Add support for other character types in filenames (fixes #6)

[MHebes]

Allows unicode filenames on windows, which uses UTF-16.

[MHebes]

Hi @sreiter, incorporated your requested changes from #6 (comment)

[sreiter]

Thanks so much for adjusting the template parameter name!
I just saw that you removed a method from the public interface. I think that's a bit unfortunate, since some users may already rely on that method. It would be great if you would consider my comment and reintroduce it - ideally togehter with an additional wstring overload.
Thank's a lot!

[sreiter]

Sorry, I didn't see this last time. But it seems that you removed this method? I think it is in the public interface, so some users may rely on it. Unless there's a good reason to remove it, I would thus rather keep it. And if we have an std::string overload, I think it would even make sense to add a second overload with const std::wstring&, as you did in the constructor...

[MHebes]

Hi @sreiter, sorry for the delay in getting this PR updated.

Since it was originally made, I realized that these two macros are preventing this from working perfectly:

  #define STL_READER_THROW(msg) {std::stringstream ss; ss << msg; throw(std::runtime_error(ss.str()));}
  #define STL_READER_COND_THROW(cond, msg)  if(cond){std::stringstream ss; ss << msg; throw(std::runtime_error(ss.str()));}

These are problematic with wchar_t e.g.

          STL_READER_THROW("ERROR while reading from " << /* const wchar_t* */ filename <<
            ": vertex not specified correctly in line " << lineCount);

where if filename was a const wchar_t*, it would print as a pointer value instead of a string in <C++20.

In C++20, they actually deleted operator<<(basic_ostream<char, Traits>&, wchar_t) so this would actually become a compiler error with this PR when TChar = wchar_t.

Options:

1. Rely on a macro to manually switch types around (UNICODE, _UNICODE, or STL_READER_USE_WCHAR) or use something like WideCharToMultiByte) that converts to UTF8 when needed
2. Use C++17's filesystem::path, which automatically accepts either character type and provides a operator<< for stringstream, or
3. Drop this PR, I'm perfectly fine with just maintaining a fork.

Please let me know how you'd like to proceed!

[sreiter]

Hi @MHebes, great to hear from you. Maybe a small overloaded function which returns the utf8 string for wchar__t and the plain string for the classic char would be simple enough. I will try to look into it to reproduce the issue.
Thanks 😃