Looking for guidance on module benchmarking

[barabo]

I'm building a module and I am looking for ideas to evaluate the performance of different functions. I'm interested in both time and memory usage, but mostly time.

For more context, I'm writing a series of functions which will be useful for converting healthcare data from json (FHIR) to a tabular format (OMOP).

I am already concerned that I'll need to profile my implementations of some of these helper functions, and selectively target my hot spots. Since I'm also learning the subtleties of the language, I expect I'll make some bad decisions until I've internalized all the features.

Having something like pprof, perf, gprof, etc would be fine.

PS> Thank you for this wonderful language!!

[wader]

Hey, I haven't done much detailed benchmarking of jq code but when i've needed it i've done it via https://github.com/itchyny/gojq and the golang tooling for profiling, a bit awkward but you can usually get some hints what is slow. I guess something similar can be done with C tooling. Might also be tricky if jq and gojq differ in TCO etc. I did do a hacky opcode level profiler for gojq some years ago, maybe i can digg it up or recreate it if your interested.

Otherwise for rough benchmarking i usually just use time or https://github.com/sharkdp/hyperfine etc to reproduce down to a minimal case of what is slow.

I'll try think of more things but one thing that has bitten me is: avoid concat arrays and strings, instead collect and then join etc.

And i agree jq is a wonderful language :)

[barabo]

I may check out gojq at some point, thanks for the links. I haven't heard of hyperfine.