[
https://issues.apache.org/jira/browse/ARROW-10848?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17389041#comment-17389041
]
Weston Pace commented on ARROW-10848:
-------------------------------------
> I see that it is possible to cast to date32 from timestamp. I would think
> that the timestamp_parsers or whatever it is that you can define now should
> also be applied for any column declared to be date or time type. I could
> envision having more involved column-level parsing instructions as a feature
> in the CSV reader too, but that's probably more than we want to tackle now.
If it can be handled losslessly by a scanner projection then I think it is not
a candidate for a CSV option. Otherwise you could use the same "utility"
rationale to copy the entire scanner API into the CSV reader. Since
timestamp->date32 is lossless I agree with Antoine that this should be closed.
> [C++] CSV ISO-8601 date and timestamp short form
> ------------------------------------------------
>
> Key: ARROW-10848
> URL: https://issues.apache.org/jira/browse/ARROW-10848
> Project: Apache Arrow
> Issue Type: Improvement
> Components: C++
> Reporter: Maciej
> Priority: Major
> Labels: csv
> Fix For: 6.0.0
>
>
> Arrow supp{color:#172b4d}orts ISO-8601 for date and timestamp parsing but
> doesn't support short form of them. E.g.{color}
> {code:java}
> 19990108
> or
> 19990108 040506
> {code}
> Examples taken from:
> [https://www.postgresql.org/docs/12/datatype-datetime.html]
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
