https://bugs.koha-community.org/bugzilla3/show_bug.cgi?id=41287 --- Comment #21 from David Cook <dcook@prosentient.com.au> --- (In reply to David Cook from comment #20)
So I assume that this had reverted the changes in 36947 but I can see that they're still there. Thanks for that, Nick :3
It could be interesting to try optimising the locale sort more...
I did find one optimization (In reply to David Cook from comment #20)
So I assume that this had reverted the changes in 36947 but I can see that they're still there. Thanks for that, Nick :3
It could be interesting to try optimising the locale sort more...
There is a trick for doing the unicode comparison faster by precomputing the sort keys using the inherited getSortKey() method from Unicode::Collate and then doing a bytewise sort on that, but it's only maybe 2-3 times faster than the original method, so probably not significantly faster. And I haven't actually confirmed if it's 100% effective. (Looking at the flame graph, it's surprising that it's faster considering how so much of Unicode::Collate::cmp is spent on Unicode::Collate:;getSortKey) One idea is to do a hybrid model that mixes sstringwise and unicode but could be messy. Anyway, I think that this solution seems practical and if anyone needs a speed boost with the Unicode sorting we can look into that further. -- You are receiving this mail because: You are watching all bug changes.