Refactor: better stack traces for profiling

[krlmlr]

This allows replacing %>% with |> which gives much much better stack traces for profiling. Also replaced one call to map2() which is very high in the chain, also for clearer stack traces.

For #759.

[codecov-commenter]

Codecov Report

Merging #980 (dbb867e) into main (bbc4eef) will increase coverage by 0.01%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##             main     #980      +/-   ##
==========================================
+ Coverage   90.23%   90.24%   +0.01%     
==========================================
  Files          47       47              
  Lines        2694     2697       +3     
==========================================
+ Hits         2431     2434       +3     
  Misses        263      263              

Impacted Files	Coverage Δ
R/nest.R	100.00% <100.00%> (ø)
R/roxygen-examples-parse.R	96.82% <100.00%> (-0.05%)	⬇️
R/transform-files.R	97.29% <100.00%> (+0.03%)	⬆️

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

[lorenzwalthert]

Looks good. Also note there is #937. Do youw awant to merge this PR as is or also perfrom the substitution as part of this PR? A limitation of using pre-commit with GitHub Actions is pushes by pre-commit (to fix problems) don't trigger other CI runs. To avoid that, either force push the changes afterwards, create an emtpy bump commit or install pre-commit locally so it's ran on every commit, i.e. on macos, I use $ brew install pre-commit for the system installation and then $ pre-commit install in this package directory.

[krlmlr]

We want styler to run on R < 4.1, I'm doing the substitution temporarily.

Thanks for the hints on pre-commit. It's unfortunate that automated pushes don't trigger a CI/CD rerun.

[github-actions]

This is how benchmark results would change (along with a 95% confidence interval in relative change) if c0932f5 is merged into main:

•   :ballot_box_with_check:cache_applying: 32.3ms -> 32.2ms [-1.74%, +1.15%]
•   :ballot_box_with_check:cache_recording: 1.51s -> 1.51s [-1.09%, +0.76%]
•   :ballot_box_with_check:without_cache: 4s -> 4s [-0.56%, +0.64%]

Further explanation regarding interpretation and methodology can be found in the documentation.

[lorenzwalthert]

Thanks for the hints on pre-commit. It's unfortunate that automated pushes don't trigger a CI/CD rerun.

yes. That’s not a pre-commit specific problem I think. And the reason why there is https://pre-commit.ci (but that integration is not (yet) activated for r-lib)

[krlmlr]

I don't understand the test failures here.

[lorenzwalthert]

That's a speed test that ensure caching works as expected from a performance point of view. It's an outlier, so re-runnig the test again should pass. We could also lift the threshold a bit. Up to you. Happy to merge this.

[github-actions]

This is how benchmark results would change (along with a 95% confidence interval in relative change) if bb30612 is merged into main:

•   :ballot_box_with_check:cache_applying: 26.8ms -> 26.8ms [-0.64%, +0.58%]
•   :ballot_box_with_check:cache_recording: 1.25s -> 1.24s [-1.47%, +0.14%]
•   :ballot_box_with_check:without_cache: 3.3s -> 3.3s [-0.6%, +0.23%]

Further explanation regarding interpretation and methodology can be found in the documentation.

[lorenzwalthert]

Looks good. I can fix failing test later, they are just very specific benchmarks that we want to keep an eye on.