PySizer - a memory profiler for Python

PySizer is a memory usage profiler for Python code.

Warning: I haven't had much time to work on PySizer recently. Another (better-maintained) project with the same aim is Heapy.

Downloading

You can get version 0.1.1 in tar.gz or zip format.

There's a win32 build for Python 2.4 here.

0.1.1 has bugfixes to 0.1 - in particular, a nasty bug which meant that copy and filterd were more or less broken. Also, it should work under Python 2.3 (but it hasn't been tested much) and under Windows, though I quietly put those two into 0.1 anyway.

The source is kept using Subversion at http://codespeak.net/svn/user/nick8325/sizer, and you can browse the source here too. The SVN head is broken at the moment, though.

Features

• You can make a snapshot of all reachable objects at a given time, organised as a tree (well, a graph, since there are cycles).
• For any given object, you can find out how much space it takes up, what objects it references and so on.

  With a patched version of Python, you can also find out what stack of function calls created an object, and what objects were created by each stack of calls.

• You can collect objects into groups. For example, you can group each object according to the module it appears to come from. Then you can treat the group as a single object.
• You can filter objects, find the amount of space used by instances of each type, find objects which appeared from one snapshot to the next, find the biggest objects/types/groups, and other things.

Requirements

See here.

Documentation

There's a tutorial here.

There's some automatically-generated documentation here.

Both of these come with the release.

You can find some information about the internals of the profiler here.

Bugs, feature requests, problems, anything else

You can contact me at nick.smallbone@gmail.com.