Re: [boost] Why shadowing?

From: David Abrahams (dave_at_[hidden])
Date: 2005-06-27 08:24:04

Haakon Humberset <humbe_at_[hidden]> writes:

> Hi..
>
> I've introduced boost as a utility libraries for the projects I work on.
> However, all of the people working on these projects compile code using
> the -Wshadow option. This generates lots of pages with warnings that
> clutter their compilation and makes it very hard for them to see what the
> real warnings and errors are in our code.
>
> I don't mind shadowing in the cpp compiled in boost, but shadowing in the
> header files are bad since these will show up for every piece of code we
> compile that uses boost.
>
> Consequently, I'm altering our boost library to remove the shadowing
> warnings we see.. This is a tad boring, but it works.. Only trouble is
> that when we at some point want a newer version of boost, we have to do it
> all over again. Any chance boost can be compiled with shadow warnings,
> making the boost developers not use shadowing?
>
> I don't see any reason to actually use shadowing. Having a local variable
> and a class member with the same name is just ground for confusion and
> subtle errors. From the code I've seen fixing names of local variables, I
> see that many variables have had an underscore added to the name to
> prevent shadowing already, why not use this consistently?

The use of member function shadowing (not involving variables at all)
is a fundamental design technique in the Boost Iterator library
(e.g. iterator_facade et. al.).

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com