batch spec YAML reference

Sourcegraph Batch Changes use batch specs to define batch changes.

This page is a reference guide to the batch spec YAML format in which batch specs are defined. If you’re new to YAML and want a short introduction, see “Learn YAML in five minutes.”

name

The name of the batch change, which is unique among all batch changes in the namespace. A batch change’s name is case-preserving.

Examples

name: update-go-import-statements
name: update-node.js

description

The description of the batch change. It’s rendered as Markdown.

Examples

description: This batch change changes all `fmt.Sprintf` calls to `strconv.Iota`.
description: |
  This batch change changes all imports from

  `gopkg.in/sourcegraph/sourcegraph-in-x86-asm`

  to

  `github.com/sourcegraph/sourcegraph-in-x86-asm`  

on

The set of repositories (and branches) to run the batch change on, specified as a list of search queries (that match repositories) and/or specific repositories.

Examples

on:
  - repositoriesMatchingQuery: lang:go fmt.Sprintf("%d", :[v]) patterntype:structural
  - repository: github.com/sourcegraph/sourcegraph

on.repositoriesMatchingQuery

A Sourcegraph search query that matches a set of repositories (and branches). Each matched repository branch is added to the list of repositories that the batch change will be run on.

See “Code search” for more information on Sourcegraph search queries.

Examples

on:
  - repositoriesMatchingQuery: file:README.md -repo:github.com/sourcegraph/src-cli
on:
  - repositoriesMatchingQuery: lang:typescript file:web const changesetStatsFragment

on.repository

A specific repository (and branch) that is added to the list of repositories that the batch change will be run on.

A branch attribute specifies the branch on the repository to propose changes to. If unset, the repository’s default branch is used. If set, it overwrites earlier values to be used for the repository’s branch.

Examples

on:
  - repository: github.com/sourcegraph/src-cli
on:
  - repository: github.com/sourcegraph/sourcegraph
    branch: 3.19-beta
  - repository: github.com/sourcegraph/src-cli

In the following example, the repositoriesMatchingQuery returns both repositories with their default branch, but the 3.23 branch is used for github.com/sourcegraph/sourcegraph, since it is more specific:

on:
  - repositoriesMatchingQuery: repo:sourcegraph\/(sourcegraph|src-cli)$
  - repository: github.com/sourcegraph/sourcegraph
    branch: 3.23

In this example, 3.19-beta branch is used, since it was named last:

on:
  - repositoriesMatchingQuery: repo:sourcegraph\/(sourcegraph|src-cli)$
  - repository: github.com/sourcegraph/sourcegraph
    branch: 3.23
  - repository: github.com/sourcegraph/sourcegraph
    branch: 3.19-beta

steps

The sequence of commands to run (for each repository branch matched in the on property) to produce the batch change’s changes.

Examples

steps:
  - run: echo "Hello World!" >> README.md
    container: alpine:3
steps:
  - run: comby -in-place 'fmt.Sprintf("%d", :[v])' 'strconv.Itoa(:[v])' .go -matcher .go -exclude-dir .,vendor
    container: comby/comby
  - run: gofmt -w ./
    container: golang:1.15-alpine
steps:
  - run: ./update_dependency.sh
    container: our-custom-image
    env:
      OLD_VERSION: 1.31.7
      NEW_VERSION: 1.33.0

steps.run

The shell command to run in the container. It can also be a multi-line shell script. The working directory is the root directory of the repository checkout.

steps.container

The Docker image used to launch the Docker container in which the shell command is run.

The image has to have either the /bin/sh or the /bin/bash shell.

It is executed using docker on the machine on which the Sourcegraph CLI (src) is executed. If the image exists locally, that is used. Otherwise it’s pulled using docker pull.

steps.env

Environment variables to set in the environment when running this command.

These may be defined either as an object or (in Sourcegraph 3.23 and later) as an array.

Environment object

In this case, steps.env is an object, where the key is the name of the environment variable and the value is the value.

Examples

steps:
  - run: echo $MESSAGE >> README.md
    container: alpine:3
    env:
      MESSAGE: Hello world!

Environment array

In this case, steps.env is an array. Each array item is either:

  1. An object with a single property, in which case the key is used as the environment variable name and the value the value, or
  2. A string that defines an environment variable to include from the environment src is being run within. This is useful to define secrets that you don’t want to include in the spec file, but this makes the spec dependent on your environment, means that the local execution cache will be invalidated each time the environment variable changes, and means that the batch spec file is no longer the sole source of truth intended by the Batch Changes design.

