How to initialize sf::String with UTF-8-encoded data?

[Tank]

Hey,

is it possible to initialize sf::String directly with UTF-8-encoded data?

[Laurent]

No, you must convert to UTF-32 first and then initialize your sf::String.

[Tank]

It's not a big deal, but is it somehow possible to provide a proper constructor or utility function? I have to do a lot of conversions from and to UTF-32 and UTF-8 and it's a little bit annoying to always use a temporary variable, like:

std::string utf8_source = "...";
std::basic_string<sf::Uint32> tmp;

sf::Uint8::toUtf32( source.begin(), source.end(), std::back_inserter( tmp ) );

sf::String final = tmp; // Avoid me :-(
 

I know I can write a little function doing exactly that, however with an additional constructor it can directly store to m_string of sf::String.

[Tank]

...how about providing functions like in the cute sf::Time?

sf::String string = sf::String::fromUtfX();
string.asUtfX();

[Laurent]

I can provide fromUtfX functions with templated iterators, but asUtfX() functions are more complicated since I have to choose a type to return, which is not obvious.

[Tank]

The fromUtfX with iterators sounds perfect. Issue added https://github.com/SFML/SFML/issues/196

[Laurent]

Thanks

[Nexus]

I can provide fromUtfX functions with templated iterators, but asUtfX() functions are more complicated since I have to choose a type to return, which is not obvious.

Couldn't you just require the template argument explicitly, i.e. string.asUtfX<32>()?

However I wouldn't mix "to" and "as", you already use toAnsiString() and toWideString().

[Laurent]

Couldn't you just require the template argument explicitly, i.e. string.asUtfX<32>()?

I think you misunderstood the point. When calling, for example, string.asUtf8(), what should be returned? Maybe the user will need a std::basic_string<sf::Uint32>, or a QString, or a GString, or even a MyCustomUtf8String, etc. I can handle the first one, but not the others.
However there's a chance that the target type has a generic fromUtf32 function.

However I wouldn't mix "to" and "as", you already use toAnsiString() and toWideString().

True. The inconsistency with sf::Time will have to wait for SFML 3

[Nexus]

Ok. One possibility is to use a function template with an output iterator, analog to fromUtfX(), but with T& instead of const T&. Or you directly use iterator ranges which are more flexible.

Anyway, I think there are more important things to implement in sf::String before, for example a substr() function (issue #21) or a constructor taking iterators (issue #89). The latter provides the base for many other use cases (among which is substr() and maybe also the UTF issue).

[Tank]

#89 got closed, probably Laurent forgot about it?

[Laurent]

Completely

[Tank]

lol I knew it. Will setup a cronjob that reopens all my issues in time.

[Laurent]

I closed it because it was a pull request, a new clean issue should be created (that links to #89).

[Tank]

https://github.com/SFML/SFML/issues/198