Details
-
Improvement
-
Status: In Progress
-
Minor
-
Resolution: Unresolved
-
3.1.0
-
None
-
None
Description
Spark can read compression files if there is compression codec for them. By default, Hadoop provides compressors/decompressors for bzip2, deflate, gzip, lz4 and snappy but they cannot be used directly for reading zip archives.
In general zip archives can contain multiple entries but in practice users use zip archives to store only one file. This use case is pretty often in wild.
The ticket aims to support reading of zipped CSV and JSON files in multi-line mode when each zip archive contains only one file.
For example, current approach of reading zipped CSV files looks complicated : https://docs.azuredatabricks.net/spark/latest/data-sources/zip-files.html#zip-files
