fix: vector index type shown as unknown in describe_indices

[jackye1995]

Vector index still needs to be opened to get the right type, otherwise it is shown as unknown.

[github-actions]

PR Review

Summary: Opens vector indices to determine their concrete type (e.g., IVF_PQ, IVF_HNSW_SQ) instead of showing "Unknown" in describe_indices. The root cause is that VectorIndexDetails in the protobuf doesn't store the concrete subtype, so the plugin registry lookup fails.

Performance concern (P1)

describe_indices now performs I/O (open_generic_index) sequentially for every vector index. This opens each index file from object storage to read its type. For datasets with many vector indices, especially on remote stores (S3/GCS/Azure), this could add significant latency.

Consider:

1. Running the open_generic_index calls concurrently (e.g., futures::future::try_join_all or FuturesUnordered) rather than the sequential for loop.
2. Alternatively, caching the index type in the manifest/metadata so this I/O isn't needed at all — though that's a larger change.

Missing tests (P1)

Per repo policy: "All bugfixes and features must have corresponding tests." There are no tests added for this change. A test that creates a vector index (e.g., IVF_PQ) and asserts describe_indices returns the correct index_type string would be valuable and would prevent regressions.

Minor: variable shadowing replaced but field_ids still used after rename

The rename from field_ids to field_ids_vec is fine, but note that in the new code block, field_ids (the original &example_metadata.fields reference on line 444) is still used to look up the column name via field_ids.first(). This works correctly since field_ids still refers to the original &Vec<i32> — just noting it's a bit subtle that both field_ids and field_ids_vec coexist. No bug here.

[justinrmiller]

looks good, my only concern is if the collect gets too large, but that also applies to the previous version.

[codecov]

Codecov Report

❌ Patch coverage is 86.44068% with 8 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
rust/lance/src/index.rs	86.44%	7 Missing and 1 partial ⚠️

📢 Thoughts on this report? Let us know!

[wjones127]

FYI this will also be fixed by #6099, but that will take longer to merge.

[jackye1995]

FYI this will also be fixed by #6099, but that will take longer to merge.

Nice, totally agree that's the right way. I guess the question is do we think this is a regression so we should have this in place first. But if we are not patching it, then it does not really make much sense