Re: adding dependency injection to groovy (was Re: [groovy-user] grails: Groovy class loader and Spring)

Yeah; its kinda scary messing with 'new'. Though I was wondering if  
we only supported this kind of operation for abstract classes or  
interfaces; but maybe thats too confusing. Maybe a new keyword is  
cleaner...

On 4 Jul 2005, at 13:47, Jeremy Rayner wrote:

>> // lets create an instance of an interface...
>> def f = new Cheese()
>>
>> which is pretty wacky - but what that'd do is try and ask the
>> Injector to make an instance of Cheese - kinda like Spring's
>> BeanContext.getBeanForType(Class) and the similar operation in Pico.
>> Thoughts?
>>
>
> +1 overall. In fact I mentioned this to you last year James 
>
> This would change the rules about 'new' always meaning...
> 1) space allocation for the new instance to hold the fields.
> 2) initialisation of fields to default values.
> 3) calling the appropriate constructor.
>
> And instead...?
> - skipping all three stages and obtaining from pico et all?
> - skipping some of the stages?
>
> How loose would the 'new' contract become?
> - would you always get the same reference back (singleton)
> - does 'new' have to be threadsafe?
> - would this muck up garbage collection algorithms?
>
> should we use a different keyword e.g. make, so that the heap  
> allocation
>  vs dependancy injection usage is obvious...
> e.g.
> def f = make Cheese()
>
> See chapter 12 section 5 of
> http://java.sun.com/docs/books/jls/download/langspec-3.0.pdf for
> discussion of Java on the topic of 'new' execution...
>
> cheers,
> jez.
> -- 
> http://javanicus.com/blog2
>

James
-------
http://radio.weblogs.com/0112098/