Snippets Groups Projects

[SPARK-15985][SQL] Eliminate redundant cast from an array without null or a map without null

Kazuaki Ishizaki authored 8 years ago

## What changes were proposed in this pull request?

This PR eliminates redundant cast from an `ArrayType` with `containsNull = false` or a `MapType` with `containsNull = false`.

For example, in `ArrayType` case, current implementation leaves a cast `cast(value#63 as array<double>).toDoubleArray`. However, we can eliminate `cast(value#63 as array<double>)` if we know `value#63` does not include `null`. This PR apply this elimination for `ArrayType` and `MapType` in `SimplifyCasts` at a plan optimization phase.

In summary, we got 1.2-1.3x performance improvements over the code before applying this PR.
Here are performance results of benchmark programs:
```
  test("Read array in Dataset") {
    import sparkSession.implicits._

    val iters = 5
    val n = 1024 * 1024
    val rows = 15

    val benchmark = new Benchmark("Read primnitive array", n)

    val rand = new Random(511)
    val intDS = sparkSession.sparkContext.parallelize(0 until rows, 1)
      .map(i => Array.tabulate(n)(i => i)).toDS()
    intDS.count() // force to create ds
    val lastElement = n - 1
    val randElement = rand.nextInt(lastElement)

    benchmark.addCase(s"Read int array in Dataset", numIters = iters)(iter => {
      val idx0 = randElement
      val idx1 = lastElement
      intDS.map(a => a(0) + a(idx0) + a(idx1)).collect
    })

    val doubleDS = sparkSession.sparkContext.parallelize(0 until rows, 1)
      .map(i => Array.tabulate(n)(i => i.toDouble)).toDS()
    doubleDS.count() // force to create ds

    benchmark.addCase(s"Read double array in Dataset", numIters = iters)(iter => {
      val idx0 = randElement
      val idx1 = lastElement
      doubleDS.map(a => a(0) + a(idx0) + a(idx1)).collect
    })

    benchmark.run()
  }

Java HotSpot(TM) 64-Bit Server VM 1.8.0_92-b14 on Mac OS X 10.10.4
Intel(R) Core(TM) i5-5257U CPU  2.70GHz

without this PR
Read primnitive array:                   Best/Avg Time(ms)    Rate(M/s)   Per Row(ns)   Relative
------------------------------------------------------------------------------------------------
Read int array in Dataset                      525 /  690          2.0         500.9       1.0X
Read double array in Dataset                   947 / 1209          1.1         902.7       0.6X

with this PR
Read primnitive array:                   Best/Avg Time(ms)    Rate(M/s)   Per Row(ns)   Relative
------------------------------------------------------------------------------------------------
Read int array in Dataset                      400 /  492          2.6         381.5       1.0X
Read double array in Dataset                   788 /  870          1.3         751.4       0.5X
```

An example program that originally caused this performance issue.
```
val ds = Seq(Array(1.0, 2.0, 3.0), Array(4.0, 5.0, 6.0)).toDS()
val ds2 = ds.map(p => {
     var s = 0.0
     for (i <- 0 to 2) { s += p(i) }
     s
   })
ds2.show
ds2.explain(true)
```

Plans before this PR
```
== Parsed Logical Plan ==
'SerializeFromObject [input[0, double, true] AS value#68]
+- 'MapElements <function1>, obj#67: double
   +- 'DeserializeToObject unresolveddeserializer(upcast(getcolumnbyordinal(0, ArrayType(DoubleType,false)), ArrayType(DoubleType,false), - root class: "scala.Array").toDoubleArray), obj#66: [D
      +- LocalRelation [value#63]

== Analyzed Logical Plan ==
value: double
SerializeFromObject [input[0, double, true] AS value#68]
+- MapElements <function1>, obj#67: double
   +- DeserializeToObject cast(value#63 as array<double>).toDoubleArray, obj#66: [D
      +- LocalRelation [value#63]

== Optimized Logical Plan ==
SerializeFromObject [input[0, double, true] AS value#68]
+- MapElements <function1>, obj#67: double
   +- DeserializeToObject cast(value#63 as array<double>).toDoubleArray, obj#66: [D
      +- LocalRelation [value#63]

== Physical Plan ==
*SerializeFromObject [input[0, double, true] AS value#68]
+- *MapElements <function1>, obj#67: double
   +- *DeserializeToObject cast(value#63 as array<double>).toDoubleArray, obj#66: [D
      +- LocalTableScan [value#63]
```

