JingsongLi opened a new pull request, #1136:
URL: https://github.com/apache/paimon-rust/pull/1136

   ## Purpose
   
   PyPaimon already builds BTree/Bitmap and vindex indexes on data-evolution 
tables with deletion vectors, but its Rust delegation rejects those tables. 
Enable the existing Rust index builders for this case.
   
   ## Changes
   
   - Remove the deletion-vector rejection from scalar and vindex build 
validation.
   - Preserve the existing physical row-ID reads without deletion files, 
matching Java `GlobalIndexBuilderUtils.splitByContiguousRowRange` and 
`createShardIndexedSplits`. Query readers apply deletion vectors from the query 
snapshot.
   - Document that invariant where index-build splits are constructed. No new 
public APIs or dependencies.
   - Add end-to-end core regressions for BTree, Bitmap and IVF-flat, including 
partial/all-row deletion, both vector training modes, incremental builds, 
deletion between preparation and commit, and Top-1 when the nearest physical 
vector was deleted.
   
   ## Validation
   
   - Confirmed both new core tests failed on main with the explicit 
deletion-vector rejection before the fix.
   - `cargo test -p paimon --locked --features fulltext --lib 
index_build_builder`: 154 passed, 2 existing ignored.
   - Final focused deletion-vector regressions: 2 passed (eight build 
configurations, including Top-1).
   - `cargo +1.98.0 clippy --locked --all-targets --workspace --features 
fulltext,vortex -- -D warnings`: passed.
   - `cargo fmt --all --check`: passed.
   - Locally rebuilt Python binding and ran related REST/native index suites: 
291 passed, including every PyPaimon vindex family.
   - Independent Rust/Java and Python integration review: no actionable P1/P2 
findings.
   
   The accompanying PyPaimon PR retains Rust `main` in CI and depends on this 
change.
   


-- 
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