You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@spark.apache.org by "Andres Perez (JIRA)" <ji...@apache.org> on 2016/06/08 17:53:21 UTC
[jira] [Created] (SPARK-15825) sort-merge-join gives invalid
results when joining on a tupled key
Andres Perez created SPARK-15825:
------------------------------------
Summary: sort-merge-join gives invalid results when joining on a tupled key
Key: SPARK-15825
URL: https://issues.apache.org/jira/browse/SPARK-15825
Project: Spark
Issue Type: Bug
Components: SQL
Environment: spark 2.0.0-SNAPSHOT
Reporter: Andres Perez
{noformat}
import org.apache.spark.sql.functions
val left = List("0", "1", "2").toDS()
.map{ k => ((k, 0), "l") }
val right = List("0", "1", "2").toDS()
.map{ k => ((k, 0), "r") }
val result = left.toDF("k", "v").as[((String, Int), String)].alias("left")
.joinWith(right.toDF("k", "v").as[((String, Int), String)].alias("right"), functions.col("left.k") === functions.col("right.k"), "inner")
.as[(((String, Int), String), ((String, Int), String))]
{noformat}
When broadcast joins are enabled, we get the expected output:
{noformat}
(((0,0),l),((0,0),r))
(((1,0),l),((1,0),r))
(((2,0),l),((2,0),r))
{noformat}
However, when broadcast joins are disabled (i.e. setting spark.sql.autoBroadcastJoinThreshold to -1), the result is incorrect:
{noformat}
(((2,0),l),((2,-1),))
(((0,0),l),((0,-313907893),))
(((1,0),l),((null,-313907893),))
{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