daml/BAZEL.md
2022-03-16 23:26:10 +01:00

39 KiB

Bazel User Guide

This document explains how to use the Bazel build system on the Daml repository from a users perspective. I.e. assuming the project you are working on has already been ported to Bazel.

This guide does not cover how to port a new project to Bazel. Please refer to the Bazel JVM Porting Guide if you intend to port a JVM project to Bazel.

Setup

This section goes through the required steps for a basic but fully functioning setup of the Bazel build system for work on the Daml repository. Additional setup as for the IntelliJ integration is listed in its own section below.

Bazel Executable

Bazel is incorporated in the dev-env. If the dev-env is setup correctly and dev-env/bin is in your $PATH, then Bazel should be ready to use.

Build and Test

Once setup is complete, you can build the whole repository with the following command.

bazel build //...

You can run all hermetic tests in the repository with the following command.

bazel test //...

Bazel Core Concepts

If you are unfamiliar with Bazel it is recommended that you read the official Concepts and Terminology guide. Here we will only provide a brief overview which may serve as a refresher.

In short, the daml repository is a Bazel workspace. It contains a WORKSPACE file, which defines external dependencies. The workspace contains several packages. A package is a directory that contains a BUILD.bazel or BUILD file. Each package holds multiple targets. Targets are either files under the package directory or rules defined in the BUILD.bazel file. You can address a target by a label of the form //path/to/package:target. For example, //ledger/sandbox-on-x:sandbox-on-x. Here sandbox-on-x is a target in the package ledger/sandbox-on-x. It is defined in the file ledger/sandbox-on-x/BUILD.bazel using da_scala_library as shown below.

da_scala_library(
    name = "sandbox-on-x",
    srcs = glob(["src/main/scala/**/*.scala"]),
    resources = glob(["src/main/resources/**/*"]),
    scala_deps = [
        "@maven//:com_github_scopt_scopt",
        "@maven//:com_typesafe_akka_akka_actor",
        "@maven//:com_typesafe_akka_akka_stream",
    ],
    tags = ["maven_coordinates=com.daml:sandbox-on-x:__VERSION__"],
    visibility = [
        "//visibility:public",
    ],
    deps = [
        ...list of deps
    ],
)

The arguments to da_scala_library are called attributes. These define the name of the target, the sources it is compiled from, its dependencies, etc. Note, that Bazel build rules are hermetic. I.e. only explicitly declared dependencies will be available during execution. In particular, if a rule depends on additional data files, then these have to be declared dependencies as well. For example using the resources or the data attributes. The details depend on the rule in question.

The following rules are commonly used in this repository. For Scala projects da_scala_library, da_scala_test_suite, da_scala_binary. For Java projects java_library, java_test_suite, java_binary. For Daml projects daml.

Labels can point to a specific target, or to a set of targets using a wild-card. The following wild-card patterns are recognized.

  • Ellipsis (//some/package/...): All rule targets within or underneath some/package.
  • All (//some/package:all): All rule targets within some/package.
  • Star (//some/package:*): All rule or file targets within some/package.

So far we have talked about targets defined in the current workspace. Bazel also has a notion of external workspaces. Targets in external workspaces are labelled as @workspace_name//path/to/package:target.

Targets have a visibility attribute that determines which other targets can depend on it. Targets can be

  • private (//visibility:private) Only targets in the same package can depend on it.
  • visible to specific packages (//some/package:__pkg__) Only targets within //some/package can depend on it.
  • visible to sub-packages (//some/package:__subpackages__) Only targets within or underneath //some/package can depend on it.
  • public (//visibility:public) Any target in any package can depend on it.

Visibility should be kept as strict as possible to help maintain a clean dependency graph.

Bazel files are written in a language called Starlark. It is very similar to Python. However, Starlark programs cannot perform arbitrary input and output, and build files are not allowed to use control structures (for, if, etc.), or define functions. These restrictions are in place to ensure hermeticity.

IntelliJ Integration

Make sure to go through the Bazel setup section and to familiarize yourself with Bazel's core concepts as explained in the sections above, before you proceed to the IntelliJ integration.

