I noticed that while the final results are filtered to display only the files that have an actual diff, the Vale command itself runs against the entire directory and THEN filters. This is considerably slower than it used to be (execution time upwards of 90 seconds).
We really only care about the specific files in the PR (ranging between 1-20 files on average).
I have a workaround of supplying an array of files of the form '["input1", "input2"]', but it was nicer when I used to be able to use __onlyModified without having to implement the workaround.
I'm in the process of upgrading our Vale action to use reviewdog through this PR: https://github.com/cockroachdb/docs/pull/15366
I noticed that while the final results are filtered to display only the files that have an actual diff, the Vale command itself runs against the entire directory and THEN filters. This is considerably slower than it used to be (execution time upwards of 90 seconds).
We really only care about the specific files in the PR (ranging between 1-20 files on average).
I have a workaround of supplying an array of files of the form
'["input1", "input2"]', but it was nicer when I used to be able to use__onlyModifiedwithout having to implement the workaround.