Examples

This example is functionally the same as the object example above:

steps:
  - run: echo $MESSAGE >> README.md
    container: alpine:3
    env:
      - MESSAGE: Hello world!

This example pulls in the USER environment variable and uses it to construct the line that will be appended to README.md:

steps:
  - run: echo $MESSAGE from $USER >> README.md
    container: alpine:3
    env:
      - MESSAGE: Hello world!
      - USER

For instance, if USER is set to adam, this would append Hello world! from adam to README.md.

steps.files

Files to create on the host machine and mount into the container when running steps.run.

steps.files is an object, where the key is the name of the file inside the container and the value is the content of the file.

Examples

steps:
  - run: cat /tmp/my-temp-file.txt >> README.md
    container: alpine:3
    files:
      /tmp/my-temp-file.txt: Hello world!
steps:
  - run: cat /tmp/global-gitignore >> .gitignore
    container: alpine:3
    files:
      /tmp/global-gitignore: |
        # Vim
        *.swp
        # JetBrains/IntelliJ
        .idea
        # Emacs
        *~
        \#*\#
        /.emacs.desktop
        /.emacs.desktop.lock
        .\#*
        .dir-locals.el        

steps.outputs

Output variables that are set after the steps.run command has been executed. These variables are available in the global outputs namespace as outputs.<name> template variables in the run, env, and outputs properties of subsequent steps, and the changesetTemplate. Two steps with the same output variable name will overwrite the previous contents.

Examples

steps:
  - run: yarn upgrade
    container: alpine:3
    outputs:
      # Set output `friendlyMessage`
      friendlyMessage:
        value: "Hello there!"
steps:
  - run: echo "Hello there!" >> message.txt && cat message.txt
    container: alpine:3
    outputs:
      friendlyMessage:
        # `value` supports templating variables and can access the just-executed
        # step's stdout/stderr.
        value: "${{ step.stdout }}"
steps:
  - run: echo "Hello there!"
    container: alpine:3
    outputs:
      stepOneOutput:
        value: "${{ step.stdout }}"
  - run: echo "We have access to the output here: ${{ outputs.stepOneOutput }}"
    container: alpine:3
    outputs:
      stepTwoOutput:
        value: "here too: ${{ outputs.stepOneOutput }}"
steps:
  - run: cat .goreleaser.yml >&2
    container: alpine:3
    outputs:
      goreleaserConfig:
        value: "${{ step.stderr }}"
        # Specifying a `format` tells Sourcegraph CLI how to parse the value before
        # making it available as a template variable.
        format: yaml

changesetTemplate:
  # [...]
  body: |
    The `goreleaser.yml` defines the following `before.hooks`:
    ${{ outputs.goreleaserConfig.before.hooks }}    

steps.outputs.<name>.value

The value the output should be set to.

steps.outputs.<name>.format

The format of the corresponding steps.outputs.<name>.value. When this is set to something other than text, it will be parsed as the given format.

Possible values: text, yaml, json. Default is text.

steps.if

Condition to check before executing the step. If the value of the if: attribute is true (boolean) or "true" (string) then the step is executed in the given repository (or workspace, in case workspaces are used). Otherwise the step is skipped.

As an optimization, the Sourcegraph CLI tries to evaluate the condition before starting to execute any steps. If the condition can be evaluated ahead of time and the result of the evaluation is false then the execution of the step won’t be attempted for the repository, which leads to better cache utilization.

Ahead-of-time evaluation is possible if the condition contains only static data. Example: if: ${{ eq repository.name "github.com/my-org/my-repo" }}. The repository name is known before the execution of the steps, so evaluation succeeds and Sourcegraph CLI will not include the given step in the list of steps to execute for repositories that don’t have the matching name. That in turn allows the modification of this step’s run attribute, for example, without invalidating the cache for the repositories in which it’s never executed.

