RegEx in C: Match pattern with a string in the middle

Question

So, I'm trying to minimize boolean expressions using RegEx in C.

I would need a regular expresion to match XXX AND XXX, where "XXX" is the same thing on both ends, for example, I need to match:

((p OR q) AND (p OR q))

but not match

(p AND (p OR q))

because the right side of the AND is not exacly the one on the left, so I can replace it with only

(p OR q)

I think I got the replacing bit, but I need the regular expresion to match things like the one stated.

Thanks :)

What about `((p OR q) AND (p OR q) AND (p OR q))` ? I think that you need to build your own parser, not that it's impossible in PCRE ... — HamZa, Dec 20 '13 at 22:43
You want to match to same pattern strings separated by an `AND` ? — sanjeev mk, Dec 20 '13 at 22:43
@HamZa ((p OR q) AND (p OR q) AND (p OR q)) would reduce the first time to ((p OR q) AND (p OR q)). and a second time to ((p OR q)) — Florius, Dec 20 '13 at 22:44

Casimir et Hippolyte · Accepted Answer · 2013-12-20T23:16:05.590

You can use something like:

(.+?)(?: AND \1)+

then you need only to replace with the content of the first capturing group.

Note that you can add (?s) at the begining of the pattern to allow the dot to match newlines.

Here is a more waterproof pattern:

(?s)(.+?)(?:\s+AND\s+\1)+

To remove possible parenthesis around an expression, you can use:

(?s)(\(\s*)(.+?)(?:\s+AND\s+\2)+(?(1)\s*\))

Note that you must apply this replacement several times (until there is no more matches) to obtain the result you want, since you can have overlapping matches. Example:

(((p OR q) AND (p OR q)) AND ((p OR q) AND (p OR q)))

score 0 · Answer 2 · answered Dec 20 '13 at 22:50

0

Regular expressions are used to describe strings that belong to the class of language named Regular Language .

In particular, strings of the form WxW aren't regular. You can prove this using the Pumping Lemma for Regular Languages . That is, same pattern strings separated by x . In fact, the pattern you want to describe is one of the key examples of how to prove a language non-regular.

The pattern you describe, can be however generated by a Context Free Grammar (equivalently, a push down automata). A parser, specifically.

Since WxW doesn't belong to the class of Regular Languages, there can't be a regular expression describing it.

answered Dec 20 '13 at 22:50

sanjeev mk

4,276
6
44
69

Not the one who downvoted, but I thought to add a small notice. Regular expressions that you learn at university aren't the same in real life. Basically, when you encounter PCRE, .NET regex and many others you could just throw away all what you have learned. Regex are no more "regular". A nice reading is this article [*The true power of regex*](http://nikic.github.io/2012/06/15/The-true-power-of-regular-expressions.html) – HamZa Dec 20 '13 at 23:00
@HamZa Thanks! Didn't realize that by using programming language constructs, regex can describe non-regular languages too. – sanjeev mk Dec 20 '13 at 23:02

score 0 · Answer 3 · answered Dec 20 '13 at 23:03

As Casimir mentioned, the RegEx I was looking for was:

(.+) AND \\1

I got these results using that patern:

[INPUT] -> [OUTPUT]
r AND (q AND q) -> r AND q
r AND ((q) AND (q)) -> r AND (q)
r AND ((q OR w) AND (q or w)) -> r AND (q OR w)
r AND ((q OR w) AND (q or t)) -> r AND ((q OR w) AND (q or t))

Thank you :)

RegEx in C: Match pattern with a string in the middle

3 Answers3