bazel-diff

bazel-diff is a command line tool for Bazel projects that allows users to determine the exact affected set of impacted targets between two Git revisions. Using this set, users can test or build the exact modified set of targets.

bazel-diff offers several key advantages over rolling your own target diffing solution

1. bazel-diff is designed for very large Bazel projects. We use Java Protobuf's parseDelimitedFrom method alongside Bazel Query's streamed_proto output option. These two together allow you to parse Gigabyte or larger protobuf messages. We have tested it with projects containing tens of thousands of targets.
2. We avoid usage of large command line query lists when interacting with Bazel, issue here. When you interact with Bazel with thousands of query parameters you can reach an upper maximum limit, seeing this error bash: /usr/local/bin/bazel: Argument list too long. bazel-diff is smart enough to avoid these errors.
3. bazel-diff has been tested with file renames, deletions, and modifications. Works on bzl files, WORKSPACE files, BUILD files and regular files

Track the feature request for target diffing in Bazel here

This approach was inspired by the following BazelConf talk by Benjamin Peterson.

There are simpler and faster ways to approximate the affected set of targets. However an incorrect solution can result in a system you can't trust, because tests could be broken at a commit where you didn't select to run them. Then you can't rely on green-to-red (or red-to-green) transitions and lose much of the value from your CI system as breakages can be discovered later on unrelated commits.

Prerequisites

• Git
• Bazel 3.3.0 or higher
• Java 8 JDK or higher (Bazel requires this)

Getting Started

To start using bazel-diff immediately, simply clone down the repo and then run the example shell script:

git clone https://github.com/Tinder/bazel-diff.git
cd bazel-diff
./bazel-diff-example.sh WORKSPACE_PATH BAZEL_PATH START_GIT_REVISION END_GIT_REVISION

Here is a breakdown of those arguments:

• WORKSPACE_PATH: Path to directory containing your WORKSPACE file in your Bazel project.
  • Note: Your project must use Git for bazel-diff to work!
• BAZEL_PATH: Path to your Bazel executable
• START_GIT_REVISION: Starting Git Branch or SHA for your desired commit range
• END_GIT_REVISION: Final Git Branch or SHA for your desired commit range

You can see the example shell script in action below:

Open bazel-diff-example.sh to see how this is implemented. This is purely an example use-case, but it is a great starting point to using bazel-diff.

How it works

bazel-diff works as follows

• The previous revision is checked out, then we run generate-hashes. This gives us the hashmap representation for the entire Bazel graph, then we write this JSON to a file.

• Next we checkout the initial revision, then we run generate-hashes and write that JSON to a file. Now we have our final hashmap representation for the Bazel graph.

• We run bazel-diff on the starting and final JSON hash filepaths to get our impacted set of targets. This impacted set of targets is written to a file.

Build Graph Distance Metrics

bazel-diff can optionally compute build graph distance metrics between two revisions. This is useful for understanding the impact of a change on the build graph. Directly impacted targets are targets that have had their rule attributes or source file dependencies changed. Indirectly impacted targets are that are impacted only due to a change in one of their target dependencies.

For each target, the following metrics are computed:

• target_distance: The number of dependency hops that it takes to get from an impacted target to a directly impacted target.
• package_distance: The number of dependency hops that cross a package boundary to get from an impacted target to a directly impacted target.

Build graph distance metrics can be used by downstream tools to power features such as:

• Only running sanitizers on impacted tests that are in the same package as a directly impacted target.
• Only running large-sized tests that are within a few package hops of a directly impacted target.
• Only running computationally expensive jobs when an impacted target is within a certain distance of a directly impacted target.

To enable this feature, you must generate a dependency mapping on your final revision when computing hashes, then pass it into the get-impacted-targets command.

git checkout BASE_REV
bazel-diff generate-hashes [...]

git checkout FINAL_REV
bazel-diff generate-hashes --depEdgesFile deps.json [...]

bazel-diff get-impacted-targets --depEdgesFile deps.json [...]

