Regex problem

gvanto

I have quite a strange regex problem, the following regex will match "5/2006" but not "5/2006 " no matter what i try and do to it ... ! :-(

$re = "((((?:(?:[0-2]?\d{1})|(?:[3][0,1]{1}))[-:\/.])?(?:[0]?[1-9]|[1][012])[-:\/.](?:(?:[1]{1}\d{1}\d{1}\d{1})|(?:[2]{1}\d{3})|(?:\d{1}\d{1})))(?![\d])$|((?:[0]?[1-9]|[1][012])[-:\/.](?:(?:[0-2]?\d{1})|(?:[3][0,1]{1}))[-:\/.](?:(?:[1]{1}\d{1}\d{1}\d{1})|(?:[2]{1}\d{3})|(?:\d{1}\d{1})))(?![\d]))";

$t = "5/2006 ";

if(preg_match("/$re/i", $t, $matches)>0) {
				$this->debug("matches", $matches);
			}
			else {
				$this->debug("", "no matches");
			}

Any advice on how to get this to match with ANYTHING on either side of the date, would be greatly appreciated!

G
regex newbie

Weedpacket

Quite a lot of that could be simplified: [3] is equivalent to 3, and {1} doesn't do anything, for example 🙂. And is that supposed to be [0,1] or [01]?

...It might be easier if you stated exactly what it's supposed to match.

Maybe it's just a matter of removing that '$' in the middle, there, but ... well, it's possible to use regular expressions to test if a number is divisible by 7, but there are much more sensible ways of going about it.

if(preg_match("/(\d\d?)([.-\/])(?:(\d\d?)\2)?(\d{4})/", $string, $m))
{
	if(count($m)==4)
	{
		$month = $m[1];
		$year = $m[3];
		if($month<1 || $month>12) return false;
		// Other tests, e.g. checkdate($month, 1, $year);
	}
	else
	{
		$day = $m[1];
		$month = $m[3];
		$year = $m[4];
		if($day<1 || $day>31) return false;
		if($month<1 || $month>12) return false;
		// Other tests, e.g. checkdate($month, $day, $year);
	}
	return true;
}

gvanto

Hi again Weedpacket,

I am trying to match ANY of these (no matter what they are surrounded with):

dd/mm/yy
dd/mm/yyyy
mm/dd/yy
mm/dd/yyyy
mm/yy
mm/yyyy

No need to account for no of days in Feb or anything ...

(where dd can be 'd', 'mm' can be 'm')

I got someone else to write this regex ... and only now realizing it doesn't quite work properly ...

Thanks for your help, much appreciated !

Gvanto

NogDog

<?php
$mon = '(0?[1-9]|1[012])';
$day = '(0?[1-9]|[12][0-9]|3[01])';
$year = '\d\d(\d\d)?';
$re = "#\b($mon/$day|($day/)?$mon)/$year\b#";

$text = <<<EOD
12/31/2009 is a date. So is 31/12/2009 or 01/01/2008.
However, 13/32/09 is not a date, though 9/9/09 and 08/99 are.
EOD;

preg_match_all($re, $text, $matches);

echo "<pre>".print_r($matches[0], 1)."</pre>";

nrg_alpha

Weedpacket;10909041 wrote:

if(preg_match("/(\d\d?)([.-\/])(?:(\d\d?)\2)?(\d{4})/", $string, $m))

The placement of the non escaped dash in a character class is crucial. Referring specifically to ([.-\/])

Since the dash is in the middle, this creates a range from . to / (much like [a-z] creates a range from, well, a to z). If this is not the intent, the non escaped dash should be the very first or very last character within the character class.

gvanto

Awesome thanks NogDog for an elegant solution, works perfectly!

many many thanks!
G

NogDog;10909062 wrote:

<?php
$mon = '(0?[1-9]|1[012])';
$day = '(0?[1-9]|[12][0-9]|3[01])';
$year = '\d\d(\d\d)?';
$re = "#\b($mon/$day|($day/)?$mon)/$year\b#";

$text = <<<EOD
12/31/2009 is a date. So is 31/12/2009 or 01/01/2008.
However, 13/32/09 is not a date, though 9/9/09 and 08/99 are.
EOD;

preg_match_all($re, $text, $matches);

echo "<pre>".print_r($matches[0], 1)."</pre>";

Weedpacket

Of course, by that standard, 31/04/2007 is a date.

gvanto

Hey WP,
yeah, for this case that doesn't really matter, so it's all good! :-)

Weedpacket

nrg_alpha;10909065 wrote:
The placement of the non escaped dash in a character class is crucial. Referring specifically to ([.-\/]).

