[
https://issues.apache.org/jira/browse/ARROW-5345?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Wes McKinney updated ARROW-5345:
--------------------------------
Fix Version/s: (was: 0.14.0)
> [C++] Relax Field hashing in DictionaryMemo
> -------------------------------------------
>
> Key: ARROW-5345
> URL: https://issues.apache.org/jira/browse/ARROW-5345
> Project: Apache Arrow
> Issue Type: Improvement
> Components: C++
> Reporter: Wes McKinney
> Priority: Major
>
> Follow up to ARROW-3144
> Currently we associate dictionaries with a hash table mapping a Field's
> memory address to a dictionary id. This poses an issue if two RecordBatches
> are equal (equal field names, equal types) but were instantiated separately.
> We don't have a hash function in C++ for Field so we should consider
> implementing one and using that instead (if it is not too expensive) so that
> same but "different" (different C++ objects) won't blow up in the user's face
> with an unintuitive error (this did in fact occur once in the Python test
> suite, not sure exactly why it wasn't a problem before, I think it worked "by
> accident")
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
