[
https://issues.apache.org/jira/browse/HUDI-1376?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Wenning Ding updated HUDI-1376:
-------------------------------
Description:
When updating a Hudi table through Spark datasource, it will use the schema of
the input dataframe as the schema stored in the commit files. Thus, when
upserted with rows containing metadata columns, the upsert commit file will
store the metadata columns schema in the commit file which is unnecessary for
common cases. And also this will bring an issue for bootstrap table.
Since metadata columns is not used during the Spark datasource processing, we
can drop those columns in the beginning of Spark datasource.
was:
When updating a Hudi table through Spark datasource, it will use the schema of
the input dataframe as the schema stored in the commit files. Thus, when
upserted with rows containing metadata columns, the upsert commit file will
store the metadata columns schema in the commit file which is unnecessary for
common cases. And also this will bring an issue for bootstrap table.
Since the schema of metadata columns is always the same, we should remove the
schema of metadata columns in the commit file for any insert/upsert/... action.
> Drop metadata columns before Spark datasource processing
> ---------------------------------------------------------
>
> Key: HUDI-1376
> URL: https://issues.apache.org/jira/browse/HUDI-1376
> Project: Apache Hudi
> Issue Type: Bug
> Reporter: Wenning Ding
> Assignee: Wenning Ding
> Priority: Major
> Labels: pull-request-available
>
> When updating a Hudi table through Spark datasource, it will use the schema
> of the input dataframe as the schema stored in the commit files. Thus, when
> upserted with rows containing metadata columns, the upsert commit file will
> store the metadata columns schema in the commit file which is unnecessary for
> common cases. And also this will bring an issue for bootstrap table.
> Since metadata columns is not used during the Spark datasource processing, we
> can drop those columns in the beginning of Spark datasource.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