Setup

If you use the IntelliJ IDE you should install the Bazel integration plugin provided by Google.

Not every version of Intellij works with the Bazel plugin. Intellij IDEA 2021.2.4 has been verified to work correctly with the Bazel plugin. It is advisable to use the JetBrains Toolbox to manage installations of Intellij IDEA.

Follow the installation instructions in the official documentation. In short: Install the plugin from within the IDE (Settings > Plugins > Marketplace, and search for 'Bazel'). Multiple Bazel plugins exist, make sure to select the Bazel plugin referencing ij.bazel.build.

If the correct plugin does not exist in the list, then your IntelliJ version might be too recent, and the Bazel plugin might not have been upgraded to support it, yet. Check for version compatibility on the JetBrains plugin page.

Set the Bazel binary in the Bazel plugin settings, at Preferences -> Bazel Settings -> Bazel binary location to point to the Bazel binary in the dev-env/bin directory.

Side note: code formatting

While not relevant to setup IntelliJ to work with Bazel, if you plan to work on Java code you are also advised to install the google-java-format plugin. This plugin integrates the IntelliJ code formatting actions with the Google Java Style Guide, which is enforced on CI with the google-java-format tool.

Importing a project

To import a Bazel project into IntelliJ select "Import Bazel Project" in the welcome dialog, or File > Import Bazel Project in the editor window. In the import dialog under "Workspace:" enter the path to the Daml repository root.

The Bazel IntelliJ integration uses a project view file to define the list of directories and targets to make accessible in IntelliJ and to control other aspects of the project. Refer to the official documentation for a detailed description.

Choose the "Generate from BUILD file" option and select the BUILD.bazel file of the project that you will be working on. Then, click on "Next".

The following dialog allows you to define the project name, or infer it, and to set the location of the project data directory. It also allows you to modify the default project view file. The default should have the following structure:

directories:
  .

# Automatically includes all relevant targets under the 'directories' above
derive_targets_from_directories: true

targets:
  # If source code isn't resolving, add additional targets that compile it here

additional_languages:
  # Uncomment any additional languages you want supported
  # ...

Make sure to add Scala, or other languages that you require, to the additional_languages section. The section will be pre-populated with a list of comments specifying the automatically detected supported languages.

Also ensure that you install and setup the Scala plugin. If you'd like to work with all directories, we recommend the following project view configuration, which stops IntelliJ from indexing the Bazel cache, and avoids rebuilding the documentation:

directories:
  .
  -docs

# Automatically includes all relevant targets under the 'directories' above
derive_targets_from_directories: true

targets:
  # If source code isn't resolving, add additional targets that compile it here

additional_languages:
  javascript
  python
  scala
  typescript

However, you can also provide an allowed list of directories for a faster experience.

If you wish to define a specific set of targets to work, then you can list these in the targets section. This is not usually necessary, as they will be derived automatically.

If you choose to limit the directories, you might end up with a project view file looking like this:

directories:
  ledger/sandbox
  # ...

# Automatically includes all relevant targets under the 'directories' above
derive_targets_from_directories: true

targets:
  # If source code isn't resolving, add additional targets that compile it here

additional_languages:
  scala

Click "Next" once you are ready. You will be able to modify the project view file whenever you like, so don't worry too much.

The first import of the project might fail due to a resolution error of the bazel binary. In order to solve this, configure the Bazel plugin settings with the location of the bazel binary, by setting PreferencesBazel SettingsBazel binary location to ./dev-env/bin/bazel.

Now, re-trigger a sync of the workspace (IntelliJ Action: Sync project with BUILD files). This process will take a while.

Configuring the JDK in IntelliJ

Daml downloads the version of the JDK it uses from Nix. A symlink will be created by the dev-env utilities (make sure you've set these up) in dev-env/jdk.

