yihua commented on PR #698:
URL: https://github.com/apache/hudi-rs/pull/698#issuecomment-6379819463

   Thanks for the thorough review @linliu-code. All the concerns are addressed 
at `f86e4638`: both blocking items, the three "worth fixing before merge" 
items, and the non-blocking ones, with two small follow-ups noted at the end. 
Details below.
   
   **Blocking 1 (Lance).** Went with a fourth option: Lance stays on by 
default, with no feature gate. Every Lance release is still on arrow 58, so 
`crates/core` now also depends on Lance's arrow (renamed `arrow-array` / 
`arrow-schema` 58) and crosses batches through the Arrow C data interface in 
`file_group/base_file/lance_arrow.rs`. `--all-features` builds, the Lance tests 
run unconditionally, and the Python wheel and cxx bridge keep Lance. With no 
feature gate, items 3 and 4 no longer apply, and the `protoc` comment you 
flagged is accurate again. Bridge overhead and the reasoning are in the PR 
description.
   
   **Blocking 2 (pyo3).** Migrated: `downcast` to `cast`, 
`PyCapsule::new_with_value`, and `from_py_object` on the cloned pyclasses.
   
   **5 (`locations`).** Took your suggestion, with a plan test that rejects two 
locations.
   
   **Non-blocking.** `crates/test` no longer sets `default-features = false`. 
The `resolver.rs` fallthrough went away with the feature gate. Tracking issue 
for `ParquetObjectReader` is #843, named from the code comment. Tests were 
added for the `locations` handling, for routing sizeless base files away from 
the parquet source, and for the Lance bridge round trips. Two items I have not 
changed: `HudiDataSource::options()` still returns the options exactly as the 
caller passed them (`Debug` keeps omitting it; I can add a doc note that 
credentials must reach workers out of band), and `HudiScanExec` still 
implements `partition_statistics` rather than `statistics_from_inputs`, which I 
would leave for a follow-up since behavior is unchanged.
   
   CI runs on this head now.
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to