Examples

steps:
  # `if:` is true, step always executes.
  - if: true
    run: echo "name of repository is ${{ repository.name }}" >> message.txt
    container: alpine:3
steps:
  # `if:` is a static string that's not "true", step never executes.
  - if: "random string"
    run: echo "name of repository is ${{ repository.name }}" >> message.txt
    container: alpine:3
steps:
  # `if:` uses templating to check for repository name and produce a "true". Only runs in github.com/sourcegraph/automation-testing
  - if: ${{ eq repository.name "github.com/sourcegraph/automation-testing" }}
    run: echo "hello from automation-testing" >> message.txt
    container: alpine:3
steps:
  # `if:` uses glob pattern to match repository name and produce "true" on match.
  - if: ${{ matches repository.name "*sourcegraph-testing*" }}
    run: echo "name contains sourcegraph-testing" >> message.txt
    container: alpine:3
steps:
  # First step prints to standard out and saves to outputs
  - run:  if [[ -f "go.mod" ]]; then echo "true"; else echo "false"; fi
    container: alpine:3
    outputs:
      goModExists:
        value: ${{ step.stdout }}

  # `if:` uses the just-set `outputs.goModExists` value as condition
  - if: ${{ outputs.goModExists }}
    run: go fmt ./...
    container: golang
steps:
  # `if:` checks for path, in case steps are executed in workspace.
  - if: ${{ eq steps.path "sub/directory/in/repo" }}
    run: echo "hello workspace" >> workspace.txt
    container: golang

importChangesets

An array describing which already-existing changesets should be imported from the code host into the batch change.

Examples

importChangesets:
  - repository: github.com/sourcegraph/sourcegraph
    externalIDs: [13323, "13343", 13342, 13380]
  - repository: github.com/sourcegraph/src-cli
    externalIDs: [260, 271]

importChangesets.repository

The repository name as configured on your Sourcegraph instance.

importChangesets.externalIDs

The changesets to import from the code host. For GitHub this is the pull request number, for GitLab this is the merge request number, for Bitbucket Server this is the pull request number.

changesetTemplate

A template describing how to create (and update) changesets with the file changes produced by the command steps.

This defines what the changesets on the code hosts (pull requests on GitHub, merge requests on Gitlab, …) will look like.

Examples

changesetTemplate:
  title: Replace equivalent fmt.Sprintf calls with strconv.Itoa
  body: This batch change replaces `fmt.Sprintf("%d", integer)` calls with semantically equivalent `strconv.Itoa` calls
  branch: batch-changes/sprintf-to-itoa
  commit:
    message: Replacing fmt.Sprintf with strconv.Iota
    author:
      name: Lisa Coder
      email: [email protected]
  published: false
changesetTemplate:
  title: Update rxjs in package.json to newest version
  body: This pull request updates rxjs to the newest version, `6.6.2`.
  branch: batch-changes/update-rxjs
  commit:
    message: Update rxjs to 6.6.2
  published: true
changesetTemplate:
  title: Run go fmt over all Go files
  body: Regular `go fmt` run over all our Go files.
  branch: go-fmt
  commit:
    message: Run go fmt
    author:
      name: Anna Wizard
      email: [email protected]
  published:
    # Do not meddle in the affairs of wizards, for they are subtle and quick to anger.
    - git.istari.example/*: false
    - git.istari.example/anna/*: true

changesetTemplate.title

The title of the changeset on the code host.

changesetTemplate.body

The body (description) of the changeset on the code host. If the code supports Markdown you can use it here.

changesetTemplate.branch

The name of the Git branch to create or update on each repository with the changes.

changesetTemplate.commit

The Git commit to create with the changes.

changesetTemplate.commit.message

The Git commit message.

changesetTemplate.commit.author

The name and email of the Git commit author.

Examples

changesetTemplate:
  commit:
    author:
      name: Alan Turing
      email: [email protected]

changesetTemplate.published

Whether to publish the changeset. This may be a boolean value (ie true or false), 'draft', or an array to only publish some changesets within the batch change. This may also be omitted, in which case the publication state will be controlled through the Sourcegraph UI, and will default to unpublished (that is, the same as specifying false).

An unpublished changeset can be previewed on Sourcegraph by any person who can view the batch change, but its commit, branch, and pull request aren’t created on the code host.

When published is set to draft a commit, branch, and pull request / merge request are being created on the code host in draft mode. This means:

  • On GitHub the changeset will be a draft pull request.
  • On GitLab the changeset will be a merge request whose title is be prefixed with 'WIP: ' to flag it as a draft merge request.
  • On BitBucket Server draft pull requests are not supported and changesets published as draft won’t be created.

A published changeset results in a commit, branch, and pull request being created on the code host.

