Optimising checkouts for speed

[alister]

I have a large and somewhat busy repository (now over 80,000 commits), that I'm writing a workflow for - getting the files that are newly created or edited in a PR, and then I'll run some coding standards on the appropriate files.

The first two steps are:

    steps:
      - uses: actions/checkout@v4

      - name: Get changed files
        id: changed-files
        uses: tj-actions/changed-files@v44

The initial checkout for my own PR takes ~23seconds, but the tj-actions/changed-files@v44 step is taking a little more than 3 minutes for a (fairly recent - couple of days old) pull-request in the repo.

Is there a way optimise the fetching of commits, so that it does not take as much time? I've been, for example, trying to find a way to only checkout the main branch where it is about to start the PR's branch, but haven't had much luck yet.

2024-08-09T17:46:35.2869172Z ##[group]changed-files
2024-08-09T17:46:35.2998707Z Using local .git directory
2024-08-09T17:46:35.3597401Z Running on a pull_request (synchronize) event...
2024-08-09T17:46:35.3598926Z Repository is shallow, fetching more history...
   # This step is what takes most of the time:
2024-08-09T17:49:36.3396444Z Completed fetching more history.
2024-08-09T17:49:36.3606293Z ##[warning]Unable to find merge base between a23c5f3be0921a758dcdef134292668ad165d158 and 3d35949773b4955f493c7f60cd662d896690f109
2024-08-09T17:49:36.3618213Z Merge base is not in the local history, fetching remote target branch...
2024-08-09T17:49:37.1158904Z Retrieving changes between a23c5f3be0921a758dcdef134292668ad165d158 (main) → 3d35949773b4955f493c7f60cd662d896690f109 (feature/add-qa-tools)
2024-08-09T17:49:37.1219193Z All Done!
2024-08-09T17:49:37.1220222Z ##[endgroup]

Thanks!

[alister]

I did manage to speed up the initial checkout, and also the running of the tj-actions/changed-files action - with a Sparse Checkout. By not retrieving most of the large files (many whole-file replacements and vendored libraries) the initial checkout has gone from 21s to 8 seconds, and the 'changed-files' step from 180 secs to just 8 seconds because it already has everything it should need.

Here are sample steps that achieve this speedup:

on: [pull_request]  
name: Main  
jobs:  
  phptools-for-changed-files:
    name: CS-Fixer-Tool
    runs-on: ubuntu-latest  
    steps:  
      - uses: actions/checkout@v4  
        with:  
          sparse-checkout-cone-mode: false  
          sparse-checkout: |  
            /*  
            /.*  
            !/*/  
            /.github
            '/Makefile'  
            /src/  
            !/src/css/  
            !/src/js/  
            !/src/other-big-directories/  
            /tests/  
  
      - name: Get changed files  
        id: changed-files  
        uses: tj-actions/changed-files@v44  

      # now run a command based on ${{ steps.changed-files.outputs.all_changed_files }}