This will produce an impacted targets json list with target label, target distance, and package distance:

[
  {"label": "//foo:bar", "targetDistance": 0, "packageDistance": 0},
  {"label": "//foo:baz", "targetDistance": 1, "packageDistance": 0},
  {"label": "//bar:qux", "targetDistance": 1, "packageDistance": 1}
]

CLI Interface

bazel-diff Command

Usage: bazel-diff [-hvV] [COMMAND]
Writes to a file the impacted targets between two Bazel graph JSON files
  -h, --help      Show this help message and exit.
  -v, --verbose   Display query string, missing files and elapsed time
  -V, --version   Print version information and exit.
Commands:
  generate-hashes       Writes to a file the SHA256 hashes for each Bazel
                          Target in the provided workspace.
  get-impacted-targets  Command-line utility to analyze the state of the bazel
                          build graph

generate-hashes command

Usage: bazel-diff generate-hashes [-hkvV] [--[no-]useCquery] [-b=<bazelPath>]
                                  [--[no-]excludeExternalTargets]
                                  [--[no-]includeTargetType]
                                  [--contentHashPath=<contentHashPath>]
                                  [-s=<seedFilepaths>] -w=<workspacePath>
                                  [-co=<bazelCommandOptions>]...
                                  [--cqueryCommandOptions=<cqueryCommandOptions>
                                  ]...
                                  [--fineGrainedHashExternalRepos=<fineGrainedHa
                                  shExternalRepos>]...
                                  [-so=<bazelStartupOptions>]... <outputPath>
Writes to a file the SHA256 hashes for each Bazel Target in the provided
workspace.
      <outputPath>        The filepath to write the resulting JSON to.
                            If not specified, the JSON will be written to STDOUT.

                            By default the JSON schema is a dictionary of target => SHA-256 values.
                            Example:
                                 {
                                    "//cli:bazel-diff_deploy.jar":  "4ae310f8ad2bc728934e3509b6102ca658e828b9cd668f79990e95c6663f9633"
                                    ...
                                 }

                            If --includeTargetType is specified, the JSON schema will include the target type (SourceFile/Rule/GeneratedFile)
                            Example:
                                {
                                  "//cli:src/test/resources/fixture/integration-test-1.zip": "SourceFile#c259eba8539f4c14e4536c61975457c2990e090067893f4a2981e7bb5f4ef4cf",
                                  "//external:android_gmaven_r8": "Rule#795f583449a40814c05e1cc5d833002afed8d12bce5b835579c7f139c2462d61",
                                  "//cli:bazel-diff_deploy.jar": "GeneratedFile#4ae310f8ad2bc728934e3509b6102ca658e828b9cd668f79990e95c6663f9633",
                                  ...
                                }
      --[no-]excludeExternalTargets
                          If true, exclude external targets. This must be
                            switched on when using `--noenable_workspace` Bazel
                            command line option. Defaults to `false`.
      --[no-]includeTargetType
                          Whether include target type in the generated JSON or not.
                            If false, the generate JSON schema is: {"<target>": "<sha256>"}
                            If true, the generate JSON schema is: {"<target>": "<type>#<sha256>"
      -tt, --targetType=<targetType>
                          The type of targets to filter, available options are SourceFile/Rule/GeneratedFile
                          Only works if the JSON was generated with `--includeTargetType` enabled.
                          If not specified, all types of impacted targets will be returned.
      -b, --bazelPath=<bazelPath>
                          Path to Bazel binary. If not specified, the Bazel
                            binary available in PATH will be used.
      -co, --bazelCommandOptions=<bazelCommandOptions>
                          Additional space separated Bazel command options used
                            when invoking `bazel query`
      --contentHashPath=<contentHashPath>
                          Path to content hash json file. It's a map which maps
                            relative file path from workspace path to its
                            content hash. Files in this map will skip content
                            hashing and use provided value
      --cqueryCommandOptions=<cqueryCommandOptions>
                          Additional space separated Bazel command options used
                            when invoking `bazel cquery`. This flag is has no
                            effect if `--useCquery`is false.
      --fineGrainedHashExternalRepos=<fineGrainedHashExternalRepos>
                          Comma separate list of external repos in which
                            fine-grained hashes are computed for the targets.
                            By default, external repos are treated as an opaque
                            blob. If an external repo is specified here,
                            bazel-diff instead computes the hash for individual
                            targets. For example, one wants to specify `maven`
                            here if they user rules_jvm_external so that
                            individual third party dependency change won't
                            invalidate all targets in the mono repo. Note that
                            when `--useCquery` is used, the canonical repo name
                            must be provided but with single `@`, e.g.
                            `@rules_jvm_external~~maven~maven`
  -h, --help              Show this help message and exit.
      --ignoredRuleHashingAttributes=<ignoredRuleHashingAttributes>
                          Attributes that should be ignored when hashing rule
                            targets.
  -k, --[no-]keep_going   This flag controls if `bazel query` will be executed
                            with the `--keep_going` flag or not. Disabling this
                            flag allows you to catch configuration issues in
                            your Bazel graph, but may not work for some Bazel
                            setups. Defaults to `true`
  -s, --seed-filepaths=<seedFilepaths>
                          A text file containing a newline separated list of
                            filepaths, each of these filepaths will be read and
                            used as a seed for all targets.
      -so, --bazelStartupOptions=<bazelStartupOptions>
                          Additional space separated Bazel client startup
                            options used when invoking Bazel
      --[no-]useCquery    If true, use cquery instead of query when generating
                            dependency graphs. Using cquery would yield more
                            accurate build graph at the cost of slower query
                            execution. When this is set, one usually also wants
                            to set `--cqueryCommandOptions` to specify a
                            targeting platform. Note that this flag only works
                            with Bazel 6.2.0 or above because lower versions
                            does not support `--query_file` flag.
  -v, --verbose           Display query string, missing files and elapsed time
  -V, --version           Print version information and exit.
  -w, --workspacePath=<workspacePath>
                          Path to Bazel workspace directory.

