Re: [PATCH] unicode: update the width tables to Unicode 16

Hi Junio,

On Tue, 17 Sep 2024, Junio C Hamano wrote:

Oh, but I obviously was not suggesting as crude a thing as to scrape it unconditionally, and certainly not daily. No, I was thinking about something checking the `Last-Modified:` header and only acting upon updated Unicode definitions, and checking for updates only on a weekly basis. Something along these lines:

	```yml
	name: update Unicode definitions

	on:
	  schedule:
	    - cron: '1 15 * * 4' # 3:01pm on Wednesdays
	  workflow_dispatch:

	jobs:
	  update-repo-variable:
	    if: vars.UNICODE_LAST_MODIFIED != ''
	    runs-on: ubuntu-latest
	    steps:
	      - id: check
		run: |
		  set -x
		  latest_update="$(curl -I https://www.unicode.org/Public/UCD/latest/ucd/UCD.zip |
		    sed -n 's/^Last-Modified: //p')" &&
		  if test '${{ vars.UNICODE_LAST_MODIFIED }}' = "$latest_update"
		  then
		    echo "result=skip" >>$GITHUB_OUTPUT
		    exit 0
		  fi
		  echo "result=$latest_update" >>$GITHUB_OUTPUT
	      - if: steps.check.outputs.result != 'skip'
		run: echo ::notice::_Now_ we scrape and do stuff
	      - if: steps.check.outputs.result != 'skip'
		env:
		  GH_TOKEN: ${{ secrets.UNICODE_LAST_MODIFIED_PAT }}
		run: |
		  gh api -X PATCH \
		    repos/$GITHUB_REPOSITORY/actions/variables/UNICODE_LAST_MODIFIED \
		    -f value='${{ steps.check.outputs.result }}'

This would use the repository variable `UNICODE_LAST_MODIFIED` to store the `Last-Modified:` value that was last seen (and implicitly act as the knob to prevent running in forks: if the variable is not yet set, the job will be skipped).

Sadly, to update the repository variable, we cannot use `permissions:` because the workflow syntax does not offer the `variables` scope. Therefore a Personal Access Token would need to be stored as a repository secret. I used a fine-grained token in my tests whose sope was Repository > Variables: read-write.

Ciao, Johannes