[boost] Re: Formal Review: FOREACH macro

From: Edward Diener (eddielee_at_[hidden])
Date: 2005-05-02 19:34:16

Andrei Alexandrescu (See Website For Email) wrote:
> Paul Mensonides wrote:
> 
>> I'll reiterate something that you said a while back (paraphrasing).  The
>> greatest harm that the macro mechanism (in C/C++) has caused is that 
>> it has kept
>> people from realizing the benefits of syntactic abstraction.
>>
>> Furthermore, whether something is or isn't a macro is irrelevant.  It 
>> is whether
>> it provides a useful abstraction that is documentable.  As a macro, 
>> the name is
>> not going to collide with anything because of all-caps and the BOOST_ 
>> prefix.
>> Similarly, it isn't going to accidentally expand in any code.  As I've 
>> said
>> before (paraphrasing), macro invocations are nothing like function 
>> calls except
>> in appearance, and nearly all other macro-related problems stem from 
>> developers
>> viewing them as if they are.  Macros' "return values" are code 
>> snippets, not
>> values.  E.g.
>>
>> #define max(a, b) ((a) < (b) ? (b) : (a))
>>
>> The problem here is not that 'max' has a problem with multiple 
>> evaluation; the
>> problem is that typical developers view 'max' as returning either 'a' 
>> or 'b'.
>> In other words, whether an abstraction is a macro is an important 
>> documentation
>> point, and the documentation of an abstraction, regardless of what it 
>> is (macro,
>> function, class, etc.), dictates the entirety of how it is to be used.
> 
> 
> Agreed. In the case of max, it's the syntactic similarity with a 
> function call that lures people into making certain assumptions. Same 
> goes about non-macro entities such as auto_ptr - "a = b;" looks like 
> something that ain't.
> 
> By the way, it was FOREACH that inspired me to figure out a simple 
> solution to max, which then Eric improved. It's basically like this:
> 
> #define max(a, b) (max_fn(true ? (a) : (b), false ? (a) : (b)))
> 
> which presto, nice-o passes the appropriate types to a template max_fn 
> function that now can be smart about returning lvalues and so on.

You have stumped me here. Why does one not do:

#define max(a, b) (max_fn((a),(b))

instead ?