From: David Abrahams (dave_at_[hidden])
Date: 2003-12-07 15:09:27
Robert Ramey <ramey_at_[hidden]> writes:
> I am working on a problem for which the most natural
> solution is to derive from the boost::transform_iterator.
Deriving from a working iterator type is almost never the most natural
solution to anything, because usually the resulting type will not be
an iterator since its operator++ will have the wrong return type.
> Unfortunately, it does not seem possible to do this.
> The transform iterator does not include a derived
> class in its list of parameters as it would need in
> order to invoke most derived functions.
> Is there any reason why the design prohibits
It doesn't prohibit it; it just doesn't encourage it.
Probably the best reason is that it wasn't an important use case, and
making it easy to build derived iterators would make
transform_iterator harder to use.
> Does anyone have any suggestion on how to get the same effect?
Not sure what effect you want. Can you describe the iterator you're
trying to build?
> I believe that this observation applies to all the other specialized
> adaptors as well.
I'm not convinced it's a design problem yet.
-- Dave Abrahams Boost Consulting www.boost-consulting.com
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk