Re: Review: Function library

From: williamkempf_at_[hidden]
Date: 2001-06-07 10:37:03

--- In boost_at_y..., "Peter Dimov" <pdimov_at_m...> wrote:
> From: "John Maddock" <John_Maddock_at_c...>
> 
> > The relevent files can be found at:
> >
> > http://groups.yahoo.com/group/boost/files/Callback/function-05-
09.zip
> > http://groups.yahoo.com/group/boost/files/Callback/function-05-
09.tgz
> 
> Note: it's -19, not -09.
> 
> I think that Function should be accepted.

I second this.  I'm still evaluating and testing this library 
(incorporating it into Boost.Threads actually) still, so I may have 
further comments yet, but initial reaction is that it's a very good 
solution.

> Comments:
> 
> * allocator_test.cpp generates remark 'Access control not 
specified, public
> by default' on
> 
> template<typename T>
> struct counting_allocator : std::allocator<T>
> 
> and similar code under Intel C++ 5.0.
> 
> * There are no tests for the numbered variants, e.g. for function2<>

Since the numbered variants are ultimately used by the unnumbered 
variant I'm not sure that tests are needed.  Examples would be 
useful, however.

Comments of my own:

* For symetry and least surprise it seems that set() should have a 
variant taking a nil_t.  If this is done I question whether or not 
clear() should exist, but it's not a big deal to leave it.

* Also for symetry there should be an operator!=(nil_t).

Bill Kempf