RunSnakeRun is a small
GUI utility that allows you to view (Python) cProfile or Profile profiler dumps in a sortable GUI view. It allows
you to explore the profiler information using a "square map" visualization or sortable tables of data. It also
(experimentally) allows you to view the output of the Meliae "memory analysis" tool using the same basic visualisations.
RunSnakeRun is a simple program, it doesn't provide all the bells-and-whistles of a program like KCacheGrind, it's intended to allow for profiling your Python programs, and just your Python programs. What it does provide, for profile viewing:
For Meliae memory-dump viewing, it provides:
You will need to have all of wxPython, SquareMap and RunSnakeRun installed to use RunSnakeRun. You may also need the "python-profiler" package for your platform, which provides the pstats view. You will likely want to use your platform wxPython package (i.e. a pre-built binary). For Debian/Ubuntu distributions the prerequisite setup looks like this:
apt-get install python-profiler python-wxgtk2.8 python-setuptools
RunSnakeRun and SquareMap will install well in a VirtualEnv if you would like to keep them isolated (normally you do not
want to use the --no-site-packages flag if you are doing this). I recommend this approach rather than
using easy_install directly on your Linux/OS-X host.
virtualenv runsnake
source runsnake/bin/activate
If you already have Python setuptools installed (a.k.a. easy_install), you should be able to install the Python packages with:
easy_install SquareMap RunSnakeRun
You will require a modern wxPython (e.g. 2.8) and Python 2.x (e.g. 2.5 through 2.7)
installation. The setup will create a script named "runsnake" on Linux machines which launches the profile viewer.
On OS-X machines a wrapper script runsnake is created that runs the runsnake32 executable with a flag to tell Python
to use the 32-bit implementation (for wxPython compatibility). On Win32 machines, a Scripts\runsnake.exe executable
is created. If you have added your scripts directory to the PATH then this will be available from the command-line.
If you are new to profiling you may wish to check out:
To use cProfile to capture your application's profile data, either using the command-line, like so:
$ python -m cProfile -o <outputfilename> <script-name> <options>
Or in code, like so:
import cProfile
command = """reactor.run()"""
cProfile.runctx( command, globals(), locals(), filename="OpenGLContext.profile" )
To view the results of your run:
python runsnake.py OpenGLContext.profile
There will be a brief delay as the application is created and begins the loading process, then you should see something like this:
Click on any column title to sort by that property within that list. Select a record in the left-most list view to see a breakdown of that record in the right-side list views. Choose the appropriate view on the right via the tabs. You can resize the borders between the lists and square-map views. You can select a package/module/function hierarchic view via the menus. You can also toggle use of percentage displays there.
Note: this feature is considered experimental, the memory consumed loading even a tiny meliae
dump is enormous, so real-world programs will make RunSnakeRun quite slow and require a very large amount of RAM (far
more than the process being viewed).
To install Meliae, you will need a working C extension compilation environment (Meliae uses a Cython extension):
easy_install meliae
Now instrument your application to be able to trigger a memory dump at the moment you would like to capture, like so:
from meliae import scanner
scanner.dump_all_objects( filename ) # you can pass a file-handle if you prefer
The memory dump will generally be quite large (e.g. 2MB to describe an application with 200KB of user-controllable memory usage (i.e. not the interpreter itself)) and for any real application will take an extremely long time to load (multiple minutes for 16MB dumps).
$ runsnakemem <filename>
The Meliae loader in RunSnakeRun performs the following simplifications:
Even with those simplifications, however, the program is tracking most ints, strings, tuples, lists, etc. separately, which
uses a large amount of RAM and slows down the GUI substantially.
RunSnakeRun is reasonably stable. I don't tend to do much work on it, as it tends to just work. My (personal) current wish list for the project follows:
If you have an idea, feel free to check out the code and implement the new feature. I'm certainly willing to entertain new features or bug-fix requests as well. The code is available in bzr here:
bzr branch lp:~mcfletch/squaremap/trunk squaremap
cd squaremap
python setup.py develop
bzr branch lp:~mcfletch/profile_viewer/trunk profile_viewer
cd profile_viewer
python setup.py develop
You can contact me directly if you'd like to contribute. Or you can just set up a bzr branch on LaunchPad and request a merge.
This is just a listing of things that either I or others have requested as features:
RunSnakeRun is by no means a comprehensive tool-set for profiling, you may want to have any or all of these other tools available for your profiling needs: