Some regex flavors (Perl, PCRE, Oniguruma, Boost) only support fixed-length lookbehinds, but offer the
\K feature, which can be used to simulate variable-length lookbehind at the start of a pattern. Upon encountering a
\K, the matched text up to this point is discarded, and only the text matching the part of the pattern following
\K is kept in the final result.
Is equivalent to:
In general, a pattern of the form:
(subpattern A)\K(subpattern B)
Ends up being similar to:
(?<=subpattern A)(subpattern B)
Except when the B subpattern can match the same text as the A subpattern - you could end up with subtly different results, because the A subpattern still consumes the text, unlike a true lookbehind.