Closed acxz closed 2 years ago
I believe this is an appropriate change as maintaining compat with older julia code for a benchmarks suite is not really in the best interest.
Does using Compat
bring a significant performance penalty? If not, I would expect that it is useful to keep it around so that the benchmark can be run on older versions (e.g. in case one wants to rebuild the historical trend using a consistent process across multiple Julia versions, or in case the CI machines change, etc.)
By the way, cross-linking to https://github.com/JuliaLang/Microbenchmarks/pull/31 which introduced the Compat
module (cc @johnfgibson).
keep it around so that the benchmark can be run on older versions
unless there is a script that does this, I wouldn't support adding it back in. Only code that is used should be included, otherwise just leads to tech debt. If you or anyone else makes a PR that adds a script/plot to compare across julia versions then I'll be more than happy to merge it.
Understood. Although I in this matter I have the opposite stance, I defer to the people actually maintaining this repository to make the decision.
I am not able to contribute such a script myself, but I worry that removing Compat will make such a contribution even harder (since it would have to reintroduce it on top of the script to run the benchmark across multiple Julia versions).
removes usage of the Compat package
resolves CI errors
I believe this is an appropriate change as maintaining compat with older julia code for a benchmarks suite is not really in the best interest.
this PR also updates julia to latest stable (1.7.3) and python to latest release (3.10)