Skip to content
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.

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

Fix: Offset planning in hash joins #16540

Merged
merged 5 commits into from
Aug 7, 2024
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
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
32 changes: 31 additions & 1 deletion go/test/endtoend/vtgate/vitess_tester/join/join.test
Original file line number Diff line number Diff line change
Expand Up @@ -25,6 +25,15 @@ CREATE TABLE `t3`
CHARSET utf8mb4,
COLLATE utf8mb4_unicode_ci;

CREATE TABLE `t4`
(
`id` bigint unsigned NOT NULL AUTO_INCREMENT,
`col` int unsigned NOT NULL,
PRIMARY KEY (`id`)
) ENGINE InnoDB,
CHARSET utf8mb4,
COLLATE utf8mb4_unicode_ci;

insert into t1 (id, name)
values (1, 'A'),
(2, 'B'),
Expand All @@ -43,7 +52,28 @@ values (1, 'A'),
(4, 'B'),
(5, 'B');

insert into t4 (id, col)
values (1, 1),
(2, 2),
(3, 3);

-- wait_authoritative t1
-- wait_authoritative t2
-- wait_authoritative t3
select 42 from t1 join t2 on t1.id = t2.t1_id join t3 on t1.id = t3.id where t1.name or t2.id or t3.name;
select 42
from t1
join t2 on t1.id = t2.t1_id
join t3 on t1.id = t3.id
where t1.name
or t2.id
or t3.name;

# Complex query that requires hash join underneath a memory sort and ordered aggregate
select 1
from t1
join t2 on t1.id = t2.t1_id
join t4 on t4.col = t2.id
left join (select t4.col, count(*) as count from t4 group by t4.col) t3 on t3.col = t2.id
where t1.id IN (1, 2)
group by t2.id, t4.col;

8 changes: 8 additions & 0 deletions go/test/endtoend/vtgate/vitess_tester/join/vschema.json
Original file line number Diff line number Diff line change
Expand Up @@ -31,6 +31,14 @@
"name": "hash"
}
]
},
"t4": {
"column_vindexes": [
{
"column": "id",
"name": "hash"
}
]
}
}
}
Expand Down
2 changes: 1 addition & 1 deletion go/vt/vtgate/planbuilder/operators/apply_join.go
Original file line number Diff line number Diff line change
Expand Up @@ -293,7 +293,7 @@ func (aj *ApplyJoin) AddWSColumn(ctx *plancontext.PlanningContext, offset int, u
func (aj *ApplyJoin) planOffsets(ctx *plancontext.PlanningContext) Operator {
if len(aj.Columns) > 0 {
// we've already done offset planning
return aj
return nil
}
for _, col := range aj.JoinColumns.columns {
// Read the type description for applyJoinColumn to understand the following code
Expand Down
2 changes: 1 addition & 1 deletion go/vt/vtgate/planbuilder/operators/dml_with_input.go
Original file line number Diff line number Diff line change
Expand Up @@ -114,7 +114,7 @@ func (d *DMLWithInput) planOffsets(ctx *plancontext.PlanningContext) Operator {
}
}
d.BvList = bvList
return d
return nil
}

var _ Operator = (*DMLWithInput)(nil)
23 changes: 1 addition & 22 deletions go/vt/vtgate/planbuilder/operators/hash_join.go
Original file line number Diff line number Diff line change
Expand Up @@ -326,20 +326,9 @@ func (hj *HashJoin) addColumn(ctx *plancontext.PlanningContext, in sqlparser.Exp
inOffset = op.AddColumn(ctx, false, false, aeWrap(expr))
}

// we turn the
// we have to turn the incoming offset to an outgoing offset of the columns this operator is exposing
internalOffset := offsetter(inOffset)

// ok, we have an offset from the input operator. Let's check if we already have it
// in our list of incoming columns

for idx, offset := range hj.ColumnOffsets {
if internalOffset == offset {
return idx
}
}

hj.ColumnOffsets = append(hj.ColumnOffsets, internalOffset)

return len(hj.ColumnOffsets) - 1
}

Expand Down Expand Up @@ -434,17 +423,7 @@ func (hj *HashJoin) addSingleSidedColumn(

// we have to turn the incoming offset to an outgoing offset of the columns this operator is exposing
internalOffset := offsetter(inOffset)

