Enable push-down of constant join conditions in outer-joins #15760
Conversation
This file contains 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
mkleen
force-pushed
the
mkleen/constant-outer-joins
branch
5 times, most recently
from
3977a0d
to
4e86d49
Compare
mkleen
force-pushed
the
mkleen/constant-outer-joins
branch
from
4e86d49
to
544df0c
Compare
mkleen
force-pushed
the
mkleen/constant-outer-joins
branch
from
544df0c
to
d54d92c
Compare
mkleen
commented
Apr 8, 2024
| " │ └ Collect[doc.t2 | [cluster_id] | (kind = 'bar')]", | ||
| " └ Rename[cluster_id, kind] AS temp", | ||
| " └ Collect[doc.t2 | [cluster_id, kind] | true]" | ||
| "Eval[id, reference]", |
There was a problem hiding this comment.
This is a good example of why this change makes sense. The constant join condition from the left-join gets extracted and pushed down and the outer-join can be converted to an inner-join which leads to a more efficient query plan using solely hash-joins.
+------------------------------------------------------+----------------------------------------------------------------------------------------+
| STEP | QUERY PLAN |
+------------------------------------------------------+----------------------------------------------------------------------------------------+
| Initial logical plan | Eval[id, reference] (rows=unknown) |
| | └ Join[LEFT | ((cluster_id = id) AND (kind = 'bar'))] (rows=unknown) |
| | ├ Join[INNER | ((cluster_id = id) AND (kind = 'bar'))] (rows=unknown) |
| | │ ├ Filter[(reference = 'bazinga')] (rows=0) |
| | │ │ └ Join[INNER | (subscription_id = id)] (rows=unknown) |
| | │ │ ├ Collect[doc.t3 | [id, reference] | true] (rows=unknown) |
| | │ │ └ Collect[doc.t1 | [subscription_id, id] | true] (rows=unknown) |
| | │ └ Collect[doc.t2 | [cluster_id, kind] | true] (rows=unknown) |
| | └ Rename[cluster_id, kind] AS temp (rows=unknown) |
| | └ Collect[doc.t2 | [cluster_id, kind] | true] (rows=unknown) |
| optimizer_extract_constant_join_conditions_to_filter | Eval[id, reference] (rows=0) |
| | └ Filter[(kind = 'bar')] (rows=0) |
| | └ Join[LEFT | (cluster_id = id)] (rows=unknown) |
| | ├ Join[INNER | ((cluster_id = id) AND (kind = 'bar'))] (rows=unknown) |
| | │ ├ Filter[(reference = 'bazinga')] (rows=0) |
| | │ │ └ Join[INNER | (subscription_id = id)] (rows=unknown) |
| | │ │ ├ Collect[doc.t3 | [id, reference] | true] (rows=unknown) |
| | │ │ └ Collect[doc.t1 | [subscription_id, id] | true] (rows=unknown) |
| | │ └ Collect[doc.t2 | [cluster_id, kind] | true] (rows=unknown) |
| | └ Rename[cluster_id, kind] AS temp (rows=unknown) |
| | └ Collect[doc.t2 | [cluster_id, kind] | true] (rows=unknown) |
| optimizer_rewrite_filter_on_outer_join_to_inner_join | Eval[id, reference] (rows=unknown) |
| | └ Join[INNER | (cluster_id = id)] (rows=unknown) |
| | ├ Join[INNER | ((cluster_id = id) AND (kind = 'bar'))] (rows=unknown) |
| | │ ├ Filter[(reference = 'bazinga')] (rows=0) |
| | │ │ └ Join[INNER | (subscription_id = id)] (rows=unknown) |
| | │ │ ├ Collect[doc.t3 | [id, reference] | true] (rows=unknown) |
| | │ │ └ Collect[doc.t1 | [subscription_id, id] | true] (rows=unknown) |
