PG1204 opened a new issue, #23474:
URL: https://github.com/apache/datafusion/issues/23474
### Describe the bug
### Describe the bug:
A MAP {'key': NULL} literal in a VALUES list fails type coercion when other
rows in the same list have a concrete value type. The bare NULL gets inferred
as Null type instead of being coerced to the value type of the surrounding rows.
Surfaced while writing sqllogictest coverage in #23406 for #15428. Broader
context: NULL literal type inference across a VALUES list works for scalar
types (e.g. VALUES (1), (NULL) coerces NULL to Int64), but the same inference
does not extend into the value type of a Map literal. An explicit CAST(NULL AS
BIGINT) is currently required as a workaround, which is what map.slt uses.
### To Reproduce:
Run against DataFusion main (also reproduces on DataFusion CLI v54.0.0):
CREATE TABLE t AS VALUES
(MAP {'k1': 1, 'k2': 2}),
(MAP {'k1': NULL});
Actual error:
Error during planning: Inconsistent data type across values list at row 1
column 0. Was Map("entries": non-null Struct("key": non-null Utf8, "value":
Int64), unsorted) but found Map("entries": non-null Struct("key": non-null
Utf8, "value": Null), unsorted)
Workaround: an explicit cast works:
CREATE TABLE t AS VALUES
(MAP {'k1': 1, 'k2': 2}),
(MAP {'k1': CAST(NULL AS BIGINT)});
### Expected behaviour:
The bare NULL in MAP {'k1': NULL} should be coerced to the value type of the
other rows in the VALUES list (Int64 in this case), so the CREATE TABLE
succeeds with a Map(Utf8, Int64) column type where the second row has a NULL
value.
### To Reproduce
_No response_
### Expected behavior
_No response_
### Additional context
_No response_
--
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]
