Re: [groovy-dev] Re: Closure syntax [[ was [groovy-dev] Pragmatic Groovy Development and Project Management ]]

On 26 Jan 2005, at 11:48, Martin C. Martin wrote:

>
>
> Russel Winder wrote:
>> On Tue, 2005-01-25 at 15:51 -0500, Mike Spille wrote:
>> [ . . . ]
>>> If it were up to me, a closure in Groovy would look very much like a 
>>> super-stripped-down anonymous inner class:
>>>
>>>         2.times (def() { println ($it);});
>>>         c = def(params) { body };
>>>
>>> But it's pointless to argue this because I'll be voted down.  So, 
>>> like I said, I could live with:
>> Not necessarily.
>> Given that this example of closure syntax has been such a big issue, I
>> have been thinking about the syntax and my own reactions.  I think I 
>> am
>> changing my ideas about this.  In particular I wonder if the syntax 
>> for
>> a closure (which is after all an unnamed function definition) should 
>> be
>> the same as a function. (OK very Algol68 and not very Ruby but then 
>> this
>> is Groovy 
>
> The simple and common uses of closures resemble statement blocks more 
> than functions, and people relate to them that way.  Compare:
>
> for (x in mylist) {
>    println x
> }
>
> and
>
> mylist.each { x |
>    println x
> }
>
> In fact, I suspect it's easier to explain a closure as a delta of a 
> block than a function, because of the scoping rules: a closure is a 
> block that returns a value, and where "return" plays the role of 
> "break."  The last one causes enough confusion to be in the FAQ.

Agreed.

> As for arguments: in a block, it's the enclosing context that 
> introduces variables (e.g. x in the for loop above), whereas in 
> closures, it's the closure syntax itself that introduces the variable. 
>  When the closure is only used once there's not a lot of difference.
>
> What about changing blocks to be closures?  So that the two examples 
> above compile to the same thing?

Agreed - I the semantics of the language should imply they're the same 
thing, even if the implementation may decide to do different things in 
the bytecode.

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