From: David Abrahams (dave_at_[hidden])
Date: 2002-07-26 06:43:07
David Abrahams * Boost Consulting
dave_at_[hidden] * http://www.boost-consulting.com
----- Original Message -----
From: "Beman Dawes" <bdawes_at_[hidden]>
To: <boost_at_[hidden]>; "boost" <boost_at_[hidden]>;
Sent: Friday, July 26, 2002 7:16 AM
Subject: Re: [boost] Iterator Adaptor Poll
> At 04:46 PM 7/25/2002, David Abrahams wrote:
> >However, enough people have had problems approaching this library that
> >thought it would be useful to ask about what would make a gentle
> >introductory example. Beman recently sent me the enclosed, but it seems
> >define such an oddball toy iterator that I'm not sure it would be a
> >good motivating example.
> ia_experiment.cpp is definitely oddball and definitely a toy. It was
> written to help me understand why I was getting compile errors from a
> Someone else can probably come up with an introductory example which is
> less oddball.
> But I'd like to defend the general principle of "toy" introductory
> When someone knows nothing about a topic, a toy example is the best
> introduction. It is "known good", and can be used to verify that
> compilations using the library work as expected. It can be modified to
> the effect of changes. Because it is a toy, it is totally
> non-threatening. It makes the reader feel superior, rather than like a
> The "C Programming Language" and "C++ Programming Language" books both
> started with a little "hello, world" program, very much a toy. Although
> simple, it starts the ball rolling with something that works and is
> understandable almost every reader.
> In the case of ia_experiment.cpp, it illustrates the relationship between
> the finished iterator, the Base class, and the Policies class with
> clarity IMO than documentation in any form other than C++ code.
Sure, I'm all for toys, as in "this doesn't really do everything I'd want
from a real example, but it shows me how the system works". The problem I
have with ia_experiment.cpp is that my first reaction was "why would anyone
want an "iterator" which does *that*?" The fact that it iterates over only
two elements and stores its value internally is pretty weird. I think most
people who have only seen the standard iterators would have a hard time
imagining that this is also an iterator. Though technically, it is an
iterator, it's so different in nature from most iterators that it seems
like it would be completely foreign.
What about a singly-linked-list iterator example? Wouldn't that show
everything you're illustrating, but more transparently?
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk