GH-39454: [JS] Support LargeList

[kylebarron]

Rationale for this change

Support large list type in JS bindings.

With the addition of LargeBinary and LargeUtf8, I'd like to get LargeList in as well, so that I don't have to special case any large-offset types (as I currently do) when handling FFI to Arrow JS

cc @domoritz@trxcllnt

What changes are included in this PR?

Mimics the previous PR to add large binary support. #39258

Are these changes tested?

Yes

Are there any user-facing changes?

Yes, new LargeList support.

• Closes: [JS] Support LargeList arrow-js#70
• GitHub Issue: ARROW-193: typos "int his" fix to "in this" #70

[github-actions]

⚠️ GitHub issue apache/arrow-js#70has been automatically assigned in GitHub to PR creator.

[domoritz]

Thanks for the pull request. Maybe we can try to get this into arrow 15. The code freeze is Monday, though.

[kylebarron]

I am struggling a bit with casting between bigint and number for indexing

[domoritz]

Indexing should always be done with Numbers because arrays and buffers cannot be large than Number right now. We have a safe method to cast (and throw an error if it doesn't work).

[kylebarron]

We have a safe method to cast (and throw an error if it doesn't work).

Can you point to an example using this?

[domoritz]

For example

arrow/js/src/visitor/set.ts

Line 131 in 38af258

constx=bigIntToNumber(valueOffsets[index]);

.

I also just merged a pull request to remove getByteLength so you will need to merge/rebase.

[domoritz]

Can you wrap up this pull request? It looks like it's pretty close and would nicely round out support in arrow js.

[kylebarron]

I took a quick look back at this and it feels like there's quite a few places where offsets are used and are assumed to be Number, and it's hard to tell how much work would be left

[domoritz]

I already implemented support for large offsets (e.g. for large binary and large utf-8). You should be able to copy what I did there. Is there anything specific you are looking at?

[kylebarron]

It seems like there are a whole lot of errors in the docker tests.

LargeList support is a "nice to have" for me, but ultimately I just don't have a use case that really needs this. In lonboard I'm constructing the data myself and can ensure I have i32 offsets. In parquet-wasm, arrow-js-ffi, and geoarrow-wasm, I can ensure the user has i32 offsets. i64 offsets would mostly support polars integration I suppose.

Sorry but this is kinda low on my long list of tasks. Much higher is prototyping lonboard-mosaic integration!

[domoritz]

What docker tests? You can just run yarn test -t src locally.

LargeList support is a "nice to have" for me, but ultimately I just don't have a use case that really needs this.

Yeah, it's mostly about completeness for me. I'll see whether I can put it on my queue to wrap up this pull request.

[kylebarron]

What docker tests? You can just run yarn test -t src locally.

I was just referring to the Execute Docker Build step of CI that's the name of the failing stage.

Yeah, it's mostly about completeness for me

With the new view types, completeness is a moving target 🥲

[github-actions]

⚠️ GitHub issue #39454 has no components, please add labels for components.

[thisisnic]

Closing this PR as the Javascript codebase has now moved to https://github.com/apache/arrow-js