support for ICU collation

[eroux]

Hello,

I'm very interested in trying duckdb-wasm, but one potentially blocking issue is the support for ICU collation since I'm working primarily with Tibetan, for which a simple collation doesn't make much sense. I couldn't find if ICU collation was available or not though, is it?

I'm quite new to wasm so I don't know if it's usable or not, but the JS API Intl.Collator('bo'); works well for Tibetan

Thanks!

[carlopi]

DuckDB documentation seems to suggest ICU collation are available via the ICU extension, but ICU integration (and in general for dynamic extension loading) support has yet to land, but it's close.

I would probably suggest to get started run some experiment on DuckDB command line interface, if the syntax you need is supported there (after INSTALL icu; LOAD icu; steps), then support in duckdb-wasm will follow shortly.

Another suggestion would be joining https://discord.duckdb.org, where I am sure you can met with others that experimented in this direction.

[carlopi]

Hi @eroux, we have put up a very experimental deployment at https://shellwip.duckdb.org/ that include loading of extensions.

LOAD icu;

works, and functions appears to be exported, I don't see any particular blocker for this to work, but also has not been tested in any extensive way, so hard to say.

It would be cool if you can give it a go and if you have any feedback.

Here (https://github.com/duckdb/duckdb-wasm-wip#readme) there are some explanations, current phase is understanding whether this path for extension loading is viable and works on fixing outstanding issues.

[eroux]

this is wonderful, thanks a lot!! I still haven't decided if I'll use duckdb yet, but that's very clearly a strong argument in favor of it. I'll keep the issue updated if I end up testing it on our database

[carlopi]

Great!

I will be closing this, centralising discussion about extension loading here: #1202.

But please do re-open if you have more to add.