github-actions[bot] commented on code in PR #67904:
URL: https://github.com/apache/doris/pull/67904#discussion_r3999418949
##########
fe/fe-connector/fe-connector-iceberg/src/main/java/org/apache/doris/connector/iceberg/IcebergConnectorMetadata.java:
##########
@@ -499,17 +500,24 @@ public ConnectorTableSchema getTableSchema(
schema = table.schema();
}
}
- return buildTableSchema(iceHandle.getTableName(), table, schema, true);
+ String specId =
snapshot.getProperties().get(PARTITION_SPEC_ID_PROPERTY);
Review Comment:
[P1] Honor the cached schema ID on the snapshot-less arm before combining it
with this cached spec. `beginQuerySnapshot` now emits a nonnegative schema ID
even when `snapshotId` is -1, and the new eager latest materialization calls
this overload. On a warm cache hit in a REST vended-credential catalog (where
raw table caching is disabled), a schema-only ALTER of a still-empty table
reloads the live table, but the `currentSnapshot() == null` branch above
selects live `table.schema()`; `getColumnHandles(..., snapshot)` repeats that
bypass. The supposedly exact S1 pin therefore binds S2 and pairs it with cached
P1, then the same unchanged cache makes these selectors flip back to S1 after
the first append. Please resolve any nonnegative snapshot schema ID through
`table.schemas()` regardless of `currentSnapshot()`, keep the same missing-ID
fallback in both schema and handle paths, and add a warm-cache schema-evolution
test that leaves the table snapshot-less (ideally also covering the
first-append transition).
--
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]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
