Control color via an environment variable

[jimhester]

No description provided.

[gaborcsardi]

Thanks! I think we can simplify this, and get rid of the env var. rcmdcheck would just use the regular crayon color detection, plus it would turn it on for GHA (for itself only). Then it would turn color off for the checked package explicitly, in a similar way that testthat creates a reproducible context.

This will improve rcmdcheck in general, and we also don't need an env var.

[jimhester]

Well, it is not just GitHub Actions that has this problem, GitLab has the same problem, which is why I used an environment variable so other places could control it.

I don't think rcmdcheck should really mess with the color in the actual R CMD check process, either turning it explicitly off or on.

[gaborcsardi]

Well, it is not just GitHub Actions that has this problem, GitLab has the same problem, which is why I used an environment variable so other places could control it.

We can detect that as well. But we can also have an env var, I don't actually mind.

I don't think rcmdcheck should really mess with the color in the actual R CMD check process, either turning it explicitly off or on.

I think the reproducible context in testthat is really nice, and even base R is doing some of this when it is avoiding fancy quotes. I think it is the right thing to do to make R CMD check reproducible.

[jimhester]

I just worry about introducing something into rcmdcheck that makes its behavior differ from R CMD check run on the command line

[gaborcsardi]

I just worry about introducing something into rcmdcheck that makes its behavior differ from R CMD check run on the command line

Yeah, you are right, I realized this as well.

It would be nice to have some system that would standardize the context for the examples, that works for R CMD check itself. For the tests, the new testthat context is great.

[jimhester]

I guess one option would be to have crayon detect R_TESTS https://github.com/wch/r-source/search?q=R_TESTS&unscoped_q=R_TESTS? That should always be set when R CMD check runs the tests.

For examples maybe it could check if the CheckExEnv (https://github.com/wch/r-source/blob/1349e5a0fd8a70ccef3a234527aca7495c4d9ff0/share/R/examples-header.R#L3) is on the search path?

[gaborcsardi]

OK, this is in good shape. Let me fix the CI for the main branch first. Then I'll some docs for this, and we can merge it.

[codecov-commenter]

Codecov Report

Merging #126 (a068355) into master (cd0f365) will decrease coverage by 0.56%.
The diff coverage is 66.66%.

@@            Coverage Diff             @@
##           master     #126      +/-   ##
==========================================
- Coverage   76.07%   75.51%   -0.57%     
==========================================
  Files          18       18              
  Lines         861      878      +17     
==========================================
+ Hits          655      663       +8     
- Misses        206      215       +9     

Impacted Files	Coverage Δ
R/comparison.R	49.12% <ø> (ø)
R/error.R	0.00% <ø> (ø)
R/print.R	93.93% <ø> (ø)
R/styles.R	68.96% <62.50%> (-31.04%)	⬇️
R/comparison-summary.R	100.00% <100.00%> (ø)
R/utils.R	82.25% <100.00%> (+0.29%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update cd0f365...a068355. Read the comment docs.

[gaborcsardi]

Thanks much!