> Thank you for reporting that. > Interesting: that progression (time vs. increasing N) is clearly quadratic > or worse when using a multibyte locale, but is linear with LC_ALL=C. > I suspect when you run "locale", it reports something like en_US.utf8.
Yes, it's a UTF8 locale. I hadn't thought of that and I should have realised, given that that's what the main change was in grep v2.19. > I.e., if you have no need for multi-byte matching, set LC_ALL=C, and that > idiom will be very quick, even for a million lines: Yes, you're totally right there. Thanks! > Currently, I am not planning even to investigate this for the imminent > release. I totally agree! Cheers, Steve.
