Aggregate with FILTER clause and alias errors

[jcrist]

Describe the bug

See the following code:

import datafusion

print(f"datafusion: {datafusion.__version__}")

ctx = datafusion.SessionContext(
    datafusion.SessionConfig(
        {"datafusion.sql_parser.dialect": "PostgreSQL"}
    ).with_information_schema(True)
)

ctx.from_pydict({"a": [1, 3, 5]}, name="test")

# A query with no alias works
good = "SELECT MIN(a) FILTER (WHERE a > 1) FROM test"
print(ctx.sql(good))

# A query with an IN filter works
bad = "SELECT MIN(a) FILTER (WHERE a IN (1, 2)) AS x FROM test"
print(ctx.sql(bad))

# But one with an inequality and an alias fails
bad = "SELECT MIN(a) FILTER (WHERE a > 1) AS x FROM test"
print(ctx.sql(bad))

And results:

datafusion: 38.0.1
DataFrame()
+-------------+
| MIN(test.a) |
+-------------+
| 3           |
+-------------+
DataFrame()
+---+
| x |
+---+
| 1 |
+---+
Traceback (most recent call last):
  File "/home/jcristharif/Code/ibis/bug.py", line 23, in <module>
    print(ctx.sql(bad))
Exception: Internal error: Input field name MIN(test.a) does not match with the projection expression MIN(test.a) FILTER (WHERE test.a > Int64(1)).
This was likely caused by a bug in DataFusion's code and we would welcome that you file an bug report in our issue tracker

All 3 queries do seem to work in versions < 38.0.1, so the bug appears to have been introduced in that version.

To Reproduce

No response

Expected behavior

No response

Additional context

No response

[Lordworms]

take

[Lordworms]

The problem here is we added a check like

but for a plan like

Projection: MIN(test.a) FILTER (WHERE test.a > Int64(1)) AS x
  Aggregate: groupBy=[[]], aggr=[[MIN(test.a) FILTER (WHERE test.a > Int64(1))]]
    TableScan: test

the corresponding name is different like

we should change the generated name in Aggregate

Since the way to generate AggregateExpr is adding other information on suffix, I think we could using string.start_with to check the name here.