| | │ └ Collect[doc.t2 | [cluster_id, kind] | true] (rows=unknown) |
| | └ Filter[(kind = 'bar')] (rows=0) |
| | └ Rename[cluster_id, kind] AS temp (rows=unknown) |
| | └ Collect[doc.t2 | [cluster_id, kind] | true] (rows=unknown) |
| optimizer_rewrite_join_plan | Eval[id, reference] (rows=unknown) |
| | └ HashJoin[(cluster_id = id)] (rows=unknown) |
| | ├ Join[INNER | ((cluster_id = id) AND (kind = 'bar'))] (rows=unknown) |
| | │ ├ Filter[(reference = 'bazinga')] (rows=0) |
| | │ │ └ Join[INNER | (subscription_id = id)] (rows=unknown) |
| | │ │ ├ Collect[doc.t3 | [id, reference] | true] (rows=unknown) |
| | │ │ └ Collect[doc.t1 | [subscription_id, id] | true] (rows=unknown) |
| | │ └ Collect[doc.t2 | [cluster_id, kind] | true] (rows=unknown) |
| | └ Filter[(kind = 'bar')] (rows=0) |
| | └ Rename[cluster_id, kind] AS temp (rows=unknown) |
| | └ Collect[doc.t2 | [cluster_id, kind] | true] (rows=unknown) |
| optimizer_extract_constant_join_conditions_to_filter | Eval[id, reference] (rows=unknown) |
| | └ HashJoin[(cluster_id = id)] (rows=unknown) |
| | ├ Filter[(kind = 'bar')] (rows=0) |
| | │ └ Join[INNER | (cluster_id = id)] (rows=unknown) |
| | │ ├ Filter[(reference = 'bazinga')] (rows=0) |
| | │ │ └ Join[INNER | (subscription_id = id)] (rows=unknown) |
| | │ │ ├ Collect[doc.t3 | [id, reference] | true] (rows=unknown) |
| | │ │ └ Collect[doc.t1 | [subscription_id, id] | true] (rows=unknown) |
| | │ └ Collect[doc.t2 | [cluster_id, kind] | true] (rows=unknown) |
| | └ Filter[(kind = 'bar')] (rows=0) |
| | └ Rename[cluster_id, kind] AS temp (rows=unknown) |
| | └ Collect[doc.t2 | [cluster_id, kind] | true] (rows=unknown) |
| optimizer_move_filter_beneath_join | Eval[id, reference] (rows=unknown) |
| | └ HashJoin[(cluster_id = id)] (rows=unknown) |
| | ├ Join[INNER | (cluster_id = id)] (rows=unknown) |
| | │ ├ Filter[(reference = 'bazinga')] (rows=0) |
| | │ │ └ Join[INNER | (subscription_id = id)] (rows=unknown) |
| | │ │ ├ Collect[doc.t3 | [id, reference] | true] (rows=unknown) |
| | │ │ └ Collect[doc.t1 | [subscription_id, id] | true] (rows=unknown) |
| | │ └ Filter[(kind = 'bar')] (rows=0) |
| | │ └ Collect[doc.t2 | [cluster_id, kind] | true] (rows=unknown) |
| | └ Filter[(kind = 'bar')] (rows=0) |
| | └ Rename[cluster_id, kind] AS temp (rows=unknown) |
| | └ Collect[doc.t2 | [cluster_id, kind] | true] (rows=unknown) |
| optimizer_rewrite_join_plan | Eval[id, reference] (rows=unknown) |
| | └ HashJoin[(cluster_id = id)] (rows=unknown) |
| | ├ HashJoin[(cluster_id = id)] (rows=unknown) |
| | │ ├ Filter[(reference = 'bazinga')] (rows=0) |
| | │ │ └ Join[INNER | (subscription_id = id)] (rows=unknown) |
| | │ │ ├ Collect[doc.t3 | [id, reference] | true] (rows=unknown) |
| | │ │ └ Collect[doc.t1 | [subscription_id, id] | true] (rows=unknown) |
| | │ └ Filter[(kind = 'bar')] (rows=0) |
| | │ └ Collect[doc.t2 | [cluster_id, kind] | true] (rows=unknown) |
| | └ Filter[(kind = 'bar')] (rows=0) |
| | └ Rename[cluster_id, kind] AS temp (rows=unknown) |
| | └ Collect[doc.t2 | [cluster_id, kind] | true] (rows=unknown) |
