test.data.table() in non-English R session

[MichaelChirico]

Closes #3039
Closes #630

This is an approach that would work to solve #3039.

But, it feels a bit hacky/manual, so filing this as a test balloon for now.

Less-than-ideal alternative would be something like:

was_translated = if (tryCatch(get(x, envir = new.env()), error = identity)$message == "object 'x' not found") FALSE else TRUE

then branch on if (was_translated) for subsequent gettext-dependent tests...

[codecov]

Codecov Report

Merging #3553 into master will increase coverage by <.01%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master    #3553      +/-   ##
==========================================
+ Coverage   97.55%   97.55%   +<.01%     
==========================================
  Files          66       66              
  Lines       12667    12672       +5     
==========================================
+ Hits        12357    12362       +5     
  Misses        310      310

Impacted Files	Coverage Δ
R/test.data.table.R	100% <100%> (ø)	⬆️

---

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 8719a1f...d5d1cf7. Read the comment docs.

[jangorecki]

I would go the other way, during test.data.table force R to use english error messages. Much less to support and we still test exactly what we need.

[MichaelChirico]

@jangorecki I believe that's what testthat tried but gave up due to caching:

r-lib/testthat#565 (comment)

[jangorecki]

In such case I would close #3039 as won't fix, it unnecessarily adds complexity. We don't want to test error translations. Eventually we can fill R bugzilla issue for providing such capability.

[MichaelChirico]

lgtm

[MichaelChirico]

nice idea. excellent.

[mattdowle]

Great. I did try (again) to change the language to English from within test.data.table() but it quickly got complicated and OS-specific. Plus with the link you found about caching the translations. Yep so I figured it would be cleaner to inform user how to change their language and rerun if they want the full-strength test, otherwise just checking the number of errors/warnings is still pretty fine.

[MichaelChirico]

yep... the number of errors relying on base R translations is pretty small so I thought quitting test.data.table() over a bad locale to be a bit much. this is v good

…

On Fri, May 17, 2019, 8:18 AM Matt Dowle ***@***.***> wrote: ***@***.**** commented on this pull request. ------------------------------ In R/test.data.table.R <#3553 (comment)> : > @@ -60,6 +60,13 @@ test.data.table <- function(verbose=FALSE, pkg="pkg", silent=FALSE, with.other.p cat("Running", fn, "\n") env = new.env(parent=.GlobalEnv) assign("testDir", function(x) file.path(fulldir, x), envir=env) + + # are R's messages being translated to a foreign language? #3039, #630 + txt = eval(parse(text="tryCatch(mean(not__exist__), error = function(e) e$message)"), envir=.GlobalEnv) + foreign = txt != "object 'not__exist__' not found" + if (foreign) cat("\n**** This R session's language is not English. Each test will still check that the correct number of errors and/or\n", Great. I did try (again) to change the language to English from within test.data.table() but it quickly got complicated and OS-specific. Plus with the link you found about caching the translations. Yep so I figured it would be cleaner to inform user how to change their language and rerun if they want the full-strength test, otherwise just checking the number of errors/warnings is still pretty fine. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#3553?email_source=notifications&email_token=AB2BA5IFUF6LTPS47FH4KXLPVX2VHA5CNFSM4HLZLFKKYY3PNVWWK3TUL52HS4DFWFIHK3DMKJSXC5LFON2FEZLWNFSXPKTDN5WW2ZLOORPWSZGOBY5CWNI#discussion_r284942179>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AB2BA5KX6Y2EA7OCW3KR4SDPVX2VHANCNFSM4HLZLFKA> .