You are viewing a plain text version of this content. The canonical link for it is here.
Posted to reviews@spark.apache.org by GitBox <gi...@apache.org> on 2020/02/20 13:36:59 UTC
[GitHub] [spark] iRakson opened a new pull request #27648: SPARK-24994 :
Support filter pushdown for short and byte
iRakson opened a new pull request #27648: SPARK-24994 : Support filter pushdown for short and byte
URL: https://github.com/apache/spark/pull/27648
<!--
Thanks for sending a pull request! Here are some tips for you:
1. If this is your first time, please read our contributor guidelines: https://spark.apache.org/contributing.html
2. Ensure you have added or run the appropriate tests for your PR: https://spark.apache.org/developer-tools.html
3. If the PR is unfinished, add '[WIP]' in your PR title, e.g., '[WIP][SPARK-XXXX] Your PR title ...'.
4. Be sure to keep the PR description updated to reflect all changes.
5. Please write your PR title to summarize what this PR proposes.
6. If possible, provide a concise example to reproduce the issue for a faster review.
7. If you want to add a new configuration, please read the guideline first for naming configurations in
'core/src/main/scala/org/apache/spark/internal/config/ConfigEntry.scala'.
-->
### What changes were proposed in this pull request?
If we have a shortType column in our table, filter will not be pushed down for the following query:
`select * from t where id = 2;`
pushed filter ->
`PushedFilters: [IsNotNull(id)], ReadSchema: struct<id:smallint>`
2 is within short range still filter is not pushed down.
But it pushes filter if we cast the literal to smallint:
`select * from t where id = cast(2 as smallint) ; `
pushed filter ->
` PushedFilters: [IsNotNull(id), EqualTo(id,2)], ReadSchema: struct<id:smallint>`
After these changes, there is no need to explicitly convert literals to short/byte for filter pushdown:
`Select * from t where id = 2;`
pushed filter ->
`PushedFilters: [IsNotNull(id), EqualTo(id,2)], ReadSchema: struct<id:smallint>`
<!--
Please clarify what changes you are proposing. The purpose of this section is to outline the changes and how this PR fixes the issue.
If possible, please consider writing useful notes for better and faster reviews in your PR. See the examples below.
1. If you refactor some codes with changing classes, showing the class hierarchy will help reviewers.
2. If you fix some SQL features, you can provide some references of other DBMSes.
3. If there is design documentation, please add the link.
4. If there is a discussion in the mailing list, please add the link.
-->
### Why are the changes needed?
These changes are needed because whenever we want to pushdown filters for short/byte value we have to explicitly cast them to short/byte even if the value is well within short/byte range.
<!--
Please clarify why the changes are needed. For instance,
1. If you propose a new API, clarify the use case for a new API.
2. If you fix a bug, you can clarify why it is a bug.
-->
### Does this PR introduce any user-facing change?
Yes, now users dont need to explicitly cast literals to short/byte for filter pushdown.
<!--
If yes, please clarify the previous behavior and the change this PR proposes - provide the console output, description and/or an example to show the behavior difference if possible.
If no, write 'No'.
-->
### How was this patch tested?
Manually. UT will be added soon.
<!--
If tests were added, say they were added here. Please make sure to add some test cases that check the changes thoroughly including negative and positive cases if possible.
If it was tested in a way different from regular unit tests, please clarify how you tested step by step, ideally copy and paste-able, so that other reviewers can test and check, and descendants can verify in the future.
If tests were not added, please describe why they were not added and/or why it was difficult to add.
-->
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
[GitHub] [spark] HyukjinKwon commented on a change in pull request #27648:
[SPARK-24994][SQL] : Support filter pushdown for short and byte without
explicit casting
Posted by GitBox <gi...@apache.org>.
HyukjinKwon commented on a change in pull request #27648: [SPARK-24994][SQL] : Support filter pushdown for short and byte without explicit casting
URL: https://github.com/apache/spark/pull/27648#discussion_r382488098
##########
File path: sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/DataSourceStrategy.scala
##########
@@ -443,31 +443,63 @@ object DataSourceStrategy {
case expressions.EqualTo(Literal(v, t), a: Attribute) =>
Some(sources.EqualTo(a.name, convertToScala(v, t)))
+ case expressions.EqualTo(Cast(c, dt, tz), Literal(v, t)) if c.isInstanceOf[Attribute] =>
+ Some(sources.EqualTo(c.asInstanceOf[Attribute].name, convertToScala(v, t)))
+ case expressions.EqualTo(Literal(v, t), Cast(c, dt, tz)) if c.isInstanceOf[Attribute] =>
+ Some(sources.EqualTo(c.asInstanceOf[Attribute].name, convertToScala(v, t)))
Review comment:
cc @liancheng FYI
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
[GitHub] [spark] iRakson commented on a change in pull request #27648:
[SPARK-24994][SQL] : Support filter pushdown for short and byte without
explicit casting
Posted by GitBox <gi...@apache.org>.
