匹配模式内的模式

Question

I'm trying to match any bracketed items within <sup> tags.

My regular expression is being too greedy, starting with the first <sup> tag and ending at the last </sup> tag.

/<sup\b[^>]*>(.*?)\[(.*?)\](.*?)<\/sup>/

Example html:

<sup>[this should be gone]</sup>
<sup>but this should stay</sup>
<sup>this should [ also stay</sup>
[and this as well]
<sup><a href="#">[but this should definitely go]</a></sup>

Any idea why?

Thanks!

EDIT: I suppose these answers make sense. I've got much of the HTML parsed without regex; I just figured that this particular example would work with regex because it would do the following:

1. see the first <sup> tag
2. find the first instance of </sup>
3. search the inside for (wild)(bracket)(wild)(closing bracket)(wild)

Answer 1

You really can't do this. It's impossible to parse HTMl with regular expressions, because regular expressions can only match regular languages; these languages are a simpler subset of the actual languages we use. One very common non-regular language is the Dyck language of balanced brackets; it's impossible to match correctly nested parentheses with regular expressions. And HTML, if you think about it, is the same as this, with tags replacing parentheses. Thus, matching (a) correctly nested sup tags is impossible, and (b) matching balanced braces is impossible. I don't use PHP myself, but I know it has access to an HTML DOM; I'd recommend using that instead. Then, filter through that for every sup tag, and check each one's inner text. If you only want to catch tags whose inner text is just [...], where the ... does not contain square brackets, you can use ^\[[^\]]+\]$ as your regex; if you want real nesting, more complicated checking is necessary.