires the compilation of all downstream projects (its dependencies).

The previous logs show Bloop has compiled both foo-test and foo. This is expected given that foo is a dependency of foo-test and foo wasn't already compiled.

Compile upstream projects

The --cascade flag allows you to change the public signature of a project (say, foo) and have bloop compile all the transitive projects depending on foo to detect compilation errors.

Assuming we have not compiled foo and foo-test before, cascade compilation of foo will compile foo and projects depending on foo; in our build, only foo-test.

→ bloop compile --cascade foo
Compiling foo (1 Scala source)
Compiled foo (13176ms)
Compiling foo-test (1 Scala source)
Compiled foo-test (578ms)

We say that cascade compilation compiles projects upstream because it's the inverse of compiling projects downstream. Use --cascade whenever you know you're going to work in a subset of your build graph but you don't remember the projects that are part of it.

Clean the caches

Clean the compilation caches in your build with:

→ bloop clean foo-test
→

This operation is uncommon but might be useful if the incremental compiler state is causing spurious errors. clean does not remove class files, it only cleans the incremental compilation state. If you want to remove the class files, do it manually.

By default, bloop clean only cleans the cache of a given project. Run clean with --propagate to clean the cache of downstream projects.

→ bloop clean foo-test --propagate
→

Compiling again foo-test should trigger the compilation of both foo and foo-test.

→ bloop compile foo-test
Compiling foo (1 Scala source)
Compiled foo (290ms)
Compiling foo-test (1 Scala source)
Compiled foo-test (455ms)

Test

Once foo and foo-test are compiled, let's test them:

→ bloop test foo
CubeCalculatorTest:
- CubeCalculator.cube
Execution took 19ms
1 tests, 1 passed
All tests in CubeCalculatorTest passed

===============================================
Total duration: 19ms
All 1 test suites passed.
===============================================

We use bloop test foo instead of bloop test foo-test out of habit even though only foo-test defines test cases. In fact, Bloop interprets bloop test foo as a shortcut for bloop test foo-test if foo-test exists in the build.

Test downstream projects

bloop test only runs tests of a concrete project. You can run tests for your project and all your dependencies with the --propagate flag:

→ bloop test foo --propagate
CubeCalculatorTest:
- CubeCalculator.cube
Execution took 16ms
1 tests, 1 passed
All tests in CubeCalculatorTest passed

===============================================
Total duration: 16ms
All 1 test suites passed.
===============================================

As expected, the output is the same as bloop test foo because foo has no dependencies defining test sources.

Test upstream projects

Like compile foo --cascade, test foo --cascade:

  1. triggers the compilation of foo and all projects depending on foo (with their dependencies)
  2. runs tests of foo and all projects depending on foo (without their dependencies)

In our case, testing foo with or without --cascade will compile and test the same number of projects because our build graph is simple enough and only foo-test defines tests.

→ bloop test foo --cascade
CubeCalculatorTest:
- CubeCalculator.cube
Execution took 16ms
1 tests, 1 passed
All tests in CubeCalculatorTest passed

===============================================
Total duration: 16ms
All 1 test suites passed.
===============================================

However, given:

  1. two new projects bar and bar-test
  2. two dependencies;
    • one between foo-test and bar-test
    • another one between bar-test and bar

test foo --cascade will:

  1. compile foo, foo-test, bar and bar-test
  2. test foo (skipping it, as it has no tests) and foo-test
→ bloop test foo --cascade
CubeCalculatorTest:
- CubeCalculator.cube
Execution took 16ms
1 tests, 1 passed
All tests in CubeCalculatorTest passed

===============================================
Total duration: 16ms
All 1 test suites passed.
===============================================

Enabling --cascade for test execution is a powerful tool to run tests on all projects that could be possibly affected by a change in a project.

Test only a specific test suite

You can test only a specific test suite with --only or -o:

→ bloop test foo -o CubeCalculatorTest
CubeCalculatorTest:
- CubeCalculator.cube
Execution took 17ms
1 tests, 1 passed
All tests in CubeCalculatorTest passed

===============================================
Total duration: 17ms
All 1 test suites passed.
===============================================

Note that:

  1. You can repeat the -o test arguments as many times you want to run different test suite at once.
  2. You can use * as part of the test suite name to match more than one suite directly.

Pass test arguments to test

Pass arguments to the test framework after --:

→ bloop test foo -- -h /disk/foo/target/html
CubeCalculatorTest:
- CubeCalculator.cube
Execution took 11ms
1 tests, 1 passed
All tests in CubeCalculatorTest passed

===============================================
Total duration: 11ms
All 1 test suites passed.
===============================================

Test framework arguments are unique per test frameworks and parsed independently. Passing arguments specific to one test framework will work only if:

  1. Your test argument is a system property (-Dkey=value)
  2. Your test project uses only one test framework
  3. You have a project with tests of different test frameworks (e.g. Scalacheck and Scalatest) but you're filtering out tests that are part of only one framework (for example, via --only)

To specify test arguments per test framework, you can add the arguments in your build definition or in the test field of the bloop configuration files.

Test only a specific test case

The mechanism to test only one single case in a suite/specification depends largely on the test framework. A few examples:

  1. JUnit: bloop test foo -o CubeCalculatorTest -- "*a single test case*"
  2. ScalaTest: bloop test foo -o CubeCalculatorTest -- -z "a single test case"
  3. utest: bloop test foo -o CubeCalculatorTest -- "CubeCalculatorTest.a single test case"

Consult the documentation of your test framework to know how to run an exact or fuzzy match for a test.

