From: Dirk Gerrits (dirk_at_[hidden])
Date: 2003-03-03 12:59:56
Robert Klarer wrote:
> There was some discussion among the extensions subgroup of the C++ standard
> committee about the question of how the language might be extended to allow string
> literals to be used as template arguments. That discussion was inconclusive.
> It seems to me that the root of the problem is that C-style strings, and hence
> string literals, are not first-class builtin types in C++. I've been
> experimenting with one possible solution to that problem: I've (partially)
> implemented a "static_string" library and uploaded it to the 'Files' section of
> Boost's Yahoo Group. Here is the abstract from the (also partial) introductory
> "The static_string C++ library is an alternative to both string literals and the
> standard C++ type const std::basic_string<char>. Any operation that can be
> performed upon a const std::basic_string<char> object at runtime can be performed
> by the static_string library. Furthermore, the static_string library implements a
> number of metafunctions that allow these operations to be performed at program
> compile time. The static_string library is significantly more efficient in its use
> of both time and space than const std::basic_string<char>."
> The static_string library allows strings to be represented by types, so that they
> can be used as template arguments.
> The syntax for using static_string is awkward, but a language extension that would
> make a library like static_string convenient to use might be worth considering as
> an alternative to the wholesale introduction of
> string-literals-as-template-arguments to the core language.
> A woefully incomplete version of the static_string library can be found at:
I guess it's not possible to generate the character sequence out of a
string literal with the Boost Preprocessor library? That'd make this a
whole lot more convenient to use.
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk