[plt-dev] PLT sickness

I think this is what Robby meant, but I don't see that much effort
duplicated between FastTest and Redex.

Yes, both let you generate random numbers, strings, and symbols, but
the emphasis is different in each case. FastTest gives you a ton of
control over the distribution of these values because it's a general
purpose library, and sometimes you really want your nats from a
Poisson distribution. In contrast, number generation is almost an
afterthought in Redex. Constants like numbers and strings are the
boring part of a Redex model, so Redex's test generation isn't very
creative when choosing them. Redex devotes maybe 75 LOC to this task,
and that's probably far too many -- I bet we'd do just as well with an
even stupider strategy, given the particular domain.

The majority of Redex's test generation code is devoted to deriving
generators from grammars. Robby has used Redex's test generator to
test 2htdp/image, so maybe there's a place for something like this in
a more general library like FastTest, but no one has wasted energy
duplicating this work yet.

On Wed, Jan 27, 2010 at 4:26 PM, Robby Findler
<robby at eecs.northwestern.edu> wrote:
> Carl's random testing framework and Casey's work in redex are largely
> complementary. And I've been banging my head against the testing
> framework one as you well know.
>
> Robby
>
> On Wednesday, January 27, 2010, Matthias Felleisen <matthias at ccs.neu.edu> wrote:
>>
>> After quickly reading through the 'random testing' thread, I am once again reminded that PLT suffers from many duplications of efforts:
>>
>>  -- unit testing frameworks
>>  -- random testing frameworks
>>
>> and probably more. Why?_________________________________________________
>>  For list-related administrative tasks:
>>  http://list.cs.brown.edu/mailman/listinfo/plt-dev
>>
> _________________________________________________
>  For list-related administrative tasks:
>  http://list.cs.brown.edu/mailman/listinfo/plt-dev
>
>