Really? Does it behave differently in Perl?
It looks like non-overlapping makes more sense. Suppose you tokenize a string, if overlapping is allowed than the same character would occur in multiple tokens, which is unusual at best. On Jan 14, 2010, at 2:49 PM, Raul Miller <[email protected]> wrote: 9!:14'' j602/2008-03-03/16:45 rxmatches does not provide all matches. require'regex' '(1x)|(_1)' rxmatches '_1x' 0 2 _1 0 0 2 '(1x)' rxmatches '_1x' 1 2 1 2 If matches would overlap, some of them are not considered. This nonoverlapping behavior might be desirable in some cases but winds up being a problem in other cases. (Specifically, this breaks an implementation of http://rosettacode.org/wiki/Markov_Algorithm) -- Raul ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
