Question

My question is fairly straightforward, even if the purpose it will serve is pretty complicated. I will use a simple example:

AzzAyyAxxxxByyBzzB

So normally I would want to get everything between A and B. However, because some of the content between the first A and the last B (one pair) contains additional AB pairs I need to push back the end of the match. (Not sure if that last part made sense).

So what I’m looking for is some RegEx that would allow me to have the following output:

Match 1
  Group 1: AzzAyyAxxxxByyBzzB
  Group 2: zzAyyAxxxxByyBzz

Then I would match it again to get:

Match 2
  Group 1: AyyAxxxxByyB
  Group 2: yyAxxxxByy

Then finally again to get:

Match 3
  Group 1: AxxxxB
  Group 2: xxxx

Obviously if I try (A(.*?)B) on the whole input I get:

Match x
  Group 1: AzzAyyAxxxxB
  Group 2: zzAyyAxxxx

Which is not what I’m looking for 🙂

I hope this makes sense. I understand if this can’t be done in RegEx, but I thought I would ask some of you regex wizards before I give up on it and try something else. Thanks!

Additional Info:

The project I’m working on is written in Java.

One other problem is that I’m parsing a document which could contain something like this:

AzzAyyAxxxxByyBzzB
Here is some unrelated stuff
AzzAyyAxxxxByyBzzB
AzzzBxxArrrBAssssB

And the top AB pairs needs to be separate from the bottom AB pairs

Answer 1

You made your regex explicitly ungreedy by using the ?. Just leave it out and the regex will consume as much as possible before matching the B:

(A(.*)B)

However, in general nested structures are beyond the scope of regular expressions. In a case like this:

AxxxByyyAzzzB

You would now also match from the first A to the last B. If this is possible in your scenario, you might be better of going through the string yourself character-by-character and counting As and Bs to figure out which ones belong together.

EDIT:

Now that you have updated the question and we figured this out in the comments, you do have the problem of multiple consecutive pairs. In this case, this cannot be done with a regex engine that does not support recursion.

However you can switch to matching from the inside out.

A([^AB]*)B

This will only get innermost pairs, because there can be neither an A nor a B between the delimiters. If you find it, you can then remove the pair and continue with your next match.