Re: [boost] metactrl header [was Smart Pointer...]

From: Kevlin Henney (kevlin_at_[hidden])
Date: 2000-07-22 13:08:58

In message <4.3.2.7.2.20000721155055.02350d00_at_[hidden]>, Beman
Dawes <beman_at_[hidden]> writes
[...]
>How about something a bit simpler:
>
>     template<bool Pred> struct meta_assert
>        { int assertion_failure_in_meta_assert[Pred ? 1 : -1]; };
>
>     template<> struct meta_assert<true> {};
>
>This actually works for BCC, MWCW, and MSVC compilers.  The specialization 
>has to be the <true> flavor for some of them.  Steve Cleary introduced me 
>to the negative subscript trick, IIRCC.  Kevlin Henney has used a similar 
>(or maybe the same?) technique in other boost postings.

I've used a bunch of different techniques, but with arrays it has tended
to be 0 rather than a -ve number. Problem with 0 is that some compilers
allow this as an array size :-( So -ve numbers are definitely the way to
go.

>The usage inside a generator (where b is a compile time expression 
>convertible to bool) is:
>
>    meta_assert< b > use_name_as_error_message_surogate;
>
>It could also be used as a base class, but I haven't tested that.

Would perhaps be better if reworked to ensure that it was an empty base
class. Can't remember, did we try statics, typedefs, etc?

>Would this be sufficient for template metaprogramming?

This looks good, and the name works for me.
____________________________________________________________

  Kevlin Henney                   phone:  [phone]
  Curbralan Ltd                   mobile: [phone]
  kevlin_at_[hidden]            fax:    [phone]
____________________________________________________________