Add support for specifying jobs to wait for by name

[m-roberts]

This has not been tested, but should achieve the desired behaviour.

Closes #474

[kachick]

Thanks for your suggestion and addressing as this PR!

I have a plan to provide the feature with workflow + job specifier. Could you check the my review?

[kachick]

As I understand it, job names might be duplicated if they are defined in another workflow. So I want to specify combined workflow file name and job name when providing this feature.

I guess #475 (comment) is a way to get them.

[kachick]

Using Array.prototype.includes() is a loop in a loop. So please prefer Set.prototype.has() as

wait-other-jobs/src/github-api.ts

Line 104 in bc780bf

ownJobIDs.has(summary.source.id) ? [] : [summary]

[m-roberts]

ChatGPT answer that addresses your points

---

To implement the feature of waiting for only a subset of jobs, you can follow these steps:

1. Modify the action.yml file to include the new workflow-job-map input:

# action.yml
...
inputs:
  ...
  workflow-job-map:
    description: 'JSON object that maps workflow names to arrays of job names to wait for'
    required: false
    default: '{}'

2. Update the src/main.ts file to parse the workflow-job-map input and create a Map with Sets for each workflow:

// src/main.ts
...
const workflowJobMapInput = JSON.parse(getInput('workflow-job-map', { required: true, trimWhitespace: true }));
const workflowJobMap = new Map<string, Set<string>>();

for (const workflowName in workflowJobMapInput) {
  workflowJobMap.set(workflowName, new Set(workflowJobMapInput[workflowName]));
}
...

3. Update the fetchOtherRunStatus function signature in src/github-api.ts to accept the workflowJobMap:

export async function fetchOtherRunStatus(
  // eslint-disable-next-line @typescript-eslint/prefer-readonly-parameter-types
  octokit: Parameters<typeof fetchRunSummaries>[0],
  params: Parameters<typeof fetchRunSummaries>[1],
  ownJobIDs: Readonly<Set<JobID>>,
  workflowJobMap: Map<string, Set<string>>,
): Promise<Report> {
  ...
}

4. Modify the fetchOtherRunStatus function in src/github-api.ts to filter the checkRunSummaries based on the workflowJobMap:

const checkRunSummaries = await fetchRunSummaries(octokit, params);
const filteredCheckRunSummaries = workflowJobMap.size === 0
  ? checkRunSummaries
  : checkRunSummaries.filter(summary => {
      const workflowName = summary.source.name.split('/')[0].trim();
      const jobName = summary.source.name.split('/')[1].trim();
      return (
        workflowJobMap.has(workflowName) &&
        workflowJobMap.get(workflowName)!.has(jobName)
      );
    });

const otherRelatedRuns = filteredCheckRunSummaries.flatMap<CheckRunsSummary>((summary) =>
  ownJobIDs.has(summary.source.id) ? [] : [summary]
);
...

5. Pass the workflowJobMap to the fetchOtherRunStatus function in the run() function of src/main.ts:

// src/main.ts
...
const report = await fetchOtherRunStatus(
  octokit,
  { ...repositoryInfo, ref: commitSha },
  ownJobIDs,
  workflowJobMap,
);
...

Now the action will only wait for the specified subset of jobs within the given workflows. If the workflow-job-map input is not provided or is empty, the action will wait for all jobs by default.

Users can now specify workflow and job names like this:

- uses: kachick/wait-for-other-jobs@v1
  with:
    github-token: ${{ secrets.GITHUB_TOKEN }}
    workflow-job-map: '{"Workflow1": ["Job1", "Job2"], "Workflow2": ["JobA"]}'

[kachick]

• Providing option as JSON map sounds like a reasonable way to me 👍
• Currently workflow name does not exist in our variables and constants. So the following code will NOT work.

      const workflowName = summary.source.name.split('/')[0].trim();
      const jobName = summary.source.name.split('/')[1].trim();

I guess #475 (comment) is a way to get them.

This is why I referred to another comment. Using GraphQL API might be a better way to get workflow names. Using only REST APIs requires many send requests.