Thank you for that information ! This would finally be a reason to upgrade I guess :).
Cheers, Sebastian On Mon, Apr 13, 2009 at 5:05 AM, Ian Jones <[email protected]> wrote: > > Yes this has been reported before and a known issue to Autodesk and > claimed-fixed in 2009 and to my testing thus far it has been. > > On Sat, Apr 11, 2009 at 8:43 AM, Byron <[email protected]> wrote: > > > > ... and I truly hope that whatever I found out is just my mistake, > > already solved and not as terrible as it appears, so here it comes: > > Whenever you make a call to maya.cmds.<commandname>, an arglist and a > > your list of kwargs will be stored as tuples and held. These tuples > > will never be collected or deleted, and thus consume memory. > > This starts becoming an issue if you call a command a few 10 thousands > > of times. > > > > I could verify this issue on Maya8.5 64 linux and Maya2008 OSX. > > > > I assume that someone in the c-part of the cmds wrap does not > > decrement the refcount on the tuples created, thus python keeps > > thinking they are in use. The refcount of these tuples 3(-1) in Maya > > 8.5 and 4(-1) in Maya2008. > > > > Is this a known bug ? If not it would be great if anyone with a nice > > support contract could bring it up :). > > > > Cheers, > > Sebastian ( being so incredibly disappointed right now ) > > > > The Code : > > Makes 10 runs and calls simple mel commands 1000 times, checking the > > garbage collectors object count after each run. > > The collection cannot get ahold of these objects being continuously > > added to the collector. > > As these commands are read-only, the undo-queue cannot be involved. > > If you rewrite the loop so that it uses the pymel API wrap for > > instance, the gc's object count will not raise. > > If you want to crank up the numbers, you will see that this really > > consumes memory once the object count hits a million and if you issue > > a little more complex commands. > > > > <pre> > > import sys > > import gc > > assert gc.isenabled() > > import maya.cmds as cmds > > > > for run in range( 10 ): > > print "%i: Pre-cmds execution: %i" % ( run, len( gc.get_objects() > ) ) > > for i in range( 1000 ): > > cmds.listRelatives( "persp", shapes = 1, ni = 1 ) > > cmds.ls( sl=1 ) > > print "Post-cmds execution: %i" % len( gc.get_objects() ) > > > > # the garbage collection cannot get ahold of these objects > > assert gc.collect() == 0 > > > > # lets check the refcount on these tuples > > lobj = gc.get_objects()[-1] > > print "%s.refcount = %i" % ( lobj, sys.getrefcount( lobj ) ) > > > > # just to be sure: yes, its our arguments we passed to the cmds > > for obj in gc.get_objects()[-50:]: print obj > > </pre> > > > > > > > > > --~--~---------~--~----~------------~-------~--~----~ http://groups.google.com/group/python_inside_maya -~----------~----~----~----~------~----~------~--~---
