github-actions[bot] commented on code in PR #68165:
URL: https://github.com/apache/doris/pull/68165#discussion_r4044650304
##########
regression-test/suites/external_table_p0/remote_doris/test_remote_doris_variant_select.groovy:
##########
@@ -104,8 +105,21 @@ suite("test_remote_doris_variant_select", "p0,external") {
);
"""
- qt_sql """
+ // Keep nested-path coverage non-empty before comparing catalog reads with
the source table.
+ check_sqls_result_equal """
+ select id from
`internal`.`${db_name}`.`test_remote_doris_variant_select_t`
+ where cast(v['b'] as double) >= 0 order by id
+ """, "select 11"
+ check_sqls_result_equal """
+ select id from
`internal`.`${db_name}`.`test_remote_doris_variant_select_t`
+ where cast(v['c']['c2'] as int) > 1 order by id
+ """, "select 12"
+
+ // Compare with the source instead of golden text tied to an older VARIANT
representation.
+ check_sqls_result_equal """
Review Comment:
The parity comparisons do not prove that rows 13-15 still provide the
malformed/escaped STRING-to-VARIANT coverage this fixture intends. With
`use_arrow_flight=false`, the catalog relation is rebound to a
`RemoteOlapTable`/`makeOlapScan`, so both sides read the same persisted table;
if those implicit casts stop producing the intended top-level strings (or their
bytes change), both queries still agree and this coverage disappears silently.
The two fixed checks above only establish the nested `b` and `c.c2` rows.
Please keep the parity checks, but add a compact fixed source-side precondition
for ids 13-15—for example `(id, variant_type(v), HEX(CAST(v AS STRING)))`—so
type and exact payload are pinned without restoring the serializer-dependent
81-line golden file.
--
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]
