Re: [Jython-dev] Magic instantiation of Runnable

[Samuele P. <ped...@bl...>]

At 00:38 05.03.2004 +0100, Michael Haggerty wrote:
>Wouldn't it be nice if Runnable-like objects could be auto-instantiated
>from callable Jython objects, similar to event listeners?  I'm thinking
>of something like
>
>     def f():
>       # do something here
>
>     EventThread.invokeLater(Runnable(run=f))
>
>or even
>
>     EventThread.invokeLater(run=f)
>
>Obviously it would be easy to write a Jython wrapper class with a
>callable as a constructor argument, but a "magical" solution that would
>apply generally to other similar interfaces would be clearer and save a
>lot of typing.
>
>I couldn't find a discussion of this sort of thing in the archives but I
>can't be the first person to wish it were possible...
>
>Does such a thing sound practicable?  If the experts see no obvious
>problems I might poke around to see if I can figure out how to implement
>this within Jython.

the problem is that the syntax has already a meaning for concrete classes,

for X a concrete Java class

X(**kws) === mk(X,**kws)

where

def mk(cl,**kws):
   x = cl()
   for name,val in kws.items():
     setattr(x,name,val)
   return x

now your proposal is for A an interface (maybe an abstract class) to have:

A(**kws) === subclassnmk(A,**kws)

where

subclassmk(A,**kws):
   class Anonym(A):
     pass
   a = Anonym()
   for name,val in kws.items():
     setattr(a,name,val)
   return a

now one can argue that right now A(**kws) and A() for an interface or 
abstract class are just an error,
but the problem is that the subclassmk semantics would make sense also for 
a concrete class, so IMO the syntax cannot be reused directly because it 
would be confusing.

It's easy enough to write

new = subclassmk

and use it as:

EventThread.invokeLater(new(Runnable,run=f))

and this way it works unambiguously even for A concrete.

regards.