Re: [Boost-users] algorithms and member functions

From: Simon Buchan (simon_at_[hidden])
Date: 2005-09-22 04:22:56

Foster Gareth wrote:
>>What's wrong with:
>>std::for_each
>>(	as.begin
>>,	as.end()
>>,	bool some_function_keyword(A& a)
>>	{	return a.wee() & SOME_FLAG;
>>	}
>>);
>>
> 
> 
> I think the previous poster (whos comments where appreciated) implied that
> while some languages (he mentioned Lisp and Perl I think, and I mentioned
> Java) could/do support this, C++ would struggle to. It would make things
> hairy for the compiler because a function definition is not preceeded by any
> keyword.
> 
A normal function, no, but I believe (as I should, I wrote it! :) that 
this syntax shouldn't have any problems. The keyword, (for the sake of 
argument, 'inline') in place of where a function name would be in a 
function definition, means an expression returning a type that the 
function call syntax is legal on. (In this case, say, 'bool(A&)', for 
lexically binding functions (where it refers to objects in it's 
enclosing scope), 'inline<bool(A&)>', as it would behave like a functor)
Having statements inside an expression is a New Thing for C++, though. 
What happens if you declare another inline function inside your inline 
function, and return it as the result? (function composition) A lot of 
kick-ass functional programming, thats what!
> I think you can also do things a bit like this in java ... Interesting
> perhaps that there is a keyword involved in this case, would this still bog
> down compilers so much as to outweight the advantages of the sytax style
> (assuming there are any, I'm happy to be argued out of thiking this kind of
> coding is neat).
> 
> button_click_event.set_handler	(
> 		new class : public button_click_event_handler {
> 			void on_click() {
> 				std::cout << "anon_class::on_click() :
> printing wee" << std::endl; }
> 		}
> 	);
> 
Java (recently) has 'Anonymous Classes', which are kind of cool, but 
syntaticly obese for most inline functions... Never used them myself, 
though.
> 
>>I don't know what a good function keyword name would be.
>>Obviously: fn, function, lambda are in common use everywhere
>>(and fn is too short for a C++ keyword! :D)
>>Reuse 'inline'?
> 
> 
> I suppose inline would make sense, but it might be a bit of a hack to have
> it used in two senses. I wonder if you can do this stuff in Delphi, they
> have a function keyword.
> 
Well, so is 'template', 'typename', 'class' (sorta), (famously) 
'static', and soon to be 'auto', so no biggy there. Anyway, inline is 
used far less nowadays, (although not as little as auto :D) since 
compilers are normally better than programmers at picking inline'able 
functions.
> 
> Thesaurus  :)
> 
> "
> function
> 
> noun
> 
>    1. The proper activity of a person or thing: job, purpose, role, task.
> See do/not do.
>    2. A large or important social gathering: affair, celebration, festivity,
> fete, gala, occasion, party, soiree. Informal do. Slang bash. See group,
> work/play.
> 
> verb
> 
>    1. To react in a specified way: act, behave, operate, perform, work. See
> action/inaction.
>    2. To perform a function effectively: go, operate, run, take, work. See
> thrive/fail/exist.
>    3. To perform the duties of another: act, officiate, serve. See do/not
> do, substitute.
> "
> 
> Cheers for the responses anyway,
> 
> Gaz
The last thing C++ needs is another keyword ;)