Wrong aggregation arguments error.

[jgoday]

Which issue does this PR close?

Closes #496.

Rationale for this change

Check the arguments coerced to an aggregation and return a DataFusionError::Plan if no arguments can be associated with the function call.
Error message will display something like

Aggregate error. Invalid or wrong number of arguments passed to aggregate 'COUNT(DISTINCT )'

What changes are included in this PR?

Checks if there are some valid coerced arguments to call an aggregation function in create_aggregate_expr (datafusion/src/physical_plan/aggregates.rs)

Are there any user-facing changes?

[codecov-commenter]

Codecov Report

Merging #505 (9db2e5f) into master (28b0dad) will increase coverage by 0.16%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master     #505      +/-   ##
==========================================
+ Coverage   75.92%   76.08%   +0.16%     
==========================================
  Files         154      155       +1     
  Lines       26195    26555     +360     
==========================================
+ Hits        19889    20205     +316     
- Misses       6306     6350      +44     

Impacted Files	Coverage Δ
datafusion/src/physical_plan/aggregates.rs	91.35% <100.00%> (+0.33%)	⬆️
datafusion/tests/sql.rs	99.28% <100.00%> (+<0.01%)	⬆️
datafusion/src/optimizer/utils.rs	48.22% <0.00%> (-1.78%)	⬇️
...ta/rust/core/src/serde/physical_plan/from_proto.rs	38.65% <0.00%> (-0.99%)	⬇️
datafusion/src/physical_plan/hash_join.rs	85.52% <0.00%> (-0.93%)	⬇️
...sta/rust/core/src/serde/logical_plan/from_proto.rs	35.91% <0.00%> (-0.26%)	⬇️
...ista/rust/core/src/serde/physical_plan/to_proto.rs	50.31% <0.00%> (-0.16%)	⬇️
datafusion/src/physical_plan/planner.rs	80.19% <0.00%> (-0.14%)	⬇️
datafusion/src/logical_plan/builder.rs	90.04% <0.00%> (-0.05%)	⬇️
python/src/dataframe.rs	0.00% <0.00%> (ø)
... and 9 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 28b0dad...9db2e5f. Read the comment docs.

[alamb]

Can we please also add a test (to sql.rs) that exercises this code path?

[jgoday]

@alamb Added this simple test (trying to create a physical plan with the example that you posted in the issue).

async fn test_aggregation_with_bad_arguments() -> Result<()> {
    let mut ctx = ExecutionContext::new();
    register_aggregate_csv(&mut ctx)?;
    let sql = "SELECT COUNT(DISTINCT) FROM aggregate_test_100";
    let logical_plan = ctx.create_logical_plan(&sql).?;
    let physical_plan = ctx.create_physical_plan(&logical_plan);
    assert!(physical_plan.is_err());
    Ok(())
}

[alamb]

Looks good to me. Thank you @jgoday

[alamb]

Thanks @jgoday -- this is much nicer than panic! ing :)