unnest errors in conjunction with SELECT *

[ahirner]

Describe the bug

One cannot include all colums in addition to the unnested column. Consequently, excluding some of all columns also stopped working in 42.

To Reproduce

Create table with regular and to be unnested columns:

CREATE TABLE d AS VALUES
('x', 1, [named_struct('a', 1, 'b', 2)]),
('y', 2, [named_struct('a', 3, 'b', 4), named_struct('a', 5, 'b', 6)]);

Query unnested structs but also all others.

SELECT unnest(column3), * FROM d;

Error:

Schema error: No field named "*". Valid fields are "UNNEST(d.column3)", d.column1, d.column2, d.column3, "UNNEST(d.column3)", d.column1, d.column2, d.column3.

When excluding some via except:

SELECT unnest(column3), * except (column3, column1) FROM d;

.. the error becomes:

Schema error: No field named "* EXCEPT (column3, column1)". Valid fields are "UNNEST(d.column3)", d.column2, "UNNEST(d.column3)", d.column2.

Expected behavior

The qeries do work in datafusion 41:

+-------------------+---------+---------+------------------------------+
| unnest(d.column3) | column1 | column2 | column3                      |
+-------------------+---------+---------+------------------------------+
| {a: 1, b: 2}      | x       | 1       | [{a: 1, b: 2}]               |
| {a: 3, b: 4}      | y       | 2       | [{a: 3, b: 4}, {a: 5, b: 6}] |
| {a: 5, b: 6}      | y       | 2       | [{a: 3, b: 4}, {a: 5, b: 6}] |
+-------------------+---------+---------+------------------------------+

..with except (column3, column1):

+-------------------+---------+
| unnest(d.column3) | column2 |
+-------------------+---------+
| {a: 1, b: 2}      | 1       |
| {a: 3, b: 4}      | 2       |
| {a: 5, b: 6}      | 2       |
+-------------------+---------+

Additional context

It seems a to be a parsing issue, or perhaps loosing track of field names (#12560)?

[alamb]

Thanks for the report @ahirner -- the only obvious change to unnest in 42 was #11797 from @jayzhan211

Hopefully someone can look into this soon

[ahirner]

Or it's related to more optimal wildcard expansion? #11681
(can't bisect right now)

[hailelagi]

take (after a few hours of debugging, this is a little bit beyond me)

[goldmedal]

take

[goldmedal]

Or it's related to more optimal wildcard expansion? #11681 (can't bisect right now)

I have confirmed that it's a missed case of #11681.
Thanks @ahirner for reporting it. I'll fix it as soon as possible.