[RESOLVED] help with preg_replace

mikawhat

As i'm completely useless with regular expressions i'll need some assistance... I'm getting some data which contain links and i want to remove those links by deleting the entire <a> tag or by changing the content inside the href="..." to something like javascript:void(0).

Could anybody advice? Thanks!

mikawhat

well i found something that works, but i have no idea what it does....
$link = eregi_replace('<a href=("|\')?([^ "\'])("|\')>([^<])</a>', '\4', $link);

nrg_alpha

$link = preg_replace('#<a [^>]+hef[^>]+>.*?</a>#si', '', $link);
// substitute $link with actual string variable name

@:
I would not use ereg anymore, as it is depreciated as of PHP 5.3 and will not be included within the core as of version 6. I would recommend learning PCRE instead.

LatecomerX

mikawhat;10920331 wrote:
well i found something that works, but i have no idea what it does....
$link = eregi_replace('<a href=("|\')?([^ "\'])("|\')>([^<])</a>', '\4', $link);

It basically changes "<a href="http://example.com/">Link 1</a>" into "Link 1" - the matched text is replaced by the text found in the fourth set of parentheses (\4).

And as suggested by nrg_alpha, it will be better to use preg_* (I don't remember why, but I did some Googling on this a long time ago). The preg equivalent for this would be:

$link = preg_replace('#<a href=("|\')?([^ "\'])("|\')>([^<])</a>#i', '$4', $link);

which is almost identical, except that we add a boundary to the pattern (I used '#'), an 'i' modifier to tell Preggy to be case-insensitive, and backreferences are in the format of $n instead of \n (but iirc, either is accepted in preg).

Live example:
http://phpieceofcake.com/view.php?id=1246664069360625

nrg_alpha

LatecomerX;10920372 wrote:
...he preg equivalent for this would be:
$link = preg_replace('#<a href=("|\')?([^ "\'])("|\')>([^<])</a>#i', '$4', $link);

Granted, this pattern can cause issues if there is an anchor tag as such:

<a class="whatever" href="http://www.whatever.com"><strong>link</strong></a>

Notice the snag(s)? With a pattern declaring <a href... this means that the href attribute has to come first, otherwise, that pattern will fail. Not only that, but there can only be the href attribute, followed by >

The second snag is ([^<]*)</a>. Notice in my example the link text is surrounded in an html tag.. this too will cause the pattern to fail. My previous post escapes these issues altogether.. so long as it is an anchor tag with an href attribute, it will replace it.

Also, character classes are more efficient than alternations.. so something like ("|\')? creates issues of slower speeds (ok, admittedy we won't perceive the difference...), and creates an unecessary capture. Simply using ["\']? acheives the same thing, only slightly quicker and without the overhead of capuring. If using the alternation method without capturing, simply use (?:"|\')? instead (but the character class is still better suited in this case).

LatecomerX

nrg_alpha;10920373 wrote:
Granted, this pattern can cause issues if there is an anchor tag as such:
<a class="whatever" href="http://www.whatever.com"><strong>link</strong></a>
Notice the snag(s)? With a pattern declaring <a href... this means that the href attribute has to come first, otherwise, that pattern will fail. Not only that, but there can only be the href attribute, followed by >

The second snag is ([^<]*)</a>. Notice in my example the link text is surrounded in an html tag.. this too will cause the pattern to fail. My previous post escapes these issues altogether.. so long as it is an anchor tag with an href attribute, it will replace it.

Also, character classes are more efficient than alternations.. so something like ("|\')? creates issues of slower speeds (ok, admittedy we won't perceive the difference...), and creates an unecessary capture. Simply using ["\']? acheives the same thing, only slightly quicker and without the overhead of capuring. If using the alternation method without capturing, simply use (?:"|\')? instead (but the character class is still better suited in this case).

I agree with your points above, and would definitely recommend mika to use your near-perfect regex (change that "hef" typo to "href"). But I'm just explaining what his regex does (which is what he asked), and am pointing out how can one convert a ereg pattern to a preg one.

Btw, nice tip on the non-capturing parentheses syntax. I wanted to do something like that for quite a few times in the past but couldn't really figure it out (character classes only work when you're matching single characters).

mikawhat

Great. Thank you guys!

nrg_alpha

LatecomerX;10920374 wrote:
I agree with your points above, and would definitely recommend mika to use your near-perfect regex (change that "hef" typo to "href"). But I'm just explaining what his regex does (which is what he asked), and am pointing out how can one convert a ereg pattern to a preg one.

Sorry, my previous comment wasn't geared towards you specifically, but rather in general for the pattern's inherent problems (I probably should have explained that earlier).

Oops... good call on the typo.. my bad (yeah, that typo would also break the pattern). ;^P

LatecomerX; wrote:
...(character classes only work when you're matching single characters).

Indeed.. that's why in this case I would have used the character class instead of using alternations (since the alternation makes use of one character or the other kind of thing).. Once we start getting into a sequence of characters, the character class obviously won't cut it.

So much preg, so little time 😉