type coercion error in clickbench50 query

[askalt]

Describe the bug

Extended clickbench query (benchmarks/queries/clickbench/extended/q6.sql) plan is failed with the error:

Internal error: Expect TypeSignatureClass::Native(LogicalType(Native(String), String)) but received NativeType::Binary, DataType: BinaryView.

To Reproduce

Run datafusion-cli:

> CREATE EXTERNAL TABLE hits STORED AS PARQUET LOCATION "../benchmarks/data/hits_partitioned"

> SELECT COUNT(*) AS ShareCount FROM hits WHERE "IsMobile" = 1 AND "MobilePhoneModel" LIKE 'iPhone%' AND "SocialAction" = 'share' AND "SocialSourceNetworkID" IN (5, 12) AND "ClientTimeZone" BETWEEN -5 AND 5 AND regexp_match("Referer", '\/campaign\/(spring|summer)_promo') IS NOT NULL AND CASE WHEN split_part(split_part("URL", 'resolution=', 2), '&', 1) ~ '^\d+$' THEN split_part(split_part("URL", 'resolution=', 2), '&', 1)::INT ELSE 0 END > 1920 AND levenshtein(CAST("UTMSource" AS STRING), CAST("UTMCampaign" AS STRING)) < 3;

type_coercion
caused by
Internal error: Expect TypeSignatureClass::Native(LogicalType(Native(String), String)) but received NativeType::Binary, DataType: BinaryView.
This issue was likely caused by a bug in DataFusion's code. Please help us to resolve this by filing a bug report in our issue tracker: https://github.com/apache/datafusion/issues
$ 

Expected behavior

Query is planned successfully.

[askalt]

Reduced example:

> create table t (b bytea);

> select count(*) from t where split_part(b, 'resolution=', 2)::INT > 0;

type_coercion
caused by
Internal error: Expect TypeSignatureClass::Native(LogicalType(Native(String), String)) but received NativeType::Binary, DataType: Binary.
This issue was likely caused by a bug in DataFusion's code. Please help us to resolve this by filing a bug report in our issue tracker: https://github.com/apache/datafusion/issues

[lemorage]

take

[askalt]

It seems, binary_as_string configuration fixes:

set datafusion.execution.parquet.binary_as_string = true

Because SplitPartFunc explicitly wants string:

impl SplitPartFunc {
    pub fn new() -> Self {
        Self {
            signature: Signature::coercible(
                vec![
                    Coercion::new_exact(TypeSignatureClass::Native(logical_string())),
                    Coercion::new_exact(TypeSignatureClass::Native(logical_string())),
                    Coercion::new_implicit(
                        TypeSignatureClass::Native(logical_int64()),
                        vec![TypeSignatureClass::Integer],
                        NativeType::Int64,
                    ),
                ],
                Volatility::Immutable,
            ),
        }
    }
}

The question is: "why did it even work at one time without this setting?". Investigated that it worked before refactoring
1ce4b51

There was cast to utf-8 in the plan:

split_part(CAST(hits.URL AS Utf8), CAST(Utf8("resolution=") AS Utf8View), Int64(2))

So maybe we should set this configuration by default for query-planner benchmark.

[askalt]

Note: it causes query-planner benchmark fail:

#19792 (comment)

[lemorage]

Maybe it would be better to add the implicit coercion in SplitPartFunc::new() there?

[askalt]

Maybe it would be better to add the implicit coercion in SplitPartFunc::new() there?

Sounds good to me.