[
https://issues.apache.org/jira/browse/SPARK-33369?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Gengliang Wang updated SPARK-33369:
-----------------------------------
Description:
For all the v2 data sources which are not FileDataSourceV2, Spark always infers
the table schema/partitioning on DataframeWriter.save().
The inference of table schema/partitioning can be expensive. However, there is
no such trait or flag for indicating a V2 source can use the input DataFrame's
schema on DataframeWriter.save(). We can resolve the problem by adding a new
expected behavior for the method TableProvider.supportsExternalMetadata():
When TableProvider.supportsExternalMetadata() is true, Spark will use the input
Dataframe's schema in DataframeWriter.save() and skip schema/partitioning
inference.
was:
For all the v2 data sources which are not FileDataSourceV2, Spark always infer
the table schema/partitioning on DataframeWriter.save().
Currently, there is no such trait or flag for indicating a V2 source can use
the input DataFrame's schema on DataframeWriter.save(). We can resolve the
problem by adding a new expected behavior for the method
TableProvider.supportsExternalMetadata():
when TableProvider.supportsExternalMetadata() is true, Spark will use the input
Dataframe's schema in DataframeWriter.save() and skip schema/partitioning
inference.
> Skip schema inference in DataframeWriter.save() if table provider supports
> external metadata
> --------------------------------------------------------------------------------------------
>
> Key: SPARK-33369
> URL: https://issues.apache.org/jira/browse/SPARK-33369
> Project: Spark
> Issue Type: New Feature
> Components: SQL
> Affects Versions: 3.1.0
> Reporter: Gengliang Wang
> Assignee: Gengliang Wang
> Priority: Major
>
> For all the v2 data sources which are not FileDataSourceV2, Spark always
> infers the table schema/partitioning on DataframeWriter.save().
> The inference of table schema/partitioning can be expensive. However, there
> is no such trait or flag for indicating a V2 source can use the input
> DataFrame's schema on DataframeWriter.save(). We can resolve the problem by
> adding a new expected behavior for the method
> TableProvider.supportsExternalMetadata():
> When TableProvider.supportsExternalMetadata() is true, Spark will use the
> input Dataframe's schema in DataframeWriter.save() and skip
> schema/partitioning inference.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
