Support Avg distinct for float64 type

[qazxcdswe123]

Which issue does this PR close?

• related: avg(distinct) support #2408
• related Support Avg distinct #15356

Rationale for this change

The original PR is splited into 2 parts, one for float64 type and one for float64+decimal type

What changes are included in this PR?

1. mv DistinctSumAccumulator to common so that it can be used in Float64DistinctAvgAccumulator
2. implement Float64DistinctAvgAccumulator using DistinctSumAccumulator
3. tested in aggregate.slt

Are these changes tested?

• tested in aggregate.slt

Are there any user-facing changes?

No

[Omega359]

the sqlite tests need updating prior to this issue being pushed to main. Confirmed test failures, here are a few examples:

External error: query is expected to fail with error:
(regex) DataFusion error: Execution error: avg\(DISTINCT\) aggregations are not available
but got error:
DataFusion error: Arrow error: Invalid argument error: number of columns(2) must match number of fields(3) in schema
[SQL] SELECT - COUNT ( * ) * AVG ( DISTINCT + col3 ) AS col3 FROM tab0 WHERE NOT NULL <= + + col3
at ../../datafusion-testing/data/sqlite/index/random/100/slt_good_1.slt:11104
External error: query is expected to fail, but actually succeed:
[SQL] SELECT DISTINCT - COUNT ( * ) * AVG ( DISTINCT + 89 ) col4 FROM tab0 AS cor0 WHERE NULL NOT IN ( CAST ( NULL AS INTEGER ) + - 15 )
at ../../datafusion-testing/data/sqlite/index/random/1000/slt_good_8.slt:17760
External error: query is expected to fail, but actually succeed:
[SQL] SELECT - AVG ( DISTINCT 27 ) + - MAX ( DISTINCT col1 ) AS col4 FROM tab0 WHERE NOT NULL <> - + 90
at ../../datafusion-testing/data/sqlite/index/random/1000/slt_good_0.slt:6040
External error: query is expected to fail with error:
(regex) DataFusion error: Execution error: avg\(DISTINCT\) aggregations are not available
but got error:
DataFusion error: Arrow error: Invalid argument error: number of columns(1) must match number of fields(2) in schema
[SQL] SELECT ALL - AVG ( DISTINCT 48 ) AS col4 FROM tab0 WHERE NOT - col4 BETWEEN ( NULL ) AND CAST ( 2 AS INTEGER )
at ../../datafusion-testing/data/sqlite/index/random/1000/slt_good_7.slt:38250

[alamb]

I merged up from main and will use the new github action to run extended tests

[alamb]

Run extended tests

[Omega359]

Run extended tests

I see it did trigger but I somehow was expecting feedback in the comments

[Omega359]

Looks like it failed? https://github.com/apache/datafusion/actions/runs/14139465370/job/39618247236

extended test

[Omega359]

I regenerated the sqlite tests again and I think an issue I'm seeing with them is actually caused or triggered by this PR.

Here is an example:

query error DataFusion error: Arrow error: Invalid argument error: number of columns\(1\) must match number of fields\(2\) in schema
SELECT84*+-61*+AVG ( DISTINCT ( 70 ) ) AS col2 FROM tab4 WHERE ( NULL ) BETWEEN NULLAND col3

If this is in fact a bug not caused by this PR a new issue should be filed. I myself do not have the time to diagnosis the cause of this unfortunately.

[qazxcdswe123]

I regenerated the sqlite tests again and I think an issue I'm seeing with them is actually caused or triggered by this PR.

Here is an example:

query error DataFusion error: Arrow error: Invalid argument error: number of columns\(1\) must match number of fields\(2\) in schema
SELECT84*+-61*+AVG ( DISTINCT ( 70 ) ) AS col2 FROM tab4 WHERE ( NULL ) BETWEEN NULLAND col3

If this is in fact a bug not caused by this PR a new issue should be filed. I myself do not have the time to diagnosis the cause of this unfortunately.

sadly I'm working on my undergrad thesis project at this time and do not have time to investigate this either 😢 , might be back around mid april

[alamb]

sadly I'm working on my undergrad thesis project at this time and do not have time to investigate this either 😢 , might be back around mid april

Good luck with your project / thesis!