[
https://issues.apache.org/jira/browse/FLINK-4795?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15564695#comment-15564695
]
Chesnay Schepler commented on FLINK-4795:
-----------------------------------------
Thank you for reporting this. As a general remark it is always helpful to
include the error message you got.
You are correct that _map should be called instead.
read_csv not supporting nested tuples is intended to be consistent with the
java & scala counterparts.
> CsvStringify crashes in case of tuple in tuple, t.e. ("a", True, (1,5))
> -----------------------------------------------------------------------
>
> Key: FLINK-4795
> URL: https://issues.apache.org/jira/browse/FLINK-4795
> Project: Flink
> Issue Type: Bug
> Components: Python API
> Reporter: Yakov Goldberg
>
> CsvStringify crashes in case of tuple in tuple, t.e. ("a", True, (1,5))
> Looks like, mistyping in CsvStringify._map()
> def _map(self, value):
> if isinstance(value, (tuple, list)):
> return "(" + b", ".join([self.map(x) for x in value]) + ")"
> else:
> return str(value)
> self._map() should be called
> But this will affect write_csv() and read_csv().
> write_csv() will work automatically
> and read_csv() should be implemented to be able to read Tuple type.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
