[jira] [Commented] (SPARK-25343) Extend CSV parsing to Dataset[List[String]]

["Hyukjin Kwon (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/SPARK-25343?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16605439#comment-16605439 ] 

Hyukjin Kwon commented on SPARK-25343:
--------------------------------------

The workaround should be like:

{code}
val ds: Dataset[List[String]] = ...
val csv = ds.mapPartition { iter =>
  val writer = new CsvWriter(...)
  iter.map { line => writer.writeRowToString(line) }
}
val df = spark.read.csv(csv.as[String])
{code}

If performance doesn't matter much but still works.

> Extend CSV parsing to Dataset[List[String]]
> -------------------------------------------
>
>                 Key: SPARK-25343
>                 URL: https://issues.apache.org/jira/browse/SPARK-25343
>             Project: Spark
>          Issue Type: Improvement
>          Components: Spark Core
>    Affects Versions: 2.3.1
>            Reporter: Frank Kemmer
>            Priority: Minor
>
> With the cvs() method it is currenty possible to create a Dataframe from Dataset[String], where the given string contains comma separated values. This is really great.
> But very often we have to parse files where we have to split the values of a line by very individual value separators and regular expressions. The result is a Dataset[List[String]]. This list corresponds to what you would get, after splitting the values of a CSV string at the separators.
> It would be great, if the csv() method would also accept such a Dataset as input especially given a target schema. The csv parser usually casts the separated values against the schema and can sort out lines where the values of the columns do not fit with the schema.
> This is especially interesting with PERMISSIVE mode and a column for corrupt records which then should contain the input list of strings as a dumped JSON string.
> This is the functionality I am looking for and I think the CSV parser is very close to it.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@spark.apache.org
For additional commands, e-mail: issues-help@spark.apache.org