iRakson commented on a change in pull request #27648: [SPARK-24994][SQL] : Support filter pushdown for short and byte without explicit casting
URL: https://github.com/apache/spark/pull/27648#discussion_r385685847
##########
File path: sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/Optimizer.scala
##########
@@ -1802,3 +1803,85 @@ object OptimizeLimitZero extends Rule[LogicalPlan] {
empty(ll)
}
}
+
+/**
+ * Removes Cast expressions from predicates to push down the filters. This will allow filter push
+ * down for byte and short columns without explicitly casting literals.
Review comment:
In this solution i am not generalising casting. This PR only deals with Byte and Short columns and Values are only downcasted in case they are within byte/short range. Otherwise whole Predicate is left unchanged.
Also before pushing i am explicitly checking whether these values can be pushed down or not.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
[GitHub] [spark] iRakson commented on issue #27648: SPARK-24994 : Support
filter pushdown for short and byte
Posted by GitBox <gi...@apache.org>.
iRakson commented on issue #27648: SPARK-24994 : Support filter pushdown for short and byte
URL: https://github.com/apache/spark/pull/27648#issuecomment-589026464
cc @cloud-fan @HyukjinKwon
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
[GitHub] [spark] cloud-fan commented on issue #27648: [SPARK-24994][SQL] :
Support filter pushdown for short and byte without explicit casting
Posted by GitBox <gi...@apache.org>.
cloud-fan commented on issue #27648: [SPARK-24994][SQL] : Support filter pushdown for short and byte without explicit casting
URL: https://github.com/apache/spark/pull/27648#issuecomment-589567927
So the goal is to push down `cast(short_col as int) = int_literal`? I agree this can be a common case, but I think we should optimize this filter before data source filter push down.
If the `int_literal` is beyond short range, then this predicate is either null (if the column is null) or false. So we can rewrite it to `If(IsNull(short_col), null, false)`.
If the `int_literal` is within short range, we can rewrite it to `short_col = short_literal` by casting the int literal to short first.
When we push down filters to data source, the predicates already get rid of the cast at the column side.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
[GitHub] [spark] HyukjinKwon commented on a change in pull request #27648:
[SPARK-24994][SQL] : Support filter pushdown for short and byte without
explicit casting
Posted by GitBox <gi...@apache.org>.
HyukjinKwon commented on a change in pull request #27648: [SPARK-24994][SQL] : Support filter pushdown for short and byte without explicit casting
URL: https://github.com/apache/spark/pull/27648#discussion_r386773397
##########
File path: sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/Optimizer.scala
##########
@@ -1802,3 +1803,85 @@ object OptimizeLimitZero extends Rule[LogicalPlan] {
empty(ll)
}
}
+
+/**
+ * Removes Cast expressions from predicates to push down the filters. This will allow filter push
+ * down for byte and short columns without explicitly casting literals.
Review comment:
The approach basically looks same. Also, we should better generalise it so other data sources can support. The current approach seems difficult to extend.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
[GitHub] [spark] maropu commented on issue #27648: [SPARK-24994][SQL] :
Support filter pushdown for short and byte without explicit casting
Posted by GitBox <gi...@apache.org>.
maropu commented on issue #27648: [SPARK-24994][SQL] : Support filter pushdown for short and byte without explicit casting
URL: https://github.com/apache/spark/pull/27648#issuecomment-589512300
> 2 is within short range still filter is not pushed down.
> But it pushes filter if we cast the literal to smallint:
> select * from t where id = cast(2 as smallint) ;
Is that pushdown legal? if `id` is short, `id = 2` is parsed as `cast(id as int) = 2`. So, I feel that the semantics are different between them.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
[GitHub] [spark] HyukjinKwon commented on a change in pull request #27648:
[SPARK-24994][SQL] : Support filter pushdown for short and byte without
explicit casting
Posted by GitBox <gi...@apache.org>.
HyukjinKwon commented on a change in pull request #27648: [SPARK-24994][SQL] : Support filter pushdown for short and byte without explicit casting
URL: https://github.com/apache/spark/pull/27648#discussion_r385148275
##########
File path: sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/Optimizer.scala
##########
@@ -1802,3 +1803,85 @@ object OptimizeLimitZero extends Rule[LogicalPlan] {
empty(ll)
}
}
+
+/**
+ * Removes Cast expressions from predicates to push down the filters. This will allow filter push
+ * down for byte and short columns without explicitly casting literals.
