[
https://issues.apache.org/jira/browse/ARROW-4390?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16911685#comment-16911685
]
Neal Richardson commented on ARROW-4390:
----------------------------------------
See [https://haven.tidyverse.org/articles/semantics.html] for reference.
The naive approach would be to call {{haven::as_factor()}} on any "labelled"
class objects before passing them from R to Arrow, but this is lossy and it can
be handled by the user outside of the {{arrow}} package, so I'm not sure
there's value in adding that here.
The best solution might be some kind of extension type that embellishes
dictionary elements with additional properties. We might be able to do
something useful with the missing data bitmask to capture the multiple types of
missingness, short of creating an extension type. But it's hard for me to
reason about the right thing to do here until there's more aggregation methods
exposed from Arrow. So I'm putting this one back down.
> [R] Serialize "labeled" metadata in Feather files, IPC messages
> ---------------------------------------------------------------
>
> Key: ARROW-4390
> URL: https://issues.apache.org/jira/browse/ARROW-4390
> Project: Apache Arrow
> Issue Type: Improvement
> Components: R
> Reporter: Wes McKinney
> Assignee: Neal Richardson
> Priority: Major
> Fix For: 0.15.0
>
>
> see https://github.com/apache/arrow/issues/3480
--
This message was sent by Atlassian Jira
(v8.3.2#803003)
