SEZ9 commented on PR #11206:
URL: https://github.com/apache/seatunnel/pull/11206#issuecomment-6622091296

   Thanks for the detailed self-review on 
`c3d11fc64757f5d96a0b752dcb8ccffff93fdafb`. Checking my earlier points against 
`95b4946ee39` and `c3d11fc6475`:
   
   **Addressed / on the right track**
   - **F1 (identifier injection):** The negative E2E fixtures now exercise both 
the semicolon and backtick cases with distinct sanitized names, so the Avro 
schema-name collision from run 41164176211 can't recur, and rejection is 
enforced by the allowlist in `MySqlCatalogTableUtils`. Please keep the comment 
explaining why the two names must differ.
   - **F6 (checkpoint state / rollback):** Recording the downgrade restriction 
in the incompatible-change doc covers the rollback half. Can you confirm (or 
point me to the test) that restoring from a pre-change checkpoint with no 
runtime tables in `MultiTableState` still works unchanged?
   - **F7 (docs):** Adding the JDBC driver option makes the EN/ZH example 
runnable. Remaining ask: the `exactly_once = false` / `generate_sink_sql = 
true` constraints for runtime-created tables should also be stated in the JDBC 
sink documentation, not only on the MySQL-CDC source page.
   
   **Needs clarification**
   - **F3 (default value):** Your review describes the feature as opt-in via 
`scan.binlog.newly-added-table.enabled = true` and default-off, while my 
earlier finding was based on `scan.newly-added-table.enabled` defaulting to 
`true`. Please confirm the final option name and default, and make sure the 
option definition and docs agree. Default-off is what I'd want here.
   
   **Still open (no production code changed since `d4faf1ea1aa`)**
   - **F2:** Define what happens when a runtime `CreateTableEvent` reaches a 
sink where `supportsNewlyCreatedTable()` is false — fail fast with a clear 
error rather than dropping rows or hitting a null writer. A short Javadoc 
contract on `SupportMultiTableSinkWriter` plus a unit test would settle this.
   - **F4:** Please describe (or add a test for) how a source-side CREATE is 
scoped so it can only produce DDL for targets matching the configured sink 
routing.
   - **F5:** Clarify the retry/idempotency behavior when runtime writer 
creation throws `IOException` — is the job failed, or is creation retried 
before the table's rows are processed?
   - **F8:** The documented allowlist `[A-Za-z_][A-Za-z0-9_$]*` still rejects 
legal MySQL names starting with a digit. Either widen the pattern or call out 
the limitation explicitly in the docs.
   
   Once F2/F4/F5 have a stated contract and F3/F7/F8 are tidied up, I'm happy 
to take another pass.
   
   <!-- streview-comment:2269 -->


-- 
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