BurntSushi/ripgrep

consider adding an `rg:` prefix all error messages

shoffmeister opened this issue · 1 comments

shoffmeister commented

Please tick this box to confirm you have reviewed the above.

  • I have a different issue.

What version of ripgrep are you using?

ripgrep 13.0.0
-SIMD -AVX (compiled)
+SIMD +AVX (runtime)

How did you install ripgrep?

dnf install ripgrep

What operating system are you using ripgrep on?

Fedora Rawhide (40)

Describe your bug.

When running any ripgrep command on a clone of https://invent.kde.org/frameworks/syntax-highlighting, stderr will emit output

./autotests/input/.gitignore: line 7: error parsing glob '/**/a?[a-z12!][]][!][:upper:]][^[=a=]]\?[\\incomplete?*': unclosed character class; missing ']'
./autotests/input/.gitignore: line 8: error parsing glob '!/**/a?[a-z][]][!][:upper:]][^[=a=]]\?[\\incomplete?*': unclosed character class; missing ']'

What are the steps to reproduce the behavior?

  • git clone https://invent.kde.org/frameworks/syntax-highlighting.git
  • cd syntax-highlighting
  • rg anything 2>stderr.txt
  • cat stderr.txt

What is the actual behavior?

./autotests/input/.gitignore: line 7: error parsing glob '/**/a?[a-z12!][]][!][:upper:]][^[=a=]]\?[\\incomplete?*': unclosed character class; missing ']'
./autotests/input/.gitignore: line 8: error parsing glob '!/**/a?[a-z][]][!][:upper:]][^[=a=]]\?[\\incomplete?*': unclosed character class; missing ']'

What is the expected behavior?

"Something else" would be better, but to me it is not clear what the most appropriate behaviour is:

From what I can tell, ripgrep prints to stderr internal diagnostics about what it found in the specific .gitignore file. I can see these options to improve the user experience in this corner case:

a) don't print those diagnostics by default; it's not within the responsibilities of ripgrep to diagnose the content of .gitignore files. Possibly add an option to turn diagnostics on, to allow for trouble-shooting

b) expand the text of the emitted diagnostic which clarifies that the origin is ripgrep itself - when running without redirection, that is not immediately obvious. Example: "rg: configuring search with ./autotests/input/.gitignore: line 7: ..."

FWIW, I believe the above .gitignore patterns are exactly as intended, as they seem to act as input into a test, not as "meaningful" .gitignore files. To that extent, here, that is a very fringe corner case. There may be other instances where ripgrep may print.

BurntSushi commented

You can use --no-ignore-messages to suppress that specific output without suppressing anything else. I don't think they should be suppressed by default because it could otherwise result in a silent failure to filter a search as one might expect.

I am fond of the idea of adding an rg: prefix to the error message, though, I suspect it should probably be added to all error messages.

  • 👍1