Re: [Boost-users] [mpl] Iterator Concept Checking?

Subject: Re: [Boost-users] [mpl] Iterator Concept Checking?
From: David Abrahams (dave_at_[hidden])
Date: 2010-10-30 20:57:19

Hi Stirling,

At Sat, 30 Oct 2010 17:55:35 -0400,
Stirling Westrup wrote:
> 
> I am writing some algorithms to operate on MPL sequences, and I would
> like to use some form of concept checking to ensure that the types
> passed in that are supposed to be iterators are, in fact, iterators.
> 
> The MPL library doesn't seem to have is_XXX_iterator<T> checks, and I
> have run into some difficulties in determining how to write them. The
> MPL documents state that a random_access_iterator is a 'refinement' of
> a bidirectional_iterator which is a 'refinement' of a
> forward_iterator, but its not clear what refinement means in this
> context.

http://www.generic-programming.org/about/intro/concepts.php#refinement

> I initially assumed that this meant that a random_access_iterator_tag
> was derived from a bidirectional_iterator_tag which was derived from a
> forward_iterator_tag. 

That's a useful trick if you're calling runtime functions, but it
wouldn't be useful for MPL as written.

> This does not appear to be the case, so the obvious test for an
> iterator (checking that the class contains a member named 'category'
> that is a type derived from the iterator tag of the least acceptable
> iterator) will not work.
> 
> I am sure I could cobble together something that would work by
> examining the internals of the library, but I would far rather have
> some guidance on the correct way to see if an iterator matches a given
> iterator model, within the API of the library.
> 
> Does anyone have any suggestions?

I'm not sure exactly how to help, because I can't quite tell where
you're stuck, so let me ask a few questions:

0. Which iterator concept do you want to check for?
1. Do you understand the requirements of that concept?
2. Do you understand how to test if those requirements are met?

-- 
Dave Abrahams
BoostPro Computing
http://www.boostpro.com