Re: [Gzz] Jython tests

[Tuomas Lukka]

On Sat, May 18, 2002 at 06:01:13PM +0200, B. Fallenstein wrote:
> Tuomas Lukka wrote:
> >I think I'd prefer just making test scripts like Perl's... They're
> >a good example of how to make tests simple using character streams.
> 
> Test scripts are what I'd like: creating a new test should be as simple 
> as creating a new file and putting a few statements in it.

Agreed.

> >Each test prints out lines of the form
> >
> >13
> >ok 5
> >not ok 6
> >
> >where the first number is the number of test cases to expect, then the
> >next numbers are in a sequence. All other output is ignored (can be used
> >for debugging).
> 
> This, on the other hand, does not sound very useful to me. I certainly 
> don't want to write 'print "ok 5"' statements in my source code-- and 
> change them every time I add a new test case in the middle or something.

Yes, that's a problem. On the other hand, it's a really simple approach.

> (Forgive my use of Ly-ish metasyntax.) In this scheme, the test engine 
> would load the module and extract the tests by reflection (easy in 
> Python). Then it would call tearDown(), re-load the module for every 
> test, run the test and tearDown() again.

Yes, that could work.

> The point is that we'd be using the module namespace for the test 
> fixture variables, meaning each test function would have instant access 
> to them; and by not placing the setUp() code in a method, we'd avoid 
> having to declare the variables in there as 'global'.
> 
> In fact, I think we could create a wrapper that encapsulates these as a 
> PyUnit class...

Sounds very good.

        Tuomas