Noel Welsh wrote:
> Why isn't this test framework compositional?  The spec
> makes explicit mention of this, but doesn't justify it.

I think it's on the proponents of "compositionality" to
justify the need for it.  What expactly are you trying to achieve
with "compositionality", and why aren't functions good enough?

I believe the proposal is "compositional" in that it separates:
(a) test cases;
(b) test-runners;
(c) and framework/api implementation.
These are all independent.

> It feels really...old-fashioned to program in this way, and to
> my mind un-Schemely.

I might argue that using functions is more "Schemely" that creating
lots of "test-case objects".

Before you can argue that "lack of compositionality" is a flaw, I
think you need to provide use cases to illustrate this.  Do you
have an important functionality in mind that would be difficult
to express using the proposal?  You have to balance these against
what I think are my proposal's main advantage: easy-to-write and
compact test-cases, in addition to portability.
--
	--Per Bothner
xxxxxx@bothner.com   http://per.bothner.com/