[Numpy-discussion] Avoiding messy code in Python

Chris Barker - NOAA Federal chris.barker at noaa.gov
Wed Jun 12 12:27:58 EDT 2013

> I have some code more or less 500 lines, but very messy code. All codes
> containing several functions are in one module, besides, without
> documentation and testing.
> Could anyone give me some advice to organize my messy code in an accurate
> style including test function as well?

This is a really broad, non-numpy specific question. You'd best read
up on good programming practices in general. But you've answered a
good portion of your questions already:

>All codes containing several functions are in one module

Nothing wrong with that -- functions that are likely to be used
together belong in one module -- and while Python supports namespaces
(one honking great idea), pythonic style supports keeping those
namespaces flat (flat is better than nested)

So do you need to do any different stucture? I'd look for two things:

1) the DRY principle (Don't Repeat Yourself) -- if you have
essentially the same code in multiple places, refactor that out into
utility functions, or classes, or...

2) Do classes make sense? Look for patterns like:
    function_a(data_structure_1, some, other, arguments)
    function_b(data_structure_1, some, other, arguments)
    function_c(data_structure_1, some, other, arguments)

that is the C procedural style -- nothing wring with that, but it may
be better suited to a class -- the class holds your data structure and
the functions that operate on it in one place.

"if a class has only two methods, and one of them is __init__ -- you
don't need a class":


also see:


For why you would want to use classes...

> without documentation and testing.

This one is easy: add documentation and testing!

put docstings in your functions and classes -- if the package is big
enough to warrant it, use Sphinx to document it -- it can extract the
docstrings, and give you a nice framework for writing up more
extensive docs.

This sounds like a collection of functions -- you want unit tests --
read up on unit testing (and test-driven development) and the write

I suggest nose or pytest (I prefer pytest, it's got some nice
features, but either is fine). You can also use the stdlib unittest,
but I found it overly wordy and klunky.



Christopher Barker, Ph.D.

Emergency Response Division
NOAA/NOS/OR&R            (206) 526-6959   voice
7600 Sand Point Way NE   (206) 526-6329   fax
Seattle, WA  98115       (206) 526-6317   main reception

Chris.Barker at noaa.gov

More information about the NumPy-Discussion mailing list