Note: --useCquery flag may not work with very large repos due to limitation of Bazel. You may want to fallback to use normal query mode in that case. See bazelbuild/bazel#17743 for more details.

What does the SHA256 value of generate-hashes represent?

generate-hashes is a canonical SHA256 value representing all attributes and inputs into a target. These inputs are the summation of the rule implementation hash, the SHA256 value for every attribute of the rule and then the summation of the SHA256 value for all rule_inputs using the same exact algorithm. For source_file inputs the content of the file are converted into a SHA256 value.

get-impacted-targets command

Usage: bazel-diff get-impacted-targets [-v] -fh=<finalHashesJSONPath>
                                       -o=<outputPath>
                                       -tt=<targetType>
                                       -sh=<startingHashesJSONPath>
Command-line utility to analyze the state of the bazel build graph
      -fh, --finalHashes=<finalHashesJSONPath>
                  The path to the JSON file of target hashes for the final
                    revision. Run 'generate-hashes' to get this value.
      -o, --output=<outputPath>
                  Filepath to write the impacted Bazel targets to, newline
                    separated
      -sh, --startingHashes=<startingHashesJSONPath>
                  The path to the JSON file of target hashes for the initial
                    revision. Run 'generate-hashes' to get this value.
      -tt, --targetType=<targetType>
                  The type of targets to filter, available options are SourceFile/Rule/GeneratedFile
                  Only works if the JSON was generated with `--includeTargetType` enabled.
                  If not specified, all types of impacted targets will be returned.
      -v, --verbose
                  Display query string, missing files and elapsed time

Installing

Integrate into your project (recommended)

First, add the following snippet to your project:

WORKSPACE Snippet

load("@bazel_tools//tools/build_defs/repo:http.bzl", "http_jar")

http_jar(
    name = "bazel_diff",
    urls = [
        "https://github.com/Tinder/bazel-diff/releases/download/6.1.0/bazel-diff_deploy.jar",
    ],
    sha256 = "5d90de4561afd1e711bc62956560a9dfcbb4454bd6b209d6e68272b65c3cb50a",
)