TO configure IntelliJ to use this JDK:

  1. Open the Project Structure window.
  2. Under Platform Settings, select SDKs.
  3. Press the plus button and select "Add JDK".
  4. Choose the dev-env/jdk directory.
  5. Name it "Daml JDK" or something similar.
  6. Ensure there's sources attached under the Sourcepath tab. If not, add them. Press the plus button and select dev-env/jdk/lib/openjdk/src.zip.
  7. Open Project SettingsProject.
  8. Select the Daml JDK from the Project SDK list.

Overview over Bazel IntelliJ Integration

The IntelliJ interface should largely look the same as under SBT. However, the main menu will have an additional entry for Bazel, and a Bazel toolbar is provided for quick access to common tasks.

The most commonly required operations are described below. Refer to the plugin documentation for further information.

Sync Project

If you modified a project BUILD.bazel file, or the project view file, then click the "Sync Project with BUILD Files" button in the Bazel toolbar, or the Sync > Sync Project with BUILD Files entry in the Bazel menu to synchronize IntelliJ with those changes.

Modify Project View

Click Project > Open Local Project View File in the Bazel menu to open and modify the current project view file. You may need to sync the project for your changes to take effect.

Build the Project

Click Build > Compile Project in the Bazel menu to build the whole project. Click Build > Compile "CURRENT FILE" to compile only the current file.

Run or Debug an Executable or Test

Click on the drop-down menu in the Bazel tool bar and select the entry Bazel run <your target> or Bazel test <your target>. If the executable or test you wish to run or debug is not in the list then follow instructions on adding a run configuration below first. Come back here when ready.

The selected entry is a run configuration. Click the green arrow in the Bazel toolbar to run the executable or test. Click the green bug icon to debug the executable or test.

Adding a Run Configuration

If you wish to add an executable or test to the run configurations, then the simplest and most consistent way is to add the target to the project view file and sync the project.

Otherwise, click on the drop-down menu in the Bazel tool bar and select "Edit Configurations...". Click the "+" in the upper left corner and select "Bazel Command" from the list. This will add a new entry to the "Bazel Command" sub-tree. Fill in the fields in the pane on the right side to define the run configuration: Give a suitable name, select the Bazel target, and choose the Bazel command (run, test, etc.). If applicable, you can also define Bazel command-line flags or command-line flags to the executable. Click on "Apply", or "OK" to add the run configuration.

Attaching sources to scala library

If you do not have the Scala library sources linked (you only see the decompiled sources), you can attach it manually by selecting the Choose sources... button on the yellow bar at the top, and selecting scala-library...-src.jar.

Known Issues

Missing folders in project tree

The "Project" pane contains a tree-view of the folders in the project. The "Compact Middle Packages" feature can cause intermediate folders to disappear, only showing their children in the tree-view. The workaround is to disable the feature by clicking on the gear icon in the "Project" pane and unchecking "Compact Middle Packages". Refer to the issue tracker for details.

Rerun failed tests is broken

IntelliJ allows to rerun only a single failed test-case by the click of a button. Unfortunately, this feature does not work with the Bazel plugin on Scala test-cases. Please refer to the issue tracker for details.

Troubleshooting

'tools.dade-exec-nix-tool' not found

If you get the error

error: attribute 'dade-exec-nix-tool' in selection path 'tools.dade-exec-nix-tool' not found

in the bazel console during project import, make sure that PreferencesBazel SettingsBazel binary location points to ./dev-env/bin/bazel rather than to ./dev-env/lib/dade-exec-nix-tool (as Intellij might have expanded the former to the latter).
If that doesn't help try starting IntelliJ from the root of the daml repository by calling idea ..

Bazel Command Reference

The following sections briefly list Bazel commands for the most common use-cases. Refer to the official Bazel documentation for more detailed information.

Building Targets

  • Build all targets

    bazel build //...
    
  • Build an individual target

    bazel build //ledger/sandbox-on-x:app
    

Running Tests

  • Execute all tests

    bazel test //...
    
  • Execute a test suite

    bazel test //ledger/sandbox-on-x:sandbox-on-x-unit-tests
    
  • Show test output

    bazel test //ledger/sandbox-on-x:sandbox-on-x-unit-tests --test_output=streamed
    

    Test outputs are also available in log files underneath the convenience symlink bazel-testlogs or bazel-out/*/testlogs.

