0

I came across the following regex at work. What does it do?

,(?=(?:[^\"]*\"[^\"]*\")*(?![^\"]*\"))

To understand it, I split it into following parts

  • , = Match everything that has ,

  • (?= = Followed by

  • (?:[^\"]*\"[^\"]*\")* = Anything which does not match ", followed by ", followed by anything which does not match ", followed by ". For example, 1111"aaaaa"

  • (?![^\"]*\") = BUT not followed by anything that does not " and matches "

In other words, match anything that has , followed by either 11111"111" OR followed by ""

The use case where the above expression is used, simply for tokenizing a string, separated by ,, but I am assuming the author built for some thing more generic.

Can anyone provide a more simpler explanation than above?

The above expression is used to assign expression to boost::regex().

UPDATE: Actually, it is searching for the "," commas, with the following constraint

  1. It is okay that there are even number of " following the comma

  2. BUT, It is NOT okay to have a single " following the comma

For example consider the string: a, "h,w", 23

The first "," is matched, because it has following even number of " ("h,w")

The second "," in between "h,w" is NOT matched because of the second expression (?![^\"]*\") which states that a "," should not be followed by a single "

Finally the last "," matches.

The final output would be 2 , ,

Jimm
  • 8,165
  • 16
  • 69
  • 118
  • I tweaked the formatting of your question to make it less ambiguous and easier to read. Most importantly, the outermost `("` and `")` were *not* part of the regex. Also, be aware that the second lookahead is *inside* the first one. Please check and see if I got the rest of it right. – Alan Moore Dec 01 '12 at 07:11

1 Answers1

2

It looks like it will match any comma, but only if there are an even number of " characters after that comma.

, - A comma.

(?= - Followed by...

(?:[^\"]*\"[^\"]*\")* - Any string ending in a " mark and containing an even total number of " marks, or the empty string,

(?![^\"]*\") - and there is no other " mark later on in the string.

) to close the (?=.

This could be useful if we already know that the entire input string has an even total of " characters, there's no such thing as nesting or escaping quotes, and commas between quote marks should not be treated as delimiters. For example, given the input

25,"Hello, world!","More text",123.45

the regex should not match the comma between Hello and world, but should match the other three commas.

aschepler
  • 70,891
  • 9
  • 107
  • 161
  • What is the use of the last (?![^\"]*\") . What do i stand to lose, if this last part is missing – Jimm Dec 01 '12 at 06:20
  • The first part of the lookahead matches an even number of quotes, but it doesn't guarantee that that's all there are. There could still be one more quote in the remainder of the string, so you have to check for that. That's what the inner lookahead does, so it is essential. – Alan Moore Dec 01 '12 at 07:43