strmatch and backslash

[Jean-Claude Arbaut]

There seems to be something weird about strmatch, and I just can't find out what. I'm trying to match file names, and I stumbled on a problem that reduces to this:

. di strmatch("a/*","*/*")
1

. di strmatch("a/b","*/*")
1

. di strmatch("a\*","*\*")
1

. di strmatch("a\b","*\*")
0

Why is the last one 0? I suspected some kind of special use of backslash (like in C to quote characters), but Stata doesn't do that usually, and anyway di strmatch("a\b","*\\*") also returns 0. And I found nothing about backslash in strmatch in the documentation.

There is the same problem with:

. di strmatch("a\?","*\?")
1

. di strmatch("a\b","*\?")
0

There is a workaround:

. di regexm("a\b",".*\\.*")
1

And here the backslash is doubled because it's the quoting character for regular expressions (e.g., to match the character "*" use "\*").

Has anyone an explanation for this behaviour of strmatch?

[daniel klein]

I have no answer. I can say that

. di strmatch("a\b","*\\*")
1

used to work in Stata 11 and 12.

Beste
Daniel