How can I match occurrences of string not in another string using regular expressions?

Question

I'm trying to match all occurances of "string" in something like the following sequence except those inside @@

as87dio u8u u7o @string@ ou os8 string os u

i.e. the second occurrence should be matched but not the first

Can anyone give me a solution?

Answer 1

You can try:

(?:[^@])string(?:[^@])

Answer 2

You can use negative lookahead and lookbehind:

(?<!@)string(?!@)

EDIT

NOTE: As per Marks comments below, this would not match @string or string@.

Answer 3

OK,

If you want to NOT match a character you put it in a character class (square brackets) and start it with the ^ character which negates it, for example [^a] means any character but a lowercase 'a'.

So if you want NOT at-sign, followed by string, followed by another NOT at-sign, you want

[^@]string[^@]

Now, the problem is that the character classes will each match a character, so in your example we'd get " string " which includes the leading and trailing whitespace. So, there's another construct that tells you not to match anything, and that is parens with a ?: in the beginning. (?: ). So you surround the ends with that.

(?:[^@])string(?:[^@])

OK, but now it doesn't match at the start of string (which, confusingly, is the ^ character doing double-duty outside a character class) or at the end of string $. So we have to use the OR character | to say "give me a non-at-sign OR start of string" and at the end "give me an non-at-sign OR end of string" like this:

(?:[^@]|^)string(?:[^@]|$)

EDIT: The negative backward and forward lookahead is a simpler (and clever) solution, but not available to all regular expression engines.

Now a follow-up question. If you had the word "astringent" would you still want to match the "string" inside? In other words, does "string" have to be a word by itself? (Despite my initial reaction, this can get pretty complicated :) )