-
Notifications
You must be signed in to change notification settings - Fork 28.5k
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
[SPARK-23973][SQL] Remove consecutive Sorts #21072
Changes from 1 commit
6ba4186
ff7d412
ac03bed
1d6ca1e
e7391f3
e2f4d4d
File filter
Filter by extension
Conversations
Jump to
Diff view
Diff view
There are no files selected for viewing
Original file line number | Diff line number | Diff line change |
---|---|---|
|
@@ -736,12 +736,15 @@ object EliminateSorts extends Rule[LogicalPlan] { | |
} | ||
|
||
/** | ||
* Removes Sort operation if the child is already sorted | ||
* Removes redundant Sort operation. This can happen: | ||
* 1) if the child is already sorted | ||
* 2) if the next operator is a Sort itself | ||
*/ | ||
object RemoveRedundantSorts extends Rule[LogicalPlan] { | ||
def apply(plan: LogicalPlan): LogicalPlan = plan transform { | ||
case Sort(orders, true, child) if SortOrder.orderingSatisfies(child.outputOrdering, orders) => | ||
child | ||
case s @ Sort(_, _, Sort(_, _, child)) => s.copy(child = child) | ||
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. Thanks for doing this! It might be useful to generalise this to any pair of sorts separated by 0 or more projections or filters. I did this for my SPARK-23975 PR, see: henryr@bb992c2#diff-a636a87d8843eeccca90140be91d4fafR322 What do you think? There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. yes, it makes sense. I will do, thanks. |
||
} | ||
} | ||
|
||
|
Original file line number | Diff line number | Diff line change |
---|---|---|
|
@@ -98,4 +98,11 @@ class RemoveRedundantSortsSuite extends PlanTest { | |
val correctAnswer = groupedAndResorted.analyze | ||
comparePlans(optimized, correctAnswer) | ||
} | ||
|
||
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. Could you add a test which explicitly confirms that sort.limit.sort is not simplified? I know the above two tests cover that case, but it's good to have one dedicated to testing this important property. |
||
test("remove two consecutive sorts") { | ||
val orderedTwice = testRelation.orderBy('a.asc).orderBy('b.desc) | ||
val optimized = Optimize.execute(orderedTwice.analyze) | ||
val correctAnswer = testRelation.orderBy('b.desc).analyze | ||
comparePlans(optimized, correctAnswer) | ||
} | ||
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. Can you add a test for three consecutive sorts? Two is the base case, three will help us show the inductive case :) |
||
} |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
nit: now it's more efficient to do
transformDown
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
isn't it the same?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
assume the plan is
If we do
transformUp
, we hit the rule 3 times, which has some unnecessary transformation(OtherPlan
is transformed 3 times). If it'stransformDown
, it's one-shot.There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
yes, but I saw that
transfrom
actually doestransformDown
. Anyway, I see that this might change and here we best havetransformDown