github-actions[bot] commented on code in PR #68466:
URL: https://github.com/apache/doris/pull/68466#discussion_r4089318095
##########
regression-test/suites/external_table_p0/adbc/test_adbc_scan_edge.groovy:
##########
@@ -124,6 +124,8 @@ suite("test_adbc_scan_edge", "p0,external") {
CREATE CATALOG ${catalogName} PROPERTIES (
"type" = "adbc",
"driver_url" = "${driverPath}",
+ -- The loopback source is Doris even when vendor detection is
unavailable.
+ "sql_dialect" = "doris",
Review Comment:
[P2] Preserve coverage for automatic Doris vendor discovery. With this
addition, all 23 reachable Flight SQL catalogs in this directory set
`sql_dialect`, and `AdbcDialectSelector.resolve` returns before
`probeVendorName`; the only unconfigured Flight catalog is the deliberately
unreachable negative fixture. The later quoted-identifier block still says the
dialect is chosen from the source vendor, but no regression now exercises the
successful `getInfo(VENDOR_NAME)`/dense-union path, while existing unit tests
cover only configured selection, fallback, or direct registry lookup. Please
keep a targeted unconfigured Doris test (or add a unit/native test that feeds a
successful vendor-info result through `readVendorName`) so this production path
remains covered.
##########
regression-test/suites/external_table_p0/adbc/test_adbc_predicate_pushdown.groovy:
##########
@@ -208,13 +210,8 @@ suite("test_adbc_predicate_pushdown", "p0,external") {
pushes("amount > 20.50", "`amount` >")
pushes("big = 9223372036854775807", "9223372036854775807")
pushes("d = '2024-01-01'", "DATE '2024-01-01'")
- // The one comparison family this connector will not push, and the
reason is that this source's
- // datetime column arrives as TIMESTAMPTZ: an instant. By the time the
literal reaches the
- // dialect it has been converted to UTC, and standard SQL's TIMESTAMP
'...' spelling carries no
- // zone, so the source would read that UTC wall clock as its own local
time. East of UTC that
- // merely widens the match; west of UTC it drops rows the query
wanted, and a scan cannot get
- // back rows the source never sent. sameAsSource below is what says
the ANSWER is still right.
- pushesNothing("ts > '2024-01-01 00:00:00'")
+ // DATETIME stays timezone-free through Flight SQL, so a TIMESTAMP
literal preserves its value.
+ pushes("ts > '2024-01-01 00:00:00'", "`ts` > TIMESTAMP '2024-01-01
00:00:00'")
Review Comment:
[P2] Exercise this DATETIME behavior on the single-statement path too. This
catalog uses `partitioned_read = required`, and every corrected DATETIME
type/value check in this PR is likewise partitioned; the existing
required-vs-disabled comparison below omits `ts`. These modes obtain their
streams through different driver APIs (`ConnectionReadPartition` versus
`StatementExecuteQuery`) before the shared Arrow materializer, so the new
assertion can pass while the direct stream still reports a zoned schema or
different wall-clock value. Please add the timestamp predicate to the
disabled-path comparison and cover a direct DATETIME value/type read (ideally
the existing 0/3/6 precisions).
--
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]
