rymurr commented on a change in pull request #7753:
URL: https://github.com/apache/arrow/pull/7753#discussion_r454427669
##########
File path: python/pyarrow/jvm.py
##########
@@ -43,9 +43,14 @@ def jvm_buffer(arrowbuf):
pyarrow.Buffer
Python Buffer that references the JVM memory.
"""
+ import jpype
address = arrowbuf.memoryAddress()
size = arrowbuf.capacity()
- return pa.foreign_buffer(address, size, arrowbuf.asNettyBuffer())
+
+ # TODO: why can we not use arrowbuf as the base?
Review comment:
@wesm ref counting is a possibility. Though I don't see anywhere where
the ref count of the netty java object is explicitly referenced.
When the python buffer object is garbage collected there is nothing (as far
as I can tell) to collect the java based buffer. When the Python object is
collected it will decrement the python reference count of the java object and
eventually GC it but we would still have to manually close the Java side buffer
regardless of it being wrapped in netty or not. So we would have a memory leak
regardless if we don't carefully clean up buffers. Unless jpype does something
cool and undocumented w/ Netty buffers.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
