Backreferences \1

Backreferences match the previously captured subexpression. Subexpressions are marked with parentheses and numbered from 1 to 9. Use \1 to refer to the leftmost subexpression, \2 to refer to the second subexpression, and so on.

(\d+) \1	Find two equal numbers with a space between them (e.g., 123 123)
\<(\w+) \1\>	Find repeating words (e.g., the the)
<a href=("|')http://[^'"]+\1>	An external link (in single or double quotes)

Non-capturing group

Sometimes you don't want to capture a subexpression, but only to group a part of your regex with parentheses. In this case, you can use a non-capturing group. Type (?: ) instead of the usual parentheses.

(http|ftp)://[^" >]+	Find http:// or ftp:// URL
(?:http|ftp)://[^" >]+	Ditto (without capturing “http or ftp”)

If you don't have any references to the “http or ftp” subexpression, then the second example (with non-capturing group) is more appropriate. You will be unable to refer to the subexpression by using \1, because the subexpression will not be saved.

Capturing and non-capturing groups can be mixed in a regex:

Search for	Replace to	Explanation
(?:http|ftp)://([^" >]+)	\1	Find an URL and remove the protocol from it

Backreferences only refer to the usual, capturing parentheses; the non-capturing groups are skipped. So, \1 refers to ([^" >]+), not to (?:http|ftp).

Non-capturing groups are useful in complex regular expressions because the number of subexpressions is limited to 9 (from \1 to \9), so you may want to capture only what is needed.

(?:abc)+

Find “abc” repeating one or more times

In this example, it makes no sense to capture the subexpression, because it's always equal to “abc”. So, the non-capturing group is a better choice here.