  • Do not cache test results

    bazel test //ledger/sandbox-on-x:sandbox-on-x-unit-tests --nocache_test_results
    
  • Execute a specific Scala test-suite class

    bazel test //ledger/participant-integration-api:participant-integration-api-tests_test_suite_src_test_suite_scala_platform_store_dao_JdbcLedgerDaoPostgresqlSpec.scala
    
  • Execute a test with a specific name

    bazel test \
    //ledger/participant-integration-api:participant-integration-api-tests_test_suite_src_test_suite_scala_platform_store_dao_JdbcLedgerDaoPostgresqlSpec.scala \
      --test_arg=-t \
      --test_arg="JdbcLedgerDao (divulgence) should preserve divulged contracts"
    
  • Execute tests that include a string in their name

    bazel test \
    //ledger/participant-integration-api:participant-integration-api-tests_test_suite_src_test_suite_scala_platform_store_dao_JdbcLedgerDaoPostgresqlSpec.scala \
      --test_arg=-z \
      --test_arg="preserve divulged"
    

    More broadly, for Scala tests you can pass through any of the args outlined in http://www.scalatest.org/user_guide/using_the_runner, separating into two instances of the --test-arg parameter as shown in the two examples above.

Running Executables

  • Run an executable target

    bazel run //ledger/sandbox-on-x:app
    
  • Pass arguments to an executable target

    bazel run //ledger/sandbox-on-x:app -- --help
    

Running a REPL

  • Run a Scala REPL which has the library on the classpath:

    bazel run //ledger/ledger-on-memory:ledger-on-memory_repl
    

Querying Targets

The Bazel query language is described in detail in the official Bazel documentation. This section will list a few common use-cases. Filters like filter or kind accept regular expressions. Query expressions can be combined using set operations like intersect or union.

  • List all targets underneath a directory

    bazel query //ledger/...
    
  • List all library targets underneath a directory

    bazel query 'kind("library rule", //ledger/...)'
    
  • List all Scala library targets underneath a directory

    bazel query 'kind("scala.*library rule", //ledger/...)'
    
  • List all test-suites underneath a directory

    bazel query 'kind("test_suite", //ledger/...)'
    
  • List all test-cases underneath a directory

    bazel query 'tests(//ledger/...)'
    
  • List all Java test-cases underneath a directory

    bazel query 'kind("java", tests(//ledger/...))'
    
  • List all Scala library dependencies of a target

    bazel query 'kind("scala.*library rule", deps(//ledger/sandbox-on-x:app))'
    
  • Find available 3rd party dependencies

    bazel query 'attr(visibility, public, filter(".*scalaz.*", //3rdparty/...))'
    

Dependency Graphs

Bazel queries can also output dependency graphs between the targets that the query includes. These can then be rendered using Graphviz.

  • Graph all Scala library dependencies of a target

    bazel query --noimplicit_deps 'kind(scala_library, deps(//ledger/sandbox-on-x:app))' --output graph > graph.in
    dot -Tpng < graph.in > graph.png
    

    The --noimplicit_deps flag excludes dependencies that are not explicitly listed in the BUILD file, but that are added by Bazel implicitly, e.g. the unused dependency checker added by rules_scala.

Help

  • List available commands

    bazel help
    
  • Show help on a Bazel command

    bazel help build
    
  • Show details on each option

    bazel help build --long
    

Continuous Build

By continuous build we mean the ability to watch the repository for source file changes and rebuild or rerun targets when any relevant files change. The dev-env provides the tool ibazel for that purpose. Similar to Bazel it can be called with the commands build, test, or run on a specific target. It will perform the command and determine a list of relevant source files. Then, it will watch these files for changes and rerun the command on file change. For example:

ibazel test //ledger/sandbox-on-x:sandbox-on-x-unit-tests

Note, that this interacts well with Bazel's test result caching (which is activated by default). In the above example the outcome of tests whose sources didn't change will already be cached by Bazel and the tests won't be repeated.