| optimizer_move_filter_beneath_join | Eval[id, reference] (rows=unknown) |
| | └ HashJoin[(cluster_id = id)] (rows=unknown) |
| | ├ HashJoin[(cluster_id = id)] (rows=unknown) |
| | │ ├ Join[INNER | (subscription_id = id)] (rows=unknown) |
| | │ │ ├ Filter[(reference = 'bazinga')] (rows=0) |
| | │ │ │ └ Collect[doc.t3 | [id, reference] | true] (rows=unknown) |
| | │ │ └ Collect[doc.t1 | [subscription_id, id] | true] (rows=unknown) |
| | │ └ Filter[(kind = 'bar')] (rows=0) |
| | │ └ Collect[doc.t2 | [cluster_id, kind] | true] (rows=unknown) |
| | └ Filter[(kind = 'bar')] (rows=0) |
| | └ Rename[cluster_id, kind] AS temp (rows=unknown) |
| | └ Collect[doc.t2 | [cluster_id, kind] | true] (rows=unknown) |
| optimizer_rewrite_join_plan | Eval[id, reference] (rows=unknown) |
| | └ HashJoin[(cluster_id = id)] (rows=unknown) |
| | ├ HashJoin[(cluster_id = id)] (rows=unknown) |
| | │ ├ HashJoin[(subscription_id = id)] (rows=unknown) |
| | │ │ ├ Filter[(reference = 'bazinga')] (rows=0) |
| | │ │ │ └ Collect[doc.t3 | [id, reference] | true] (rows=unknown) |
| | │ │ └ Collect[doc.t1 | [subscription_id, id] | true] (rows=unknown) |
| | │ └ Filter[(kind = 'bar')] (rows=0) |
| | │ └ Collect[doc.t2 | [cluster_id, kind] | true] (rows=unknown) |
| | └ Filter[(kind = 'bar')] (rows=0) |
| | └ Rename[cluster_id, kind] AS temp (rows=unknown) |
| | └ Collect[doc.t2 | [cluster_id, kind] | true] (rows=unknown) |
| optimizer_merge_filter_and_collect | Eval[id, reference] (rows=unknown) |
| | └ HashJoin[(cluster_id = id)] (rows=unknown) |
| | ├ HashJoin[(cluster_id = id)] (rows=unknown) |
| | │ ├ HashJoin[(subscription_id = id)] (rows=unknown) |
| | │ │ ├ Collect[doc.t3 | [id, reference] | (reference = 'bazinga')] (rows=unknown) |
| | │ │ └ Collect[doc.t1 | [subscription_id, id] | true] (rows=unknown) |
| | │ └ Filter[(kind = 'bar')] (rows=0) |
| | │ └ Collect[doc.t2 | [cluster_id, kind] | true] (rows=unknown) |
| | └ Filter[(kind = 'bar')] (rows=0) |
| | └ Rename[cluster_id, kind] AS temp (rows=unknown) |
| | └ Collect[doc.t2 | [cluster_id, kind] | true] (rows=unknown) |
| optimizer_merge_filter_and_collect | Eval[id, reference] (rows=unknown) |
| | └ HashJoin[(cluster_id = id)] (rows=unknown) |
| | ├ HashJoin[(cluster_id = id)] (rows=unknown) |
| | │ ├ HashJoin[(subscription_id = id)] (rows=unknown) |
| | │ │ ├ Collect[doc.t3 | [id, reference] | (reference = 'bazinga')] (rows=unknown) |
| | │ │ └ Collect[doc.t1 | [subscription_id, id] | true] (rows=unknown) |
| | │ └ Collect[doc.t2 | [cluster_id, kind] | (kind = 'bar')] (rows=unknown) |
| | └ Filter[(kind = 'bar')] (rows=0) |
| | └ Rename[cluster_id, kind] AS temp (rows=unknown) |
| | └ Collect[doc.t2 | [cluster_id, kind] | true] (rows=unknown) |
| optimizer_move_filter_beneath_rename | Eval[id, reference] (rows=unknown) |
| | └ HashJoin[(cluster_id = id)] (rows=unknown) |
| | ├ HashJoin[(cluster_id = id)] (rows=unknown) |
| | │ ├ HashJoin[(subscription_id = id)] (rows=unknown) |
| | │ │ ├ Collect[doc.t3 | [id, reference] | (reference = 'bazinga')] (rows=unknown) |
| | │ │ └ Collect[doc.t1 | [subscription_id, id] | true] (rows=unknown) |
