Re: [boost] Re: Tuples Library Formal Review

From: Beman Dawes (bdawes_at_[hidden])
Date: 2001-06-23 09:11:46

At 08:53 AM 6/23/2001, Kostya Altukhov wrote:

 >Beman Dawes wrote:
 >
 >> >bind(f, _1, _2);
 >> >one-based indexing is much more intuitive.
 >>
 >>My memory of past discussions in the C++ committee was a strong 
preference
 >
 >>for consistent 0 based indexing even in the cases where it is less
 >>intuitive.
 >>
 >Why than std::pair has 'first' and 'second' instead of 'nullth' and 
'first'
 >(and why tuple of two variables should use different scheme than
 >std::pair)?
 >
 >Why do we have std::bind1st and std::bind2nd instead of std::bind0th
 >and std::bind1st (and I think when we use tuples for binding,
 >the scheme should be consistent with standard binders)?

Very good point!

So the rule actually seems something like:

    For by-value indexes such as selecting an element in a container, use 0 
as the base.  Examples: std::vector or std::string.

    For by-name ordinals such as naming operands, arguments, or structure 
members, use 1 as the base.  Examples: std::pair.first, std::bind1st.

On that basis, bind( f, _1, _2 ) seems the correct choice.  It might be 
worth mentioning the above rules in the rationale.

--Beman