Gabriel39 opened a new pull request, #68778: URL: https://github.com/apache/doris/pull/68778
### What problem does this PR solve? Related PR: #68615; https://github.com/zhangstar333/lance-c/pull/3 Reopening a Lance dataset can recreate its underlying object store. The per-instance UUID namespace then prevents the new reader from finding cached data, so repeated RowIDFetcher queries read from the origin again even with a shared session. Revert the per-instance namespace portion of the Foyer patch and restore the original v1 store-prefix/path keys. Keep the optimization that avoids rewriting unchanged size records. Update the repeated-scan regression to require an immediate cache hit and add a multi-fragment `take_rows` regression covering close/reopen cycles and returned row values. This restores the original keying limitation: different endpoints with identical store prefixes and paths are not isolated within one cache owner. The dependency README documents using separate cache owners/directories for those backends. Validation on the pinned dependency with the updated patch: - Reproduced before the change: the second open/read/close cycle requested 35,816 origin bytes; after the revert, both subsequent cycles request zero origin bytes. - `cargo test --locked`: 495 tests passed; the three opt-in native consumer tests were not rerun. - `cargo clippy --locked --all-targets -- -D warnings` and `cargo fmt -- --check` passed. - The Doris Lance patch lifecycle harness passed fresh extraction, idempotence, source refresh, and patch-failure checks; `git diff --check` passed. - Full Doris compilation and BE integration remain for CI. No FE or C/C++ source files are changed. ### Release note Restore Lance data-cache reuse when datasets are reopened between queries. ### Check List (For Author) - [x] Unit Test - [x] Behavior changed: reopened datasets reuse store-prefix/path cache entries. - [x] Documentation updated in the dependency README carried by the patch. ### Check List (For Reviewer who merge this PR) - [ ] Confirm the release note - [ ] Confirm test cases - [ ] Confirm document - [ ] Add branch pick label -- 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
