You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@spark.apache.org by "gurmukh singh (JIRA)" <ji...@apache.org> on 2016/09/01 22:53:20 UTC
[jira] [Comment Edited] (SPARK-17211) Broadcast join produces
incorrect results on EMR with large driver memory
[ https://issues.apache.org/jira/browse/SPARK-17211?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15456867#comment-15456867 ]
gurmukh singh edited comment on SPARK-17211 at 9/1/16 10:52 PM:
----------------------------------------------------------------
Thanks Davies.
Can see the issue with the offsets:
It picks some random values beyond offset and then tries to do a join on rows, which will obviosuly return NULL as those rows are not there:
scala> val bc=sc.broadcast(df1)
scala> bc.value.view.take(10).foreach(println)
(abc,1)
(bcd,2)
(dcb,5)
scala> bc.value.view.toDF
res13: org.apache.spark.sql.DataFrame = [_1: string, _2: int]
scala> bc.value.view.toDF("gid", "id")
res14: org.apache.spark.sql.DataFrame = [gid: string, id: int]
scala> bc.value.view.toDF("gid", "id").show()
+---+---+
|gid| id|
+---+---+
|abc| 1|
|bcd| 2|
|dcb| 5|
+---+---+
scala> df1.join(bc.value.view.toDF("gid", "id"), "id").show()
+---+---+----+
| id|gid| gid|
+---+---+----+
| 1|123|null|
| 2|234|null|
| 5|432|null|
+---+---+----+
scala> df1.join(bc.value.view.toDF("gidyy", "id"), "id").show()
+---+---+-----+
| id|gid|gidyy|
+---+---+-----+
| 1|123| null|
| 2|234| null|
| 5|432| null|
+---+---+-----+
scala> bc.value.view.toDF("gid", "id").show()
+---+---+
|gid| id|
+---+---+
|abc| 1|
|bcd| 2|
|dcb| 5|
+---+---+
scala> df1.show()
+---+---+
|gid| id|
+---+---+
|123| 1|
|234| 2|
|432| 5|
+---+---+
scala> df1.join(bc.value.view.toDF("gidyy", "id"), "id").show()
+---+---+-----+
| id|gid|gidyy|
+---+---+-----+
| 1|123| null|
| 2|234| null|
| 5|432| null|
+---+---+-----+
scala> bc.value.view.toDF("gid", "id").join(df1, "id").show()
+-------+----+---+
| id| gid|gid|
+-------+----+---+
|6513249|null|123| <----- Look at the "id" where is it picking from ? is is fetching some out of address locations, which it should not be
|6579042|null|234|
|6447972|null|432|
+-------+----+---+
scala> bc.value.view.toDF("gidy", "id").join(df1, "id").show()
+-------+----+---+
| id|gidy|gid|
+-------+----+---+
|6513249|null|123|
|6579042|null|234|
|6447972|null|432|
+-------+----+---+
was (Author: gurmukhd):
Thanks Davies.
Can see the issue with the offsets:
*** Interestingly ****
It picks some random values beyond offset and then tries to do a join on rows, which will obviosuly return NULL as those rows are not there:
scala> val bc=sc.broadcast(df1)
scala> bc.value.view.take(10).foreach(println)
(abc,1)
(bcd,2)
(dcb,5)
scala> bc.value.view.toDF
res13: org.apache.spark.sql.DataFrame = [_1: string, _2: int]
scala> bc.value.view.toDF("gid", "id")
res14: org.apache.spark.sql.DataFrame = [gid: string, id: int]
scala> bc.value.view.toDF("gid", "id").show()
+---+---+
|gid| id|
+---+---+
|abc| 1|
|bcd| 2|
|dcb| 5|
+---+---+
scala> df1.join(bc.value.view.toDF("gid", "id"), "id").show()
+---+---+----+
| id|gid| gid|
+---+---+----+
| 1|123|null|
| 2|234|null|
| 5|432|null|
+---+---+----+
scala> df1.join(bc.value.view.toDF("gidyy", "id"), "id").show()
+---+---+-----+
| id|gid|gidyy|
+---+---+-----+
| 1|123| null|
| 2|234| null|
| 5|432| null|
+---+---+-----+
scala> bc.value.view.toDF("gid", "id").show()
+---+---+
|gid| id|
+---+---+
|abc| 1|
|bcd| 2|
|dcb| 5|
+---+---+
scala> df1.show()
+---+---+
|gid| id|
+---+---+
|123| 1|
|234| 2|
|432| 5|
+---+---+
scala> df1.join(bc.value.view.toDF("gidyy", "id"), "id").show()
+---+---+-----+
| id|gid|gidyy|
+---+---+-----+
| 1|123| null|
| 2|234| null|
| 5|432| null|
+---+---+-----+
scala> bc.value.view.toDF("gid", "id").join(df1, "id").show()
+-------+----+---+
| id| gid|gid|
+-------+----+---+
|6513249|null|123| <----- Look at the "id" where is it picking from ? is is fetching some out of address locations, which it should not be
|6579042|null|234|
|6447972|null|432|
+-------+----+---+
scala> bc.value.view.toDF("gidy", "id").join(df1, "id").show()
+-------+----+---+
| id|gidy|gid|
+-------+----+---+
|6513249|null|123|
|6579042|null|234|
|6447972|null|432|
+-------+----+---+
> Broadcast join produces incorrect results on EMR with large driver memory
> -------------------------------------------------------------------------
>
> Key: SPARK-17211
> URL: https://issues.apache.org/jira/browse/SPARK-17211
> Project: Spark
> Issue Type: Bug
> Components: PySpark
> Affects Versions: 2.0.0
> Reporter: Jarno Seppanen
> Assignee: Davies Liu
>
> Broadcast join produces incorrect columns in join result, see below for an example. The same join but without using broadcast gives the correct columns.
> Running PySpark on YARN on Amazon EMR 5.0.0.
> {noformat}
> import pyspark.sql.functions as func
> keys = [
> (54000000, 0),
> (54000001, 1),
> (54000002, 2),
> ]
> keys_df = spark.createDataFrame(keys, ['key_id', 'value']).coalesce(1)
> keys_df.show()
> # +--------+-----+
> # | key_id|value|
> # +--------+-----+
> # |54000000| 0|
> # |54000001| 1|
> # |54000002| 2|
> # +--------+-----+
> data = [
> (54000002, 1),
> (54000000, 2),
> (54000001, 3),
> ]
> data_df = spark.createDataFrame(data, ['key_id', 'foo'])
> data_df.show()
> # +--------+---+
> # | key_id|foo|
> # +--------+---+
> # |54000002| 1|
> # |54000000| 2|
> # |54000001| 3|
> # +--------+---+
> ### INCORRECT ###
> data_df.join(func.broadcast(keys_df), 'key_id').show()
> # +--------+---+--------+
> # | key_id|foo| value|
> # +--------+---+--------+
> # |54000002| 1|54000002|
> # |54000000| 2|54000000|
> # |54000001| 3|54000001|
> # +--------+---+--------+
> ### CORRECT ###
> data_df.join(keys_df, 'key_id').show()
> # +--------+---+-----+
> # | key_id|foo|value|
> # +--------+---+-----+
> # |54000000| 2| 0|
> # |54000001| 3| 1|
> # |54000002| 1| 2|
> # +--------+---+-----+
> {noformat}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@spark.apache.org
For additional commands, e-mail: issues-help@spark.apache.org