[
https://issues.apache.org/jira/browse/SPARK-38314?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Yaohua Zhao updated SPARK-38314:
--------------------------------
Description:
Selecting and then writing df containing hidden file metadata column
`_metadata` into a file format like `parquet`, `delta` will still keep the
internal `Attribute` metadata information. Then when reading those `parquet`,
`delta` files again, it will actually break the code, because it wrongly thinks
user data schema named `_metadata` is a hidden file source metadata column.
Reproducible code:
{code:java}
// prepare a file source df
df.select("*", "_metadata")
.write.format("parquet").save(path)
spark.read.format("parquet").load(path)
.select("*").show(){code}
was:
Selecting and then writing df containing hidden file metadata column
`_metadata` into a file format like `parquet`, `delta` will still keep the
internal `Attribute` metadata information. Then when reading those `parquet`,
`delta` files again, it will actually break the code, because it wrongly thinks
user data schema named `_metadata` is a hidden file source metadata column.
Reproducible code:
```
// prepare a file source df
df.select("*", "_metadata")
.write.format("parquet").save(path)
spark.read.format("parquet").load(path)
.select("*").show()
```
> Fail to read parquet files after writing the hidden file metadata in
> --------------------------------------------------------------------
>
> Key: SPARK-38314
> URL: https://issues.apache.org/jira/browse/SPARK-38314
> Project: Spark
> Issue Type: Bug
> Components: SQL
> Affects Versions: 3.2.1
> Reporter: Yaohua Zhao
> Priority: Major
>
> Selecting and then writing df containing hidden file metadata column
> `_metadata` into a file format like `parquet`, `delta` will still keep the
> internal `Attribute` metadata information. Then when reading those `parquet`,
> `delta` files again, it will actually break the code, because it wrongly
> thinks user data schema named `_metadata` is a hidden file source metadata
> column.
>
> Reproducible code:
> {code:java}
> // prepare a file source df
> df.select("*", "_metadata")
> .write.format("parquet").save(path)
> spark.read.format("parquet").load(path)
> .select("*").show(){code}
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
