JingsongLi commented on code in PR #11163:
URL: https://github.com/apache/paimon/pull/11163#discussion_r4643622272
##########
docs/docs/primary-key-table/merge-engine/aggregation.mdx:
##########
@@ -268,6 +268,10 @@ count of distinct event IDs.
- `error`: Throw an exception if the nested-key contains null values,
because primary key fields must not be null.
Use `fields.<field-name>.nested-sequence-field=seq0,seq1,...` to control the
update sequence of a nested table, you must configure
`fields.<field-name>.nested-key` when using it.
+ Multi-field sequences are compared lexicographically in configuration order
(nulls first). A nested ROW is overwritten when its sequence is greater
+ than the accumulated one; when sequences are equal, overwrite behavior is
controlled by
+ `fields.<field-name>.nested-sequence-overwrite-on-equal` (default `true`).
Set it to `false` to overwrite only when the new sequence is strictly greater.
Review Comment:
**[P1] Preserve the equal-sequence policy in PyPaimon SQL reads**
With Native 0.5.0 and two separately committed Parquet versions at the same
nested key and seq=7, the public Python table reader returns `first` with this
option set to false, but `pypaimon.sql.SQLContext` with a registered catalog
returns `second` for `SELECT id, updates FROM t`. Both readers return `second`
under explicit true. SQLContext and the `pypaimon sql` CLI use the Rust
SQLContext directly and never encounter the new Python table-read gate. Please
implement the option in that runtime or reject reads of incompatible tables
before returning a different result.
--
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]