Regex has problems with poitive lookbehind

Question

I am currently trying to create a regex that strips unecessary quotation marks from HTML tags. The regex will be used in PHP code.

<input type="image" src="/flags/en.png" alt="English" title="English" name="en" class="screen selected" />

converts to

<input type=image src="/flags/en.png" alt=English title=English name=en class="screen selected" />

I have come up with this regex and replacement:

/(?<=<(?:[^>]+?\s)?)([\w-]+=)"([\w-]+)"(?=(?:\s[^>]+)?>)/g
$1$2

The problem is that the positive lookbehind does not allow quantifiers (See http://regex101.com/ as a reference.).

So I thought I modify the pattern a little bit like this:

/(<(?:[^>]+?\s)?)([\w-]+=)"([\w-]+)"((?:\s[^>]+)?>)/g
$1$2$3$4

Now it's valid but it only strips one set of quotes from each tag.

How do I acomplish this?

Smaller HTML. I do this to speed up my webpage. It is not 100% necessary but I want to do it since I already started it! — BrainStone, Aug 13 '14 at 21:20
this is insane microopertimisation (no idea how to spell that) — , Aug 13 '14 at 21:25
Call me insane or OCD but considering that this could easily save several KiB of code I think it is worth my time! *(My current opimizations saved me something around 10%. I did this optimization by hand and it saved an additional 2%. That makes indeed a differnece.)* — BrainStone, Aug 13 '14 at 21:28
ok well lets do this properly then: rename all the files to a single letter "/flags/en.png" to just "f" then mod rewrite rules; replace ever line break\tab; you don't need those optional elements in the input such as title at all, so strip them ... — , Aug 13 '14 at 21:32
I'm trying to **losslessly** optimize the HTML site without changing it's content. Striping elemets like the title attribute would change the page itself *(how the user sees it)* and have an impact on the usability. — BrainStone, Aug 13 '14 at 21:59

score 1 · Accepted Answer · answered Aug 13 '14 at 21:31

1

Try the following:

$pattern = '/(<(?:[^>]+?\s)?)([\w-]+=)"([\w-]+)"((?:\s[^>]+)?>)/';
$replacement = '$1$2$3$4';
$subject = '<input type="image" src="/flags/en.png" alt="English" title="English" name="en" class="screen selected" />';

while(preg_match($pattern, $subject)){
    $subject = preg_replace($pattern, $replacement, $subject);
}
var_dump($subject);

answered Aug 13 '14 at 21:31

paolo

2,528
3
17
25

This works. Thank you. I wonder why I didn't think of this before! – BrainStone Aug 13 '14 at 21:38

Erlesand · Answer 2 · 2014-08-13T21:43:18.680

0

Probably won't save much, but here you go :)

 $string = '<input type="image" src="/flags/en.png" alt="English" title="English" name="en" class="screen selected" />'; 
 echo preg_replace('/="([a-z]+)"/i', '=$1', $string);

Output:

<input type=image src="/flags/en.png" alt=English title=English name=en class="screen selected" />

edited Aug 13 '14 at 21:43

answered Aug 13 '14 at 21:31

Erlesand

1,525
12
16

This can't work since there is some quoted text outside of HTML tags. These need to stay there! – BrainStone Aug 13 '14 at 21:39
Then the edited version ought to work if you want to keep it simple :) – Erlesand Aug 13 '14 at 21:44
Still there is a good chance that this might strip some quotes from outside the HTML tags... – BrainStone Aug 13 '14 at 21:44
If you were to type something such as ="blah" it would replace it with =blah, but fair enough :) – Erlesand Aug 13 '14 at 21:46
1

I think your solution is great in terms of simplicity but @paolo's solution is far more reliable! – BrainStone Aug 13 '14 at 21:55

Regex has problems with poitive lookbehind

2 Answers2