New style mapscript test

[Howard Butler]

On Aug 10, 2007, at 10:24 AM, Sean Gillies wrote:

> Hi all,
>
> On 1 May, I added Python mapscript tests of loading map objects  
> from mapfile strings. Yesterday Howard expressed his concern that  
> the doctest style of the new test makes the already uncomfortable  
> Python testing framework worse. I disagree.

My concern is that this is the addition of yet another testing method/ 
framework.  Another smaller one is that a doctest that can ingest  
text files is only available by default in Python 2.4 and above.  I  
agree that the doctests are cleaner, but to work with and add to  
them, I have to learn the magic of writing python doctests (I do know  
how to do that, but not everyone does, and others have been adding to  
the Python MapScript tests...).

Port all of the Python MapScript tests to use doctest so we have a  
clean, unified framework?  If someone steps forward to do that, sign  
me up.  On the other hand, introduce yet another testing method that  
provides partial and different test coverage than the established  
existing framework?  Don't complain when few bother to run them...

Don't get me wrong, I like doctest's style much better than what we  
have now in ./swig/python/tests.  I just think that we should go all  
or nothing (or go most of the way so there are well-established  
patterns of how to do things so people can fill in the rest).  We  
already have two ways of running and writing MapServer tests.  Having  
three doesn't improve our situation.

Automated regression testing for MapServer needs a champion (again).   
I don't have the time or motivation to rewrite everything into  
doctests.  I have done my best to keep what we do have running and  
relevant (and so have others).  If someone comes up with a patch that  
rewrites most of the tests in doctest so someone with casual  
familiarity can easily write and run the tests, I would gladly  
support ditching the current x-unit style ones.  Otherwise, I don't  
think a single test of new functionality in a new testing framework  
that our current testing framework doesn't test goes far enough.

Howard