c21 opened a new pull request #29342:
URL: https://github.com/apache/spark/pull/29342
<!--
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?
<!--
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.
-->
Add support for full outer join inside shuffled hash join. Currently if the
query is a full outer join, we only use sort merge join as the physical
operator. However it can be CPU and IO intensive in case input table is large
for sort merge join. Shuffled hash join on the other hand saves the sort CPU
and IO compared to sort merge join, especially when table is large.
This PR implements the full outer join as followed:
1. Construct hash relation from build side, with extra boolean value at the
end of row to track look up information (done in
`ShuffledHashJoinExec.buildHashedRelation` and `UnsafeHashedRelation.apply`).
2. Process rows from stream side by looking up hash relation, and mark the
matched rows from build side be looked up (done in
`ShuffledHashJoinExec.fullOuterJoin`).
3. Process rows from build side by iterating hash relation, and filter out
rows from build side being looked up already (done in
`ShuffledHashJoinExec.fullOuterJoin`).
TODO: codegen for full outer shuffled hash join can be implemented in
another followup PR.
### Why are the changes needed?
<!--
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.
-->
As implementation in this PR, full outer shuffled hash join will have
overhead to iterate build side twice (once for building hash map, and another
for outputting non-matching rows), and iterate stream side once. However, full
outer sort merge join needs to iterate both sides twice, and sort the large
table can be more CPU and IO intensive. So full outer shuffled hash join can be
more efficient than sort merge join when stream side is much more larger than
build side.
For example query below, full outer SHJ saved 30% wall clock time compared
to full outer SMJ.
```
def shuffleHashJoin(): Unit = {
val N: Long = 4 << 22
withSQLConf(
SQLConf.SHUFFLE_PARTITIONS.key -> "2",
SQLConf.AUTO_BROADCASTJOIN_THRESHOLD.key -> "20000000") {
codegenBenchmark("shuffle hash join", N) {
val df1 = spark.range(N).selectExpr(s"cast(id as string) as k1")
val df2 = spark.range(N / 10).selectExpr(s"cast(id * 10 as string)
as k2")
val df = df1.join(df2, col("k1") === col("k2"), "full_outer")
df.noop()
}
}
}
```
```
Running benchmark: shuffle hash join
Running case: shuffle hash join off
Stopped after 2 iterations, 15839 ms
Running case: shuffle hash join on
Stopped after 5 iterations, 28969 ms
Java HotSpot(TM) 64-Bit Server VM 1.8.0_181-b13 on Mac OS X 10.15.4
Intel(R) Core(TM) i9-9980HK CPU @ 2.40GHz
shuffle hash join: Best Time(ms) Avg Time(ms)
Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
shuffle hash join off 7555 7920
516 2.2 450.3 1.0X
shuffle hash join on 5731 5794
62 2.9 341.6 1.3X
```
### Does this PR introduce _any_ user-facing change?
<!--
Note that it means *any* user-facing change including all aspects such as
the documentation fix.
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 possible, please also clarify if this is a user-facing change compared to
the released Spark versions or within the unreleased branches such as master.
If no, write 'No'.
-->
No.
### How was this patch tested?
<!--
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.
-->
Added unit test in `JoinSuite.scala`.
----------------------------------------------------------------
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:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