| | │ └ Collect[doc.t2 | [cluster_id, kind] | (kind = 'bar')] (rows=unknown) |
| | └ Rename[cluster_id, kind] AS temp (rows=0) |
| | └ Filter[(kind = 'bar')] (rows=0) |
| | └ Collect[doc.t2 | [cluster_id, kind] | true] (rows=unknown) |
| optimizer_merge_filter_and_collect | Eval[id, reference] (rows=unknown) |
| | └ HashJoin[(cluster_id = id)] (rows=unknown) |
| | ├ HashJoin[(cluster_id = id)] (rows=unknown) |
| | │ ├ HashJoin[(subscription_id = id)] (rows=unknown) |
| | │ │ ├ Collect[doc.t3 | [id, reference] | (reference = 'bazinga')] (rows=unknown) |
| | │ │ └ Collect[doc.t1 | [subscription_id, id] | true] (rows=unknown) |
| | │ └ Collect[doc.t2 | [cluster_id, kind] | (kind = 'bar')] (rows=unknown) |
| | └ Rename[cluster_id, kind] AS temp (rows=unknown) |
| | └ Collect[doc.t2 | [cluster_id, kind] | (kind = 'bar')] (rows=unknown) |
| Final logical plan | Eval[id, reference] (rows=unknown) |
| | └ HashJoin[(cluster_id = id)] (rows=unknown) |
| | ├ HashJoin[(cluster_id = id)] (rows=unknown) |
| | │ ├ HashJoin[(subscription_id = id)] (rows=unknown) |
| | │ │ ├ Collect[doc.t3 | [id, reference] | (reference = 'bazinga')] (rows=unknown) |
| | │ │ └ Collect[doc.t1 | [subscription_id, id] | true] (rows=unknown) |
| | │ └ Collect[doc.t2 | [cluster_id] | (kind = 'bar')] (rows=unknown) |
| | └ Rename[cluster_id] AS temp (rows=unknown) |
| | └ Collect[doc.t2 | [cluster_id] | (kind = 'bar')] (rows=unknown) |
+------------------------------------------------------+----------------------------------------------------------------------------------------+
EXPLAIN 14 rows in set (0.026 sec)
mfussenegger
approved these changes
Apr 8, 2024
server/src/main/java/io/crate/planner/optimizer/rule/ExtractConstantJoinConditionsToFilter.java
Outdated
Show resolved
Hide resolved
|
@mfussenegger I made a second iteration to handle outer joins properly. Outer-joins only apply constant join conditions on their non-preserved side and get ignored on the other side:
Therefore we can only extract the filter in these cases on top of the join. The pushdown took that already into account, but we have to be aware of that also on the filter creation. |
|
The test is still flaky, i will put this on hold. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Summary of the changes / Why this improves CrateDB
This is based on @jeeminso previous work on joins.
At the current state we have the optimizer rule
optimizer_move_constant_join_conditions_beneath_joinwhich pushes down constant join conditions beyond an inner-join:becomes:
We also have already
optimizer_move_filter_beneath_joinwhich pushes filters beyond inner/outer/cross/nested joins:becomes:
This pr changes
optimizer_move_constant_join_conditions_beneath_joinso the constant join condition is extracted to a filter on top of the join which can then be pushed byoptimizer_move_filter_beneath_join:becomes:
to:
The benefit of this change is:
Checklist
docs/appendices/release-notes/<x.y.0>.rstfor user facing changessql_featurestable for user facing changesdocs/appendices/release-notes/<x.y.0>.rst(E.g. AdminUI)