> -----Original Message-----
> From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On
> Behalf Of Arthur
> Sent: Wednesday, October 19, 2005 8:19 PM
> To: 'Kirby Urner'; 'Mark Engelberg'; [email protected]
> Subject: Re: [Edu-sig] Python as a first language for computer sciencist
> 
> Love disagreeing with Kirby.
> 
> When do we get to play, to hack around.


Testing appropriate to a particular application needs to be thought of
creatively, not as formula. 

My test suite for PyGeo is very much my own, but it is there. And the
feedback - in conformity with the thinking behind the application - is
purely visual.  If I have a point that is supposed to be harmonic to 3
others on a line, I have a test that constructs the harmonic point
"classically", and I test to see that the harmonic calculated algebraically
coincides with it. And so on.  The test suite has in fact gotten fairly
elaborate - I did find it essential once the application had reached a
certain level of complexity - especially since I am constantly refactoring
and may not be able to see up front all the implications of a change.

But it took some time before that was true, and the testing grew organically
with the application.

My point being that I think you need to be pretty deep into an application
before you can understand what kind of testing is most appropriate for it,
and getting at it too early is its own kind of premature optimization.

I am certainly not saying it is not important.  In fact, in my way, quite
the opposite.

Art


_______________________________________________
Edu-sig mailing list
[email protected]
http://mail.python.org/mailman/listinfo/edu-sig

Reply via email to