Skip to content

Removed a rule that turned Shuffle into Map breaking some TPCDS queries #6982

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
pavelvelikhov merged 5 commits into from
Jul 30, 2024
Merged

Removed a rule that turned Shuffle into Map breaking some TPCDS queries #6982

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
7651b50
Added tpcds 87 unit test
pavelvelikhov Jul 23, 2024
f50d056
Removed a broken rule that removed HashShuffle
pavelvelikhov Jul 30, 2024
26ebfe2
Commented out TCPDS64, still broken
pavelvelikhov Jul 30, 2024
5359d9f
Removed some debug code
pavelvelikhov Jul 30, 2024
c23d6d3
Cannonized clickbench
pavelvelikhov Jul 30, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
9 changes: 0 additions & 9 deletions ydb/core/kqp/opt/physical/kqp_opt_phy.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -88,8 +88,6 @@ class TKqpPhysicalOptTransformer : public TOptimizeTransformerBase {
AddHandler(0, &TCoTake::Match, HNDL(PropagatePrecomuteTake<false>));
AddHandler(0, &TCoFlatMap::Match, HNDL(PropagatePrecomuteFlatmap<false>));

AddHandler(0, &TDqCnHashShuffle::Match, HNDL(BuildHashShuffleByKeyStage));

AddHandler(0, &TCoAggregateCombine::Match, HNDL(ExpandAggregatePhase));
AddHandler(0, &TCoAggregateCombineState::Match, HNDL(ExpandAggregatePhase));
AddHandler(0, &TCoAggregateMergeState::Match, HNDL(ExpandAggregatePhase));
Expand Down Expand Up @@ -253,13 +251,6 @@ class TKqpPhysicalOptTransformer : public TOptimizeTransformerBase {
return output;
}

TMaybeNode<TExprBase> BuildHashShuffleByKeyStage(TExprBase node, TExprContext& ctx) {
auto output = DqBuildHashShuffleByKeyStage(node, ctx, {});
DumpAppliedRule("BuildHashShuffleByKeyStage", node.Ptr(), output.Ptr(), ctx);
return TExprBase(output);
}


TMaybeNode<TExprBase> ExpandAggregatePhase(TExprBase node, TExprContext& ctx) {
auto output = ExpandAggregatePeepholeImpl(node.Ptr(), ctx, TypesCtx, KqpCtx.Config->HasOptUseFinalizeByKey(), false);
DumpAppliedRule("ExpandAggregatePhase", node.Ptr(), output, ctx);
Expand Down
42 changes: 42 additions & 0 deletions ydb/core/kqp/ut/join/data/queries/tpcds87.sql
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,42 @@
PRAGMA TablePathPrefix='/Root/test/ds';

-- NB: Subquerys
$bla1 = (select distinct
COALESCE(c_last_name,'') as c_last_name,
COALESCE(c_first_name,'') as c_first_name,
COALESCE(cast(d_date as date), cast(0 as Date)) as d_date
from store_sales as store_sales
cross join date_dim as date_dim
cross join customer as customer
where store_sales.ss_sold_date_sk = date_dim.d_date_sk
and store_sales.ss_customer_sk = customer.c_customer_sk
and d_month_seq between 1221 and 1221+11);

$bla2 = ((select distinct
COALESCE(c_last_name,'') as c_last_name,
COALESCE(c_first_name,'') as c_first_name,
COALESCE(cast(d_date as date), cast(0 as Date)) as d_date
from catalog_sales as catalog_sales
cross join date_dim as date_dim
cross join customer as customer
where catalog_sales.cs_sold_date_sk = date_dim.d_date_sk
and catalog_sales.cs_bill_customer_sk = customer.c_customer_sk
and d_month_seq between 1221 and 1221+11)
union all
(select distinct
COALESCE(c_last_name,'') as c_last_name,
COALESCE(c_first_name,'') as c_first_name,
COALESCE(cast(d_date as date), cast(0 as Date)) as d_date
from web_sales as web_sales
cross join date_dim as date_dim
cross join customer as customer
where web_sales.ws_sold_date_sk = date_dim.d_date_sk
and web_sales.ws_bill_customer_sk = customer.c_customer_sk
and d_month_seq between 1221 and 1221+11));

-- start query 1 in stream 0 using template query87.tpl and seed 1819994127
select count(*)
from $bla1 bla1 left only join $bla2 bla2 using (c_last_name, c_first_name, d_date)
;

