[
https://issues.apache.org/jira/browse/ARROW-14319?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17429651#comment-17429651
]
Roee Shlomo commented on ARROW-14319:
-------------------------------------
I suspect that a better approach would be to create a new module and keep
pyarrow.jvm as is:
# Backward compatibility seems like a challenge. There must be a reference
provided to org.apache.arrow.c so ArrowSchema, ArrowArray and the various
import/export functions would be available on the python side. In addition, all
C data interface methods require an allocator as a parameter. These are not
provided in the current pyarrow.jvm API.
# The current pyarrow.jvm module works with a pure java build of Arrow Java,
while the C data interface requires building a small JNI library. Unless you
rely on end users to build the Java jar on their own, packaging the JNI lib
will be required for all platforms targeted by pyarrow.
> [Python] Rewrite pyarrow.jvm using the C data interface
> -------------------------------------------------------
>
> Key: ARROW-14319
> URL: https://issues.apache.org/jira/browse/ARROW-14319
> Project: Apache Arrow
> Issue Type: Task
> Components: Java, Python
> Reporter: Antoine Pitrou
> Priority: Major
> Fix For: 7.0.0
>
>
> The {{pyarrow.jvm}} is currently a custom-written bridge between PyArrow and
> Arrow Java, with limited datatype support. Now that Java implements the C
> data interface (see ARROW-12965), we should be able to simplify the code
> while making it more general.
> Also, we should reenable the {{conda-python-jpype}} build somewhere, for
> example in the Crossbow nightly builds.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
