[
https://issues.apache.org/jira/browse/ARROW-341?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15775305#comment-15775305
]
Wes McKinney commented on ARROW-341:
------------------------------------
I thought a bit about this. The easiest thing to do would be to install all of
the libpyarrow header files under the pyarrow build tree, similar to how NumPy
has {{numpy/core/include}} and the helper function {{numpy.get_include()}}. We
would need an equivalent "import hook" like "import_array" which ensures that
the right shared libraries have been loaded at runtime.
> [Python] Making libpyarrow available to third parties
> -----------------------------------------------------
>
> Key: ARROW-341
> URL: https://issues.apache.org/jira/browse/ARROW-341
> Project: Apache Arrow
> Issue Type: New Feature
> Components: Python
> Reporter: Wes McKinney
>
> We may want to link against this library in pandas 2.0 -- [~xhochy] what do
> you think? One option is that we could move the Python glue into the C++
> library and create a leaf lib {{libarrow_python.so}}.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