Refer to the project README for more information.

Haskell in Bazel

In Bazel terminology, a directory containing a BUILD.bazel file is called a "package". Packages contain targets and BUILD.bazel files are where targets are defined. Mostly we are concerned in our BUILD.bazel files with writing rules to produce specific Haskell derived files (artifacts) from Haskell source file inputs. Of these, most are libraries, some are executables and some are tests.

For Haskell, most BUILD.bazel files begin with a variation on the following:

load( "//bazel_tools:haskell.bzl",
      "da_haskell_library", "da_haskell_executable","da_haskell_test" )

This directive loads from the //bazel_tools package, the rules da_haskell_library for building libraries, da_haskell_binary for building executables and da_haskell_test for building test-suites. The da_* rules are DA specific overrides of the upstream rules_haskell rules. Their API docs can be found in //bazel_tools/haskell.bzl, or by executing the bazel-api-docs tool from dev-env. They mostly behave like the upstream rules, just adding some defaults, and adding a hackage_deps attribute (more on this below) for convenience.

Library : da_haskell_library

One specific library in the daml-foundations stack is daml-ghc-compiler. Here's a synopsis of its definition.

da_haskell_library(
    name = "daml-ghc-compiler",
    srcs = glob([
        "src/**/*.hs",
    ]),
    src_strip_prefix = "src",
    deps = [
        "//compiler/daml-lf-ast",
        "//compiler/daml-lf-proto",
        ...
    ],
    hackage_deps = [
        "base",
        "bytestring",
        ...
    ],
    visibility = ["//visibility:public"],
)

To build this single target from the root of the Daml repository, the command would be:

bazel build //compiler/damlc/daml-compiler

since the BUILD.bazel that defines the target is in the compiler/damlc sub-folder of the root of the DA repository and the target name is damlc.

Let's break this definition down:

  • name: A unique name for the target;
  • srcs: A list of Haskell source files;
  • src_stip_prefix: Directory in which the module hierarchy starts;
  • deps: A list of in-house Haskell or C library dependencies to be linked into the target;
  • hackage_deps: A list of external Haskell (Hackage) libraries to be linked into the target;
  • visibility: Define whether depending on this target by others is permissible.

Note the use of the Bazel glob function to define the srcs argument allowing us to avoid having to enumerate all source files explicitly. The ** part of the shown glob expression is Bazel syntax for any sub-path. Read more about glob here.

The deps argument in the above invocation can be interpreted as linking the libraries defined by the list of targets provided on the right hand side (when we say "package" here we mean Haskell package - i.e. library):

  • //:ghc-lib is the ghc-lib package defined in the root BUILD file,
  • //compiler/daml-lf-ast is the daml-lf-ast package defined in the daml-foundations/daml-compiler/BUILD.bazel file (that is, //compiler/daml-lf-ast is shorthand for //compiler/daml-lf-ast:daml-lf-ast)
  • Similarly, //nix/third-party/proto3-suite is the Haskell library //nix/third-party/proto3-suite:proto3-suite defined in the file nix/third-party/proto3-suite/BUILD.bazel.

The hackage_deps argument details those Haskell packages (from Hackage) that the daml-ghc-compiler target depends upon. In this case that is base, bytestring and some other packages not shown. Finally, visibility is set to public so no errors will result should another target attempt to link daml-ghc-compiler. [Note : Public visibility means that any other target from anywhere can depend on the target. To keep the dependency graph sane, its a good idea to keep visibility restrictive. See here for more detail.]

Executable : da_haskell_binary

Here's the synopsis of the rule for the executable daml-ghc:

da_haskell_binary (
  name = "daml-ghc",
  srcs = glob (["src/DA/Cli/**/*.hs", "src/DA/Test/**/*.hs"])
  src_strip_prefix = "DA",
  main_function = "DA.Cli.GHC.Run.main",
  hackage_deps = [ "base", "time", ...],
  data = [
    "//compiler/damlc/pkg-db"
    , ...
  ],
  deps = [
      ":daml-ghc-compiler"
    , "//:ghc-lib"
    , "//compiler/daml-lf-ast"
    , ...
  ]
  , visibility = ["//visibility:public"]
)

Haskell binaries require a definition of the distinguished function main. The main_function argument allows us to express the qualified module path to the definition of main to use.

