Simple query fails with column types must match schema types

[adriangb]

Describe the bug

Found in our production system using datafusion internals, but reproducible in datafusion-cli:

SELECT
    'foo' AS text,
    arrow_cast('2024-01-01T00:00:00Z'::timestamptz, 'Timestamp(Microsecond, Some("UTC"))') AS ts
GROUP BY ts, text

Gives:

Arrow error: Invalid argument error: column types must match schema types, expected Timestamp(Microsecond, Some("UTC")) but found Timestamp(Microsecond, None) at column index 0

To Reproduce

No response

Expected behavior

No response

Additional context

I bisected it with this script:

pushd datafusion-cli && cargo run --bin datafusion-cli -- --file ../q.sql && popd

And then:

git checkout c0ca4b4  # current HEAD
git bisect start
git bisect bad
git checkout 88f58bf # 43.0, I manually verified it is okay
git bisect good
git bisect run ./test.sh  # the above script

The result points to ecc04d4 / #13457

[adriangb]

cc @jonathanc-n @jayzhan211 @alamb

[alamb]

@jayzhan211 is on the case ❤️

• Set timezone for group column timestamp type #13535

[alamb]

I think this issue was introduced in

• feat: Support faster multi-column grouping ( GroupColumn) for Date/Time/Timestamp types #13457

Which has not been released yet