[
https://issues.apache.org/jira/browse/ARROW-16020?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17511946#comment-17511946
]
Antoine Pitrou commented on ARROW-16020:
----------------------------------------
In any case, if you iterate in values in Python one by one, it is going to be
slow.... Is your custom format not Arrow-compatible?
> [Python] Provide access to buffers underlying scalars
> ------------------------------------------------------
>
> Key: ARROW-16020
> URL: https://issues.apache.org/jira/browse/ARROW-16020
> Project: Apache Arrow
> Issue Type: New Feature
> Components: Python
> Reporter: Kyle Kavanagh
> Priority: Major
>
> I'm building a process to take data from pyarrow Tables and write their data
> a memory mapped file in a custom format. Currently, I iterate through the
> arrow table and must call as_py() only to convert the python value to bytes
> and write to the memory mapped file. If the pyarrow scalar API provided a
> view over the underlying storage, I could simply memcopy the values from the
> arrow buffer into the mmap buffer.
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
