Testing 6

[rahul799]

This PR will test
Different mixes of R vs other files, including confirming that it does nothing when no R files are changed

[rahul799]

/style

[infotroph]

The next invocation of the style command will find only .prettierignore changed. Since no R files have changed, there's nothing for styler to do. Let's see what happens!

[infotroph]

/style

[infotroph]

Answer, styler runs, finds no changes to make, and exits reporting success. Nothing is committed.

The only downside I see here is that you need to go check the actions tab to tell whether the run is still in progress or completed without finding any changes. But then again most people won't be requesting a styler run when there's nothing to style, so this is probably not a big deal.

[infotroph]

The next invocation of the styler command will find no R files changed, but should notice and fix one misedited Rd file. Let's see what happens!

[infotroph]

/style

[infotroph]

Answer: Styler runs but doesn't commit the change! It took me a while to track down why:

• Issue 1: Initial list of changed files includes the changed base/utils/PEcAn.Rd, but only *.R and *.Rmd are passed in to files_to_style.txt. files_to_style.txt is the only artifact passed to the check step, so the check step never gets the information that base/utils needs to be Roxygenized. To fix this, the check step should be given the full list of changed files.

• Issue 2: But In this case, the process of checking models/fates and models/gday prompts Make to re-Roxygenize base/utils anyway! But the change isn't committed because the command

  if [[ $(git diff --name-only --cached) != "" ]]; then  git commit -m 'automated documentation update' ; fi
  

  returns [[: not found. To fix this, we should either use [ instead of the bashism [[, or explicitly execute the command in bash rather than the default shell.

[infotroph]

The next invocation tests again after fixing issue 2 ([[: not found) from above. If it's fixed correctly, the next commit should revert the changes to PEcAn.Rd.

Note also that each time we invoke styler it looks at the set of all files changed in this pull request, not just the files changed in the most recent commit like I was assuming above. But of course the style changes to *.R have already been committed by previous styler invocations, so we should only see changes in the newly touched files.

[infotroph]

/style

[infotroph]

/style

(Apparently I didn't actually push the [[ fix before)

[infotroph]

/style

[infotroph]

Result: In this case the mangled Rd file gets fixed, but by accident (because utils gets rebuilt when make .doc/models/fates runs) rather than because of the change. This should be fixed eventually, but "change to Rd file with no corresponding change to R files in the same package" is a weird enough corner case that it needn't stop us from merging this now.