Re: [boost] Filesystem i18n mini-review

From: Martin (adrianm_at_[hidden])
Date: 2005-07-13 13:56:43

> Since boost::filesystem::basic_path is trying to stay in sync with 
> std::basic_string, I'd like to at least consider all the overloads for 
> operations provided by std::basic_string.

Don't understand what you mean.

The String is just the internal storage type in the path, similar to 
char/wchar for basic_string. basic_string doesn't have an interface using the 
internal storage except c-strings "const char*".

I think the path interface should always have a basic_string interface 
regardless of the internal storage type.

One solution could be to have a second template argument which specifies the 
interface type.

template <class StorageStringT, class InterfaceStringT = basic_string<typename 
StorageStringT::value_type> >
class path {
  path(const InterfaceStringT& str) { append(str.begin(), str.end()); }
  InterfaceStringT file_string() const { ... }
  const StorageStringT& string() const { return m_path; }
...
};

> Anyhow, the consistent (and sensible) set for path would be:
> 
>      const path &
>      const string &
>      const char *
>      InputIterator, InputIterator

The last one could be confusing since on a path you iterate over path elements 
and not characters.

string s("/abc/def");
path p1(s.begin(), s.end());
path p2(p1.begin(), p1.end()); // should this work?