Yeah, I knew there was something screwy about that bit, but I kept putting it down to the fact I was missing a candidate delimiter. For some now-unfathomable reason I was too lazy at the time to go back and look at the original; just as I was too lazy to look up that quote of Zawinski's about regular expressions.

Speaking of delimiters, is the range of options still desirable?

gvanto

Oooh yeah, good point. the Germans use '.' between dates!

NogDog

gvanto;10909135 wrote:
Oooh yeah, good point. the Germans use '.' between dates!

Take 2:

<?php
$m = '(0?[1-9]|1[012])';
$d = '(0?[1-9]|[12][0-9]|3[01])';
$y = '\d\d(\d\d)?';
$s = '[-./]';
$re = "#\\b($m(?P<x>$s)$d(?P=x)$y|$d(?P<y>$s)$m(?P=y)$y|$m$s$y)\\b#";

$text = <<<EOD
12/31/2009 is a date. So is 31-12-2009 or 01.01.2008.
However, 13/32/09 is not a date, though 9/9/09 and 08-99 are.
EOD;

preg_match_all($re, $text, $matches);

echo "<pre>".print_r($matches[0], 1)."</pre>";

nrg_alpha

Weedpacket;10909134 wrote:
Yeah, I knew there was something screwy about that bit, but I kept putting it down to the fact I was missing a candidate delimiter. For some now-unfathomable reason I was too lazy at the time to go back and look at the original; just as I was too lazy to look up that quote of Zawinski's about regular expressions.

hehe.. I know the feeling. Laziness alleviates issues, thus stress 😃

Weedpacket wrote:
Speaking of delimiters, is the range of options still desirable?

Sorry, I'm not sure I follow. Are you asking what range of characters can or cannot be delimiters? (if so, any non alphanumeric, non whitespace ASCII character [except a backslash]).

Weedpacket

It's just that the original regexp allowed :, ., /, or - to be used to separate the components of the date; but that list appears to have shrunk down to just /.

gvanto

NogDog;10909140 wrote:

Take 2:

<?php
$m = '(0?[1-9]|1[012])';
$d = '(0?[1-9]|[12][0-9]|3[01])';
$y = '\d\d(\d\d)?';
$s = '[-./]';
$re = "#\\b($m(?P<x>$s)$d(?P=x)$y|$d(?P<y>$s)$m(?P=y)$y|$m$s$y)\\b#";

$text = <<<EOD
12/31/2009 is a date. So is 31-12-2009 or 01.01.2008.
However, 13/32/09 is not a date, though 9/9/09 and 08-99 are.
EOD;

preg_match_all($re, $text, $matches);

echo "<pre>".print_r($matches[0], 1)."</pre>";

Thanks NogDog but I get for this one:

Warning: preg_match() [function.preg-match]: Unknown modifier ']'

gvanto

PS: I am using this in conjunction with other regex's ... (none similar to this one though).

NogDog

I copy-and-pasted my "take 2" code and it ran fine for me. 😕

nrg_alpha

gvanto;10909377 wrote:
Warning: preg_match() [function.preg-match]: Unknown modifier ']'

I tested Nog's take-2 and it worked as well. The warning you are getting is is most likely because you are using a certain character as an opening delimiter, but have the equivalent of a closing delimiter character somewhere in your pattern [other than the actual closing delimiter]. This in turn causes the regex engine to see that equivalent closing delimiter character as indeed the final actual closing delimiter and thus all the remaining characters after that as modifiers, which will quickly land you into trouble...

some example could be something like:

preg_match('[(\d+)[^]]+]', $text, $match);

preg_match('%(\d+)[^%+]%', $text, $match);

This gives the + sign an unknown modifier, as the first ] (or % in the second example) character in the pattern causes the regex engine to treat it as the closing delimiter (thus, characters that follow must be legal modifiers (such as i, s, u etc...). The + sign is not a legal modifier, hence the error:

Warning: preg_match() [function.preg-match]: Unknown modifier '+'

The best solution to unknown modifiers is to use characters for delimiters that are not found within your pattern. I would suggest stuff like #, ! or ~, as these are 'generally safer' to use. Otherwise, you will need to escape seeming closing character(s) within your actual pattern.

So the solution to the second example would be:

preg_match('%(\d+)[^\%+]%', $text, $match);

gvanto

Hey yeah thanks it does work Nogdog, only not when I combine it with a few other regex's but obviously this cant be your regex's fault ... 😃

Thanks much for your help! I've learnt much here