-- end query 1 in stream 0 using template query87.tpl
25 changes: 23 additions & 2 deletions ydb/core/kqp/ut/join/kqp_join_order_ut.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -251,6 +251,27 @@ Y_UNIT_TEST_SUITE(KqpJoinOrder) {
ExplainJoinOrderTestDataQuery("queries/tpcds61.sql", StreamLookupJoin);
}

void ExecuteJoinOrderTestDataQueryWithStats(const TString& queryPath, const TString& statsPath, bool useStreamLookupJoin) {
auto kikimr = GetKikimrWithJoinSettings(useStreamLookupJoin, GetStatic(statsPath));
auto db = kikimr.GetTableClient();
auto session = db.CreateSession().GetValueSync().GetSession();

CreateSampleTable(session);

/* join with parameters */
{
const TString query = GetStatic(queryPath);

auto result = session.ExecuteDataQuery(query,TTxControl::BeginTx().CommitTx()).ExtractValueSync();

UNIT_ASSERT_VALUES_EQUAL(result.GetStatus(), EStatus::SUCCESS);
}
}

Y_UNIT_TEST_TWIN(TPCDS87, StreamLookupJoin) {
ExecuteJoinOrderTestDataQueryWithStats("queries/tpcds87.sql", "stats/tpcds1000s.json", StreamLookupJoin);
}

Y_UNIT_TEST_TWIN(TPCDS88, StreamLookupJoin) {
ExplainJoinOrderTestDataQuery("queries/tpcds88.sql", StreamLookupJoin);
}
Expand Down Expand Up @@ -314,13 +335,13 @@ Y_UNIT_TEST_SUITE(KqpJoinOrder) {
);
}

/*
/*
Y_UNIT_TEST_TWIN(OverrideStatsTPCDS64, StreamLookupJoin) {
JoinOrderTestWithOverridenStats(
"queries/tpcds64.sql", "stats/tpcds1000s.json", "join_order/tpcds64_1000s.json", StreamLookupJoin
);
}
*/
*/

Y_UNIT_TEST_TWIN(OverrideStatsTPCDS78, StreamLookupJoin) {
JoinOrderTestWithOverridenStats(
Expand Down
44 changes: 0 additions & 44 deletions ydb/library/yql/dq/opt/dq_opt_phy.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1190,50 +1190,6 @@ TExprBase DqBuildShuffleStage(TExprBase node, TExprContext& ctx, IOptimizationCo
.Done();
}

NNodes::TExprBase DqBuildHashShuffleByKeyStage(NNodes::TExprBase node, TExprContext& ctx, const TParentsMap& /*parentsMap*/) {
if (!node.Maybe<TDqCnHashShuffle>()) {
return node;
}
auto cnHash = node.Cast<TDqCnHashShuffle>();
auto stage = cnHash.Output().Stage();
if (!stage.Program().Body().Maybe<TCoExtend>()) {
return node;
}
auto extend = stage.Program().Body().Cast<TCoExtend>();
TNodeSet nodes;
for (auto&& i : stage.Program().Args()) {
nodes.emplace(i.Raw());
}
for (auto&& i : extend) {
if (nodes.erase(i.Raw()) != 1) {
return node;
}
}
if (!nodes.empty()) {
return node;
}
TExprNode::TListType nodesTuple;
for (auto&& i : stage.Inputs()) {
if (!i.Maybe<TDqCnUnionAll>()) {
return node;
}
auto uAll = i.Cast<TDqCnUnionAll>();
nodesTuple.emplace_back(ctx.ChangeChild(node.Ref(), 0, uAll.Output().Ptr()));
}
auto stageCopy = ctx.ChangeChild(stage.Ref(), 0, ctx.NewList(node.Pos(), std::move(nodesTuple)));
auto output =
Build<TDqOutput>(ctx, node.Pos())
.Stage(stageCopy)
.Index().Build("0")
.Done();
auto outputCnMap =
Build<TDqCnMap>(ctx, node.Pos())
.Output(output)
.Done();

return TExprBase(outputCnMap);
}

TExprBase DqBuildFinalizeByKeyStage(TExprBase node, TExprContext& ctx,
const TParentsMap& parentsMap, bool allowStageMultiUsage)
{
Expand Down
1 change: 0 additions & 1 deletion ydb/library/yql/dq/opt/dq_opt_phy.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -55,7 +55,6 @@ NNodes::TExprBase DqBuildShuffleStage(NNodes::TExprBase node, TExprContext& ctx,

NNodes::TExprBase DqBuildFinalizeByKeyStage(NNodes::TExprBase node, TExprContext& ctx,
const TParentsMap& parentsMap, bool allowStageMultiUsage = true);
NNodes::TExprBase DqBuildHashShuffleByKeyStage(NNodes::TExprBase node, TExprContext& ctx, const TParentsMap& parentsMap);

