yihua commented on PR #666: URL: https://github.com/apache/hudi-rs/pull/666#issuecomment-6483047411
Pushed three commits on top of 8f7a8609 for the structure items in the follow-up list above: - 1981756e **Complete every write through one shared sequence.** `write::complete_write` now runs the metadata table updates, lock, metadata table deltacommit, completed instant, bookkeeping and reload for append, MOR upsert, MOR delete and the COW rewrite finalizer (item 11), with `commit_metadata_bytes` building the instant's bytes once. One behavior change: the append commit's schema now leaves out the `_hoodie_*` meta fields like the other verbs and the Java writer; it recorded them when a read-back batch was appended. - c75afef7 **Fold the table version 6 append path into `write::append`.** `table/writer.rs` is gone; `append_version_6` sits next to the version 8 path with what it leaves out stated on it, and `Table::append` routes through `append_batches` for every version (item 15). Its base file is written with the table's parquet writer properties like every other base file, and the refusal names `hoodie.table.record.merge.strategy`, the config that carries `append_only`. - a4294584 **Split `write::rewrite` by concern** into `mod.rs` (public verbs), `validate.rs`, `mor.rs`, `rows.rs`, `cow.rs` and `lifecycle_tests.rs` (item 12). Code moves unchanged apart from `pub(super)` visibility. Filed separately, since one writer per process is all that is supported for now: #857 for the process-wide timezone flag and the DST ordering of minted instants (item 3). Item 14 is covered by #856. Items 13 (`append.rs` as the shared utility module) and the remaining half of 15 stay on the list above; neither touches storage bytes or layout. Verified with the full `hudi-core` suite (`--no-fail-fast`) and the `--no-default-features` clippy and test leg. The only failures are the two `gold_parity_tests` cases, which fail identically on 8f7a8609 on this macOS setup (chrono-tz formatter) and are unrelated. -- 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]
