Thanks @edmondop I'm planning to read it through soon

Hi @edmondop are you planning to proceed on this PR?

Yes I was stuck because one test doesn't pass, will need to investigate unless you have some hints to share

Yes I was stuck because one test doesn't pass, will need to investigate unless you have some hints to share

If you referring to CI test failed, it is

  left: (0, "+----+----+-------------+------------+")
 right: (0, "+----+----+-------------+-------------+")

which should probably be straightforward to fix, we can check row counts first?

Yes I was stuck because one test doesn't pass, will need to investigate unless you have some hints to share

If you referring to CI test failed, it is

  left: (0, "+----+----+-------------+------------+")
 right: (0, "+----+----+-------------+-------------+")

which should probably be straightforward to fix, we can check row counts first?

Thanks for the tip, checking for the row count effectively show that the right_join filtered return a different number of rows when using SMJ and HashJoin. What could be the root cause?

I have tried to modify the column used for the filter un-commenting the section commented below, but this breaks all the other test cases too

fn less_than_10_join_filter(schema1: Arc<Schema>, _schema2: Arc<Schema>) -> JoinFilter {
    let less_than_100 = Arc::new(BinaryExpr::new(
        Arc::new(Column::new("a", 0)),
        Operator::Lt,
        Arc::new(Literal::new(ScalarValue::from(100))),
    )) as _;
    let column_indices = vec![
        ColumnIndex {
            index: 0,
            side: JoinSide::Left,
        },
        // ColumnIndex {
        //     index: 0,
        //     side: JoinSide::Right,
        // },
    ];
    let intermediate_schema =
        Schema::new(vec![schema1.field_with_name("a").unwrap().to_owned()]);

    JoinFilter::new(less_than_100, column_indices, intermediate_schema)
}

Thanks @edmondop, would you be able to fetch the exact example where HJ and SMJ mismatches?

Details

These are the two tests that fails

failures:
    fuzz_cases::join_fuzz::test_anti_join_1k_filtered
    fuzz_cases::join_fuzz::test_right_join_1k_filtered

---- fuzz_cases::join_fuzz::test_anti_join_1k_filtered stdout ----
thread 'fuzz_cases::join_fuzz::test_anti_join_1k_filtered' panicked at datafusion/core/tests/fuzz_cases/join_fuzz.rs:404:13:
assertion `left == right` failed: SortMergeJoinExec and HashJoinExec produced different row counts
  left: 5
 right: 902

right join passes on my machine, I think I need to rebase, but also it seems easier to solve

---- fuzz_cases::join_fuzz::test_right_join_1k_filtered stdout ----
thread 'fuzz_cases::join_fuzz::test_right_join_1k_filtered' panicked at datafusion/core/tests/fuzz_cases/join_fuzz.rs:377:70:
called `Result::unwrap()` on an `Err` value: ArrowError(InvalidArgumentError("Column 'a' is declared as non-nullable but contains null values"), None)
stack backtrace:

@edmondop I think it is a solid start already, lets comment this tests for now and address them separately

I took the liberty to ignore tests

Right filtered join fuzz tests failures probably related to #10882

Fixed formatting @comphead all the checks are passing now

FYI, CI is failing on main after merge.

FYI, CI is failing on main after merge.

This should fix it #10887. I also filed #10886 cc @comphead

Thanks @ozankabak and @edmondop
Not sure why it failed, the build was green. I'm checking tests