NNodes::TExprBase DqBuildAggregationResultStage(NNodes::TExprBase node, TExprContext& ctx,
IOptimizationContext& optCtx);
Expand Down
15 changes: 6 additions & 9 deletions ...s/functional/clickbench/canondata/test.test_plans_column_/queries-original-plan-column-22
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -51,7 +51,10 @@
"PlanNodeId": 8,
"Plans": [
{
"Node Type": "Map",
"KeyColumns": [
"SearchPhrase"
],
"Node Type": "HashShuffle",
"PlanNodeId": 7,
"PlanNodeType": "Connection",
"Plans": [
Expand All @@ -73,10 +76,7 @@
"PlanNodeId": 6,
"Plans": [
{
"KeyColumns": [
"SearchPhrase"
],
"Node Type": "HashShuffle",
"Node Type": "UnionAll",
"PlanNodeId": 5,
"PlanNodeType": "Connection",
"Plans": [
Expand Down Expand Up @@ -279,10 +279,7 @@
]
},
{
"KeyColumns": [
"SearchPhrase"
],
"Node Type": "HashShuffle",
"Node Type": "UnionAll",
"PlanNodeId": 3,
"PlanNodeType": "Connection",
"Plans": [
Expand Down
15 changes: 6 additions & 9 deletions ...ts/functional/clickbench/canondata/test.test_plans_column_/queries-original-plan-column-9
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -51,7 +51,10 @@
"PlanNodeId": 8,
"Plans": [
{
"Node Type": "Map",
"KeyColumns": [
"RegionID"
],
"Node Type": "HashShuffle",
"PlanNodeId": 7,
"PlanNodeType": "Connection",
"Plans": [
Expand All @@ -73,10 +76,7 @@
"PlanNodeId": 6,
"Plans": [
{
"KeyColumns": [
"RegionID"
],
"Node Type": "HashShuffle",
"Node Type": "UnionAll",
"PlanNodeId": 5,
"PlanNodeType": "Connection",
"Plans": [
Expand Down Expand Up @@ -136,10 +136,7 @@
]
},
{
"KeyColumns": [
"RegionID"
],
"Node Type": "HashShuffle",
"Node Type": "UnionAll",
"PlanNodeId": 3,
"PlanNodeType": "Connection",
"Plans": [
Expand Down
15 changes: 6 additions & 9 deletions ydb/tests/functional/clickbench/canondata/test.test_plans_row_/queries-original-plan-row-22
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -51,7 +51,10 @@
"PlanNodeId": 8,
"Plans": [
{
"Node Type": "Map",
"KeyColumns": [
"SearchPhrase"
],
"Node Type": "HashShuffle",
"PlanNodeId": 7,
"PlanNodeType": "Connection",
"Plans": [
Expand All @@ -73,10 +76,7 @@
"PlanNodeId": 6,
"Plans": [
{
"KeyColumns": [
"SearchPhrase"
],
"Node Type": "HashShuffle",
"Node Type": "UnionAll",
"PlanNodeId": 5,
"PlanNodeType": "Connection",
"Plans": [
Expand Down Expand Up @@ -122,10 +122,7 @@
]
},
{
"KeyColumns": [
"SearchPhrase"
],
"Node Type": "HashShuffle",
"Node Type": "UnionAll",
"PlanNodeId": 3,
"PlanNodeType": "Connection",
"Plans": [
Expand Down
15 changes: 6 additions & 9 deletions ydb/tests/functional/clickbench/canondata/test.test_plans_row_/queries-original-plan-row-9
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -51,7 +51,10 @@
"PlanNodeId": 8,
"Plans": [
{
"Node Type": "Map",
"KeyColumns": [
"RegionID"
],
"Node Type": "HashShuffle",
"PlanNodeId": 7,
"PlanNodeType": "Connection",
"Plans": [
Expand All @@ -73,10 +76,7 @@
"PlanNodeId": 6,
"Plans": [
{
"KeyColumns": [
"RegionID"
],
"Node Type": "HashShuffle",
"Node Type": "UnionAll",
"PlanNodeId": 5,
"PlanNodeType": "Connection",
"Plans": [
Expand Down Expand Up @@ -113,10 +113,7 @@
]
},
{
"KeyColumns": [
"RegionID"
],
"Node Type": "HashShuffle",
"Node Type": "UnionAll",
"PlanNodeId": 3,
"PlanNodeType": "Connection",
"Plans": [
Expand Down