kosiew commented on code in PR #24972:
URL: https://github.com/apache/datafusion/pull/24972#discussion_r3986501526
##########
datafusion/sqllogictest/test_files/subquery_projection.slt:
##########
@@ -0,0 +1,57 @@
+# IN subqueries in projection use SQL three-valued logic.
+
+query BB
+WITH empty AS (SELECT 10 WHERE false)
+SELECT
+ NULL IN (SELECT * FROM empty),
+ NULL NOT IN (SELECT * FROM empty)
+FROM (SELECT 1) t;
+----
+false true
+
+query B
+SELECT 'a' IN (SELECT column1 FROM VALUES ('b'), ('c'), ('d'));
+----
+false
+
+query BBBB
+WITH vals AS (SELECT * FROM (VALUES (1), (2), (NULL)) AS t(x))
+SELECT
+ 1 IN (SELECT x FROM vals) AS found,
+ 3 IN (SELECT x FROM vals) AS unknown,
+ 3 NOT IN (SELECT x FROM vals) AS not_unknown,
+ NULL IN (SELECT x FROM vals) AS null_unknown;
+----
+true NULL NULL NULL
+
+query BB
+WITH vals AS (SELECT * FROM (VALUES (1), (2)) AS t(x))
+SELECT
+ EXISTS (SELECT x FROM vals WHERE x = 2),
+ NOT EXISTS (SELECT x FROM vals WHERE x = 3);
+----
+true true
+
+# Correlated IN distinguishes a match, a miss, a NULL-containing result,
+# and an empty result independently for each outer row.
Review Comment:
Could we also add correlated `NOT IN` execution cases here? It would be
useful to cover an inner NULL, an empty correlated result, and a NULL outer
value. The current correlated test covers `IN`, while these cases would
directly exercise the projection-specific `not(result)` path and make sure the
per-row NULL and non-empty marks behave correctly.
--
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]
