pcre returning shortest match

[redbeardmcg]

Hi everyone,

I'm having an issue splitting a string with preg_split.... Here is a sample string:

congestion >= 80

Where >= could also be <=, ==, !=, < or >

What I need to accomplish is splitting this, and put each of the three parts (item, operator, value) into an array.... Here is how I'm doing this:

$this_expr = preg_split('/(==|>=|<=|<|>|\!=)/U', $rule, -1, PREG_SPLIT_DELIM_CAPTURE);

This works great so long as the operator is one of ==, !=, <, or >.... but where the operator is >= or <= (as in this case) pcre matches ONLY the > or < so that the resultant array looks like this:

$this_expr[0] = congestion;
$this_expr[1] = >;
$this_expr[2] = =80;

Clearly the operator should be >= and the value should be only 80.... Any ideas of how I could get around this? I did try changing the order (put >= | <= after the < | > in the sub-patter) to no avail... It seems to return only the shortest match.

I could of course write a HACK (if the value starts with = append it to the operator and strip it from the value) but I am so dead set against hacking things together that it hurts me to even consider that as an option.

Thanks!

-Ryan

[Dormilich]

what about a character class?

[<>=!]{1,3}

[Atli]

Hi.

This doesn't seem to happen on my side.
I tried this code:
[code=php]<?php
header("content-type: text/plain");

$str = "congestion >= 80";
$reg = '/(==|>=|<=|<|>|\ !=)/U';

$out = preg_split($reg , $str, -1, PREG_SPLIT_DELI M_CAPTURE);

print_r($out);
?>[/code]
And I got these results:

Array
(
    [0] => congestion 
    [1] => >=
    [2] =>  80
)

Isn't that what you would expect?

P.S.
Using PHP 5.2 over here.

[redbeardmcg]

Thank you both for the responses... Dormilich, your example should work however I think you need to change [<>=!]{1,3} to [<>=!]+ to make it more greedy.

Atli, I'm running php 5.1.6 :-/

I did however realize that I am totally over-engineering this.... They are always separated by a space... Using explode(' ', $string) works more than fine!

Sometimes simple is better.

Thanks again!

-Ryan

[Dormilich]

however I think you need to change [<>=!]{1,3} to [<>=!]+ to make it more greedy.

... if you use operators longer than 4 characters... anyway REs are in greedy mode by default.

[Atli]

Dormilich, your example should work however I think you need to change [<>=!]{1,3} to [<>=!]+ to make it more greedy.

How about: '/\s*?([<>=!]{1,3}?)\s*?/U'.
Even strips away the spaces.

Anyways, glad you found a solution :)

[redbeardmcg]

You're absolutely right, I spoke too soon. Thanks again guys!

-Ryan

[Markus]

How about: '/\s*?([<>=!]{1,3}?)\s*?/U'.
Even strips away the spaces.

Anyways, glad you found a solution :)

Being slightly picky here, I apologise.

Wouldn't that allow for ==>, ==<, etc? Not sure if you are bothered about it, but if you are, you could improve it a little.

Maybe: '/\s*?([<>!]{1}[=]{1,2}?)\s*?/U'

Would the square brackets be needed around the '='? Also, would the above work? I haven't tested it.

Mark (off to read Lemony Snicket's: A Series of Unfortunate Events :)

[Atli]

Yea, that's even better. Makes sure you don't get matches on operators like <<< or >>=.

Would the square brackets be needed around the '='? Also, would the above work? I haven't tested it.

No, you don't need to create a character class for a single char.
You can just do: '={1,2}'

Mark (off to read Lemony Snicket's: A Series of Unfortunate Events :)

A very good story. One of my favorites :)
(The movie was good to... Well, how could it not; starring Jim Carrey :P)