[picocontainer-dev] Parameter names question
Paul Hammant <[email protected]>
| Newsgroups | gmane.comp.java.picocontainer.devel |
|---|---|
| Message-ID | <[email protected]> |
Folks,
See http://picocontainer.org/component-configuration.html ... section
called "Relying on parameter names" and its example - BigFatComponent
We've debated the need for the dummy components before. It troubles
me too there is not a more elegant design to prevent accidental
population of a type with something that's not for it. See
initForConfig() in DefaultPicoContainer , which it might be best for
us to dispense with.
What if we had two modes of operation, like so:
Mode 1 - Components are injected by type first, where there are more
than one type that could satisfy a ctor argument, then the parameter
name can be used to select the appropriate thing to inject.
Mode 2 - Components are injected by parameter name first, and where
there is nothing so-named that could be injected, then the parameter
will be resolved by type.
I can kinda see that working. I'm wondering how the two modes are
triggered..
Should we have a setMode(..) on DPC ? With a better name of course.
Or should we have a different container - NamedParameterPicoContainer
or NameFirstPicoContainer (yeesh)
Or should we somehow set a name-first preference for a component on
registration ..
pico.as(NAMED_PARAMS).addComponent(BigFatComponent.class)
Hmmm, now that I've typed it, that last looks attractive.
Thoughts?
- Paul
---------------------------------------------------------------------
To unsubscribe from this list please visit:
http://xircles.codehaus.org/manage_email