Regex for overlapping matches

Question

For a linguistics project I am trying to match all occurrences of one or two consonants between vowels in some text. I am trying to write a very simple matcher in PHP (preg_match_all), but once the match is consumed, it cannot match again.

The following is very simple and should do the trick, but only matches the first occurrence:

[aeiou](qu|[bcdfghjklmnprstvwxyz]{1,2})[aeiou]

In: officiosior: offi and osi are returned, but not ici because the trailing i is the first part of the match in the second match.

As far as I can tell, it's impossible to do, but is there a decent way to work around the issue?

hwnd · Accepted Answer · 2014-07-07T22:01:26.077

10

You can use a Positive Lookahead assertion to achieve this.

(?=([aeiou](?:qu|[^aeiou]{1,2})[aeiou]))

A lookahead does not consume any characters on the string. After looking, the regular expression engine is back at the same position on the string from where it started looking. From there, it can start matching again...

Explanation:

(?=                    # look ahead to see if there is:
  (                    #   group and capture to \1:
    [aeiou]            #     any character of: 'a', 'e', 'i', 'o', 'u'
    (?:                #     group, but do not capture:
      qu               #       'qu'
     |                 #      OR
      [^aeiou]{1,2}    #       any character except: 'a', 'e', 'i', 'o', 'u' 
                       #       (between 1 and 2 times)
    )                  #     end of grouping
    [aeiou]            #     any character of: 'a', 'e', 'i', 'o', 'u'
  )                    #   end of \1
)                      # end of look-ahead

Working Demo

edited Jul 07 '14 at 22:01

answered Jul 07 '14 at 20:29

hwnd

69,796
4
95
132

That works really well. Is it because the lookahead changes the pattern to zero-width? – Ryan Ward Valverde Jul 07 '14 at 20:34
Yes the Lookahead does not consume any of the string. – hwnd Jul 07 '14 at 20:43
I'm currently playing around with recursive patterns. (never did that before) It looks like this applies more here. However, I deleted my comment because you are right with "the lookahead does not consume any of the string".. +1 for that – hek2mgl Jul 07 '14 at 20:50

Regex for overlapping matches

1 Answers1