SEZ9 commented on PR #12853:
URL: https://github.com/apache/seatunnel/pull/12853#issuecomment-6963868401

   Thanks for carrying this through the rebase. This follow-up covers the 
earlier findings against the current head 
`3ea222df900f284e5c350c240674be9c667f0e85`. The latest re-review found the Java 
converters, tests and English docs byte-identical to the previously reviewed 
`11a1c7df`, so the rebase itself changes nothing here.
   
   **Settled according to the latest re-review**
   - **F3 (dialect divergence):** `OceanBaseMySqlTypeConverter` now maps 
`DECIMAL(p, s)` and `DECIMAL(p, s) UNSIGNED` to `DecimalType(p, s)` the same 
way `MySqlTypeConverter` does, so the same table no longer yields two different 
schemas.
   - **F6 (scale bound):** the clamp now applies only above precision 65 and 
scale 30, the same limits `reconvert()` uses. If the clamped scale is still 
computed twice, a local variable would tidy it up, but that is non-blocking.
   - **F7 (row converter test):** no change is needed. 
`testWideDecimalValueIsReadAndWrittenWithoutNarrowing` guards the value path 
only, because `AbstractJdbcRowConverter` reads and writes through 
`getBigDecimal` / `setBigDecimal` without looking at precision. The converter 
behaviour is guarded by `MySqlTypeConverterTest`, 
`OceanBaseMySqlTypeConverterTest` and `MysqlCreateTableSqlBuilderTest`.
   
   **Still open — I can't confirm these from the thread**
   1. **F1 / F2 (precision 39–65 downstream):** does the incompatible-changes 
entry now say explicitly that MySQL source jobs on the Spark engine, and sinks 
limited to precision 38 (e.g. MySQL -> Iceberg auto-create), are affected by 
the removal of the `(38,18)` narrowing? It should also say what users should 
do. Please point me to the wording or add it.
   2. **F4 / F5 (docs):** please confirm that `7a3c7991` and `825f69f9` update 
the MySQL source and sink data-type mapping tables in both EN and ZH. They 
should no longer document `>38 -> DECIMAL(38,18)` or `DECIMAL UNSIGNED -> p+1`. 
Please also confirm that the "scale 30" sentence in the incompatible-changes 
entry matches what `convert()` now does.
   3. **F8 (clamp test):** is there a test that feeds a precision above 65 and 
asserts the clamped precision and reduced scale? If so, please name it; if not, 
please add one for both converters.
   
   Once those three are answered, I think this is in good shape.
   
   <!-- streview-comment:2423 -->


-- 
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