Re: [Asunit-users] FlexUnit4 features and AsUnit direction

[Robert P. <in...@ro...>]

Luke,

Thanks for this investigation; I wasn't aware of all those Flex
dependencies. I agree with your overall assessment and recommendations.

I've worked with Hamcrest a fair bit and I would love to work on integrating
it into AsUnit. I'll start another thread to discuss how to do that.

Robert


On Sun, Jun 14, 2009 at 3:18 PM, Luke Bayes <lb...@pa...> wrote:

> I did some digging this weekend into FlexUnit 4 sources, implementations
> and metadata / annotations in MXMLC.
>
> One major problem I have with metadata tags is that they can only be
> discovered at runtime for public members. I imagine there are hooks in the
> compiler that make [Bindable] and [Embed] work even for
> private/protected/namespaced members, but getting into runtime, custom tags
> means we have access only to public members that are visible to
> describeType.
>
> This is a little annoying to me, because I value having private / protected
> features for set up and tear down, as well as test helpers with custom
> domain-specific assertions. It feels awkward to expose these methods as
> public everywhere.
>
> I'm also annoyed with metadata in that we don't have any reliable
> information about the 'order of declaration' which leads to ugly duplication
> like [Before(order=1)], [Before(order=2)], etc.
>
> Then there are of course the issues Robert uncovered around supporting
> metadata in Authoring CS3, which pretty much means that today - MXMLC is the
> only compiler for which these tags have value.
>
> The worst news is that the FlexUnit 3 implementation (excluding the
> printers) had pretty limited Flex framework dependencies, all of which would
> have been trivial to rip out. Unfortunately, FlexUnit 4 has tightly
> integrated the AsyncToken garbage along with Flex collections and Logging.
> Fluint (which is now also bundled) has extensive framework dependencies that
> are unlikely to be extricated without a major overhaul...
>
> My conclusions at this point are as follows:
>
> FlexUnit is now appropriately named and works only with projects that don't
> mind executing tests within a Flex application SWF.
>
> I'm not sure I really understand the true value of annotations when
> compared to inheritance... We could certainly gain many of the benefits
> (multiple setup/teardown, class-level setup/teardown, and custom assertions
> using inheritance and minor modifications to our existing code.
>
> I perceive the real value of the latest release of FlexUnit to actually be
> the following:
>
> 1) Fluint seems to do a great job of describing and handling asynchronous
> event sequences
>
> 2) Hamcrest seems to be helpful for mocking and stubbing features
>
> 3) Generators and test output integrated into Eclipse
>
> It might be a worthwhile endeavor for us to do the following:
>
> 1) Improve how we deal with asynchronous event chains using Fluint as a
> design guide.
>
> 2) Consider integrating Hamcrest, or at least look more closely at how it
> may improve our own recommended workflow.
>
> 3) Investigate writing an Eclipse plugin that lets us generate classes and
> test cases more easily within the IDE, and write a new printer/runner that
> emits results into the Flash/Flex Builder test output. This could actually
> be a separate project that anyone could contribute.
>
> I'd welcome any input on these ideas and concede that I may well be missing
> something critical in any of the above points.
>
> Anyone out there agree or disagree?
>
>
> Thanks,
>
> Luke
>
>
>
> ------------------------------------------------------------------------------
> Crystal Reports - New Free Runtime and 30 Day Trial
> Check out the new simplified licensing option that enables unlimited
> royalty-free distribution of the report engine for externally facing
> server and web deployment.
> http://p.sf.net/sfu/businessobjects
> _______________________________________________
> Asunit-users mailing list
> Asu...@li...
> https://lists.sourceforge.net/lists/listinfo/asunit-users
>
>