SEZ9 commented on PR #11493:
URL: https://github.com/apache/seatunnel/pull/11493#issuecomment-5358313724

   Thanks for the thorough re-verification pass, @DanielLeens. Given the head 
is unchanged at `8ddb8ff420f0`, I agree there's no need to re-post the full 
review — but the four previously raised findings still stand as merge blockers, 
and your re-confirmation of the silent-behaviour defects reinforces that. 
Concretely, here is what I still need before this can move forward:
   
   1. **PR11493-F1 (HIGH, Compatibility):** The silent STRING → JSON mapping 
change for Snowflake `VARIANT`, Doris `JSON`/`JSONB`, and StarRocks `JSON` 
source columns needs an explicit opt-out (e.g., a config flag to retain the 
legacy STRING mapping) plus an entry in 
`docs/en/introduction/concepts/incompatible-changes.md`. Upgrading users must 
not have pipelines break with no escape hatch.
   
   2. **PR11493-F2 (Logic):** Since the JSON logical type is physically a plain 
String, please clarify (and document in 
`docs/en/introduction/concepts/schema-feature.md`) where validation happens. If 
validation only exists at the StarRocks load boundary, we need a stated 
strategy for the other native-JSON sinks, or a documented contract that 
malformed text is the user's responsibility.
   
   3. **PR11493-F3 (Robustness):** The StarRocks JSON-column + default CSV 
Stream Load combination should be rejected at sink initialization, not at load 
time. Please add fail-fast validation and update 
`docs/en/connectors/sink/StarRocks.md` accordingly.
   
   4. **PR11493-F4 (Robustness):** The Snowflake sink silently skipping 
`PARSE_JSON` wrapping when a user-defined query is supplied defers failure to a 
mid-run server error. Either fail fast at init with a clear message, or 
document the limitation prominently in `docs/en/connectors/sink/Snowflake.md` — 
silence is not acceptable here.
   
   Additionally, you noted that two of this PR's own new E2E suites are red — 
those need to be green before merge, so please push the fixes for the above and 
let CI re-run.
   
   One housekeeping note: your comment appears to have been cut off 
mid-sentence ("every connector/format that switches on..."). Could you re-post 
the remainder so nothing from your re-verification is lost?
   
   No new scope from my side — resolving F1–F4 and getting the E2E suites 
passing is the full remaining list.
   
   <!-- streview-comment:406 -->


-- 
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