From: Mathieu Bouchard <matju@...>
Date: 2006-01-14T13:36:16+09:00
Subject: Re: Design contracts and refactoring (was Re: mathn: ugly warnings)

On Sat, 14 Jan 2006, Hugh Sasse wrote:
> On Sat, 14 Jan 2006, mathew wrote:
> > Consider '+'. Requirements are that it add any two things which constitute
> > numbers. Trouble is, you can't even test every possible pair of floating point
> > numbers, let alone the integers and rationals. Any test of N pairs that you do
> > will be statistically insignificant.
> You are not testing number theory, you are testing software.

You are testing "implementation" in general, not software in particular.  
This may include "+". The method for "+" is software, even though it uses
hardware. There's nothing saying that unit-tests can't (or shouldn't) test
hardware.

Actually it's very appropriate to write unit-tests for hardware, and very 
commonplace as well.

> Yes, you must test the brakes.  Yes you must calibrate the speedometer.  
> Software, except the Monte Carlo sort, is not statistical in nature.

You haven't computed a lot of Monte-Carlo integrals, have you? =)

  http://mathworld.wolfram.com/MonteCarloIntegration.html

> If one is expected to write tests for all types a generic method
> will support, then it is making a rod for one's own back.  There
> isn't time.

unit-tests don't have to be as flat as they are usually done. Try to write 
polymorphic tests.

 _ _ __ ___ _____ ________ _____________ _____________________ ...
| Mathieu Bouchard - t�l:+1.514.383.3801 - http://artengine.ca/matju
| Freelance Digital Arts Engineer, Montr�al QC Canada