jastrachan | 18 Nov 10:58 2004
Picon

Re: Re: def, any, var, Object


On 17 Nov 2004, at 06:15, spullara.1684899@... wrote:
> I really
> don't think anything in this language should be decided because of 
> possible
> optimizations that may be done at some time.  Look at all the trouble 
> that
> got us into in java with int/float/etc.

Agreed.

Though for me the primary issue here is should we allow the compile to 
catch more typeos by using statically typed (compile time) checking. 
This is the main driver here; to reuse some static type information to 
provide better error messages. Increasingly I can see folks using some 
static type information to (i) help integration with Java code and (ii) 
help IDEs do completion/refactoring.

So it seems silly not to use this to help improve the error messages we 
give to users. There's also a side effect which is we can generate much 
more efficient bytecode & use much less RAM (e.g. no need to use the 
MetaClass stuff quite so much). Though the primary driver is better 
compile time warnings. However we should view it as such - and try and 
maintain the exact same method invocation semantics irrespective of 
whether static or dynamic typing is used.

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


Gmane