diff options
author | Dongjoon Hyun <dongjoon@apache.org> | 2016-07-08 16:44:53 +0800 |
---|---|---|
committer | Wenchen Fan <wenchen@databricks.com> | 2016-07-08 16:44:53 +0800 |
commit | dff73bfa5e08c4c065584cfa9655a7839d28ad49 (patch) | |
tree | a32f2729fe6b5885c1b5a01286fb5ef3cfa50e6c /python/pyspark | |
parent | 5bce4580939c27876f11cd75f0dc2190fb9fa908 (diff) | |
download | spark-dff73bfa5e08c4c065584cfa9655a7839d28ad49.tar.gz spark-dff73bfa5e08c4c065584cfa9655a7839d28ad49.tar.bz2 spark-dff73bfa5e08c4c065584cfa9655a7839d28ad49.zip |
[SPARK-16052][SQL] Improve `CollapseRepartition` optimizer for Repartition/RepartitionBy
## What changes were proposed in this pull request?
This PR improves `CollapseRepartition` to optimize the adjacent combinations of **Repartition** and **RepartitionBy**. Also, this PR adds a testsuite for this optimizer.
**Target Scenario**
```scala
scala> val dsView1 = spark.range(8).repartition(8, $"id")
scala> dsView1.createOrReplaceTempView("dsView1")
scala> sql("select id from dsView1 distribute by id").explain(true)
```
**Before**
```scala
scala> sql("select id from dsView1 distribute by id").explain(true)
== Parsed Logical Plan ==
'RepartitionByExpression ['id]
+- 'Project ['id]
+- 'UnresolvedRelation `dsView1`
== Analyzed Logical Plan ==
id: bigint
RepartitionByExpression [id#0L]
+- Project [id#0L]
+- SubqueryAlias dsview1
+- RepartitionByExpression [id#0L], 8
+- Range (0, 8, splits=8)
== Optimized Logical Plan ==
RepartitionByExpression [id#0L]
+- RepartitionByExpression [id#0L], 8
+- Range (0, 8, splits=8)
== Physical Plan ==
Exchange hashpartitioning(id#0L, 200)
+- Exchange hashpartitioning(id#0L, 8)
+- *Range (0, 8, splits=8)
```
**After**
```scala
scala> sql("select id from dsView1 distribute by id").explain(true)
== Parsed Logical Plan ==
'RepartitionByExpression ['id]
+- 'Project ['id]
+- 'UnresolvedRelation `dsView1`
== Analyzed Logical Plan ==
id: bigint
RepartitionByExpression [id#0L]
+- Project [id#0L]
+- SubqueryAlias dsview1
+- RepartitionByExpression [id#0L], 8
+- Range (0, 8, splits=8)
== Optimized Logical Plan ==
RepartitionByExpression [id#0L]
+- Range (0, 8, splits=8)
== Physical Plan ==
Exchange hashpartitioning(id#0L, 200)
+- *Range (0, 8, splits=8)
```
## How was this patch tested?
Pass the Jenkins tests (including a new testsuite).
Author: Dongjoon Hyun <dongjoon@apache.org>
Closes #13765 from dongjoon-hyun/SPARK-16052.
Diffstat (limited to 'python/pyspark')
-rw-r--r-- | python/pyspark/sql/dataframe.py | 4 |
1 files changed, 2 insertions, 2 deletions
diff --git a/python/pyspark/sql/dataframe.py b/python/pyspark/sql/dataframe.py index a0ac7a9342..dd670a9b3d 100644 --- a/python/pyspark/sql/dataframe.py +++ b/python/pyspark/sql/dataframe.py @@ -464,10 +464,10 @@ class DataFrame(object): +---+-----+ |age| name| +---+-----+ - | 5| Bob| - | 5| Bob| | 2|Alice| + | 5| Bob| | 2|Alice| + | 5| Bob| +---+-----+ >>> data.rdd.getNumPartitions() 7 |