Re: [boost] Re: [iterator facade] Why is dereference const?

From: Vladimir Prus (ghost_at_[hidden])
Date: 2004-05-14 01:27:55

Neal D. Becker wrote:
> Thanks for the advice.  I think I have it now.
>
> Problem:
>
>   A function F accepts a pair of stl-style iterators as input.  A nullary
> functor G can generate an output.  Arrange for F to call G n times without
> storing to an intermediate sequence.
>
> Example: G is a random number generator.
>
> I wonder if there is a more elegant approach?  

I have a couple of questions

1. Is this really forward_traversal_iterator? E.g. if your function extracts 
data from stream, then you can't store copies of iterators and make a second 
pass over the data.

2. Why did you declared 'advance'. IIRC, this is not required from forward 
iterators.

As for improving this, I also have two ideas:

1. Some time ago I've created an utility class for creating input iterators, 
called "eof_iterator". Reimplementing function_input_iterator with it yeilds 
somewhat smaller code, see:

    http://zigzag.cs.msu.su:7813/iterators/function_input_iterator2.hpp

the eof iterator itself is at:

    http://zigzag.cs.msu.su:7813/iterators/eof_iterator.hpp

There are some drawbacks, though: the size of iterator will be bigger.

2. It's possible to create iterator which just calls functional object 
(without counting). As separate "until_iterator" would take any iterator and 
a functional object and iterate until the functional object returns true.
So, your iterator will be implemented like:

    typedef function_input_iterator<SomeFunc> it;
    class CountTo { ... };
    typedef until_iterator<it, CountTo> uit;
    copy(uit(it(some_func), CountTo(10)), uit(), ........)

I actually believe that such until_iterator can be rather usefull sometimes.

- Volodya