On Fri, 2006-01-06 at 12:31, Eivind Eklund wrote:

> First of all, I'd coordinate this with Nathaniel (author of
> Test::Unit, nathaniel / talbott.ws).

I mailed him, but it was only yesterday, so he hasn't had very much time
to reply yet. 

As you may notice, I'm chomping at the bit. I have two other projects in
the pipeline that involve both Ruby and XML. They will all rely heavily
on the XML unit test framework.

I expect to have a lot of free time in which to bury myself in Ruby code
the next two, maybe three weeks. After that, the work will probably slow
down. This is important to me, since the projects I'm working on have
very definite deadlines.

> 
> Second, I'd call it something complete different - a unique name,
> totally off the Test::Unit namespace, until it has lived for a while
> with different people using it.  If it seems like a good idea *then*,
> I'd rename it into Test::Unit::XML (again, after coordinating with
> Nathaniel).

Good advice.

> 
> The delay is to make sure that the design is sound, that you don't
> find a better and incompatible way to test shortly after, and have
> taken the main namespace.  That namespace should be for something
> that's fairly permanent.

Also good advice. In this case, the API will be fairly stable though.
There is only one assertion, assert_xml_equal. The internals may change,
but the method signature won't. I will add more assertions eventually,
but only as the need (my own or other peoples) arise.

The main design decision has been whether to subclass TestCase, or just
mix XML assertions in. Since Test::Unit seems to favor mixins, that's
the approach I'm currently taking. (It is also one of the things I want
to discuss with Nathaniel before I inflict the module on the world.)

/Henrik

--

http://www.henrikmartensson.org/  - Reflections on software development