perf: optimize json parser

[kwannoel]

For q17

Now:

Before:

1. After investigation, we should setup some benchmarks for json parser which can catch this regression. (Or ignore this comment if they already do).
2. Overall, source throughput increased to 244,000 from 225,000. As such, normalized regression may be around 6% instead of 7%.
3. @tabVersion mentioned investigating fix(parser): fetch json keys regardless of cases #9086 may be helpful.

   @tabVersionI don’t know if it is caused by this pr fix(parser): fetch json keys regardless of cases #9086
   it adds a loop for all keys when parsing every row to match the column names and the json keys ignoring cases

Here's the flamegraph:


Interactive version

[tabVersion]

bench #9195

[lmatz]

Went through the flame graphs generated in https://buildkite.com/risingwavelabs/main-cron/builds/458#0187c454-58bd-439a-bf57-6f774ea41b80

For queries q0,q1,q2,q3,q7,q8,q10,q14,q15,q17,q18,q20,q21,q22, json parser takes more time than other executors.

[kwannoel]

Likely need to optimize append_datum_n

[kwannoel]

Investigating it

[st1page]

The SourceStreamChunkRowWriter here only append one row to the ArrayBuilderImpl and it will take dynamic dispatch for each row. Maybe we can append a batch of data each time?

[kwannoel]

Another potential cost

risingwave/src/common/src/array/mod.rs

Lines 481 to 496 in 5fabd2a

	pub fn append_datum_n(&mut self, n: usize, datum: impl ToDatumRef) {
	match datum.to_datum_ref() {
	None => match self {
	$( Self::$variant_name(inner) => inner.append_n(n, None), )*
	}
	Some(scalar_ref) => match (self, scalar_ref) {
	$( (Self::$variant_name(inner), ScalarRefImpl::$variant_name(v)) => inner.append_n(n, Some(v)), )*
	(this_builder, this_scalar_ref) => panic!(
	"Failed to append datum, array builder type: {}, scalar type: {}",
	this_builder.get_ident(),
	this_scalar_ref.get_ident()
	),
	},
	}
	}

The inner match under the Some branch can also cost 3%?

[st1page]

risingwave/src/common/src/array/mod.rs

Lines 485 to 490 in 060b233

	}
	Some(scalar_ref) => match (self, scalar_ref) {
	$( (Self::$variant_name(inner), ScalarRefImpl::$variant_name(v)) => inner.append_n(n, Some(v)), )*
	(this_builder, this_scalar_ref) => panic!(
	"Failed to append datum, array builder type: {}, scalar type: {}",
	this_builder.get_ident(),

The match here leads to too many branches(O(n^2)) do you have any idea? @BugenZhao

[github-actions]

This issue has been open for 60 days with no activity.

If you think it is still relevant today, and needs to be done in the near future, you can comment to update the status, or just manually remove the no-issue-activity label.

You can also confidently close this issue as not planned to keep our backlog clean.
Don't worry if you think the issue is still valuable to continue in the future.
It's searchable and can be reopened when it's time. 😄

[st1page]

closed with #17146 and related PRs
cc @BugenZhao