The data argument in essence is a list of files needed by the target at runtime. Consult this documentation for more detail. The targets that are given on the right-hand-side are (long-hand) labels for "file-groups". Here's the one for daml-stdlib-src for example.

filegroup(
  name = "daml-stdlib-src",
  srcs = glob(["daml-stdlib/**"]),
  visibility = ["//visibility:public"]
)

Having looked at deps in the context of haskell_library there's not much more to say except note the :daml-ghc-compiler syntax for the depedency on daml-ghc-compiler. That is, targets defined in the same BUILD.bazel as the target being defined can be referred to by preceding their names with :.

Test : da_haskell_test

For an example of a test target, we turn to //libs-haskell/da-hs-base:da-hs-base-tests:

da_haskell_test(
    name = "da-hs-base-tests",
    src_strip_prefix = "src-tests",
    srcs = glob(["src-tests/**/*.hs"]),
    deps = [
        ":da-hs-base",
    ],
    visibility = ["//visibility:public"],
)

There is nothing new in the above to expound upon here! How might you invoke that single target? Simple as this:

bazel test "//libs-haskell/da-hs-base:da-hs-base-tests"

More comprehensive documentation on the bazel command can be found here.

Beyond defining targets

If your work goes beyond simply adding targets to existing BUILD.bazel files and involves things like defining toolchains and external dependencies, then this document is for you!

Scala in Bazel

In this section we will provide an overview of how Scala targets are defined in Bazel in this repository. This should provide enough information for most everyday development tasks. For more information refer to the Bazel porting guide for JVM developers (to be written as of now). For a general reference to BUILD.bazel file syntax refer to the official documentation. Note, that BUILD.bazel and BUILD are both valid file names. However, BUILD.bazel is the preferred spelling. BUILD is the older spelling and still to be found in large parts of the documentation.

Bazel targets are defined in BUILD.bazel files. For example //ledger-client/ods:ods is defined in ledger-client/ods/BUILD.bazel. First, we import the required rules and macros. For example, the following loads the da_scala_library macro defined in //bazel_tools:scala.bzl, and the daml rule defined in ledger-client/daml.bzl. The distinction of rules and macros is not important here.

load('//bazel_tools:scala.bzl', 'da_scala_library')
load('//rules_daml:daml.bzl', 'daml')

Scala Libraries

The macro da_scala_library is a convenience function that defines a Scala library and sets common compiler flags, plugins, etc. To explain it we will take an example instance and describe the individual attributes. For details refer to the rules_scala project README. For a Java rules refer to the official Bazel documentation.

da_scala_library(

  # Set the target name to 'ods'.
  name = 'ods',

  # Mark this target as public.
  # I.e. targets in other package can depend on it.
  visibility = ['//visibility:public'],

  # Define the target's source files by globbing.
  # The details of file globbing are explained here:
  # https://docs.bazel.build/versions/master/be/functions.html#glob
  srcs = glob(['src/main/**/*.scala'], exclude = [...]),

  # Define the target's resources by globbing.
  resources = glob(['src/main/resources/**/*']),

  # Define the target's dependencies.
  # These will appear in the compile-time classpath.
  # And the transient closure of `deps`, `runtime_deps`, and `exports`
  # will appear in the runtime classpath.
  deps = [
    # A third party dependency.
    '//3rdparty/jvm/ch/qos/logback:logback_classic',
    # A dependency in the same workspace.
    '//ledger-client/nanobot-framework',
    # A dependency in the same package.
    ':ods-macro'
    ...
  ],

  # Define the target's runtime dependencies.
  # These will appear only in the runtime classpath.
  runtime_deps = [...],

  # List of exported targets.
  # E.g. if something depends on ':ods', it will also depend on ':ods-macro'.
  exports = [':ods-macro'],

