[
https://issues.apache.org/jira/browse/ARROW-4091?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16918640#comment-16918640
]
Neal Richardson commented on ARROW-4091:
----------------------------------------
I don't think the C++ csv reader should have a complex default for this.
Perhaps not any strings at all (i.e. the default is that the only way to get
missing values from CSV is if you have an empty cell). The R package and
pyarrow/pandas bindings may want to have different defaults that match their
users' expectations--for example, "NA" is the default {{na.strings}} for R but
isn't how missing is spelled in every other language, so it would make sense
for R users to have that default but probably not in the C++ library. Does that
make sense?
> [C++] Curate default list of CSV null spellings
> -----------------------------------------------
>
> Key: ARROW-4091
> URL: https://issues.apache.org/jira/browse/ARROW-4091
> Project: Apache Arrow
> Issue Type: Wish
> Components: C++
> Affects Versions: 0.11.1
> Reporter: Antoine Pitrou
> Priority: Major
> Fix For: 1.0.0
>
>
> "NaN" is not null in Arrow-land (at least not for float columns?).
--
This message was sent by Atlassian Jira
(v8.3.2#803003)
