agriffis
commented
4 years ago I appreciate your efforts but I'm not going to merge this. I don't think keeping an entire historical copy of the lib in the repo is the right approach—the benefit doesn't justify the added complexity and maintenance burden.
In general, the right approach is external benchmarking over git history rather than coupling it to the repository. That would still be compatible with integrated CI since it could report regressions from prior versions. For example, how Quantstack is using ASV
With this in, Travis will run the benchmark for every build and log if the current branch is faster / slower than last released version (
v0.7.0)(Merging this to
masterfirst will allow us to compare all remaining open PRs..)