Pattern matching

ketema has asked for the wisdom of the Perl Monks concerning the following question:

Replies are listed 'Best First'.
Re: Pattern matching by Fletch (Bishop) on Jun 07, 2005 at 16:33 UTC
`perldoc File::Basename` -- We're looking for people in ATL	[reply] [d/l]
Re^2: Pattern matching by ketema (Scribe) on Jun 07, 2005 at 16:54 UTC
Thanks That's perfect and easy.	[reply]
Re: Pattern matching by ikegami (Patriarch) on Jun 07, 2005 at 16:39 UTC
`($file_name) = $path =~ /([^\\])$/;` [download] But paths can often be seperated with "/", even in Windows. Modules usually know details like this, which is why it's usually important to use them. Refer to File::Basename. `($file_name) = $path =~ m#([^\\/])$#;` [download]	[reply] [d/l] [select]
Re: Pattern matching by jZed (Prior) on Jun 07, 2005 at 16:38 UTC
While I applaud your desire to learn regexes and am sure someone will have some suggestions in that regard, I'd recommend for this task that you instead use the File::Basename module which is part of core perl and is specifically designed to find paths and filenames in a portable fashion.	[reply]
Re: Pattern matching by davidrw (Prior) on Jun 07, 2005 at 16:57 UTC
Definitely go with File::Basename, and i like ikegami's solution above better than this one, but to directly translate your question of "find the last occurrence of \\ and then give me everything from that match to the end of the string" into a regex, i offer this: `$s =~ /\\(.?)$/; # UPDATE: THIS IS WRONG!!! my $endPart = $1;` [download] Four parts: 1) matches a (escaped) backslash 2) anchor to the end of the string with $ 3) capture all the stuff ( `.` ) inbetween by surrounding w/parens 4) make that cature non-greedy with the ? -- ~~this way it implicitly makes the \\ be the furthest one to the right in the string.~~ Update: OOPS. As pointed out, that greedy usage is wrong. I would just stick with the abover method of the non-backslash character class. Though this will work if matching the beginning of the string like `/^(.?)\\/` Update 2:* I think what i originally intended was `/.+\\(.*?)$/` As for grouping and back matching in general, the Tutorials and `perldoc perlre` and starting places. Quick & dirty grouping overview: Basically the way grouping works is anything you put in parens will get captured, and the first capture will be stored in $1, the second in $2, and so on. `my $s = "This is a blurb"; if( $s =~ /(\S+)\s\S+\s\S+\s(\S+)/ ){ print "$1:$2"; # This:blurb }` [download] Two quick notes: A) Always make sure the regex matched successfully before using $1, etc. B) You can prevent parens from capturing by using ?: like this: `/(?:blah)/`	[reply] [d/l] [select]
Re^2: Pattern matching by hv (Prior) on Jun 08, 2005 at 08:46 UTC
`$s =~ /\\(.?)$/;` ... 4) make that cature non-greedy with the ? -- this way it implicitly makes the \\ be the furthest one to the right in the string.* This is incorrect: greedy or non-greedy, the regexp engine will always try to match the string at a given location any way it can before it proceeds to try at the next location. This will therefore always match the first `"\\"`, not the last. See How will my regular expression match? for more information. Hugo	[reply] [d/l] [select]
Re: Pattern matching by tlm (Prior) on Jun 07, 2005 at 16:42 UTC
This `my $last_one = ( split, /\\\\/, $string )[ -1 ]` [download] will give you what you want. Alternatively, `my $last_one = ( $string =~ /([^\\](?!.\\).)$/ )[ 0 ];` [download] comes to mind, but it looks more complicated that it ought to be (this is largely because it is trying to accommodate the possibility that the desired string contain a single embedded backslash). I prefer the first (`split`) option over the second. Update: Fixed original regexp, which incorrectly included a single backslash at the start of the returned string. the lowliest monk	[reply] [d/l] [select]
(ignore) by tlm (Prior) on Jun 07, 2005 at 16:39 UTC
Posted in error. Please disregard. See 464384 instead. the lowliest monk	[reply]


Pathologically Eclectic Rubbish Lister
	PerlMonks