> Tatsuo Ishii <t-ishii@sra.co.jp> writes:
> > ... There seems some problems existing in the
> > implementation. Considering REGEX is not so slow, maybe we should
> > employ the same design as REGEX. i.e. using wide charcters, not
> > multibyte streams...
>
> Seems like a good thing to put on the to-do list. In the meantime,
> we still have the question of whether to enable multibyte in the
> default configuration. I'd still vote YES, as these results seem
> to me to demonstrate that there is no wide-ranging performance
penalty.
> A problem confined to LIKE on long strings isn't a showstopper IMHO.
As I said, with a valid not anchored like expression the performance
difference was substantial, even for shorter strings it was 37%.
The test with "like 'aaa'" was not a good test case, and we should not
deduce anything from that.
Andreas