fix: dedup join key in degree table

[fuyufjh]

I hereby agree to the terms of the RisingWave Labs, Inc. Contributor License Agreement.

What's changed and what's your intention?

Deduplicate columns in HashJoin's degree table. Check the ch_benchmark.yaml for example.

Limitation: Note that this only applies to values. For the key, the columns are still duplicated, e.g. primary key: [ $0 ASC, $1 ASC, $0 ASC ]. This is because the HashJoin executor assumes that the join key is always the first (second, third...) columns in state/degree table's PK. As a result, fixing that may require additional work, such as passing a join_key_columns_in_state_table in proto.

Checklist

• I have written necessary rustdoc comments
• I have added necessary unit tests and integration tests
• I have added test labels as necessary. See details.
• I have added fuzzing tests or opened an issue to track them. (Optional, recommended for new SQL features Sqlsmith: Sql feature generation #7934).
• My PR contains breaking changes. (If it deprecates some features, please create a tracking issue to remove them in the future).
• All checks passed in ./risedev check (or alias, ./risedev c)
• My PR changes performance-critical code. (Please run macro/micro-benchmarks and show the results.)

• My PR contains critical fixes that are necessary to be merged into the latest release. (Please check out the details)

Documentation

• My PR needs documentation updates. (Please use the Release note section below to summarize the impact on users)

Release note

If this PR includes changes that directly affect users or other significant modifications relevant to the community, kindly draft a release note to provide a concise summary of these changes. Please prioritize highlighting the impact these changes will have on users.

[BugenZhao]

LGTM.

Since I'm not an expert in optimizer, I'm not sure whether there could be any potential issues similar to #7698 (comment).

[yuhao-su]

You may also need to fix the dk_indices_in_jk for degree table

[fuyufjh]

You may also need to fix the dk_indices_in_jk for degree table

I think not. Neither distribution key nor join key were affected. Let me add some test cases.

[yuhao-su]

Neither distribution key nor join key were affected. Let me add some test cases.

You are right. I guess what really need to be fixed is exector code. e.g. this https://github.com/singularity-data/risingwave/blob/9cd7f6486e659af72857aef7ac27247305e880c0/src/stream/src/executor/managed_state/join/mod.rs#L79

[fuyufjh]

Neither distribution key nor join key were affected. Let me add some test cases.

You are right. I guess what really need to be fixed is exector code. e.g. this https://github.com/singularity-data/risingwave/blob/9cd7f6486e659af72857aef7ac27247305e880c0/src/stream/src/executor/managed_state/join/mod.rs#L79

Seems no problem. order_key is projected with state_order_key_indices, which is from order_key_indices, which is from degree_table.pk_indices().to_vec()

Thanks for pointing this out.

The to_table_rows() (actually should be called to_degree_rows(), which builds a degree row from an incoming row), is hard to implement after this PR. This is because the projection from incoming row to degree row is missing. We need to pass this information in HashJoin's proto to make it work. This cause more trouble and I would rather close this PR.