On 03/06/2010 09:56 AM, Jim Meyering wrote:
Won't this induce a large performance penalty for a buffer with no matches but with many lines?
It shouldn't be too large, at least until PCRE uses a normal per-character scan to find possible matches (instead of something like Boyer-Moore). That has the same complexity as memchr but a higher constant in front of it.
Anyway yes, it's a matter of correctness so... Paolo