Publishing only specific changesets

To publish only specific changesets within a batch change, an array of single-element objects can be provided. For example:

published:
  - github.com/sourcegraph/sourcegraph: true
  - github.com/sourcegraph/src-cli: false
  - github.com/sourcegraph/batchutils: draft

Each key will be matched against the repository name using glob syntax. The gobwas/glob library is used for matching, with the key operators being:

Term Meaning
* Match any sequence of characters
? Match any single character
[ab] Match either a or b
[a-z] Match any character between a and z, inclusive
{abc,def} Match either abc or def

If multiple entries match a repository, then the last entry will be used. For example, github.com/a/b will not be published given this configuration:

published:
  - github.com/a/*: true
  - github.com/*: false

If no entries match, then the repository will not be published. To make the default true, add a wildcard entry as the first item in the array:

published:
  - "*": true
  - github.com/*: false

By adding a @<branch> at the end of a match-rule, the rule is only matched against changesets with that branch:

published:
  - github.com/sourcegraph/src-*@my-branch: true
  - github.com/sourcegraph/src-*@my-other-branch: true

Examples

To publish all changesets created by a batch change:

changesetTemplate:
  published: true

To publish all changesets created by a batch change as drafts:

changesetTemplate:
  published: draft

To only publish changesets within the sourcegraph GitHub organization:

changesetTemplate:
  published:
    - github.com/sourcegraph/*: true

To publish all changesets that are not on GitLab:

changesetTemplate:
  published:
    - "*": true
    - gitlab.com/*: false

To publish all changesets on GitHub as draft:

changesetTemplate:
  published:
    - "*": true
    - github.com/*: draft

To publish only one of many changesets in a repository by addressing them with their branch name:

changesetTemplate:
  published:
    - "*": true
    - github.com/sourcegraph/*@my-branch-name-1: draft
    - github.com/sourcegraph/*@my-branch-name-2: false

(Multiple changesets in a single repository can be produced, for example, per project in a monorepo or by transforming large changes into multiple changesets).

transformChanges

A description of how to transform the changes (diffs) produced in each repository before turning them into separate changeset specs by inserting them into the changesetTemplate.

This allows the creation of multiple changeset specs (and thus changesets) in a single repository.

Examples

# Transform the changes produced in each repository.
transformChanges:
  # Group the file diffs by directory and produce an additional changeset per group.
  group:
    # Create a separate changeset for all changes in the top-level `go` directory
    - directory: go
      branch: my-batch-change-go # will replace the `branch` in the `changesetTemplate`

    - directory: internal/codeintel
      branch: my-batch-change-codeintel # will replace the `branch` in the `changesetTemplate`
      repository: github.com/sourcegraph/src-cli # optional: only apply the rule in this repository
transformChanges:
  group:
    - directory: go/utils/time
      branch: my-batch-change-go-time

    # The *last* matching directory is used, not the most specific one,
    # so only this changeset would be opened.
    - directory: go/utils
      branch: my-batch-change-go-date

transformChanges.group

A list of groups to define which file diffs to group together to create an additional changeset in the given repository.

The order of the list matters, since each file diff’s filepath is matched against the directory of a group and the last match is used.

If no changes have been produced in a directory then no changeset will be created.

transformChanges.group.directory

The name of the directory in which file diffs should be grouped together.

The name is relative to the root of the repository.

transformChanges.group.branch

The branch that should be used for this additional changeset. This overwrites the changesetTemplate.branch when creating the additional changeset.

Important: the branch can not be nested under the changesetTemplate.branch, i.e. if the changesetTemplate.branch is my-batch-change then this can not be my-batch-change/my-subdirectory since git doesn’t allow that. Additionally branch names must be unique and cannot be used as arguments for multiple directory fields.

transformChanges.group.repository

Optional: the file diffs matching the given directory will only be grouped in a repository with that name, as configured on your Sourcegraph instance.

workspaces

The optional workspaces property allows users to define where projects are located in repositories and cause the steps to be executed for each project, instead of once per repository. That allows easier creation of multiple changesets in large repositories.

For each repository that’s yielded by on and matched by a workspaces.in property, Sourcegraph search is used to get the locations of the rootAtLocationOf file. Each location then serves as a workspace for the execution of the steps, instead of the root of the repository.

Important: Since multiple workspaces in the same repository can produce multiple changesets, it’s required to use templating to produce a unique changesetTemplate.branch for each produced changeset. See the examples below.

Examples

Defining JavaScript projects that live in a monorepo by using the location of the package.json file as the root for each project:

on:
  - repository: github.com/sourcegraph/sourcegraph

workspaces:
  - rootAtLocationOf: package.json
    in: github.com/sourcegraph/sourcegraph

changesetTemplate:
  # [...]

  # Since a changeset is uniquely identified by its repository and its
  # branch we need to ensure that each changesets has a unique branch name.

