Comment from: Tony Cook [Visitor]
Tony Cook

Just as a proof of concept:
https://github.com/tonycoz/Unicode-ICU/tree/master/Collator

updated benchmark:
https://github.com/tonycoz/perl_uca_sort_benchmark/tree/icu

Testing locally with ICU 4.8.1:

       Rate   uca   icu  sort
uca 7.66/s -- -91% -98%
icu 83.7/s 992% -- -82%
sort 472/s 6057% 464% --
09/20/11 @ 09:44
Comment from: robin [Member]

Great work, Tony! Can't wait to try it out!

09/29/11 @ 06:11
Comment from: Mats Peterson [Visitor]
Mats Peterson

I really like the idea of having a uniform OS independent collation algorithm for Perl. The collation orders for Windows locales seems to differ from the ones in Unix, for example. Pity it's so slow, like you said yourself...

06/10/16 @ 03:47
Comment from: Mats Peterson [Visitor]
Mats Peterson

If you're in Unix, there is Unicode::ICU::Collator, which is a wrapper around the ICU library. It is thousands of times faster than Unicode::Collate, but I have yet to compile it in Windows. It's not provided as a package in ActivePerl.

06/10/16 @ 13:40
Comment from: Mats Peterson [Visitor]
Mats Peterson

Interestingly enough, Unicode::Collate::Locale is much faster in Windows than it is in Linux. Peculiar...

06/11/16 @ 04:28
Comment from: Mats Peterson [Visitor]
Mats Peterson

Wrong, it is not. I don't know what happened there...

06/11/16 @ 04:53
Comment from: Mats Peterson [Visitor]
Mats Peterson

Alright Tony, I didn't realize until now that it was you who wrote Unicode::ICU::Collator! Thanks a bunch for this one. It's a hell of a lot faster than Unicode::Collate::Locale, and I use it whenever I can.

06/12/16 @ 01:09
« JW Player uses term "Open Source", but violates Open Source Definition rule #6Using perl instead of awk in my one-liner »