[fix](mtmv) Fix high nest level materialized view can not be rewritten, because low level mv aggregate roll up #36567
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
Thank you for your contribution to Apache Doris. Since 2024-03-18, the Document has been moved to doris-website. |
Contributor
Author
|
run buildall |
seawinde
changed the title
Contributor
|
run performance |
TPC-H: Total hot run time: 39837 ms |
TPC-DS: Total hot run time: 171735 ms |
…cause low level mv aggregate roll up
seawinde
force-pushed
the
nested_mv_not_hit_fix
branch
from
566143d to
fe79381
Compare
ClickBench: Total hot run time: 30.42 s |
Contributor
Author
|
run buildall |
TPC-H: Total hot run time: 40247 ms |
TPC-DS: Total hot run time: 172986 ms |
ClickBench: Total hot run time: 30.9 s |
morrySnow
reviewed
Jun 24, 2024
| Plan rewrittenPlan = MaterializedViewUtils.rewriteByRules(cascadesContext, | ||
| childContext -> { | ||
| Rewriter.getCteChildrenRewriter(childContext, | ||
| ImmutableList.of(Rewriter.topDown(new EliminateGroupByKey()))).execute(); |
Contributor
There was a problem hiding this comment.
why need apply this rule seperately? rewrite the whole mv plan tree could not eliminate group by key?
Contributor
Author
There was a problem hiding this comment.
because there is five group by dimension in mv, but query use four of them.
So need add projects that query used on mv plan, and run eliminate group by key again.
morrySnow
approved these changes
Jun 25, 2024
github-actions
bot
added
the
approved
Contributor
|
PR approved by at least one committer and no changes requested. |
Contributor
|
PR approved by anyone and no changes requested. |
keanji-x
approved these changes
Jun 25, 2024
dataroaring
pushed a commit
that referenced
this pull request
Jun 26, 2024
…n, because low level mv aggregate roll up (#36567) Query is aggregate, the query group by expression is less than materialzied view group by expression. when the more dimensions than queries in materialzied view can be eliminated with functional dependencies. it can be rewritten with out roll up aggregate. For example as following: mv def is CREATE MATERIALIZED VIEW mv BUILD IMMEDIATE REFRESH AUTO ON MANUAL DISTRIBUTED BY RANDOM BUCKETS 2 PROPERTIES ('replication_num' = '1') AS select l_orderkey, l_partkey, l_suppkey, o_orderkey, o_custkey, ps_partkey, cast( sum( IFNULL(ps_suppkey, 0) * IFNULL(ps_partkey, 0) ) as decimal(28, 8) ) as agg2 from lineitem_1 inner join orders_1 on lineitem_1.l_orderkey = orders_1.o_orderkey inner join partsupp_1 on l_partkey = partsupp_1.ps_partkey and l_suppkey = partsupp_1.ps_suppkey where partsupp_1.ps_suppkey > 1 group by l_orderkey, l_partkey, l_suppkey, o_orderkey, o_custkey, ps_partkey; query is as following: select l_orderkey, l_partkey, l_suppkey, o_orderkey, o_custkey, cast( sum( IFNULL(ps_suppkey, 0) * IFNULL(ps_partkey, 0) ) as decimal(28, 8) ) as agg2 from lineitem_1 inner join orders_1 on lineitem_1.l_orderkey = orders_1.o_orderkey inner join partsupp_1 on l_partkey = partsupp_1.ps_partkey and l_suppkey = partsupp_1.ps_suppkey where partsupp_1.ps_suppkey > 1 group by l_orderkey, l_partkey, l_suppkey, o_orderkey, o_custkey; we can see that query doesn't use `ps_partkey` which is in mv group by expression. Normally will add roll up aggragate on materialized view if the gorup by dimension in mv is mucher than query group by dimension. And, in this scane we can get the function dependency on `l_suppkey = ps_suppkey `. and we doesn't need to add roll up aggregate on materialized view in rewritten plan. this improve performance and is beneficial for nest materialized view rewrite.
seawinde
added a commit
to seawinde/doris
that referenced
this pull request
Jul 11, 2024
…n, because low level mv aggregate roll up (apache#36567) Query is aggregate, the query group by expression is less than materialzied view group by expression. when the more dimensions than queries in materialzied view can be eliminated with functional dependencies. it can be rewritten with out roll up aggregate. For example as following: mv def is CREATE MATERIALIZED VIEW mv BUILD IMMEDIATE REFRESH AUTO ON MANUAL DISTRIBUTED BY RANDOM BUCKETS 2 PROPERTIES ('replication_num' = '1') AS select l_orderkey, l_partkey, l_suppkey, o_orderkey, o_custkey, ps_partkey, cast( sum( IFNULL(ps_suppkey, 0) * IFNULL(ps_partkey, 0) ) as decimal(28, 8) ) as agg2 from lineitem_1 inner join orders_1 on lineitem_1.l_orderkey = orders_1.o_orderkey inner join partsupp_1 on l_partkey = partsupp_1.ps_partkey and l_suppkey = partsupp_1.ps_suppkey where partsupp_1.ps_suppkey > 1 group by l_orderkey, l_partkey, l_suppkey, o_orderkey, o_custkey, ps_partkey; query is as following: select l_orderkey, l_partkey, l_suppkey, o_orderkey, o_custkey, cast( sum( IFNULL(ps_suppkey, 0) * IFNULL(ps_partkey, 0) ) as decimal(28, 8) ) as agg2 from lineitem_1 inner join orders_1 on lineitem_1.l_orderkey = orders_1.o_orderkey inner join partsupp_1 on l_partkey = partsupp_1.ps_partkey and l_suppkey = partsupp_1.ps_suppkey where partsupp_1.ps_suppkey > 1 group by l_orderkey, l_partkey, l_suppkey, o_orderkey, o_custkey; we can see that query doesn't use `ps_partkey` which is in mv group by expression. Normally will add roll up aggragate on materialized view if the gorup by dimension in mv is mucher than query group by dimension. And, in this scane we can get the function dependency on `l_suppkey = ps_suppkey `. and we doesn't need to add roll up aggregate on materialized view in rewritten plan. this improve performance and is beneficial for nest materialized view rewrite.
morrySnow
pushed a commit
that referenced
this pull request
Jul 12, 2024
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
5 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Proposed changes
Query is aggregate, the query group by expression is less than materialzied view group by expression.
when the more dimensions than queries in materialzied view can be eliminated with functional dependencies.
it can be rewritten with out roll up aggregate.
For example as following:
mv def is
query is as following:
we can see that query doesn't use
ps_partkeywhich is in mv group by expression.Normally will add roll up aggragate on materialized view if the gorup by dimension in mv is mucher than query group by dimension.
And, in this scane we can get the function dependency on
l_suppkey = ps_suppkey. and we doesn't need to add roll up aggregate on materialized view in rewritten plan. this improve performance and is beneficial for nest materialized view rewrite.