[pymvpa] dissimilarity stuff

[Michael Hanke]

HI James,

On Sat, Apr 04, 2009 at 01:30:14PM -0400, James M. Hughes wrote:
> Hi all,
> 
> I just pushed into my branch an implementation of some stuff from  
> Kriegeskorte's representational similarity analysis paper.  Well,  
> actually it's a generic (unoptimized) implementation of a  
> dissilimarity matrix, along with a "DSMDatasetMeasure" (i.e.,  
> dissimilarity dataset measure) for using dissimilarity matrices in  
> searchlight algorithms, etc.
> 
> The code's not pretty and might break things, so please have at it!

First of all: Thanks for your contribution! I know it is much easier to
keep code on a private harddisk instead of risking public review -- I
very much appreciate that you have decided to take this path ;-)

So far I have just glanced over the code and would like to ask you for a
few modifications:

1. There is neither an example, nor any unittest for the new
   functionality. If I, or anyone else wants to start polishing your
   code they would be forced to craft a unittest first to ensure that
   any refactoring does not change the intended behavior. However, it is
   much more efficient if the original author provides these tests.

   If you want to attract users to test your code in their analysis the
   typically benefit from a simple example. If possible just make use of
   the example dataset that we ship inside the pymvpa source tarball.
   Ideally the example would generate a figure that is similar to what
   people would recognize from the literature. If you need other/more
   data -- just tell me.

   Both things should be easily doable since most likely you already
   have some test code that just needs to be turned into a proper
   unittest, if you need help with that just push the code.

2. Documentation. You have lots of valueable information in comments
   (e.g. type of input arguments, notes on behavior, ...). In the
   comments it is hardly accessible by users. Please move that
   information inside the docstrings.


Looking forward to see how it goes!


Cheers,

Michael

-- 
GPG key:  1024D/3144BE0F Michael Hanke
http://apsy.gse.uni-magdeburg.de/hanke
ICQ: 48230050