On Wed, 11 Jun 2003 09:24 pm, Jose Alberto Fernandez wrote: > Connor, > > Can you expand on why DynamicConfigurator wasn't an option for > your ScriptBaseDef class?
Actually I think it could (should) be used. I thought originally that since createElement was called at config time, I would need to know the actual class to be returned then and there. Since I was getting the script definition at execute time, the info would not be available at config-time. That reasoning is, however, bogus. The tasktype has been set before config takes place so the taskdefinition and type lookup could occur in the createElement method. It would simply be a matter or moving the code around a little and storing the elementInstances map in the ScriptDefBase class. Actually I think I should make this change as it is clearer than using the TaskContainer interface to prevent configuration. > I would love to be able to write scripting > tasks as close as the java as possible, that way one could use > scripting for prototyping and eventualy move it into classes > once the task is defined enough. (I am assuming BSH here). This is possible for javascript and, I guess, BSH. Is it possible for all BSF supported scripting languages - I don't know? You'll remember the code in Mutant worked this way. Actually the current <scriptdef> implementation does not use the setXXX style due in part to the feedback to that proposal. Check out the full thread: http://marc.theaimsgroup.com/?l=ant-dev&m=101357395019115&w=2 <scriptdef> could work that way, perhaps as an option but I think perhaps Dawid's approach is more worthwhile for task prototyping since, as Peter demonstrated, (http://marc.theaimsgroup.com/?l=ant-user&m=105523577505707&w=2) it could be used to implement other components such as selectors. Pretty cool. So perhaps Dawid's DynamicBSH thing for prototyping and <scriptdef> for simple scripts where you don't want all the setXXX machinery, which can be tedious. > > Do you provide a "src" attribute on <scriptdef/> so that the code > can be on a file? That could be good for large code. Good idea. I'll try to get around to it, and also a classpath for the classname approach to nested element creation (or just scrap that concept completely and just rely on type names). Conor --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
