Bug: Star is not treated as a literal character inside brackets

Question

Bug: Star is not treated as a literal character inside brackets

Opened this issue 7 months ago · 4 comments

The reference implementation does not treat a '*' character as a literal character inside brackets.

Current behavior

pattern = "ab[e*]cd.i";
  
// matches:
"ab[e]cd.i"
"ab[[]cd.i"
"ab[^]cd.i"
"ab[/]cd.i"
"ab[[]cd.i"
"ab[e[^/e[e]cd.i"

// does not match:
"abecd.i"
"ab*cd.i"

Expected behavior

the opposite

Answer 1 · 2024-02-11T20:54:52.000Z

Thanks for reporting! Would you please post instructions for reproducing the issue? Please make sure to include the specific commit at which you are testing. Much appreciated!

Answer 2 · 2024-02-12T18:00:06.000Z

I am testing against master branch by invoking ec_glob(pattern, string) with the pattern and the test strings from above.

Answer 3 · 2024-02-18T08:57:33.000Z

I think this is a bug, as there's really no meaning * in brackets. We really tried to mimic gitignore patterns in the beginning, and my experiment shows that gitignore would see * as a literal inside brackets. Also checked out PCRE, same behavior.

I think this is also what the spec means, but we didn't write it clearly out.

Answer 4 · 2024-02-18T12:46:20.000Z

Yes, I agree. There is another thing the spec does not clearly write out: slashes have precedence over anything else.

Meaning that a slash will "destroy" the bracket pattern.

Example: [ab/cd] will not be translated as "one of a, b, /, c, or d" but instead it will be the literal sequence [ab/cd].