ant-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Costin Manolache <cmanola...@yahoo.com>
Subject Re: Antlib... when?
Date Tue, 07 Jan 2003 00:49:37 GMT
Conor MacNeill wrote:

> Costin Manolache wrote:
>> 
>> The surprise would be why would a user (ab)use xmlns in such a construct
>> :-)
> 
> In the long run, IMOH, we want to have polymorphism - i.e. passing new
> subtypes to a task or implementations of an interface. So, while today the
> namespace doesn't make much sense (since the task defines what types it
> will accept), in the future, it will be necessary.
> 
>    <path>
>       <foo:fileset dir="." />
>    </path>
> 
> This would be a reasonable construct (once we agree on the polymorphism
> aspects) to pass a new type of fileset to an existing task.
> 
> So, IMHO, namespaces should be considered at all levels, even if the first
> pass may just be to cause an error.

What about postponing the polymorphism to Ant1.7 :-) - and let the ns settle
down a bit ? 

I'm +1 on adding the ns parameter to IntrospectionUtil and even making it 
"pluggable". DynamicConfigurator will be a big problem ( you can't add 
methods to an interface without breaking binary compat - like JDK1.4 does 
in JDBC ).

What will be the pattern for polymorphism ? Right now childs are created
with a createFoo or addFoo. Somehow you need to associate ns:element with
a method in the task. We can support a createFoo( String ns, String name ) - 
but that won't be polymorphism.
Probably addFoo( Intf i ) and some mechanism to support creating classes 
that interface - but it sounds quite complex.

Costin
implement 









--
To unsubscribe, e-mail:   <mailto:ant-dev-unsubscribe@jakarta.apache.org>
For additional commands, e-mail: <mailto:ant-dev-help@jakarta.apache.org>


Mime
View raw message