Registered: Apr 2002
Location: On a collision course with reality.
Posts: 334

How can I use regular expressions to match for any word/string that contains an expression totally dependent upon an earlier match in the same string?

For a simplified example, I would like to match all string starting and ending with the same letter (Example string: "car noon treat boys hannah", I would like the regexp that matches against "noon", "treat", and "hanna", and capturing "n", "t", and "h").

For bonus points: I would like to take this problem further though, particularly in matching html tags:

I.e.: I would like to scan my entire html document and match against all 2-part html tags, such as "<b>BOLD TEST</b>", "<a href=whater>sometext</a>", and "<img src="image.jpg>sometext</img>", capturing "b", "a", and "img".

I don't need involved help on the second part (unless someone happens to know it offhand). I'm sure that I can extrapolate the solution to the html tag matcher by knowing the solution to the first problem I posed.

thanks!

Last edited by Daniel on 01-19-2006 at 01:40 AM

Report this post to a moderator | IP: Logged

Registered: Nov 2000
Location: UK
Posts: 4737

Your first problem is palindromic in nature and there's some stuff out there for solving it(a common example is finding balanced brackets to extract function snippets from code). But assuming you split the sentence yourself into words, something as simplistic as:

code:

$word =~ /^(\w)(\w+)(\1)$/

Report this post to a moderator | IP: Logged

Registered: Apr 2002
Location: On a collision course with reality.
Posts: 334

Probably both, actually.

Thanks, I eventually figured out the solution you gave above. It was simple of course, I had just never used backreferences before.

My next problem is quite frustrating: I'm trying to match the smallest possible tag-pair that is nested inside other tag-pairs of the same kind, and may alsohave other unrelated tags inside it (so no simple [^<]* trickery).

E.g.: "<ul ><li>1<li>2<ul><li>2a<li>2b</ul><li>3</ul>"

I want to match "<ul><li>2a<li>2b</ul>". The "<ul>" is actually a literal. I'm going bald quickly from all the hair pulling...

I feel that this is simple and I am overlooking something obvious, but the best I've been able to get is "<ul><li>1<li>2<ul><li>2a<li>2b</ul>". I need some way to specify "a string that starts with the one-and-only <ul> and ends with </ul>"

Last edited by Daniel on 01-22-2006 at 04:20 AM

Report this post to a moderator | IP: Logged

Registered: Apr 2002
Location: On a collision course with reality.
Posts: 334

I figured it out. Unfortunately I had to do a look-ahead on a character by character basis, but it works:

$word =~ /<ul[^>]*?>(.(?!<\/?ul> ))*.(?=<\/ul> )<\/ul>/

Now it will match only <ul> tag-pairs that don't contain further <ul> nesting.

Report this post to a moderator | IP: Logged

Registered: Nov 2000
Location: UK
Posts: 4737

Cool. I was going to suggest a lookahead but wanted to reread the perlre perldoc to make sure it would actually do what you wanted(I think I've used lookaheads twice in my entire life, and regretted it a few months later when re-reading the code).

Report this post to a moderator | IP: Logged