Regular expression to remove HTML-style comments
 

I would like to use PHP's output buffering capability to remove HTML-style comments from web pages as they are served to the client. This would be useful for removing comments inserted by HTML editing software such as Dreamweaver, for example. I have output buffering working, but the regular expression that I'm using is causing me some headache:
return (ereg_replace("<!--.*-->", "", $buffer));

This works really well... a little TOO well. If I have a string like this:
<!--Comment #1-->Text<!--Comment#2-->
The whole string is replaced, including Text. What I would like to happen, is that Text is not replaced, just the comments on either side.

I'd appreciate any assistance with accomplishing this.
Thanks, Matt

This is because the PHP ereg family of routines are 'greedy'... :(

There is no easy way to get around it short of using a negated character class, like so:
(ereg_replace("<!--[^\-]+-->", "", $buffer))

Your best bet is to use the much more civilized PCRE library which does support non-greedy qualifiers...
/<!--.*?-->/

--
Terra
--ereg and friends need therapy--
FutureQuest

One of the problems with PHP is that it's too large, so it's easy to miss built-in-functions, like, say, strip_tags() which is specifically meant to strip HTML tags. Haven't tried it on something with comment tags, but I'd certainly hope they'd be wiped out with all the rest, and there's a good chance this function is faster than a general regex. (It's also impossible to strip some valid HTML tags using a regex, but usually it's not a big problem).

Thanks for the quick feedback :) Terra, I'll take a look at the PCRE library. Paul, that is an interesting idea, but in this case, I want to keep most of the HTML code, just not the comment tags. It looks like with strip_tags() I would have to define all the tags that I didn't want stripped (which would be everything except comments). If this is correct, then I'm not sure that strip_tags() would be faster due to the large array of valid tags that would have to be evaluated/ passed over. Were you thinking of something a little different? -Matt

Something else to watch out for, if you don't have control of the HTML doc, are nested comment tags. For example:

<!-- Begin Cmt <!-- Nested Cmt --> Hi There -->

Using a typical regex looking for the open tag and close tag will result in:

Hi There -->

Strangely enough, you will have to use a bit more than that. Since "." does not match line breaks, use:

/<!--(.|\s)*?-->/

instead, as in:

$x= preg_replace('/<!--(.|\s)*?-->/', '', $x);

Because <!-- comments
can be several lines
long -->.

Netscape and IE parse nested comments that way anyway. So it would look yucky even before you applied the regex to it. I thought of the same thing and tested it. But there probably is some oddball case we're missing.

That's why you should use the /s attribute to treat the "string" as a single line; newlines will then match the period '.'

Look for PCRE_DOTALL in http://www.zend.com/manual/pcre.pattern.modifiers.php

Thanks everyone, it now works perfectly (I think). For anyone interested in stripping comments from HTML pages, here is the code I'm using:

One of the problems with being up at 4:00 in the morning is that phrases like "HTML-style comments" are easily misread... :ididthat: