feat(array): refactor StreamChunk & introduce record iterator

[BugenZhao]

What's changed and what's your intention?

In this PR, we refactor StreamChunk with ops and data_chunk instead of unzipping the data chunk into parts, so that we can reuse some functionalities of DataChunk.

https://github.com/singularity-data/risingwave/blob/1228f9272c6e7bcce9db86f7f7c982a1bc5032e0/src/common/src/array/stream_chunk.rs#L72-L79

Then, we remove the StreamChunkRowRef introduced in #1813 and provide a record iterator based on DataChunkRowRef whose item is...

https://github.com/singularity-data/risingwave/blob/1cc6400140e40843de46f79f2a575da6989a206d/src/common/src/array/stream_chunk_iter.rs#L74-L82

Some executors can be benefited from this record-style interface. For now, we keep the row interface for compatibility.

Benchmark

test array::stream_chunk_iter::tests::bench_record_iterator                 ... bench:      45,728 ns/iter (+/- 1,268)
test array::stream_chunk_iter::tests::bench_rows_iterator                   ... bench:      39,444 ns/iter (+/- 865)
test array::stream_chunk_iter::tests::bench_rows_iterator_from_records      ... bench:      73,246 ns/iter (+/- 2,349)
test array::stream_chunk_iter::tests::bench_rows_iterator_vec_of_datum_refs ... bench:     254,524 ns/iter (+/- 11,879)

Checklist

• I have written necessary docs and comments
• I have added necessary unit tests and integration tests

Refer to a related PR or issue link (optional)

• streaming: use StreamChunk::rows to iterate on rows #1839
• array: refactor rows iterator for DataChunk like StreamChunk #1840

[codecov]

Codecov Report

Merging #1916 (59d2eb0) into main (cdfac58) will increase coverage by 0.25%.
The diff coverage is 51.61%.

@@            Coverage Diff             @@
##             main    #1916      +/-   ##
==========================================
+ Coverage   70.62%   70.88%   +0.25%     
==========================================
  Files         611      617       +6     
  Lines       80064    79990      -74     
==========================================
+ Hits        56543    56697     +154     
+ Misses      23521    23293     -228     

Flag	Coverage Δ
rust	70.88% <51.61%> (+0.25%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
src/common/src/lib.rs	100.00% <ø> (ø)
src/common/src/test_utils/test_stream_chunk.rs	40.45% <7.46%> (-28.38%)	⬇️
src/common/src/array/stream_chunk_iter.rs	57.54% <54.25%> (-41.24%)	⬇️
src/common/src/array/data_chunk_iter.rs	86.75% <100.00%> (+7.62%)	⬆️
src/common/src/array/stream_chunk.rs	89.07% <100.00%> (+5.03%)	⬆️
src/stream/src/executor/debug/update_check.rs	87.35% <100.00%> (ø)
src/stream/src/executor_v2/hop_window.rs	80.55% <100.00%> (ø)
src/stream/src/executor_v2/top_n.rs	90.50% <100.00%> (-0.06%)	⬇️
src/stream/src/executor_v2/top_n_appendonly.rs	93.26% <100.00%> (-0.06%)	⬇️
src/batch/src/executor/stream_scan.rs	0.00% <0.00%> (-45.14%)	⬇️
... and 55 more

📣 Codecov can now indicate which changes are the most critical in Pull Requests. Learn more

[yuhao-su]

LGTM!

[fuyufjh]

Generally LGTM.

And a question: what's the benefit from doing so?

[BugenZhao]

And a question: what's the benefit from doing so?

Actually the main purpose of this PR is to reuse the row-level iterator of data chunk in #1905 for stream chunk. For record, some executors may use this interface to handle "update" more elegantly, but in the future. 😁