On Tue, 26 Feb 2002, Jose Alberto Fernandez wrote: > > I would prefer you write a new <taskdef/> that worked with whatever rules > > you > > wanted to use and used your own adapter rather than growing ants base > > adapter > > > > I agree. Of course on <antlib>'s proposal you just need to add a new role with > a new Adapter and the same TaskContainer interface and presto ;-) > > <role name="start-stop-task" class="o.a.t.ant.TaskContainer" > adapter="my.funny.TaskAdapter" /> > > <start-stop-task name="tomcat-a" class="my.tomcat.Bean" />
Ok, if 2 agree I'll rewrite my code so it can be implemented by external tasks. I can't understand how <role whatever> is more intuitive then what I proposed: <taskdef name="tomcat" class="..." /> <tomcat do="start" /> IMHO having an intuitive syntax is the most important thing, but I'm used with workarounds. I'll commit a different change that makes Project.createTask() use a TaskFactory - and that will allow user code ( i.e. external tasks ) to plug in a different TaskAdapter. BTW, it seems that a pluggable TaskFactory will also solve the ClassLoader problem ( at least for me ) - since the plugin can implement whatever policy it wants. I'm still working on it, I want the new interface to work well with namespaces. Costin P.S.: all this discussion on ClassLoader and TaskAdapter is just another good argument why the so-called 'ant2 as a new codebase' idea is fundamentally wrong - I haven't seen this kind of review and debate going into any of the proposals, and I certainly want to be sure that whatever will be changed in ant2 will get the chance of a similarly rigurous analyse - change by change. -- To unsubscribe, e-mail: <mailto:[EMAIL PROTECTED]> For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>