MODULE.bazel Snippet (if you are using Bzlmod)

http_jar = use_repo_rule("@bazel_tools//tools/build_defs/repo:http.bzl", "http_jar")
http_jar(
    name = "bazel_diff",
    urls = [
        "https://github.com/Tinder/bazel-diff/releases/download/7.0.0/bazel-diff_deploy.jar"
    ],
    sha256 = "0b9e32f9c20e570846b083743fe967ae54d13e2a1f7364983e0a7792979442be",
)

WORKSPACE snippet

http_jar = use_repo_rule("@bazel_tools//tools/build_defs/repo:http.bzl", "http_jar")
http_jar(
    name = "bazel_diff",
    urls = [
        "https://github.com/Tinder/bazel-diff/releases/download/7.0.0/bazel-diff_deploy.jar"
    ],
    sha256 = "0b9e32f9c20e570846b083743fe967ae54d13e2a1f7364983e0a7792979442be",
)

load("//@bazel_diff:repositories.bzl", "bazel_diff_dependencies")

bazel_diff_dependencies()

Second, add in your root BUILD.bazel file:

load("@rules_java//java:defs.bzl", "java_binary")

java_binary(
    name = "bazel-diff",
    main_class = "com.bazel_diff.Main",
    runtime_deps = ["@bazel_diff//jar"],
)

That's it! You can now run the tool with:

bazel run //:bazel-diff

Note, in releases prior to 2.0.0 the value for the main_class attribute is just BazelDiff

Run Via JAR Release

curl -Lo bazel-diff.jar https://github.com/Tinder/bazel-diff/releases/latest/download/bazel-diff_deploy.jar
java -jar bazel-diff.jar -h

Build from Source

After cloning down the repo, you are good to go, Bazel will handle the rest

To run the project

bazel run :bazel-diff -- bazel-diff -h

Debugging (when running from source)

To run bazel-diff with debug logging, run your commands with the verbose config like so:

bazel run :bazel-diff --config=verbose -- bazel-diff -h

Build your own deployable JAR

bazel build //cli:bazel-diff_deploy.jar
java -jar bazel-bin/cli/bazel-diff_deploy.jar # This JAR can be run anywhere

Build from source in your Bazel Project

Add the following to your WORKSPACE file to add the external repositories, replacing the RELEASE_ARCHIVE_URL with the archive url of the bazel-diff release you wish to depend on:

load("@bazel_tools//tools/build_defs/repo:http.bzl", "http_archive")

http_archive(
  name = "bazel_diff",
  urls = [
        "RELEASE_ARCHIVE_URL",
    ],
    sha256 = "UPDATE_ME",
    strip_prefix = "UPDATE_ME"
)

load("@bazel_diff//:repositories.bzl", "bazel_diff_dependencies")

bazel_diff_dependencies()

load("@rules_jvm_external//:defs.bzl", "maven_install")
load("@bazel_diff//:artifacts.bzl", "BAZEL_DIFF_MAVEN_ARTIFACTS")

maven_install(
    name = "bazel_diff_maven",
    artifacts = BAZEL_DIFF_MAVEN_ARTIFACTS,
    repositories = [
        "http://uk.maven.org/maven2",
        "https://jcenter.bintray.com/",
    ],
)

Now you can simply run bazel-diff from your project:

bazel run @bazel_diff//:bazel-diff -- bazel-diff -h

Learn More

Take a look at the following bazelcon talks to learn more about bazel-diff:

• BazelCon 2023: Improving CI efficiency with Bazel querying and bazel-diff
• BazelCon 2024: Not Going the Distance: Filtering Tests by Build Graph Distance

Running the tests

To run the tests simply run

bazel test //...

Versioning

We use SemVer for versioning. For the versions available, see the tags on this repository.

License

---

Copyright (c) 2020, Match Group, LLC
All rights reserved.

Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
    * Redistributions of source code must retain the above copyright
      notice, this list of conditions and the following disclaimer.
    * Redistributions in binary form must reproduce the above copyright
      notice, this list of conditions and the following disclaimer in the
      documentation and/or other materials provided with the distribution.
    * Neither the name of Match Group, LLC nor the names of its contributors
      may be used to endorse or promote products derived from this software
      without specific prior written permission.

THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL MATCH GROUP, LLC BE LIABLE FOR ANY
DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.