Include uv in cli benchmarks

[dimbleby]

No description provided.

[ofek]

Thanks! I like the --shell=none change but I think including UV isn't meaningful because it's not in Python i.e. you might as well plug in any other CLI of a compiled language like git. Could you please remove that part and just keep the former modification?

[dimbleby]

I included the --shell=none only because hyperfine otherwise complained that shell startup was too significant in the uv benchmarking, it is otherwise pretty irrelevant.

I do not agree that comparison is only valid if the other tool is written in python. Why would users care about that? If this is measuring something useful - then surely it is just as useful for uv as for anything else?

uv is a relevant comparison where eg git is not because it is an "equivalent" (ish!) tool: just as poetry and pipenv are relevant comparisons but, say, black and ansible would not be.

Of course this benchmark will show that uv is fast, which perhaps is not the point you want to make - but that is how things are!

[ofek]

I would prefer the benchmark job be removed then, and the reference in the documentation.

[dimbleby]

sure, #1772