[jira] [Commented] (SPARK-15140) ensure input object of encoder is not null
[ https://issues.apache.org/jira/browse/SPARK-15140?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15301654#comment-15301654 ] Apache Spark commented on SPARK-15140: -- User 'cloud-fan' has created a pull request for this issue: https://github.com/apache/spark/pull/13322 > ensure input object of encoder is not null > -- > > Key: SPARK-15140 > URL: https://issues.apache.org/jira/browse/SPARK-15140 > Project: Spark > Issue Type: Improvement >Reporter: Wenchen Fan > > Current we assume the input object for encoder won't be null, but we don't > check it. For example, in 1.6 `Seq("a", null).toDS.collect` will throw NPE, > in 2.0 this will return Array("a", null). > We should define this behaviour more clearly. -- This message was sent by Atlassian JIRA (v6.3.4#6332) - To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-15140) ensure input object of encoder is not null
[ https://issues.apache.org/jira/browse/SPARK-15140?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15296805#comment-15296805 ] Michael Armbrust commented on SPARK-15140: -- I don't think you should ever get a null row back. So I think all the fields being null makes sense. > ensure input object of encoder is not null > -- > > Key: SPARK-15140 > URL: https://issues.apache.org/jira/browse/SPARK-15140 > Project: Spark > Issue Type: Improvement >Reporter: Wenchen Fan > > Current we assume the input object for encoder won't be null, but we don't > check it. For example, in 1.6 `Seq("a", null).toDS.collect` will throw NPE, > in 2.0 this will return Array("a", null). > We should define this behaviour more clearly. -- This message was sent by Atlassian JIRA (v6.3.4#6332) - To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-15140) ensure input object of encoder is not null
[ https://issues.apache.org/jira/browse/SPARK-15140?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15295313#comment-15295313 ] Wenchen Fan commented on SPARK-15140: - I'm a little worried about supporting null input object here, what should the serialized row be? a null row or a row with all fields null? > ensure input object of encoder is not null > -- > > Key: SPARK-15140 > URL: https://issues.apache.org/jira/browse/SPARK-15140 > Project: Spark > Issue Type: Improvement >Reporter: Wenchen Fan > > Current we assume the input object for encoder won't be null, but we don't > check it. For example, in 1.6 `Seq("a", null).toDS.collect` will throw NPE, > in 2.0 this will return Array("a", null). > We should define this behaviour more clearly. -- This message was sent by Atlassian JIRA (v6.3.4#6332) - To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-15140) ensure input object of encoder is not null
[ https://issues.apache.org/jira/browse/SPARK-15140?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15273100#comment-15273100 ] Michael Armbrust commented on SPARK-15140: -- The 2.0 behavior seems correct. Ideally .toDS().collect() will always round-trip the data without change. > ensure input object of encoder is not null > -- > > Key: SPARK-15140 > URL: https://issues.apache.org/jira/browse/SPARK-15140 > Project: Spark > Issue Type: Improvement >Reporter: Wenchen Fan > > Current we assume the input object for encoder won't be null, but we don't > check it. For example, in 1.6 `Seq("a", null).toDS.collect` will throw NPE, > in 2.0 this will return Array("a", null). > We should define this behaviour more clearly. -- This message was sent by Atlassian JIRA (v6.3.4#6332) - To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-15140) ensure input object of encoder is not null
[ https://issues.apache.org/jira/browse/SPARK-15140?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15271885#comment-15271885 ] Wenchen Fan commented on SPARK-15140: - cc [~marmbrus] [~lian cheng] > ensure input object of encoder is not null > -- > > Key: SPARK-15140 > URL: https://issues.apache.org/jira/browse/SPARK-15140 > Project: Spark > Issue Type: Improvement >Reporter: Wenchen Fan > > Current we assume the input object for encoder won't be null, but we don't > check it. For example, in 1.6 `Seq("a", null).toDS.collect` will throw NPE, > in 2.0 this will return Array("a", null). > We should define this behaviour more clearly. -- This message was sent by Atlassian JIRA (v6.3.4#6332) - To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org For additional commands, e-mail: issues-h...@spark.apache.org