executor: vectorize the merge join executor by vecGroupChecker

[catror]

PCP #14216

What problem does this PR solve?

Vectorize the merge join executor by vecGroupChecker, then we can process it group by group.

What is changed and how it works?

1. combine struct mergeJoinInnerTable and mergeJoinOuterTable to mergeJoinTable
2. group rows by join keys by vecGroupChecker
3. process outer table by group instead of by row

Check List

Tests

• Unit test
• Integration test

Benchmark

benchstat -delta-test none old.txt new.txt
name                                                                                                                                                old time/op    new time/op    delta
MergeJoinExec/merge_join_(outerRows:300000,_innerRows:300000,_concurency:4,_outerJoinKeyIdx:_[0_1],_innerJoinKeyIdx:_[0_1],_NeedOuterSort:false)-8     1.27s ± 0%     0.34s ± 0%  -72.87%
MergeJoinExec/merge_join_(outerRows:3000,_innerRows:300000,_concurency:4,_outerJoinKeyIdx:_[0_1],_innerJoinKeyIdx:_[0_1],_NeedOuterSort:false)-8       338ms ± 0%     172ms ± 0%  -48.96%
MergeJoinExec/merge_join_(outerRows:30,_innerRows:300000,_concurency:4,_outerJoinKeyIdx:_[0_1],_innerJoinKeyIdx:_[0_1],_NeedOuterSort:false)-8         330ms ± 0%      43ms ± 0%  -86.96%
MergeJoinExec/merge_join_(outerRows:300000,_innerRows:330000,_concurency:4,_outerJoinKeyIdx:_[0_1],_innerJoinKeyIdx:_[0_1],_NeedOuterSort:false)-8     958ms ± 0%     309ms ± 0%  -67.73%

name                                                                                                                                                old alloc/op   new alloc/op   delta
MergeJoinExec/merge_join_(outerRows:300000,_innerRows:300000,_concurency:4,_outerJoinKeyIdx:_[0_1],_innerJoinKeyIdx:_[0_1],_NeedOuterSort:false)-8    72.2MB ± 0%    58.5MB ± 0%  -18.98%
MergeJoinExec/merge_join_(outerRows:3000,_innerRows:300000,_concurency:4,_outerJoinKeyIdx:_[0_1],_innerJoinKeyIdx:_[0_1],_NeedOuterSort:false)-8      65.9MB ± 0%    58.5MB ± 0%  -11.25%
MergeJoinExec/merge_join_(outerRows:30,_innerRows:300000,_concurency:4,_outerJoinKeyIdx:_[0_1],_innerJoinKeyIdx:_[0_1],_NeedOuterSort:false)-8        64.5MB ± 0%    62.6MB ± 0%   -2.94%
MergeJoinExec/merge_join_(outerRows:300000,_innerRows:330000,_concurency:4,_outerJoinKeyIdx:_[0_1],_innerJoinKeyIdx:_[0_1],_NeedOuterSort:false)-8    72.1MB ± 0%    58.5MB ± 0%  -18.97%

name                                                                                                                                                old allocs/op  new allocs/op  delta
MergeJoinExec/merge_join_(outerRows:300000,_innerRows:300000,_concurency:4,_outerJoinKeyIdx:_[0_1],_innerJoinKeyIdx:_[0_1],_NeedOuterSort:false)-8      908k ± 0%        4k ± 0%  -99.59%
MergeJoinExec/merge_join_(outerRows:3000,_innerRows:300000,_concurency:4,_outerJoinKeyIdx:_[0_1],_innerJoinKeyIdx:_[0_1],_NeedOuterSort:false)-8       37.8k ± 0%      3.8k ± 0%  -90.08%
MergeJoinExec/merge_join_(outerRows:30,_innerRows:300000,_concurency:4,_outerJoinKeyIdx:_[0_1],_innerJoinKeyIdx:_[0_1],_NeedOuterSort:false)-8         8.72k ± 0%     3.71k ± 0%  -57.45%
MergeJoinExec/merge_join_(outerRows:300000,_innerRows:330000,_concurency:4,_outerJoinKeyIdx:_[0_1],_innerJoinKeyIdx:_[0_1],_NeedOuterSort:false)-8      908k ± 0%        4k ± 0%  -99.58%

[sre-bot]

Thanks for your contribution. If your PR get merged, you will be rewarded 3000 points.

[qw4990]

/run-unit-test

[qw4990]

Hi @catror , I add a new test for the vectorized merge join in this PR #14519, please merge it into your PR to check this PR's correctness at your convenience.
And also please fix this CI problem in the check_dev_2 test:

[2020-01-17T09:51:51.002Z] FAIL: physical_plan_test.go:784: testPlanSuite.TestAggToCopHint
[2020-01-17T09:51:51.002Z] 
[2020-01-17T09:51:51.002Z] physical_plan_test.go:824:
[2020-01-17T09:51:51.002Z]     c.Assert(planString, Equals, output[i].Best, comment)
[2020-01-17T09:51:51.002Z] ... obtained string = "TableReader(Table(t))"
[2020-01-17T09:51:51.002Z] ... expected string = "TableReader(Table(t)->HashAgg)->HashAgg->HashAgg"
[2020-01-17T09:51:51.002Z] ... case:2 sql:select /*+ AGG_TO_COP(), HASH_AGG(), USE_INDEX(t) */ distinct a from t group by a

[qw4990]

@catror Friendly ping, please resolve conflicts.

[catror]

@catror Friendly ping, please resolve conflicts.

Done.

[qw4990]

LGTM, NICE WORK, Thanks for your contribution.

[wshwsh12]

LGTM

[sre-bot]

/run-all-tests

[sre-bot]

@catror complete task #14216 and get 3000 score, currerent score 3500