search

LilithHafner / Chairmarks.jl

Benchmarks with back support
GNU General Public License v3.0
81 stars 8 forks source link

Provide longer names #122

Open LilithHafner opened 2 weeks ago

LilithHafner commented 2 weeks ago

As Charimarks.jl has gained popularity and started to get used outside of the REPL, it makes sense to move to longer names for @b and @be.

What sort of longer names do folks want?

A. @btime/@benchmark does not make sense because @b, unlike BencharmkTools.@btime, neither prints runtime data nor returns the evaluation result B. @benchmark_summary/@benchmark_full works, though very verbose, and precludes adding an even more informative version that saves more data C. @bench/@benchmark is fun and silly and concise while still being clear that we're benchmarking. It still relies on the pun: shorter name = shorter results.

I would implement this by adding long names as an option and then maybe deprecating the use of short names in packages and scripts. I do not currently plan to deprecate the use of short names at the REPL.

Alternatively, I could export the long names and publicize the short alternatives, like Unitful.

Originally posted by @LilithHafner in https://github.com/LilithHafner/Chairmarks.jl/issues/100#issuecomment-2211134260

Tortar commented 2 weeks ago

I vote C. because it seems the clearest and not-too-verbose alternative and I like the pun ;-)

Tortar commented 2 weeks ago

what about also (manipulated suggestion from the other options in #100) D. @minbench and @fullbench? Seem like not too verbose to me, and the macros are different from BenchmarkTools

alfaromartino commented 2 weeks ago

I like @minbench and @fullbench more, although I'd suggest @minibench or @litebench for the former. They keep clarity, while keeping conciseness.
Note that @bench and @benchmark are problematic for tab completion, as they start with the same letters.