Re: Zeus: Plan of Attack and Timeline

[Spencer A Marks <smarks@xxxxxxxxxxxxxxxxx>]

Hi, 

I might be albe to lead a hand in these areas:

> 2. Code I/O Classes
>    * StreamSource
>    * StreamResult
> 
> 3. Code Binding Interfaces/Classes
>    * Container
>    * Property
>    * AtomicProperty
>    * ContainerProperty

Can you supply more detailed requirements so I can better judge? 

> 
> 6. Test/Tune/Refine

I am a big advocate of writing tests as (or even before) I write the
actual class. I've had good success using JUnit (http://www.junit.org)
as a testing framework. The strategy I follow is to write a test class for 
every
concrete class. The test class tests every public method of the
concrete class. This does not provide 100% coverage, but in my
experience as been more than adequate. I have a make target that runs
all the tests, which have to pass, before I check in any code. 

I appreciate the compressed schedule, however, I don't think writing
tests along with classes takes much, if any, longer. I know there are
differing opinions here, but I think the time costs of writing the
tests up front are offset by time that would otherwise be spent in a
protected debugging cycle. This argument excludes the longer term benefits of
having tests upfront because they might not be of immediately advantage
during a rush to get things done. 
(However, longer term advantages include the ability to easily refactor
etc.) Anyway, I'll get the off the XP soapbox now. However, I'd be
happy if this project's guidelines included a quasi requirement of a
test class for every application class. How do people feel about
agreeing to something like this?



Spencer


-----------------------------------------------------------------------------
To unsubscribe from this mailing list, send email to majordomo@xxxxxxxxxxx
with the text "unsubscribe zeus" in the body of the email.
If you have other questions regarding this mailing list, send email to
the list admin at owner-zeus@xxxxxxxxxxxx