[Pymerase-devel] [ pymerase-Bugs-1256258 ] DBAPI Memory Leak

[SourceForge.net <no...@so...>]

Bugs item #1256258, was opened at 2005-08-10 16:19
Message generated for change (Comment added) made by kingb
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=505345&aid=1256258&group_id=63836

Please note that this message will contain a full copy of the comment thread,
including the initial issue submission, for this request,
not just the latest update.
Category: Output: CreateDBAPI
Group: v0.2
>Status: Closed
>Resolution: Fixed
Priority: 9
Submitted By: Brandon King (kingb)
Assigned to: Brandon King (kingb)
Summary: DBAPI Memory Leak

Initial Comment:
Pymerase DBAPI has a memory leak where if you have two
DBAPI objects (lets say A and B) and you call
A.appendB(B) and then call del on B, it won't destory
B... This actually shouldn't happen, because if your
still holding on to A, then A should still hold on to
B. But if you call del on   A and B, neither will
actually be deleted because this have references
eachother and them selves.

The solution for is too fold... First using a weakref
from each object to itself head in the right direction.
Second, a function will need to be implemented such as
B._destroySelf() to tell it you want to remove it from
memory (which is needed if you commit a large number of
Bs and you're try to free up some memory by not holding
on to all previously commited B objects). This function
_destroySelf() would tell B to go to all associated
objects and remove any referrences to itself. Then you
would call del B to really remove the object from memory.

----------------------------------------------------------------------

>Comment By: Brandon King (kingb)
Date: 2005-08-15 11:09

Message:
Logged In: YES 
user_id=552216

Fixes for bugs 1256258 and 1256271. Memory leak should now be
fixed. Ten test cases have be implemented in
tests/TestMemLeak.py... All ten pass currently. I tested for
slow down
in creating and commiting 10,000 objects to a database. It
has a 0.2
sec slow down in a 94 sec time frame. The fix includes an
object cache
which reuses objects which have already been pulled into
memory (using
weakrefs). Also, there is a new function for DBClass objects
called
_destroySelf. This function must be called before calling
del on an
DBClass object. This is only useful when you want to free up
some
memory when your done with a DBClass object, but the program
is going
to continue to run. It works by removing all references to
the class
your trying to remove so that the python ref count will be
low enough
for garbage collection to free up the memory. Calling del
without
calling _destroySelf will NOT remove the object from memory.
To go
back to the memory leaky version check out tag
'pre_aug_2005_mem_patch'.

----------------------------------------------------------------------

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=505345&aid=1256258&group_id=63836