  # We can use templating and helper functions get the `path` in which
  # the `steps` executed and turn that into a branch name:
  branch: my-multi-workspace-batch-change-${{ replace steps.path "/" "-" }}

Using templating to produce a unique branch name in repositories with workspaces and repositories without workspaces:

on:
  - repository: github.com/sourcegraph/sourcegraph
  - repository: github.com/sourcegraph/src-cli

workspaces:
  - rootAtLocationOf: package.json
    in: github.com/sourcegraph/sourcegraph

changesetTemplate:
  # [...]

  # Since the steps in `github.com/sourcegraph/src-cli` are executed in the
  # root, where path is "", we can use `join_if` to drop it from the branch name
  # if it's a blank string:
  branch: ${{ join_if "-" "my-multi-workspace-batch-change" (replace steps.path "/" "-") }}

Defining where Go, JavaScript, and Rust projects live in multiple repositories:

workspaces:
  - rootAtLocationOf: go.mod
    in: github.com/sourcegraph/go-*
  - rootAtLocationOf: package.json
    in: github.com/sourcegraph/*-js
    onlyFetchWorkspace: true
  - rootAtLocationOf: Cargo.toml
    in: github.com/rusty-org/*

changesetTemplate:
  # [...]

  branch: ${{ join_if "-" "my-multi-workspace-batch-change" (replace steps.path "/" "-") }}

Using steps.outputs to dynamically create unique branch names:

# [...]
on:
  # Find all repositories with a package.json file
  - repositoriesMatchingQuery: repohasfile:package.json

workspaces:
  # Define that workspaces have their root folder at the location of the
  # package.json files
  - rootAtLocationOf: package.json
    in: "*"

steps:
  # [... steps that produce changes ...]

  # Run `jq` to extract the "name" from the package.json
  - run:  jq -j .name package.json
    container: jiapantw/jq-alpine:latest
    outputs:
      # Set outputs.packageName to stdout of this step's `run` command.
      packageName:
        value: ${{ step.stdout }}

changesetTemplate:
  # [...]

  # Use `outputs` variables to create a unique branch name per changeset:
  branch: my-batch-change-${{ outputs.projectName }}

workspaces.rootAtLocationOf

The full name of the file that sits at the root of one or more workspaces in a given repository.

Sourcegraph code search is used to find the location of files with this name in the repositories returned by on.

For example, in a repository with the following files:

  • packages/sourcegraph-ui/package.json
  • packages/sourcegraph-test-helper/package.json

the workspace configuration

workspaces:
  - rootAtLocationOf: package.json
    in: "*"

would create two changesets in the repository, one in packages/sourcegraph-ui and one in packages/sourcegraph-test-helper.

workspaces.in

The repositories in which the workspace should be discovered.

This field supports globbing by using glob syntax. See “Publishing only specific changesets” for more information on globbing.

A repository matching multiple entries results in an error.

Examples

Match all repository names that begin with github.com/:

workspaces:
  - rootAtLocationOf: go.mod
    in: github.com/*

Match all repository names that begin with gitlab.com/my-javascript-org/ and end with -plugin:

workspaces:
  - rootAtLocationOf: package.json
    in: gitlab.com/my-javascript-org/*-plugin

workspaces.onlyFetchWorkspace

When set to true, only the folder containing the workspace is downloaded to execute the steps.

This field is not required and when not set the default is false.

Additional files — .gitignore and .gitattributes as of now — are downloaded from the location of the workspace up to the root of the repository.

For example, with the following file layout in a repository

.
├── a
│   ├── b
│   │   ├── [... other files in b ...]
│   │   ├── package.json
│   │   └── .gitignore
│   │
│   ├── [... other files in a ...]
│   ├── .gitattributes
│   └── .gitignore
│
├── [... other files in root ... ]
└── .gitignore

and this workspace configuration

workspaces:
  - rootAtLocationOf: package.json
    in: github.com/our-our/our-large-monorepo
    fetchOnlyWorkspace: true

then

  • the steps will be executed in b
  • the complete contents of b will be downloaded and are available to the steps
  • the .gitattributes and .gitignore files in a will be downloaded and put in a, but only those
  • the .gitignore files in the root will be downloaded and put in the root folder, but only that file

Examples

Only download the workspaces of specific JavaScript projects in a large monorepo:

workspaces:
  - rootAtLocationOf: package.json
    in: github.com/our-our/our-large-monorepo
    fetchOnlyWorkspace: true