  # Scalac compiler plugins to use for this target.
  # Note, that these have to be specified as JAR targets.
  # I.e. you cannot use `//3rdparty/jvm/org/scalameta/paradise_2_12_6` here.
  plugins = [
    '//external:jar/org/scalameta/paradise_2_12_6',
  ],

  # Scalac compiler options to use for this target.
  scalacopts = ['-Xplugin-require:macroparadise'],

  # JVM flags to pass to the Scalac compiler.
  scalac_jvm_flags = ['-Xmx2G'],
)

Scala Executables

Scala executables are defined using da_scala_binary. It takes most of the same attributes that da_scala_library takes. Notable additional attributes are:

  • main_class: Name of the class defining the entry-point main().
  • jvm_flags: Flags to pass to the JVM at runtime.
  • data: Files that are needed at runtime. In order to access such files at runtime you should use the utility library in com.daml.testing.BuildSystemSupport.

Scala Test Cases

Scala test-suites are defined using da_scala_test_suite. It takes most of the same attributes as da_scala_binary.

Note, that this macro will create one target for every single source file specified in srcs. The advantage is that separate test-cases can be addressed as separate Bazel targets and Bazel's test-result caching can be applied more beneficially. However, this means that test-suite source files may not depend on each other.

A single Scala test-cases, potentially consisting of multiple source files, can be defined using da_scala_test. It is preferable to always use da_scala_test_suite, and define separate testing utility libraries using da_scala_library if test-cases depend on utility modules.

Scala JMH Benchmarks

Scala benchmarks based on the JMH toolkit can be defined using the scala_benchmark_jmh macro provided by rules_scala. It supports a restricted subset of the attributes of da_scala_binary, namely: name, deps, srcs, scalacopts, resources and resource_jars.

The end result of building the benchmark is a Scala binary of the same name, which can be executed with bazel run.

Java and Scala Deployment

Bazel's builtin Java rules and rules_scala will automatically generate a fat JAR suitable for deployment for all your Java and Scala targets. For example, if you defined a Scala executable target called foo, then Bazel will generate the target foo_deploy.jar next to the regular foo.jar target. Building the foo_deploy.jar target will generate a self-contained fat JAR suitable to be passed to java -jar.

Scaladoc

We generate scaladoc targets for each da_scala_library with a _scaladoc name suffix. All scaladoc targets have a scaladoc tag which allows you to filter them out to speed up your builds. You can either do that on the command line by running e.g.

bazel build --build_tag_filters=-scaladoc //...

or if you want to never build them locally you can set this in your .bazelrc.local:

build --build_tag_filters=-scaladoc

You can still build an individual target by selecting it explicitly, e.g.

bazel build //daml-lf/validation:validation_scaladoc

Daml Packages

Daml package targets are defined using the daml rule loaded from //rules_daml:daml.bzl. To explain it we will take an example instance and describe the individual attributes.

daml(
  name = "it-daml",
  # The main Daml file. This file will be passed to damlc.
  main_src = "src/it/resources/TestAll.daml",
  # Other Daml files that may be imported by the main Daml file.
  srcs = glob(["src/it/resources/**/*.daml"]),
  # The directory prefix under which to create the DAR tree.
  target_dir = "target/scala-2.12/resource_managed/it/dars",
  # The group ID.
  group = "com.daml.sample",
  # The artifact ID.
  artifact = "test-all",
  # The package version.
  version = "0.1",
  # The package name.
  package = "com.daml.sample",
)

This will compile and package the Daml code into a DAR file under the following target, where <group-dir> is the group attribute with . replaced by /.

:<target_dir>/<group-dir>/<artifact>/<version>/<artifact>-<version>.dar,

For example:

:target/scala-2.12/resource_managed/it/dars/com/digitalasset/sample/test-all/0.1/test-all-0.1.dar

POM and SHA files will be stored in the same directory.

Additionally, this will perform Scala code generation and bundle the generated Scala modules into a source JAR available under the following target.

