On Mon, Oct 7, 2024 at 9:02 AM Shiv Iyer <shiv@xxxxxxxxxxxxx> wrote: > - As the string length increases, the performance degrades exponentially when using special characters. This is due to the collation’s computational complexity for each additional character comparison. That's a pretty interesting observation, worthy of a bug report. I don't know the details offhand but the algorithm used should be basically the same, or more likely a simpler subset, of what other collation implementations are using IIRC (don't quote me but I think it's supposed to be ISO 14651 which is essentially a subset of the UCA stuff that ICU is using (it is "aligned with" UCA DUCET), without CLDR customisations and perhaps various other complications), so it doesn't sound like it should be fundamentally required to be *more* expensive than ICU...
