wankunde opened a new pull request, #39502:
URL: https://github.com/apache/spark/pull/39502
<!--
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'.
8. If you want to add or modify an error type or message, please read the
guideline first in
'core/src/main/resources/error/README.md'.
-->
### What changes were proposed in this pull request?
**percentile** function will put all the target elements into a Hashset and
then compute the result. We can try to combine percentile functions into one
and compute the result with only one Hashset.
For example:
```
SELECT max(value1) as max_value1, percentile(value2, 0.3) as p1,
percentile(value3, 0.4) + percentile(value3, 0.5) as p2,
percentile(value2, 0.6) as p3
FROM t1
```
can be optimized to
```
SELECT max_value1, _combined_percentile_0[0] as p1, p2,
_combined_percentile_0[1] as p3
FROM (
SELECT max(value1) as max_value1,
percentile(value3, 0.4) + percentile(value3, 0.5) as p2,
percentile(value2, array(0.3, 0.6)) as _combined_percentile_0
FROM t1) as t1
```
Benchmark for this change:
```
object CollapsePercentileBenchmark extends SqlBasedBenchmark {
override def runBenchmarkSuite(mainArgs: Array[String]): Unit = {
runBenchmark("Collapse percentiles") {
spark.range(1, 3000000, 1, 100)
.selectExpr("id", "id as c1")
.createOrReplaceTempView("tab")
val benchmark = new Benchmark("Test CollapsePercentiles", 1000, output
= output)
def runQuery(): Unit = {
val df = spark.sql(
s"""SELECT id,
| percentile(c1, 0.1) as p1,
| percentile(c1, 0.2) as p2,
| percentile(c1, 0.3) as p3,
| percentile(c1, 0.4) as p4,
| percentile(c1, 0.5) as p5,
| percentile(c1, 0.6) as p6,
| percentile(c1, 0.7) as p7,
| percentile(c1, 0.8) as p8,
| percentile(c1, 0.9) as p9
|FROM tab
|GROUP BY id
|""".stripMargin)
df.noop()
}
benchmark.addCase(s"Without CollapsePercentiles rule ", numIters = 2)
{ _ =>
withSQLConf(SQLConf.OPTIMIZER_EXCLUDED_RULES.key ->
"org.apache.spark.sql.catalyst.optimizer.CollapsePercentiles") {
runQuery()
}
}
benchmark.addCase(s"With CollapsePercentiles rule ", numIters = 2) { _
=>
runQuery()
}
benchmark.run()
}
}
}
```
Benchmark result :
```
Java HotSpot(TM) 64-Bit Server VM 1.8.0_281-b09 on Mac OS X 10.16
Intel(R) Core(TM) i9-9980HK CPU @ 2.40GHz
Test CollapsePercentiles: Best Time(ms) Avg Time(ms)
Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
Without CollapsePercentiles rule 49410 49949
762 0.0 49410389.9 1.0X
With CollapsePercentiles rule 13000 13000
0 0.0 13000355.1 3.8X
```
### Why are the changes needed?
Optimize percentile functions in SQL
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
Added UT
--
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.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