Plans after this PR
```
== Parsed Logical Plan ==
'SerializeFromObject [input[0, double, true] AS value#6]
+- 'MapElements <function1>, obj#5: double
   +- 'DeserializeToObject unresolveddeserializer(upcast(getcolumnbyordinal(0, ArrayType(DoubleType,false)), ArrayType(DoubleType,false), - root class: "scala.Array").toDoubleArray), obj#4: [D
      +- LocalRelation [value#1]

== Analyzed Logical Plan ==
value: double
SerializeFromObject [input[0, double, true] AS value#6]
+- MapElements <function1>, obj#5: double
   +- DeserializeToObject cast(value#1 as array<double>).toDoubleArray, obj#4: [D
      +- LocalRelation [value#1]

== Optimized Logical Plan ==
SerializeFromObject [input[0, double, true] AS value#6]
+- MapElements <function1>, obj#5: double
   +- DeserializeToObject value#1.toDoubleArray, obj#4: [D
      +- LocalRelation [value#1]

== Physical Plan ==
*SerializeFromObject [input[0, double, true] AS value#6]
+- *MapElements <function1>, obj#5: double
   +- *DeserializeToObject value#1.toDoubleArray, obj#4: [D
      +- LocalTableScan [value#1]
```

## How was this patch tested?

Tested by new test cases in `SimplifyCastsSuite`

Author: Kazuaki Ishizaki <ishizaki@jp.ibm.com>

Closes #13704 from kiszk/SPARK-15985.

d92cd227

d92cd227 8 years ago

Name	Last commit	Last update
.github
R
assembly
bin
build
common
conf
core
data
dev
docs
examples
external
graphx
launcher
licenses
mesos
mllib-local
mllib
project
python
repl
sbin
sql
streaming
tools
yarn
.gitattributes
.gitignore
.travis.yml
CONTRIBUTING.md
LICENSE
NOTICE
README.md
pom.xml
scalastyle-config.xml

Apache Spark

Spark is a fast and general cluster computing system for Big Data. It provides high-level APIs in Scala, Java, Python, and R, and an optimized engine that supports general computation graphs for data analysis. It also supports a rich set of higher-level tools including Spark SQL for SQL and DataFrames, MLlib for machine learning, GraphX for graph processing, and Spark Streaming for stream processing.

http://spark.apache.org/

Online Documentation

You can find the latest Spark documentation, including a programming guide, on the project web page and project wiki. This README file only contains basic setup instructions.

Building Spark

Spark is built using Apache Maven. To build Spark and its example programs, run:

build/mvn -DskipTests clean package

(You do not need to do this if you downloaded a pre-built package.)

You can build Spark using more than one thread by using the -T option with Maven, see "Parallel builds in Maven 3". More detailed documentation is available from the project site, at "Building Spark". For developing Spark using an IDE, see Eclipse and IntelliJ.

Interactive Scala Shell

The easiest way to start using Spark is through the Scala shell:

./bin/spark-shell

Try the following command, which should return 1000:

scala> sc.parallelize(1 to 1000).count()

Interactive Python Shell

Alternatively, if you prefer Python, you can use the Python shell:

./bin/pyspark

And run the following command, which should also return 1000:

>>> sc.parallelize(range(1000)).count()

Example Programs

Spark also comes with several sample programs in the examples directory. To run one of them, use ./bin/run-example <class> [params]. For example:

./bin/run-example SparkPi

will run the Pi example locally.

You can set the MASTER environment variable when running examples to submit examples to a cluster. This can be a mesos:// or spark:// URL, "yarn" to run on YARN, and "local" to run locally with one thread, or "local[N]" to run locally with N threads. You can also use an abbreviated class name if the class is in the examples package. For instance:

MASTER=spark://host:7077 ./bin/run-example SparkPi

Many of the example programs print usage help if no params are given.

Running Tests

Testing first requires building Spark. Once Spark is built, tests can be run using:

./dev/run-tests

Please see the guidance on how to run tests for a module, or individual tests.

A Note About Hadoop Versions

Spark uses the Hadoop core library to talk to HDFS and other Hadoop-supported storage systems. Because the protocols have changed in different versions of Hadoop, you must build Spark against the same version that your cluster runs.

Please refer to the build documentation at "Specifying the Hadoop Version" for detailed guidance on building for a particular distribution of Hadoop, including building for particular Hive and Hive Thriftserver distributions.

Configuration

Please refer to the Configuration Guide in the online documentation for an overview on how to configure Spark.