4 ms·

Note that, at least as of GNU grep 2.14, if you don't use -i, the discrepancy doesn't show up, so it's smart enough to recognize that the UTF-8 search can be co

by comex 13y ago

Note that, at least as of GNU grep 2.14, if you don't use -i, the discrepancy doesn't show up, so it's smart enough to recognize that the UTF-8 search can be correctly performed as a byte search. I suspect the insensitive version can also be done correctly much faster, though.

acdha 13y ago
> it's smart enough to recognize that the UTF-8 search can be correctly performed as a byte search It shouldn't that simple – it'd also need to confirm that the pattern wouldn't match any combining characters or normalization would still be necessary.
- Joeri 13y ago
  The non-leading utf-8 bytes all are in an easily detected range that doesn't overlap with ascii.
  - acdha 13y ago
    Yes - it's not hard to do but it does require someone to remember to check before attempting the optimization.