reg expression help

Maiku

I was to make it so a user cannot put special characters into their user name. The only two that I want allowed are underscore(_) and space ( ). I made an exression that checks for straight alphanumeric characters, but it doesn't allow spaces of underscores:

ereg('^[[:alnum:]]+$', $data)

Can anyone help me with this? 😕

pyro

Try this:

<?PHP
$username = "test";
if (preg_match("/^[\\w ]+$/", $username)) {
	echo "username is good";
}
else {
	echo "username is bad";
}
?>

Maiku

That is exactly what I'm looking for, thank you.

Does anyone know a good website that covers how to make regular expressions?

pyro

The php.net site covers it pretty well: http://us4.php.net/pcre

Sxooter

Originally posted by Maiku
I was to make it so a user cannot put special characters into their user name. The only two that I want allowed are underscore(_) and space ( ). I made an exression that checks for straight alphanumeric characters, but it doesn't allow spaces of underscores:
ereg('^[[:alnum:]]+$', $data)
Can anyone help me with this? 😕 [/B]

ereg('^[ _[:alnum:]]+$', $data)

Note that for ranges with space, the space needs to come first, like above

pyro

The one I posted seems quite a bit simpler...

Sxooter

The one I posted is more easily understood. I.e. self-documenting.

i.e. what does [w ] mean? [:alnum:] pretty obviously means alphanumeric.

Further, more applications (i.e. postgresql) support posix regex than do perl regex, and on top of that, posix regexes don't change very much, whereas perl regexes are set to change, once again, with the introuction of perl 6.

further, posix regexs are less error prone in my experience. But to each his own. I use posix regex because I do half my development in postgresql and half in php, and I don't have to learn / remember two different systems.

Weedpacket

To each vis own, but I use PCRE regexps in PHP because they're faster and more powerful than POSIX regexps.

(Looking forward to plPHP 1.0....)

Sxooter

I here all the time how much faster perl regex is than posix regex. So, I thought I'd test it instead of just taking people's word for it.

OK, here's the script I ran to compare the performance of these two regexes:

<?php
function mt(){
        $tmp = split(" ",microtime());
        return $tmp[0]+$tmp[1];
}
$string = "Bob Jones_University*";
$tp[] = mt();
for ($i=0;$i<100000;$i++){
        $a = ereg('^[ _[:alnum:]]+$', $string);
}
$tp[] = mt();
for ($i=0;$i<100000;$i++){
        $a = preg_match("/^[\w ]+$/", $string);
}
$tp[] = mt();

print $tp[1]-$tp[0];
print "\n";
print $tp[2]-$tp[1];
print "\n";

?>

And the output I got was:

1.3014370203018 <- regex
1.0914809703827 <- perl regex

Let's face it, the performance difference isn't that great. Maybe if you're importing 1,000,000 rows you might see a 3 second difference processing the rows with pcregex versus posixregex.

In what way are perl regex generally more powerful? Just wondering, I've never spent a lot of time working on them because I consider the syntax to be confusing and it reminds me way too much of my old days wrestling with perl regex when I coded in perl.

Weedpacket

Originally posted by Sxooter
In what way are perl regex generally more powerful? Just wondering, I've never spent a lot of time working on them because I consider the syntax to be confusing and it reminds me way too much of my old days wrestling with perl regex when I coded in perl.

Oh, I've got a few examples in this thread, it might be interesting to rewrite them to use POSIX regexps. One property that I haven't used there yet is that you can turn on or off things like case-sensitivity for individual parts of a regexp: if one bit has to be case-sensitive and another doesn't, the other bit [dD][oO][eE][sS][nN]'[tT] [hH][aA][vV][eE] [tT][oO] [bB][eE] [lL][iI][kK][eE] [tT][hH][iI][sS].

Also, as noted on the POSIX [man]regex[/man] manual page, they're not binary-safe (though I haven't used them for binary data yet anyway).

And of course there is the fact that POSIX regexps lack ungreedy matching. Without that they can be exponentially longer than equivalent PCRE expressions like

#<script(\s[^>]*)?>.*?</script>#s

which would be written in POSIX ... how? (I'm too distracted right now to figure it out for myself. I'll look through my notes later.)

The only feature that is theoretically more powerful, in the sense that additional code would have to be ritten to emulate it using POSIX regexps, is the /e modifier, but that isn't so much a property of the expression syntax itself as how the results are used. But "exponentially longer" does have practical relevance (consider what that would do to running time).