RE: Quantifiers in regular expressions

[Anonymous Monk]

in reply to Quantifiers in regular expressions

As a suggestion, this is something I had trouble with when I was learning regex. Non-greedy not only matches as little as possible, but it also *doesn't backtrack*; This is sort of an example that gave me trouble until I read about backtracking in the man page.

$string = "<foo>...</foo><bar>...</bar>";

# This matches <foo></bar>, not what we want.
$string =~ /\<.*\>(.*)\</.*\>/;
# this works, i think.
$string =~ /\<.*?\>(.*?)\</.*?\>/;

Even if it doesn't work, I hope you get the idea. I'd show the exact example of what I was doing when I ran into this, but it's overly-complicated (removing parts of tags from a string, where there's a list of tags and attributes for those tags that need to be removed).

Greediness relies a lot on backtracking, so to avoid frustrating another fledgling perl coder with the same problems, it's at least worth a note wherever greediness is in issue.

[pkn]

I don't understand this: $string = "<foo>...</foo><bar>...</bar>"; # This matches <foo></bar>, not what we want. $string =~ /\<.*\>(.*)\</.*\>/; How does it match? And I thought that the matching operator was m//, not m///. What does m/// do?

[le]

Yeah, AM escapes the wrong chars. Let's put it right. If you say:

$string =~ /<.*>(.*)<\/.*>/;

on a string "<foo>...</foo><bar>...</bar>" you get everything between "<foo>" and "</bar>", because the "*" modifier is "greedy", which means it tries to match as much as possible. A "." in a regex matches anything, so ".*" matches until the end of the string. Then the rest of the regex is evaluated, done by backtracking (the regex machine is now at the end of the string and goes back one by one until it finds a match). I hope this was correct.

Update:
Damn HTML escaping :) I fixed it, so the strings actually show up.

[Anonymous Monk]

This is maching tag delimiters and their belongond text. Let`s say you have text in H1 format <h1>This is my text</h1> and you want to replace both the heading and the text with only one step. Then you`ll need such a string as the mentioned above to find a matching pair.

[Anonymous Monk]

What is that u want to searh ? what is the pattern u are searching for?

[ninja-joe]

Here's a simple modification to the example code that will show what the regex matched when you typed it in:

#!/usr/bin/perl
while(<>) {
     chomp; # chomp so this next output is pretty.
            # newlines aren't discarded when you <>
     print "\"$1\" was matched out of \"$_\"" if m/(your_pattern)/;
}

Be sure to include the parenthesis around the entire regex that way it will save what it matches in $1.

Do recall that while the regex operators are greedy by default you can suffix them with ? and they'll go to nongreedy. An example:

#!/usr/bin/perl

while(<>) {
        chomp;
        print "\"$1\" was matched out of \"$_\"\n" if m/(\w{5}?)/;
}

This will match "mywor" out of "myword"

Have fun regexing.

[Roy Johnson]

Putting a ? on a quantifier that isn't variable-length is like having any color car you want, as long as it's black. There's no difference between a greedy and a non-greedy {5}.

A greedy quantifier will take as many characters as it can, and then start backtracking until whatever follows it matches. A non-greedy one will take as few characters as it can, and then take more until whatever is supposed to follow it is found.