<name>.srcjar

For example:

it-daml.srcjar

Daml Executables

The rule daml_binary is provided to generate executable targets that execute the Daml sandbox on a given DAR package. For example:

daml_binary(
  name = "ping-pong-exec",
  dar = ':target/repository/.../PingPong-0.1.dar',
)

Such a target can then be executed as follows, where arguments after -- are passed to the Daml sandbox.

bazel run //ledger-client/nanobot-sample-app:ping-pong-exec -- --help

External Java and Scala Dependencies

External dependencies are these that are not defined and built within the local workspace, but are defined in an external workspace in some way. The most common case are Maven JAR dependencies which are fetched from Artifactory.

We distinguish direct and transitive dependencies. Direct dependencies are explicitly defined on targets in the local workspace. Most commonly on the deps, runtime_deps, exports, or plugins attributes. Transitive dependencies are introduced implicitly through direct dependencies, most commonly on another dependency's exports attribute.

All direct Scala and Java dependencies are listed explicitly in the file bazel-java-deps.bzl. Each dependency is defined by its Maven coordinates. The maven_install repository rule calls Coursier to perform transitive dependency resolution and import the required artifacts into the Bazel build.

The resolved versions are pinned in the file maven_install_2.13.json. Execute bazel run @unpinned_maven//:pin when you wish to update or add a new dependency. See rules_jvm_external for details.

Typescript in Bazel

We are using rules_typescript to build typescript projects. It works in conjunction with rules_nodejs to provide access to npm packages.

Please refer to the documentation in the above url for usage. For an example, please see compiler/daml-extension/BUILD.bazel.

Protocol buffers in Bazel

We use protocol buffers for Daml-LF and the Ledger API. The Daml-LF protocol buffer build rules can be found from //daml-lf/archive/BUILD.bazel. It produces bindings for Java and Haskell (via proto3-suite).

Bazel provides built-in rules for protocol buffer bindings for Java and C++. See the following resources for more information on its usage: Protocol Buffer Rules Blog post: Protocol Buffers in Bazel

The rules for haskell are currently ad-hoc genrules and use the proto3-suite's compile-proto-file program directly. Please refer to //daml-lf/archive/BUILD.bazel for example usage. If you find yourself writing similar rules, please take a moment to write some Starlark to abstract it out and document it here. Note that proto3-suite isn't compatible with protoc, so it is not currently possible to hook it up into the "proto_library" tooling.

Known issues

Unchanged Haskell library being rebuilt

Unfortunately, GHC builds are not deterministic. This, coupled with the way Bazel works, may lead to Haskell libraries that have not been changed to be rebuilt. If the library sits at the base of the dependency graph, it may cause a ripple effect that forces you to rebuild most of the workspace without an actual need for it (ghc-lib is one example of this).

To work around this issue you can clean the local and build cache, making sure you are fetching the GHC build artifacts from remote:

bazel clean --expunge # clean the build cache
rm -r .bazel-cache    # clean the local cache

This will also mean that changes made locally will need to be rebuilt, but it's likely that this will still result in a net positive gain on your build time.

If you are still rebuilding after this, you probably also have a poisoned Nix cache. To clear that run through the following steps:

bazel clean --expunge # clean the build cache
rm -r .bazel-cache    # clean the local cache
rm dev-env/var/gc-roots/* # Remove dev-env GC roots
rm result* # Remove GC roots you might have from previous nix-build invocations.
nix-store --gc --print-roots # View all garbage collection roots
# Verify that there is nothing from our repo or some Bazel cache.
# If you are not sure ask in #team-daml
nix-store --gc # Run garbage collection
nix-build nix -A tools -A cached --no-out-link # Build the nix derivations (they should be fetched from the cache)
bazel build //... # You should now see things being fetched from the cache

Working in environments with low or intermittent connectivity

Bazel tries to leverage the remote cache to speed up the build process but this can turn out to work against you if you are working in an environment with low or intermittent connectivity. To disable fetching from the remote cache in such scenario, you can use the --noremote_accept_cached option.