Replace a pattern only within links

Question

I've got a string like this:

Google is a <a href="http://hi.hi?xxx&yyy&zzz">web&amp;search engine</a>.

I want to replace & with & only within links, as needed by W3C validator:

Google is a <a href="http://hi.hi?xxx&amp;yyy&amp;zzz">web&amp;search engine</a>.

Could you suggest a regexp for that? Thanks!

I have edited the question, now you can see. The initial version was not describing my situation. — Dmitry Isaev, Sep 28 '12 at 05:42

score 1 · Answer 1 · edited May 23 '17 at 10:24

The official correct answer is that you should not use a regex to parse HTML. Instead, take a look at HTML-parsing libraries. This question covers your options:

How do you parse and process HTML/XML in PHP?

I suggest taking this approach. Once you use a tool like DOM to parse the HTML, you can use a simple regex to perform your replacement within the links. People will be glad to help if you have trouble.

If you do insist on using a regex for this (and it can be ok in some limited cases where the HTML content is under your control) just search this site, and you will find tons of questions in which people show how to do this.

I need a very fast solution for high-load project. The only convenient thing seems to be preg_replace(). — Dmitry Isaev, Sep 28 '12 at 05:43

score 1 · Accepted Answer · answered Sep 28 '12 at 05:51

As dan1111 noted, regexes are a brittle tool for this at best. The next problem is that you would need variable-length lookbehind assertions to get to some degree of reliability that makes me feel comfortable.

That said, it may well be that the following works well enough for you - give it a try on some data that you've backed up before:

$result = preg_replace('/&(?=[^<>]*>)/', '&amp;', $subject);

This replaces an & only if the next angle bracket is a closing angle bracket.

A nice idea. I don't have a lot of angle brackets in my texts, so in my case it should work. Thanks! — Dmitry Isaev, Sep 28 '12 at 05:56

score 0 · Answer 3 · answered Sep 27 '12 at 15:01

0

you can use a lookahead and lookbehind.

&(?<=\<a\s(href).*)(?=.*\"\>)

What this does is look for all & preceded with < href and any characters, and also any character followed by a "> and. When I tested on RegexHero it selected only the & within the link itself.

answered Sep 27 '12 at 15:01

Nick

4,302
2
24
38

..but it doesn't work in PHP because it uses variable-length lookbehind assertions, and PHP can't handle those. Also, even in .NET where they *are* supported, your solution would not work correctly because you're using `.*` which is way to unspecific. – Tim Pietzcker Sep 28 '12 at 05:46
Yes. Still cannot find a way to use it on PHP. – Dmitry Isaev Sep 28 '12 at 05:49
I'm not sure what you mean by too unspecific. It doesn't need to be very specific, since he's looking for a known and fixed pattern. He wants to change all `&` between ``. The `.*` simply tells the regex to not care how many characters are after the lookahead/lookbehind. The key parts are making sure it falls within the required strings. As far as the PHP part, I've never used PHP, so I was just looking at it how I would do it in PowerShell and posting the regex hoping it would help guide to the answer he needed. – Nick Sep 28 '12 at 15:20

Replace a pattern only within links

3 Answers3