Improve GroupBy

[sundy-li]

I hereby agree to the terms of the CLA available at: https://datafuse.rs/policies/cla/

Summary

• More generic way in AggregateFunction.
• Refactor group by, disable take block in the group by, this can avoid lots memcpy, performance can improve 30%~50%.
• Now we don't need to store "_group_keys" in the partial group by stage, using deserialize from hash key instead.
• Improve Cast performance in arrow2.
• Always convert constant column to full column before aggregation.
• ...

Changelog

• Improvement

Related Issues

Fixes #1340 #644

Test Plan

Unit Tests

Stateless Tests

[databend-bot]

Thanks for the contribution!
I have applied any labels matching special text in your PR Changelog.

Please review the labels and make any necessary changes.

[sundy-li]

/runperf

[bohutang]

/run-perf master

[databend-bot]

run performance on sha b7b370b reference on master

[sundy-li]

GroupBy is 1.7X faster.

https://test-1306375239.cos.accelerate.myqcloud.com/report/1334/b7b370b3e8336ae007a1fa143939675443c9b2f0/36b91cc0-5a19-4256-9644-6c71011831a4/performance.html

[sundy-li]

/run-perf master

[databend-bot]

run performance on sha 0a0009e reference on master

[codecov-commenter]

Codecov Report

Merging #1334 (0a0009e) into master (ad6cca6) will decrease coverage by 0%.
The diff coverage is 60%.

@@          Coverage Diff           @@
##           master   #1334   +/-   ##
======================================
- Coverage      72%     72%   -1%     
======================================
  Files         481     485    +4     
  Lines       28033   28032    -1     
======================================
- Hits        20311   20228   -83     
- Misses       7722    7804   +82     

Impacted Files	Coverage Δ
common/datavalues/src/arrays/builders/builder.rs	41% <ø> (-17%)	⬇️
common/datavalues/src/arrays/comparison.rs	37% <ø> (ø)
common/datavalues/src/arrays/kernels/iterator.rs	47% <ø> (ø)
common/datavalues/src/arrays/ops/downcast.rs	65% <ø> (ø)
common/datavalues/src/columns/common.rs	88% <ø> (+41%)	⬆️
common/datavalues/src/data_df_type.rs	11% <ø> (ø)
common/datavalues/src/macros.rs	34% <0%> (-2%)	⬇️
common/datavalues/src/series/arithmetic.rs	63% <ø> (ø)
common/datavalues/src/series/date_wrap.rs	0% <0%> (ø)
common/datavalues/src/utils.rs	24% <ø> (-20%)	⬇️
... and 60 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update ad6cca6...0a0009e. Read the comment docs.

[sundy-li]

https://test-1306375239.cos.accelerate.myqcloud.com/report/1334/0a0009ee85293be40899e94a553f9874f3b8265f/3471f870-3f4c-4562-b845-ccaef75733ea/performance.html

GroupBy is nearly 2.0X faster now!

[bohutang]

LGTM
Thank you @sundy-li

[databend-bot]

CI Passed
Reviewer Approved
Let's Merge