Details
-
Bug
-
Status: Resolved
-
Minor
-
Resolution: Fixed
-
1.8.0
-
None
Description
In Spark, I can save an RDD to the local file system using a relative path, e.g.:
rdd.saveAsNewAPIHadoopFile( "relativeRoot", classOf[Void], tag.runtimeClass.asInstanceOf[Class[T]], classOf[ParquetOutputFormat[T]], job.getConfiguration)
This leads to a crash in the ParquetFileWriter.mergeFooters(..) method since the footer paths are read as fully qualified paths, but the root path is provided as a relative path:
org.apache.parquet.io.ParquetEncodingException: /Users/stevenshe/schema/relativeRoot/part-r-00000.snappy.parquet invalid: all the files must be contained in the root relativeRoot
Attachments
Issue Links
- links to