Group by incorrectly works with alias.

[comphead]

Describe the bug
Group by incorrectly works with alias. The query fails with "Projection references non-aggregate values"
`

To Reproduce

async fn csv_query_group_by_substr() -> Result<()> {
    let ctx = SessionContext::new();
    register_aggregate_csv(&ctx).await?;
    let sql = "SELECT substr(c1, 1, 1) c1 \
        FROM aggregate_test_100 \
        GROUP BY substr(c1, 1, 1) \
        ";
    let actual = execute_to_batches(&ctx, sql).await;
    let expected = vec![
        "+----+-----------------------------+",
        "| c1 | AVG(aggregate_test_100.c12) |",
        "+----+-----------------------------+",
        "| a  | 0.48754517466109415         |",
        "| b  | 0.41040709263815384         |",
        "| c  | 0.6600456536439784          |",
        "| d  | 0.48855379387549824         |",
        "| e  | 0.48600669271341534         |",
        "+----+-----------------------------+",
    ];
    assert_batches_sorted_eq!(expected, &actual);
    Ok(())
}

Expected behavior
Test should pass

Additional context
None

[comphead]

@andygrove can this be assigned to me?

[andygrove]

@comphead I have assigned this to you. I have spent some time working on this and keep running into problems with the way we try and compute names and determine type information based on an expression and a schema. I will be interested to see what you find.

I am now going to start working on #1468 and see where that takes me. If the expressions carry their names with them then perhaps it helps with some of these issues? I am still learning and exploring ...

[comphead]

2 bugs here.

• resolve_aliases_to_exprs(&group_by_expr, &alias_map)?; which does wrong resolving and breaks expr to substr(substr(aggregate_test_100.c1,Int64(1),Int64(1)),Int64(1),Int64(1)) instead of substr(aggregate_test_100.c1,Int64(1),Int64(1))
• check_columns_satisfy_exprs that gets confused because of alias the same as col name in agg expression

[comphead]

@comphead I have assigned this to you. I have spent some time working on this and keep running into problems with the way we try and compute names and determine type information based on an expression and a schema. I will be interested to see what you find.

I am now going to start working on #1468 and see where that takes me. If the expressions carry their names with them then perhaps it helps with some of these issues? I am still learning and exploring ...

@andygrove I'm not sure this will solve all the problems. Currently the test like select count(1), count(1) from t will fail because the columns aliases are the same. Currently the name gets generated based on colname + function + types. To diff them you may want include some increment like column position, or reference to current query block hashcode, or other unique constituent

[andygrove]

@comphead I have the test passing in the WIP PR #2457

I ran into the solution whilst working on ROLLUP support

[comphead]

Thank you, I was trying to solve it in planner but your proposed solution looks more robust! Looking forward to test it