[
https://issues.apache.org/jira/browse/SPARK-17986?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Drew Robb updated SPARK-17986:
------------------------------
Description:
The SQLTransformer creates a temporary table when called, and does not delete
this temporary table. When using a SQLTransformer in a long running Spark
Streaming task, these temporary tables accumulate.
I believe that the fix would be as simple as calling
`dataset.sparkSession.catalog.dropTempView(tableName)` in the last part of
`transform`:
https://github.com/apache/spark/blob/v2.0.1/mllib/src/main/scala/org/apache/spark/ml/feature/SQLTransformer.scala#L65.
was:
The SQLTransformer creates a temporary table when called, and does not delete
this temporary table. When using a SQLTransformer in a long running Spark
Streaming task, these temporary tables accumulate.
I believe that the fix would be as simple as calling
`dataset.sparkSession.catalog.dropTempView(tableName)` in the last part of
`transform`:
https://github.com/apache/spark/blob/v2.0.1/mllib/src/main/scala/org/apache/spark/ml/feature/SQLTransformer.scala#L65.
I would be happy to attempt this fix myself if someone could validate this
issue.
> SQLTransformer leaks temporary tables
> -------------------------------------
>
> Key: SPARK-17986
> URL: https://issues.apache.org/jira/browse/SPARK-17986
> Project: Spark
> Issue Type: Bug
> Components: ML
> Affects Versions: 2.0.1
> Reporter: Drew Robb
> Priority: Minor
>
> The SQLTransformer creates a temporary table when called, and does not delete
> this temporary table. When using a SQLTransformer in a long running Spark
> Streaming task, these temporary tables accumulate.
> I believe that the fix would be as simple as calling
> `dataset.sparkSession.catalog.dropTempView(tableName)` in the last part of
> `transform`:
> https://github.com/apache/spark/blob/v2.0.1/mllib/src/main/scala/org/apache/spark/ml/feature/SQLTransformer.scala#L65.
>
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
