[Greenhouse] Memory profiling tips

[Asheesh Laroia]

Hi Dave, and anyone else interested,

Dave was noticing that the UDD import code gets killed on Dotcloud, 
presumably by running out of RAM.

As in all programming, to fix this bug, one must know where it is.

Therefore, I suggest being able to reproduce it locally. To do that, here 
is a quick Asheesh tutorial on memory profiling with Python.

Do all this on a machine with oodles of RAM; we'll use this as a way to 
get the stats on a working version of the program.

First, install 'memory_profiler' into your local machine/virtualenv. e.g.

$ pip install memory_profiler

Then, visit https://pypi.python.org/pypi/memory_profiler and create a file 
called example.com with the contents suggested in the "Usage" section. 
Then, run "python -m memory_profiler example.py" and make sure you can see 
the same table they show.

So far, so good?

Now make sure you undersetand it! Read carefully or ask questions.

Finally, temporarily hack your Django code so that the management 
command's handle() method has this @profile decorator applied to it, and 
run it via 'python -m memory_profiler manage.py my_management_command'.

Then you should get a table like that.

Is there one line that sees a massive spike in memory usage that isn't 
what you'd expect?

If so, then maybe we can make that one line perform better!

-- Asheesh.