Per my test in CentOS, the sort order of UTF-8 characters were completely different between locale C, ja_JP.UTF-8 and en_US.UTF-8.
Conversation
Replying to
I'm told JIS X 4061 offers the most useful sort order for Japanese, but unsupported by glibc: userguide.icu-project.org/collation/conc
2
Replying to
thanks for the info! that sounds nice. but I'm sure many Japanese DBA would prefer locale-C because of performance.
1
Replying to
icu gets you back abbreviated keys, so maybe it won't matter. Standard I mentioned is particularly slow, though.
1
Replying to
ok, icu sounds nice! so what about indexscan for a prefix search? that's not available in some locales for now.
1
Replying to
yes, index scans are still slower, in general. We should improve that, too. C locale makes smaller difference there, though.