chore(bench): add bench for json parser

[tabVersion]

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

What's changed and what's your intention?

Checklist For Contributors

as title, the pr focuses on comparing cases that column names that do not match json keys

- [ ] I have written necessary rustdoc comments
- [ ] I have added necessary unit tests and integration tests
- [ ] I have added fuzzing tests or opened an issue to track them. (Optional, recommended for new SQL features #7934).
- [ ] I have demonstrated that backward compatibility is not broken by breaking changes and created issues to track deprecated features to be removed in the future. (Please refer to the issue)

• All checks passed in ./risedev check (or alias, ./risedev c)

Checklist For Reviewers

• I have requested macro/micro-benchmarks as this PR can affect performance substantially, and the results are shown.

Documentation

• My PR DOES NOT contain user-facing changes.

Click here for Documentation

Types of user-facing changes

Please keep the types that apply to your changes, and remove the others.

• Installation and deployment
• Connector (sources & sinks)
• SQL commands, functions, and operators
• RisingWave cluster configuration changes
• Other (please specify in the release note below)

Release note

[codecov]

Codecov Report

Merging #9195 (6476bd2) into main (a2fe5aa) will increase coverage by 0.00%.
The diff coverage is n/a.

@@           Coverage Diff           @@
##             main    #9195   +/-   ##
=======================================
  Coverage   70.81%   70.82%           
=======================================
  Files        1200     1200           
  Lines      200245   200245           
=======================================
+ Hits       141813   141822    +9     
+ Misses      58432    58423    -9     

Flag	Coverage Δ
rust	70.82% <ø> (+<0.01%)	⬆️

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

see 3 files with indirect coverage changes

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[kwannoel]

LGTM as a starting point for benchmarking json parser. Thanks for this!

I'm curious about source::benches::json_parser.rs though. Is that benchmark still relevant then?

[tabVersion]

LGTM as a starting point for benchmarking json parser. Thanks for this!

I'm curious about source::benches::json_parser.rs though. Is that benchmark still relevant then?

Yes, it is still ok to benchmark json parser perf for general purposes. In this pr, because I suspect the regression is introduced in #9086, it focuses on comparing cases that column names that do not match json keys.

[tabVersion]

run on local machine: Apple M1 silicon, 32G RAM

ParseMatchCase/32       time:   [673.13 ms 675.36 ms 677.62 ms]

ParseMatchCase/128      time:   [638.88 ms 644.43 ms 653.78 ms]

ParseMatchCase/512      time:   [646.35 ms 652.50 ms 660.39 ms]

ParseMatchCase/1024     time:   [638.15 ms 640.67 ms 643.56 ms]

ParseMatchCase/2048     time:   [632.13 ms 633.57 ms 635.16 ms]

ParseMatchCase/4096     time:   [626.47 ms 627.26 ms 628.06 ms]

ParseMisMatchCase/32    time:   [712.50 ms 715.36 ms 718.23 ms]

ParseMisMatchCase/128   time:   [681.59 ms 685.19 ms 689.48 ms]

ParseMisMatchCase/512   time:   [669.22 ms 670.75 ms 672.17 ms]

ParseMisMatchCase/1024  time:   [669.12 ms 670.78 ms 672.34 ms]

ParseMisMatchCase/2048  time:   [671.74 ms 679.59 ms 690.52 ms]

ParseMisMatchCase/4096  time:   [675.99 ms 682.49 ms 692.14 ms]

seems there gonna be an extra 40ms+(5%) if column names do not match json keys. maybe it can explain the observation in #9113

[kwannoel]

run on local machine: Apple M1 silicon, 32G RAM

ParseMatchCase/32       time:   [673.13 ms 675.36 ms 677.62 ms]

ParseMatchCase/128      time:   [638.88 ms 644.43 ms 653.78 ms]

ParseMatchCase/512      time:   [646.35 ms 652.50 ms 660.39 ms]

ParseMatchCase/1024     time:   [638.15 ms 640.67 ms 643.56 ms]

ParseMatchCase/2048     time:   [632.13 ms 633.57 ms 635.16 ms]

ParseMatchCase/4096     time:   [626.47 ms 627.26 ms 628.06 ms]

ParseMisMatchCase/32    time:   [712.50 ms 715.36 ms 718.23 ms]

ParseMisMatchCase/128   time:   [681.59 ms 685.19 ms 689.48 ms]

ParseMisMatchCase/512   time:   [669.22 ms 670.75 ms 672.17 ms]

ParseMisMatchCase/1024  time:   [669.12 ms 670.78 ms 672.34 ms]

ParseMisMatchCase/2048  time:   [671.74 ms 679.59 ms 690.52 ms]

ParseMisMatchCase/4096  time:   [675.99 ms 682.49 ms 692.14 ms]

seems there gonna be an extra 40ms+(5%) if column names do not match json keys. maybe it can explain the observation in #9113

Hmm not entirely, but can be a contributing factor. E2e was 6% but just json parser alone regression is like 60% (11%/18%=0.6), so 5% doesnt seem to explain it fully.

Flamegraph percentages probably not too reliable though, only rough indicators. We can probably can checkout the old main branch where 11% is observed and run the json parser micro benchmarks, get a more precise figure.