// ok, we have an offset from the input operator. Let's check if we already have it
// in our list of incoming columns
for idx, offset := range hj.ColumnOffsets {
if internalOffset == offset {
return idx
}
}

hj.ColumnOffsets = append(hj.ColumnOffsets, internalOffset)

return len(hj.ColumnOffsets) - 1
}

Expand Down
9 changes: 7 additions & 2 deletions go/vt/vtgate/planbuilder/operators/offset_planning.go
Original file line number Diff line number Diff line change
Expand Up @@ -38,7 +38,6 @@ func planOffsets(ctx *plancontext.PlanningContext, root Operator) Operator {
panic(vterrors.VT13001(fmt.Sprintf("should not see %T here", in)))
case offsettable:
newOp := op.planOffsets(ctx)

if newOp == nil {
newOp = op
}
Expand All @@ -47,7 +46,13 @@ func planOffsets(ctx *plancontext.PlanningContext, root Operator) Operator {
fmt.Println("Planned offsets for:")
fmt.Println(ToTree(newOp))
}
return newOp, nil

if newOp == op {
return newOp, nil
} else {
// We got a new operator from plan offsets. We should return that something has changed.
return newOp, Rewrote("planning offsets introduced a new operator")
}
}
return in, NoRewrite
}
Expand Down
93 changes: 54 additions & 39 deletions go/vt/vtgate/planbuilder/testdata/aggr_cases.json
Original file line number Diff line number Diff line change
Expand Up @@ -6663,55 +6663,70 @@
"OrderBy": "(4|6) ASC, (5|7) ASC",
"Inputs": [
{
"OperatorType": "Join",
"Variant": "HashLeftJoin",
"Collation": "binary",
"ComparisonType": "INT16",
"JoinColumnIndexes": "-1,1,-2,2,-3,3",
"Predicate": "`user`.col = ue.col",
"TableName": "`user`_user_extra",
"OperatorType": "Projection",
"Expressions": [
"count(*) as count(*)",
"count(*) as count(*)",
"`user`.col as col",
"ue.col as col",
"`user`.foo as foo",
"ue.bar as bar",
"weight_string(`user`.foo) as weight_string(`user`.foo)",
"weight_string(ue.bar) as weight_string(ue.bar)"
],
"Inputs": [
{
"OperatorType": "Route",
"Variant": "Scatter",
"Keyspace": {
"Name": "user",
"Sharded": true
},
"FieldQuery": "select count(*), `user`.col, `user`.foo from `user` where 1 != 1 group by `user`.col, `user`.foo",
"Query": "select count(*), `user`.col, `user`.foo from `user` group by `user`.col, `user`.foo",
"Table": "`user`"
},
{
"OperatorType": "Aggregate",
"Variant": "Ordered",
"Aggregates": "count_star(0)",
"GroupBy": "1, (2|3)",
"ResultColumns": 3,
"OperatorType": "Join",
"Variant": "HashLeftJoin",
"Collation": "binary",
"ComparisonType": "INT16",
"JoinColumnIndexes": "-1,1,-2,2,-3,3,-3,3",
"Predicate": "`user`.col = ue.col",
"TableName": "`user`_user_extra",
"Inputs": [
{
"OperatorType": "SimpleProjection",
"Columns": "2,0,1,3",
"OperatorType": "Route",
"Variant": "Scatter",
"Keyspace": {
"Name": "user",
"Sharded": true
},
"FieldQuery": "select count(*), `user`.col, `user`.foo from `user` where 1 != 1 group by `user`.col, `user`.foo",
"Query": "select count(*), `user`.col, `user`.foo from `user` group by `user`.col, `user`.foo",
"Table": "`user`"
},
{
"OperatorType": "Aggregate",
"Variant": "Ordered",
"Aggregates": "count_star(0)",
"GroupBy": "1, (2|3)",
"ResultColumns": 3,
"Inputs": [
{
"OperatorType": "Sort",
"Variant": "Memory",
"OrderBy": "0 ASC, (1|3) ASC",
"OperatorType": "SimpleProjection",
"Columns": "2,0,1,3",
"Inputs": [
{
"OperatorType": "Limit",
"Count": "10",
"OperatorType": "Sort",
"Variant": "Memory",
"OrderBy": "0 ASC, (1|3) ASC",
"Inputs": [
{
"OperatorType": "Route",
"Variant": "Scatter",
"Keyspace": {
"Name": "user",
"Sharded": true
},
"FieldQuery": "select ue.col, ue.bar, 1, weight_string(ue.bar) from (select col, bar from user_extra where 1 != 1) as ue where 1 != 1",
"Query": "select ue.col, ue.bar, 1, weight_string(ue.bar) from (select col, bar from user_extra) as ue limit 10",
"Table": "user_extra"
"OperatorType": "Limit",
"Count": "10",
"Inputs": [
{
"OperatorType": "Route",
"Variant": "Scatter",
"Keyspace": {
"Name": "user",
"Sharded": true
},
"FieldQuery": "select ue.col, ue.bar, 1, weight_string(ue.bar) from (select col, bar from user_extra where 1 != 1) as ue where 1 != 1",
