Unable to perform lead/lag built in functions on List and Struct data types

[timsaucer]

Describe the bug

When you use the lead or lag built in functions and the data type is either a list or struct, you will get a panic with error Exception: Arrow error: Compute error: concat requires input of at least one array

I have root caused this to list_to_array_of_size in datafusion/common/src/scalar/mod.rs where we do not check to see if the arrays we are attempting to concat have any contents, which they will not because in WindowAggState::new() we are calling to_array_of_size(0). These calls work for primitive data, but for list data we need an additional check. I am submitting a PR to resolve the issue.

To Reproduce

Data file is a simple csv:

a,b,c
1,2,3
4,5,6
7,8,9
10,11,12

Code to reproduce:

use datafusion::{logical_expr::{expr::WindowFunction, BuiltInWindowFunction, WindowFrame, WindowFunctionDefinition}, prelude::*};

#[tokio::main]
async fn main() -> datafusion::error::Result<()> {

    let ctx = SessionContext::new();
    let mut df = ctx.read_csv("/Users/tsaucer/working/testing_ballista/lead_lag/example.csv", CsvReadOptions::default()).await?;

    df = df.with_column("array_col", make_array(vec![col("a"), col("b"), col("c")]))?;

    df.clone().show().await?;

    let lag_expr = Expr::WindowFunction(WindowFunction::new(
        WindowFunctionDefinition::BuiltInWindowFunction(
            BuiltInWindowFunction::Lead,
        ),
        vec![col("array_col")],
        vec![],
        vec![],
        WindowFrame::new(None),
        None,
    ));

    df = df.select(vec![col("a"), col("b"), col("c"), col("array_col"), lag_expr.alias("lagged")])?;

    df.show().await?;

    Ok(())
}

Results:

+----+----+----+--------------+
| a  | b  | c  | array_col    |
+----+----+----+--------------+
| 1  | 2  | 3  | [1, 2, 3]    |
| 4  | 5  | 6  | [4, 5, 6]    |
| 7  | 8  | 9  | [7, 8, 9]    |
| 10 | 11 | 12 | [10, 11, 12] |
+----+----+----+--------------+
Error: ArrowError(ComputeError("concat requires input of at least one array"), None)

Expected behavior

Expect lag to work on these structures. Here is output from the PR I will put up shortly.

+----+----+----+--------------+
| a  | b  | c  | array_col    |
+----+----+----+--------------+
| 1  | 2  | 3  | [1, 2, 3]    |
| 4  | 5  | 6  | [4, 5, 6]    |
| 7  | 8  | 9  | [7, 8, 9]    |
| 10 | 11 | 12 | [10, 11, 12] |
+----+----+----+--------------+
+----+----+----+--------------+--------------+
| a  | b  | c  | array_col    | lagged       |
+----+----+----+--------------+--------------+
| 1  | 2  | 3  | [1, 2, 3]    | [4, 5, 6]    |
| 4  | 5  | 6  | [4, 5, 6]    | [7, 8, 9]    |
| 7  | 8  | 9  | [7, 8, 9]    | [10, 11, 12] |
| 10 | 11 | 12 | [10, 11, 12] |              |
+----+----+----+--------------+--------------+

Additional context

This is the root cause for apache/datafusion-python#647

[alamb]

I saw this example and it reminded me how hard it is to create window functions with the expr api -- it would be great to make this better. See #6747