RE: [Tm4j-developers] Test code

[Kal A. <ka...@te...>]

> >>- test code for package are stored in packages parallel to the packages
> >>to test with
> >>   an test suffix. package a.b.c would have a test package a.b.ctest.
> >>
> >
> > Yes, except that the packages that implement the com.techquila.topicmap
> > interface (com.techquila.topicmap.memory and
> com.techquila.topicmap.ozone)
> > should be tested by the package com.techquila.topicmap.test -
> to validate
> > that the back-end transparency is correct.
> >
> I agree. Test code which tests implementation resides in the implementing
> package while interface test code resides in the interface package.
> you mentioned again the topicmap.test package. The parallel thing is that
> it would be named topicmaptest package. Your strategie would be test code
> in test named subpackages. I stress this a little bit to be sure we are
> talking about the same thing.
>

Oops, yes I missed that. What is the rational for using parallel packages
vs. using nested packages ? I can't see any major difference really and if
there isn't a strong reason for changing to parallel packages I would prefer
not to.

> >
> >>- one additional package as a test suite base. this provides a package
> >>for the main
> >>   test classes and for the classes which are wiring several
> test suites.
> >>
> >>
> >
> > +1 This should include plumbing to enable the selective running
> of tests.
> >
> >
> you mean selective suites? Junit tests are always selectively runnable.
>

Yep, I meant selective suites.

> >>Is this ok?
> >>
> >>
> >
> > Looks really good to me!
> >
> Ok, I will think about the test stuff. I'm trying to create some fixtures
> for general testing.
>

Cool!

> I will start with the simple tests of creating and initializing. This will
> lead to a basic test structure. Kal, when it comes to test the
> processing model
> i will need a lot of your help.
>

No problem. I'll try and be as responsive as I can, although I am going to
be mostly offline from tomorrow until Monday.

> >
> >>What about the package names of the API. Are there any plans to move the
> >>package
> >>naming from com.techquila to org.tm4j ? It isn't really important at the
> >>moment but some things should be clear to everyone who is
> >>contributing code.
> >>This includes the license thing.
> >>
> >>
> >
> > Indeed - I only recently acquired the tm4j.org domain. I've
> been planning
> > that the 0.5 release should be the last release with packages in the
> > techquila.com domain.
> >
> Ok, that sounds very good.
>
> > The license text was changed for the 0.5 release to specify The
> TM4J Project
> > as copyright owner. I would like to add a contributors list to the
> > distribution starting with the next release. Is there something
> more that
> > should be done ?
> >
>
> As long as you are managing the code repository you should declare how the
> code is to be updated. How can someone contribute code? I guess
> you and Gerd
> have write access to the cvs.
> This includes a group policy for deciding who is getting access to cvs
> and when.
> I think there is a lot of bureaucratical things to do. But before
> we really
> start doing more work this seems a little bit of unnecessary overhead.
>

I agree. I had a look at the way that the Apache Project manage this and I
think that in the long run some charter like that of the Apache Software
Foundation would be a useful thing to have.

Right now we have both yourself and Florian willing to add new code to the
TM4J codebase. If Gerd is OK with this I would be happy to grant write
access to both of you - which then gives us a reasonable sized team of
"committers" who can introduce code from other contributors into the
repository. We can then work out a process by which someone can become a
committer at some later date (I believe that a process for this is detailed
in the ASF charter...)

How does that sound ?

Cheers,

Kal