[MERGE] is_ignored improvements...

[Wayne Davison]

On Fri, May 19, 2006 at 03:51:40PM +1000, Martin Pool wrote:
> So the ?! says "don't match anything with .*/ after this point", which
> is OK but I wonder if it'd be simpler or faster to just translate '*' to [^/]*.

I was wondering the same thing.  Note also that '?' would need to be
similarly translated into "[^/]".

While poking around, I saw that there is a python function named
glob_to_re that is in distutils.filelist.  This function takes the
fnmatch.translate output and tries to turn every non-backslash-escaped
dot into "[^/]".  Unfortunately, the regex-substitution that it uses does
not translate a dot after the matching of a literal backslash (e.g. the
regex "foo\\.*" would be unchanged) and it messes up if two dots in a row
are present (e.g. "foo??" -> "foo..$" -> "foo[^/].$").  It has this code:

    pattern_re = re.sub(r'(^|[^\\])\.', r'\1[^/]', pattern_re)

I think that the following regex should actually work properly:

    pattern_re = re.sub(r'((?<!\\)(\\\\)*)\.', r'\1[^/]', pattern_re)

This will translate any dot that has an even number of backslashes in
front it (including 0).  I'll look into reporting this bug to the python
folks.

..wayne..