[Twisted-Python] General question on components, interfaces, etc.

glyph at divmod.com glyph at divmod.com
Thu Jun 15 13:17:01 MDT 2006


On Thu, 15 Jun 2006 17:24:10 +0200, Terry Jones <terry at jon.es> wrote:

>All that makes sense, but I'm not sure what it really buys me. An
>alternative would just be to document the expected interface (perhaps as an
>Interface), and then use __import__() to load the desired implementation.

That's basically what the plugin system does.  At 156 SLOC, it doesn't have room to do much else :).

>It seems that without needing adapters, the component architecture doesn't
>really buy me anything except a formalism (which I do think has value,
>though limited).

It allows you to use the same language to talk about plugins as you would about any other kind of adapter.  Effectively, the plugin system is a way to adapt your "system environment" to an object which provides a given interface.  If you use the twisted plugin system (or some other system which uses Interface objects) then it is pretty easy to factor your code such that implementations of the desired interface can be obtained implicitly through the environment or explicitly by user code passing in an implementation -- or by passing in an object that is adaptable to that interface, and then doing adaptation.

As Jim Fulton is fond of saying, the whole point of the components system is to allow you to reason about the requirements are for a given bit of code.  It also lets you translate that reasoning to code directly.  "argument x must be adaptable to IFoo" is a nice succinct way of dictating your requirements both in documentation (by giving a link to the fully-qualified name of IFoo which includes lots more documentation) and in code (by adapting to IFoo, or looking for IFoo plugins, at runtime).

>Is this thinking more or less correct? How would people go about doing what
>I outline above?

Mostly yes.  Obviously, I'd personally use twisted.plugin, because that's what I wrote it for :).




More information about the Twisted-Python mailing list