Improve count if rewrite

[taiyang-li]

Changelog category (leave one):

• Improvement

Changelog entry (a user-readable short description of the changes that goes to CHANGELOG.md):

sumIf(123, cond) -> 123 * countIf(1, cond) 
sum(if(cond, 123, 0)) -> 123 * countIf(cond)
sum(if(cond, 0, 123)) -> 123 * countIf(not(cond))

Information about CI checks: https://clickhouse.com/docs/en/development/continuous-integration/

[zhanglistar]

How much improve?

[Algunenano]

There are cases where this changes the result:

:) Select sumIf(NaN, number > 10) from numbers(10);

SELECT sumIf(nan, number > 10)
FROM numbers(10)

Query id: 3610ba57-4d00-41d7-b2ca-da2c31937b02

┌─sumIf(nan, greater(number, 10))─┐
│                               0 │
└─────────────────────────────────┘

1 row in set. Elapsed: 0.004 sec. 

:) Select NaN * countIf(number > 10) from numbers(10);

SELECT nan * countIf(number > 10)
FROM numbers(10)

Query id: b9158e73-bb1c-49e3-8e56-f555637c6d63

┌─multiply(nan, countIf(greater(number, 10)))─┐
│                                         nan │
└─────────────────────────────────────────────┘

1 row in set. Elapsed: 0.009 sec. 

[taiyang-li]

How much improve?

Query: SELECT sum(if(number % 2 == 0, 123, 0)) FROM numbers(1000000000)
Before: 1 rows in set. Elapsed: 1.233 sec. Processed 1.00 billion rows, 8.00 GB (810.95 million rows/s., 6.49 GB/s.)
After: 1 rows in set. Elapsed: 1.076 sec. Processed 1.00 billion rows, 8.00 GB (929.22 million rows/s., 7.43 GB/s.)

[taiyang-li]

@Algunenano Please notice that
Select sumIf(NaN, number > 10) from numbers(10); won't be rewritten to Select NaN * countIf(number > 10) from numbers(10);
Because NaN's type is not UInt64 or Int64.

[Avogar]

@taiyang-li please look at the Fuzzer failures, they look related to the changes

[taiyang-li]

@taiyang-li please look at the Fuzzer failures, they look related to the changes

@Avogar Thanks for reminding. They are fixed now.

[Avogar]

It will be great if you will also add the same implementation in the new analyzer. See src/Analyzer/Passes/SumIfToCountIfPass.cpp

[taiyang-li]

It will be great if you will also add the same implementation in the new analyzer. See src/Analyzer/Passes/SumIfToCountIfPass.cpp

Ok. I'll finish it later.

[taiyang-li]

It will be great if you will also add the same implementation in the new analyzer. See src/Analyzer/Passes/SumIfToCountIfPass.cpp

Done.

[taiyang-li]

@Avogar Do you have time to reivew again? Thanks very much.

[Avogar]

Sorry for long response, I was on vacation. let me review it now