"Query": "select ue.col, ue.bar, 1, weight_string(ue.bar) from (select col, bar from user_extra) as ue limit 10",
"Table": "user_extra"
}
]
}
]
}
Expand Down
113 changes: 113 additions & 0 deletions go/vt/vtgate/planbuilder/testdata/from_cases.json
Original file line number Diff line number Diff line change
Expand Up @@ -720,6 +720,119 @@
]
}
},
{
"comment": "Complex query that has hash left join underneath a memory sort and ordered aggregation",
"query": "select 1 from user join user_extra on user.id = user_extra.user_id join music on music.intcol = user_extra.col left join (select user_metadata.col, count(*) as count from user_metadata group by user_metadata.col) um on um.col = user_extra.col where user.id IN (103) group by user_extra.col, music.intcol",
"plan": {
"QueryType": "SELECT",
"Original": "select 1 from user join user_extra on user.id = user_extra.user_id join music on music.intcol = user_extra.col left join (select user_metadata.col, count(*) as count from user_metadata group by user_metadata.col) um on um.col = user_extra.col where user.id IN (103) group by user_extra.col, music.intcol",
"Instructions": {
"OperatorType": "Aggregate",
"Variant": "Ordered",
"Aggregates": "any_value(0) AS 1",
"GroupBy": "1, 4",
"ResultColumns": 1,
"Inputs": [
{
"OperatorType": "Sort",
"Variant": "Memory",
"OrderBy": "1 ASC, 4 ASC",
"Inputs": [
{
"OperatorType": "Join",
"Variant": "HashLeftJoin",
"Collation": "binary",
"ComparisonType": "FLOAT64",
"JoinColumnIndexes": "-1,-2,1,-2,-4,-1",
"Predicate": "user_extra.col = um.col",
"TableName": "music_`user`, user_extra_user_metadata",
"Inputs": [
{
"OperatorType": "Join",
"Variant": "Join",
"JoinColumnIndexes": "L:0,R:0,R:0,L:1",
"JoinVars": {
"music_intcol": 1
},
"TableName": "music_`user`, user_extra",
"Inputs": [
{
"OperatorType": "Route",
"Variant": "Scatter",
"Keyspace": {
"Name": "user",
"Sharded": true
},
"FieldQuery": "select 1, music.intcol from music where 1 != 1 group by music.intcol",
"Query": "select 1, music.intcol from music group by music.intcol",
"Table": "music"
},
{
"OperatorType": "Route",
"Variant": "EqualUnique",
"Keyspace": {
"Name": "user",
"Sharded": true
},
"FieldQuery": "select user_extra.col, user_extra.col from `user`, user_extra where 1 != 1 group by user_extra.col",
"Query": "select user_extra.col, user_extra.col from `user`, user_extra where `user`.id in (103) and user_extra.col = :music_intcol /* INT16 */ and `user`.id = user_extra.user_id group by user_extra.col",
"Table": "`user`, user_extra",
"Values": [
"103"
],
"Vindex": "user_index"
}
]
},
{
"OperatorType": "Aggregate",
"Variant": "Ordered",
"GroupBy": "(0|1)",
"ResultColumns": 1,
"Inputs": [
{
"OperatorType": "SimpleProjection",
"Columns": "0,2",
"Inputs": [
{
"OperatorType": "Aggregate",
"Variant": "Ordered",
"Aggregates": "sum_count_star(1) AS count",
"GroupBy": "(0|2)",
"ResultColumns": 3,
"Inputs": [
{
"OperatorType": "Route",
"Variant": "Scatter",
"Keyspace": {
"Name": "user",
"Sharded": true
},
"FieldQuery": "select user_metadata.col, count(*) as `count`, weight_string(user_metadata.col) from user_metadata where 1 != 1 group by user_metadata.col, weight_string(user_metadata.col)",
"OrderBy": "(0|2) ASC",
"Query": "select user_metadata.col, count(*) as `count`, weight_string(user_metadata.col) from user_metadata group by user_metadata.col, weight_string(user_metadata.col) order by user_metadata.col asc",
"Table": "user_metadata"
}
]
}
]
}
]
}
]
}
]
}
]
},
"TablesUsed": [
"user.music",
"user.user",
"user.user_extra",
"user.user_metadata"
]
}
},
{
"comment": "Straight-join (ignores the straight_join hint)",
"query": "select m1.col from unsharded as m1 straight_join unsharded as m2",
Expand Down
6 changes: 6 additions & 0 deletions go/vt/vtgate/planbuilder/testdata/vschemas/schema.json
Original file line number Diff line number Diff line change
Expand Up @@ -282,6 +282,12 @@
"column": "id",
"name": "music_user_map"
}
],
"columns": [
{
"name": "intcol",
"type": "INT16"
}
]
},
"authoritative": {
Expand Down
Loading