Review comment:
So, what you're doing right now is basically https://github.com/apache/spark/pull/8718 which was rejected.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
[GitHub] [spark] AmplabJenkins commented on issue #27648: [SPARK-24994][SQL]
: Support filter pushdown for short and byte without explicit casting
Posted by GitBox <gi...@apache.org>.
AmplabJenkins commented on issue #27648: [SPARK-24994][SQL] : Support filter pushdown for short and byte without explicit casting
URL: https://github.com/apache/spark/pull/27648#issuecomment-589027829
Can one of the admins verify this patch?
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
[GitHub] [spark] HyukjinKwon commented on a change in pull request #27648:
[SPARK-24994][SQL] : Support filter pushdown for short and byte without
explicit casting
Posted by GitBox <gi...@apache.org>.
HyukjinKwon commented on a change in pull request #27648: [SPARK-24994][SQL] : Support filter pushdown for short and byte without explicit casting
URL: https://github.com/apache/spark/pull/27648#discussion_r382487108
##########
File path: sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/DataSourceStrategy.scala
##########
@@ -443,31 +443,63 @@ object DataSourceStrategy {
case expressions.EqualTo(Literal(v, t), a: Attribute) =>
Some(sources.EqualTo(a.name, convertToScala(v, t)))
+ case expressions.EqualTo(Cast(c, dt, tz), Literal(v, t)) if c.isInstanceOf[Attribute] =>
+ Some(sources.EqualTo(c.asInstanceOf[Attribute].name, convertToScala(v, t)))
+ case expressions.EqualTo(Literal(v, t), Cast(c, dt, tz)) if c.isInstanceOf[Attribute] =>
+ Some(sources.EqualTo(c.asInstanceOf[Attribute].name, convertToScala(v, t)))
Review comment:
I don't think this is correct. It can potentially end up with a filter with having a different type of the value in the pushed filter.
If you want to handle the cast, it will end up with the same fix as https://github.com/apache/spark/pull/8718 which was virtually rejected.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
[GitHub] [spark] AmplabJenkins removed a comment on issue #27648:
[SPARK-24994][SQL] : Support filter pushdown for short and byte without
explicit casting
Posted by GitBox <gi...@apache.org>.
AmplabJenkins removed a comment on issue #27648: [SPARK-24994][SQL] : Support filter pushdown for short and byte without explicit casting
URL: https://github.com/apache/spark/pull/27648#issuecomment-589027195
Can one of the admins verify this patch?
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
[GitHub] [spark] AmplabJenkins commented on issue #27648: [SPARK-24994][SQL]
: Support filter pushdown for short and byte without explicit casting
Posted by GitBox <gi...@apache.org>.
AmplabJenkins commented on issue #27648: [SPARK-24994][SQL] : Support filter pushdown for short and byte without explicit casting
URL: https://github.com/apache/spark/pull/27648#issuecomment-589027195
Can one of the admins verify this patch?
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
[GitHub] [spark] iRakson commented on issue #27648: [SPARK-24994][SQL] :
Support filter pushdown for short and byte without explicit casting
Posted by GitBox <gi...@apache.org>.
iRakson commented on issue #27648: [SPARK-24994][SQL] : Support filter pushdown for short and byte without explicit casting
URL: https://github.com/apache/spark/pull/27648#issuecomment-591984984
This is another approach as suggested by @cloud-fan.
Rather than handling `cast(id as int)` in `translateFilter` , i have optimized the filters to remove cast expressions during optimization phase.
As of now i am removing cast for `byte/short` columns only. Whenever `int literal` is within `byte/short` range, cast expression is removed and `int literal` is casted to `byte/short`.
@cloud-fan @HyukjinKwon @maropu
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
[GitHub] [spark] iRakson commented on issue #27648: [SPARK-24994][SQL] :
Support filter pushdown for short and byte without explicit casting
Posted by GitBox <gi...@apache.org>.
iRakson commented on issue #27648: [SPARK-24994][SQL] : Support filter pushdown for short and byte without explicit casting
URL: https://github.com/apache/spark/pull/27648#issuecomment-589515318
> Is that pushdown legal? if `id` is short, `id = 2` is parsed as `cast(id as int) = 2`. So, I feel that the semantics are different between them.
Naive users might not cast `literal` to `shortType` when passing a value within short range.
Also, if there are large number of queries which deals with short and byte fields, it is cumbersome to always explicitly cast them.
In those cases users can't get push-down benefits.
In my opinion we should push the filter if `literals` are in `short`/`byte` range.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org