Pass JVM arguments to test

The configuration file of a project contains the options the tests should be run with.

To add new arguments to a test execution, append -J to every argument after --. These arguments will be passed directly to the forked virtual machine. For example, let's increase the heap size of our tests until 4gb.

→ bloop test foo -- -J-Xms2g -J-Xmx4g
CubeCalculatorTest:
- CubeCalculator.cube
Execution took 11ms
1 tests, 1 passed
All tests in CubeCalculatorTest passed

===============================================
Total duration: 11ms
All 1 test suites passed.
===============================================

To debug our tests, we can use instead -J-agentlib:jdwp=transport=dt_socket,server=y,suspend=y,address=5005.

Run an application

If our application defines one main method, we can run it:

→ bloop run foo
Hello, World!

The bloop run command only accepts a project as a command.

Run a concrete main method

Bloop complains where there is more than one main method defined in your project and you run a vanilla bloop run invocation.

Assuming we duplicate the main entrypoint with a scond CubeCalculator2, bloop emits an error:

→ bloop run foo
[E] Multiple main classes found. Expected a default main class via command-line or in the configuration of project 'foo'
[E] Use the following main classes:
[E]  * CubeCalculator2
[E]  * CubeCalculator

To fix the error, specify the main class you want to run with -m or --main:

→ bloop run foo -m foo.HelloWorld
Hello, World!

Pass JVM arguments to run

Like Pass JVM arguments to test, you can add new arguments to the execution by appending -J to every argument after --. These arguments will be passed directly to the forked virtual machine.

→ bloop run foo -- -J-Xms2g -J-Xmx4g
Hello, World!

To debug an application, pass the following JVM argument instead -J-agentlib:jdwp=transport=dt_socket,server=y,suspend=y,address=5005.

Enable file watching

When using the command-line tool, it's common to run a bloop task with file watching enabled. For example, you can run your tests in a loop with:

→ bloop test foo -w
CubeCalculatorTest:
- CubeCalculator.cube
Execution took 18ms
1 tests, 1 passed
All tests in CubeCalculatorTest passed

===============================================
Total duration: 18ms
All 1 test suites passed.
===============================================
Watching 2 directories... (press Ctrl-C to interrupt)

-w or --watch is available for compile, test and run. You can interrupt file watching by pressing Ctrl + C.

It's not recommended to run -w at the same time you use Bloop via a text editor given that concurrent actions may collide. Such actions will be handled gracefully in future bloop releases.

Run an Ammonite REPL on a project

The console commands runs an Ammonite REPL with a project's classpath.

→ bloop console foo
Compiling foo (1 Scala source)
Compiled foo (290ms)
https://repo1.maven.org/maven2/com/lihaoyi/ammonite_2.12.8/maven-metadata.xml
  No new update since 2019-05-19 04:07:48
Loading...
Compiling (synthetic)/ammonite/predef/interpBridge.sc
Compiling (synthetic)/ammonite/predef/replBridge.sc
Compiling (synthetic)/ammonite/predef/DefaultPredef.sc
Welcome to the Ammonite Repl 1.6.7-1-a44339b
(Scala 2.12.8 Java 1.8.0_202-ea)
If you like Ammonite, please support our development at www.patreon.com/lihaoyi
@

Use --ammonite-version to use a particular version of Ammonite. By default, the latest released version is used.

Compile, test and run several projects

The compile, test and run actions take several projects as arguments which allows you to process several projects in one go. For example, compile foo, bar and baz with:

→ bloop compile foo bar baz
Compiling foo (1 Scala source)
Compiled foo (290ms)
Compiling bar (1 Scala source)
Compiled bar (455ms)
Compiling baz (1 Scala source)
Compiled baz (229ms)

Summary

You've just learned the most basic bloop commands. Learn more commands with bloop --help:

→ bloop --help
bloop 2.0.10
Usage: bloop [options] [command] [command-options]


Available commands: about, autocomplete, bsp, clean, compile, configure, console, help, link, projects, run, test
Type `bloop 'command' --help` for help on an individual command
     
Type `--nailgun-help` for help on the Nailgun CLI tool.

If you're looking for the flags supported by a command such as compile, run bloop compile --help:

→ bloop compile --help
Command: compile
Usage: bloop compile <project>
  --project | -p  <value>
        The project to compile (will be inferred from remaining cli args).
  --incremental  
        Compile the project incrementally. By default, true.
  --pipeline  
        Pipeline the compilation of modules in your build. By default, false.
  --reporter  <value>
        Pick reporter to show compilation messages. By default, bloop's used.
  --watch | -w  
        Run the command when projects' source files change. By default, false.
  --config-dir | -c  <.bloop>
        File path to the bloop config directory, defaults to `.bloop` in the current working directory.
  --version | -v  
        If set, print the about section at the beginning of the execution. Defaults to false.
  --verbose  
        If set, print out debugging information to stderr. Defaults to false.
  --no-color  
        If set, do not color output. Defaults to false.
  --debug  <value>
        Debug the execution of a concrete task.

For a complete overview, visit the CLI --help reference page.

← Integration GuideCLI --help →