[
https://issues.apache.org/jira/browse/ARROW-1389?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16137457#comment-16137457
]
Philipp Moritz commented on ARROW-1389:
---------------------------------------
For the use case I described above, almost anything would work I think; I'm not
familiar which standard is the best to use. There is also the really simple
string based ones that GMP has:
https://gmplib.org/manual/I_002fO-of-Integers.html (which are probably less
efficient than VLQ)
A stop gap soluntion if we don't want to make decisions right now is to use a
StructArray of strings that encodes the integer the as a string with whatever
_PyLong_AsByteArray (see
http://svn.python.org/projects/python/tags/r23a1/Include/longobject.h) returns.
> Support arbitrary precision integers
> ------------------------------------
>
> Key: ARROW-1389
> URL: https://issues.apache.org/jira/browse/ARROW-1389
> Project: Apache Arrow
> Issue Type: Improvement
> Reporter: Philipp Moritz
> Priority: Minor
>
> For Python serialization it would be great if we had Arrow support for
> arbitrary precision integers, see the comment in
> https://github.com/apache/arrow/blob/de7c6715ba244e119913bfa31b8de46dbbd450bf/python/pyarrow/tests/test_serialization.py#L183
> Long integers are for example used in the uuid python module and having this
> would increase serialization performance for uuids and also make the code
> cleaner.
> I wonder if this is more generally useful too, any thoughts?
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
