[Paraview] Memory leak in Paraview 3.6.1(?)

[Olumide]

Utkarsh,

I typically use python to load groups (sometimes up to a dozen) of VTK 
files  generated by each run of my experiment -- opening files manually 
is very tiring. However, whenever I clear the currently displayed files 
(using the "delete all" Edit menu option) Paraview never seems to give 
back all the memory it allocated. As such, after loading and unloading 
10-20 simulations, Paraview's memory consumption is so high that I 
either have to terminate the program or it crashes.

Here's the result of my typical run. I saved a small simulation to a 
state file and tried opening it and then clearing the data. The reported 
data is from the Windows Task manager (for some weird reason the task 
manager lists that the memory allocated to the Paraview process is 
greater than the process' virtual memory!?)

Initial (new instance of Paraview):
Mem usage: 75,272Mb  VM: 56,256Mb

After loading state file:
Mem usage: 89,512Mb  VM: 70,224Mb

After deleting all objects
Mem usage: 84,476Mb  VM: 65,132Mb

Another run, this time the python scripts are used to load the two 
simulation datasets. Growth in memory usage is more aggressive:

Initial (new instance of Paraview):
Mem usage: 75,304Mb  VM: 56,276Mb

After loading data set 1 via Python shell/script:
Mem usage: 113,264Mb  VM: 88,704Mb

After deleting all objects
Mem usage: 114,456Mb  VM: 89,688Mb

After loading dataset 2 via Python shell/script:
Mem usage: 126,116Mb  VM: 101,196Mb

After deleting all objects
Mem usage: 126,284Mb  VM: 101,388Mb

I'm aware that memory is dynamically allocated for the python shell and 
GUI, and their corresponding objects are probably kept in memory to make 
future launches of the python shell faster. This would explain why some 
memory isn't returned after the initial launch. However as this should 
done just once, the memory consumption should not grow each time a 
dataset is loaded using a Python script/GUI.

Also, after deleting a dataset and a new one is loaded, the following 
the message is frequently displayed.

qDrawShadeRect: Invalid parameters
qDrawShadeRect: Invalid parameters
qDrawShadeRect: Invalid parameters
..
qDrawShadeRect: Invalid parameters
qDrawShadeRect: Invalid parameters

A few other questions/suggestions:

1. Does the python command shell have a "run python script" command? For 
example: RunPythonScript("C:/Work/Data/Foo.py"). Such a feature would be 
a productivity boost. Clicking on the Tools menu and pressing the run 
script button is okay for loading a few datasets but tiring when there 
are lots of datasets to be evaluated. Also, it would be nice if the 
python command were dockable, with an autohide feature like Visual 
studio. Again this would be productivity boost. In fact it would be nice 
if most views had an autohide feature. This would be a quick way of 
increasing the size of the view window.

2. Does Paraview 3.6.2 have a disable offset feature as 
proposed/discussed here (I've searched but I can't find one):
http://tinyurl.com/yf86vwy

3. When can we expect to be able to have an external program send data 
to a Paraview, in the manner suggested in the following thread:
http://tinyurl.com/yfdyp7w

I hope I'm not being too greedy. I believe I speak for the entire 
community when I say we appreciate all your hard work.

Thanks,

- Olumide

-------- Original Message --------
Subject: 	Re: [Paraview] Memory leak in Paraview 3.6.1(?)
Date: 	Mon, 18 Jan 2010 18:30:41 GMT
From:	Utkarsh Ayachit (utka... at kitware.com)
List:	org.paraview.paraview

Is it reproducible without using Python for creating the objects?
Easiest way to try that is create the objects using python, then save
the state out, then restart paraview and load the state file to create
the objects.

Utkarsh

On Sat, Jan 16, 2010 at 1:34 AM, Olumide <502... at web.de> wrote:

This issue persists in Paraview 3.6.2, with which I've just viewed a
relatively small data set (one polysphere and about a 40 line segments).
However, after disconnecting the server, thereby unloading the data,
Paraview's memory consumption remains at 103Mb -- that's not right. Why
isn't Paraview releasing memory from purged objects. If it matters, the
objects were imported using a python script.

- Olumide