Support parquet statistics for struct columns

[alamb]

Is your feature request related to a problem or challenge?

While working on #8294 @tustvold noted that the statistics extraction code does not do the right thing with Structs that were written to parquet

I think the usecase is something like the following query:

SELECT * 
FROM my_table
WHERE struct_column = struct('foo', 'bar')

Describe the solution you'd like

I would like the parquet statistics reading code to handle structs (I think).

Describe alternatives you've considered

No response

Additional context

No response

[alamb]

It appears that statistics for such types is ready as all nulls

[tustvold]

Yes, this is what I would expect in the absence of a column name collision. My greater concern is that the presence of the struct column will mess up the ordinals of the other columns

[alamb]

Yes, this is what I would expect in the absence of a column name collision. My greater concern is that the presence of the struct column will mess up the ordinals of the other columns

I added tests for this case here: #8294 (comment) (TLDR is it seems to do the right thing, though it would be good to get a second set of eyes)

[edmondop]

I read in statistics

/// Lookups up the parquet column by name
///
/// Returns the parquet column index and the corresponding arrow field
pub(crate) fn parquet_column<'a>(
    parquet_schema: &SchemaDescriptor,
    arrow_schema: &'a Schema,
    name: &str,
) -> Option<(usize, &'a FieldRef)> {
    let (root_idx, field) = arrow_schema.fields.find(name)?;
    if field.data_type().is_nested() {
        // Nested fields are not supported and require non-trivial logic
        // to correctly walk the parquet schema accounting for the
        // logical type rules - <https://github.com/apache/parquet-format/blob/master/LogicalTypes.md>
        //
        // For example a ListArray could correspond to anything from 1 to 3 levels
        // in the parquet schema
        return None;
    }

    // This could be made more efficient (#TBD)
    let parquet_idx = (0..parquet_schema.columns().len())
        .find(|x| parquet_schema.get_column_root_idx(*x) == root_idx)?;
    Some((parquet_idx, field))
}```

`git blame` shows @alamb as an author of those lines... I'll look into the rules.  I suppose aggregation functions will need to be updated to walk the schema correctly? 

[edmondop]

@alamb I realized the last part of the code is actually the question I have for you, apologies

I'll look into the rules. I suppose aggregation functions will need to be updated to walk the schema correctly?

[alamb]

I'll look into the rules. I suppose aggregation functions will need to be updated to walk the schema correctly?

That sounds likely.

I think the first thing to do would be to write a test case showing the incorrect / lack of behavior. Then we can work out how to solve the problem

[edmondop]

@alamb I have done some progress in getting a test, but I need guidance here:

        let (idx, _) = parquet_column(parquet_schema, &schema, "struct_col").unwrap();
        assert_eq!(idx, 0);
 
        let row_groups = metadata.row_groups();
        let iter = row_groups.iter().map(|x| x.column(idx).statistics());

The test now fails because row_groups.iter().map(|x| x.column(idx).statistics()) doesn't return the statistics of the first column of the RecordBatch, the struct, but of the first column of the struct. I can definitely create myself the statistics doing something like that:

        let datatype = DataType::Struct(
            Fields::from(vec![
                Field::new("bool_col", DataType::Boolean, false),
                Field::new("int_col", DataType::Boolean, false),
            ]),
        );
       // Comment next line, change iter above for the two fields of the struct, and build the min as a struxt
        let min_field1 = min_statistics(&datatype, iter1.clone()).unwrap();
        let min_field2 = min_statistics(&datatype, iter2.clone()).unwrap();
       let min = find out how to do this// 
        assert_eq!(
            &min,
            &expected_min,
            "Min. Statistics\n\n{}\n\n",
            DisplayStats(row_groups)
        );

but I am not sure that's the right way to proceed