Github user MaxGekk commented on the issue:
https://github.com/apache/spark/pull/22528
> Another concern here is, we have another place to control the compression
codec (where we usually delegate to HDFS libraries).
I was considering using Compressor API but its streaming nature
controverses to structure of zip archive where meta-info is located at the end
of files, and you cannot read/uncompress it sequentially block-by-block.
> It just sounds like a bandaid fix to allow one zipped file case in multi
line mode.
I believe it is better to return correct result in a case when wrong result
is returned for now (try to read zipped CSV), or to force users to use this
workaround only to read zip archives via RDD API:
https://docs.databricks.com/spark/latest/data-sources/zip-files.html#zip-files
. Especially in the case of compressed not splittable CSV, there is not big
difference how to read it in multiLine enabled or disabled.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
