Nasty nested regex?

Desdinova

OK so I'm trying to build a template engine that uses a small amount of simple logic. The issue I encounter is with somekind of if-statement, which I have like this:

{IF [SOMEVAR]}
<p>Hi, how are you?</p>
{END IF}

Fairly easy.

The issue is that these statements can be nested in each other, ideally without having a limited depth. To make it more challenging, the statements can also follow each other up.

An example of a nasty piece that doesn't come out right:

{IF [AAA]}
	<p>single AAA</p>
{END IF}
{IF [BBB]}
	<p>single BBB</p>
{END IF}

{IF [AAA]}
	<p>AAA</p>
	{IF [BBB]}
		<p>Nested BBB</p>
	{END IF}
{END IF}

I got this piece of code working so that it matches the first {END IF} after hitting an {IF. This breaks the nested IF statements.

I also got this working so that it matches the last {END IF}, this breaks even more.

The regex strings I used are these:

'/(\{IF \[)(.+)(\]\})(.+)(\{END IF\})/iUse',
'/(\{IF \[.+?\]\})(.+\{IF \[.+\]\}.*\{END IF\}.*)(\{END IF\})/ise',

I'm not much of a regex pro, so if it can be improved anywhere, that would be great.

Any help is truly appreciated.

etully

I don't know the answer... but I've always been curious about the same thing so I looked at my copy of O'Reilly's Mastering Regexp's 2nd Ed. Nothing. Then I checked the 3rd edition and there's a section about the topic. Chapter 10.5, page 475. Recursive expressions, matching nested text. Requires use of the "or" operation (the | pipe) and also the "++" operator which I don't understand yet. Sorry that I don't have an answer for you but hopefully that lead might get you started.

Desdinova

Thanks a bunch for your response. I find it fairly odd that there's so little to find on this. It seems each programming language does the same, though maybe not through regex.

I have worked around it in the meantime with an irrelevant solution for any other readers.

The usage of the OR-symbol will probably mean there's a limit on the amount of nested statements, depending on how much of those ORs you put in there. I guess. but I'm no king regex either 😉

reli4nt

I am no regex expert but I don't see how it should matter if it is nested or not so long as it is nested propoerly. You control the names of the nested sections so couldn't you use an identifier such as the top level beginning with a, the next with b and so on. Then is seems like you could:
Search through for all sections,
order them by your identifier
replace them from innermost to outermost

This way there is no need to worry about placing nested content where an outer div was replaced.

etully

The usage of the OR-symbol will probably mean there's a limit on the amount of nested statements

Actually, no. The way they've written it is very clever. By use of the * (zero or more), they are saying:

Look for OPEN followed by zero or more instances of CLOSE OR OPEN...CLOSE followed by a CLOSE. And that will return the major OPEN...CLOSE blocks in your document. Then you pass each of those blocks to a function to see if they have any nested OPEN...CLOSE blocks inside. The function keeps calling itself, over and over (that's the recursive part) until there are no more OPEN...CLOSE blocks inside each block.

So it's designed for precisely the problem you described, no matter how many levels deep your nesting goes.

It requires use of practically every advanced regexp feature; greed, laziness, alternation, possessive quantifiers - and throw in a recursive PHP function. My Regexp-foo is pretty good but this technique relies on a few functions I've never used before. I got the answer - I just need to understand it!

Desdinova

OK, Cool.

The recursive PHP function is already in place, so that shouldn't matter much.

Could you perhaps write an example from that book out here? Maybe it helps deciphering. I've been toying with it for 2 days now so I'm as hot for regex as I can be 😉 maybe I have a bright moment when I see the general idea.

etully

Unfortunately, I can't re-print the example (it's copyrighted text) and I don't have time right now to comprehend / re-explain the topic. The best I can do at the moment is point you to where I found the solution. O'Reilly's Mastering Regexp's 3rd Ed. Chapter 10.5, page 475. The solution is not in the second edition. Depending on my work day, I might have time to digest the topic later tonight or tomorrow.

nrg_alpha

As etully mentioned, it's copyrighted property. I would suggest actually purchasing the book. It really is worth it. You will gain a much better understanding of how regex works, which pitfalls to avoid, which strengths to exploit, etc..

If you want to learn regex, this book is THE resource to really help you out.