Open
Description
I have been working on adding benchmarks using ASV to the Dedupe library. I started off by copying the comment bot from your repo: https://github.com/pandas-dev/pandas/blob/main/.github/workflows/asv-bot.yml
As we started to use the bot, we found that there were many bugs with it, and I don't think your bot works in its current state. I just wanted to share some of our discoveries in case you every tried to fix your bot, so you don't have to do the same work:
- benchmark will not comment results if performance declines dedupeio/dedupe#1019
- using @actions/checkout actually checks out
main
, not the commit of the PR, since the workflow is triggered by a comment.
See https://github.com/dedupeio/dedupe/commits/main/.github/workflows/benchmark-bot.yml - funny formatting of the results comment
- In general look at the file history for our bot to see what fixes we had to implement
- You can see an example run of the bot here: start to use sklearn for ml algorithms dedupeio/dedupe#992 (comment)
Thanks for the great starting point!