[RESOLVED] Regex Lookbehind & Metacharacter Question?

[joe_C_nice]

Is it possible to use a metacharacter in a lookbehind?

Here's sample code.

$foo = "some random \t text";
preg_match_all("/(?<!\\)\w+/i", $foo, $bar);
print_r($bar);

Btw when I execute this code I get.

Warning: preg_match_all(): Compilation failed: lookbehind assertion is not fixed length at offset *

[NogDog]

Be careful with backslashes within a double-quoted string. You'll probably find it simpler to use single quotes around your regexp string.

[nrg_alpha]

You must be confusing that error message with perhaps a previous preg attempt? When I cut and paste your code and test it, I get:

Warning: preg_match_all() [function.preg-match-all]: Compilation failed: missing ) at offset 9 in C:... .. .. .

I think you mean

preg_match_all("/(?<!\\\)\w+/i", $foo, $bar); // notice the three escape slashed bundled.. 

As for your look behind assertion, you cannot have any unknown lengths..
so for example,

preg_match('#(?<=\s+)\w+#', $str, $match); 
// illegal due to the + quantifier. Cannot have quantifiers in a lookbehind assertion..

I think you can have quantifiers in look ahead assertions.. but not in look behinds..

Cheers,

NRG

[NogDog]

Within double quotes, you'll actually need 4 back-slashes:

"/(?<!\\\\)\w+/i"

PS: Actually, I just tested it, and it works with 3 or 4 back-slashes, but not with 2:

<?php
header('Content-Type: text/plain');
echo "/(?<!\\)\w+/i" . "\n";
echo "/(?<!\\\)\w+/i" . "\n";
echo "/(?<!\\\\)\w+/i" . "\n";
?>

Output:

/(?<!\)\w+/i
/(?<!\\)\w+/i
/(?<!\\)\w+/i

[joe_C_nice]

nrg_alpha the three escapes got it working. Thank you

Edit: NogDog thank you for that. In the example I provided there's a double quote but withing my code it's single quote. Didn't think it had much on an impact of the code itself.

[joe_C_nice]

Question I understand that for lookbehind the lenght cannot be unknown, but how does have '\\' or even '\\' accomplish this?

[NogDog]

The extra slashes are needed because in a PHP string literal, "\" means one literal backslash, i.e. the first slash is saying to read the next slash as a literal character and ignore any special escaping meaning it might have. So in your original example, the regexp string is parsed by PHP to be just one backslash, which within the PCRE regexp parser is then applied to the following ")" to say that the PCRE parse should treat it as a literal closing parenthesis character, rather than its special PCRE meaning of closing a pattern (in this case the look-behind assertion).

[joe_C_nice]

Makes sense, thank you for that explanation.

[NogDog]

joe_C_nice;10883475 wrote:

Makes sense, thank you for that explanation.

Glad it helped, it pretty much confused me while I was typing it. 😉