Use the standard to_string() functions for C++11

[Rogier-5]

If compiling according to a C++ version before C++11, then define
std::to_string ourselves.

Add a to_wstring version as well

[est31]

👍

[SmallJoker]

Code looks good. What about moving the functions itos (etc) into the namespace too?

[est31]

no, we shouldn't mess with the std namespace more than we should. @Rogier-5 only added the to_string functions to std because they already exist in c++11.

[nerzhul]

Warning, there is a big difference it our implementation & stdlib implementation. std::to_string throw exception when to_string failed

[Rogier-5]

I'm investigating the failing travis check. The test failures are on OSX only, in test_map_settings_manager.cpp and test_settings.cpp. However, I can't seem to find the specifications of those test cases. Am I overlooking them, or are they hidden somewhere ?

[Rogier-5]

Warning, there is a big difference it our implementation & stdlib implementation. std::to_string throw exception when to_string failed

@nerzhul From what I read, to_string only throws when the std::string constructor throws, which is (probably) only when out of memory. ostringstream::operator<< does not throw indeed, but it sets badbit if it catches an exception itself. Presumably in particular when out-of-memory. Arguably, throwing an exception in such a case is better than failing silently ?

Or am I missing something obvious, or maybe less obvious, here ?

[est31]

Well it doesn't have to be perfect, it only has to work.

[ghost]

So it seems in some implementations, std::to_string adds trailing zeros to float numbers.((float)2.4 is converted to (std::string)"2.40000")

[Rogier-5]

@dlaboratory I checked the c++11 standard, and this behaviour is actually in accordance with the standard... It says that std::to_string(float) converts according to the %f format specifier for sprintf(), which in turn is documented as using a fixed format with exactly 6 decimal digits.

It seams that std::ostream::operator<<() uses the %G format instead. Thumbs up for consistency :-(

Te be continued...

[est31]

Removing my approval.

Seems this PR should be closed? We do want the %G format.

Maybe we could rename the to_string function to something different so that nobody mixes the two up.

[Rogier-5]

The revised patch still replaces to_string() with std::to_string(), but also ensures that ftos() uses the %G format of std::ostream::operator<<().
Still testing...

[Rogier-5]

The problems with the previous version have been fixed, and all tests pass.
Please reconsider your (dis)approval.

[est31]

Ah no, overriding the behaviour of std function templates is not a good way to do things.

[Rogier-5]

@est31 I don't quite understand what you mean. Are you objecting to the overriding of function templates per sé, regardless of where they are, are you objecting because they are in std::, or are you objecting because I might be overriding standard c++ functions ?

Obviously, the functions have to be in std::, because that's where the corresponding c++11 functions are. But they are not standard functions, as the code in question is not for c++11.

If you don't like the overriding of template functions at all, I can define every necessary std::to_string() function individually (not as a template) ?

[est31]

Sorry, seems I've misunderstood something. Its good. 👍

[nerzhul]

Okay for me

[est31]

(taking @nerzhul 's "okay for me" as 👍 )