Implicit assignment to std::string fails

[Cleroth]

Put simply, when constructing a string, you can pass a basic_json:

std::string s = o["name"];

However, this does not work:

std::string s;
s = o["name"];
s = o["name"].get<std::string>(); // this works

The following compiler errors pop up:

1>g:\files\smoldy_adventure\src\client\units\unitmgr.cpp(53): error C2593: 'operator =' is ambiguous
1> g:\program files (x86)\microsoft visual studio 14.0\vc\include\xstring(1032): note: could be 'std::basic_string<char,std::char_traits,std::allocator> &std::basic_string<char,std::char_traits,std::allocator>::operator =(_Elem)'
1> with
1> [
1> _Elem=char
1> ] (compiling source file units\unitmgr.cpp)
1> g:\program files (x86)\microsoft visual studio 14.0\vc\include\xstring(1027): note: or 'std::basic_string<char,std::char_traits,std::allocator> &std::basic_string<char,std::char_traits,std::allocator>::operator =(const _Elem *)'
1> with
1> [
1> _Elem=char
1> ] (compiling source file units\unitmgr.cpp)
1> g:\program files (x86)\microsoft visual studio 14.0\vc\include\xstring(1011): note: or 'std::basic_string<char,std::char_traits,std::allocator> &std::basic_string<char,std::char_traits,std::allocator>::operator =(const std::basic_string<char,std::char_traits,std::allocator> &)' (compiling source file units\unitmgr.cpp)
1> g:\program files (x86)\microsoft visual studio 14.0\vc\include\xstring(972): note: or 'std::basic_string<char,std::char_traits,std::allocator> &std::basic_string<char,std::char_traits,std::allocator>::operator =(std::initializer_list<_Elem>)'
1> with
1> [
1> _Elem=char
1> ] (compiling source file units\unitmgr.cpp)
1> g:\program files (x86)\microsoft visual studio 14.0\vc\include\xstring(914): note: or 'std::basic_string<char,std::char_traits,std::allocator> &std::basic_string<char,std::char_traits,std::allocator>::operator =(std::basic_string<char,std::char_traits,std::allocator> &&) noexcept' (compiling source file units\unitmgr.cpp)
1> g:\files\smoldy_adventure\src\client\units\unitmgr.cpp(53): note: while trying to match the argument list '(std::string, stlx::basic_jsonstd::map,std::vector,std::string,bool,int64_t,double,std::allocator)'

Compiler: Visual Studio 2015

[nlohmann]

Thanks for reporting! I'll check the issue soon.

[nlohmann]

Unfortunately, I cannot reproduce the error with GCC or Clang. As I have no access to MSVC, I cannot dig deeper into the error. Do you have any idea what the reason could be that MSVC rejects the code?

[Cleroth]

I couldn't really find the reason. I'll check again in a bit.
Meanwhile, VS 2015 is free, and if you don't want the IDE you can get just the compiler: http://blogs.msdn.com/b/vcblog/archive/2015/11/02/announcing-visual-c-build-tools-2015-standalone-c-tools-for-build-environments.aspx

[nlohmann]

I am not using a Windows system, so trying MSVC is still difficult :-)

[ZahlGraf]

I can confirm this issue with MSVC 2015. The workaround with get<std::string>() works well for me.

[Cleroth]

Why?

[nlohmann]

I am unable to debug MSVC error messages. If you have any idea how to fix it, I would be happy.

[Cleroth]

The problem is that basic_json can convert to both a char and a std::string, so MSVC doesn't know which function to call. The only solution I've found is to prevent implicit conversion to char (ie. by adding typename = std::enable_if<!std::is_same<ValueType, char>::value>::type to operator ValueType()).
You could probably easily replicate this by creating a dummy class with two operator= templates, one which takes a char, and the other a std::string.
I'm curious as to how this works in GCC/Clang... How are they implementing the std::string::operator=?

[Furcube]

GCC 5.3.0 on linux produces similar error. Need to use explicit conversion.

[Cleroth]

Clang 3.7 also reports the error.

[nlohmann]

I fixed the code (see 457bfc2) and added test cases. The code compiles for GCC and Clang, but AppVeyor (https://ci.appveyor.com/project/nlohmann/json) fails to compile the code with MSVC. Does anyone have an idea what could be the issue?

[erichkeane]

@nlohmann : I checked out master with a VM, and get the same issue on Master, so I would say that the issue is NOT with the proposed patch, but with master itself.

[Cleroth]

The patch is on master, I believe.

[erichkeane]

@Cleroth: I mean the latest issue that he's having in AppVeyor. Even without the std::string::operator= stuff, I get the issue of an ambigious operator <<.

I filed bug #167 to comprehend the issue that is blocking this currently.

[twelsby]

This is fixed in pull #183 - see issue #167.

[nlohmann]

@twelsby, do I understand correctly that this issue was solved by PR #188?

[twelsby]

@nlohmann. yes. Actually it was fixed in 457bfc2 but that created side effects that were fixed in PR #188. As that pull has now been merged, the current situation on master is that this is fixed. You should be able to close this issue.

[nlohmann]

Thanks!

[thomasdullien]

Hey there,

it seems the issue is cropping up for me again. I am trying to compile the same project on GCC and on Clang (built from current trunk), and the unit test compilation is failing for issue 144:

/XXX/project-zero/thomasdullien/XXX/json-src/test/src/unit-regression.cpp:250:12: error:
use of overloaded operator '=' is ambiguous (with operand types 'std::string' (aka 'basic_string<char,
char_traits, allocator >') and 'value_type' (aka 'nlohmann::basic_json<std::map, std::vector,
std::__1::basic_string, bool, long, unsigned long, double, std::allocator>'))
s2 = o["name"];
~~ ^ ~~~~~~~~~
/usr/local/bin/../include/c++/v1/string:825:19: note: candidate function
basic_string& operator=(basic_string&& __str)
^
/usr/local/bin/../include/c++/v1/string:820:19: note: candidate function
basic_string& operator=(const basic_string& __str);
^
/usr/local/bin/../include/c++/v1/string:822:19: note: candidate function
basic_string& operator=(__self_view __sv) {return assign(__sv);}
^
1 error generated.
make[2]: *** [json-build/test/CMakeFiles/json_unit.dir/src/unit-regression.cpp.o] Error 1
make[1]: *** [json-build/test/CMakeFiles/json_unit.dir/all] Error 2
make: *** [all] Error 2

What's the best way to help fixing this? :) (btw, big fan of your library)

[nlohmann]

That is strange, because this regression test is run with every commit on AppVeyor, see https://ci.appveyor.com/project/nlohmann/json/history.

Edit: Of course, I meant Travis, see https://travis-ci.org/nlohmann/json/branches

[jaredgrubb]

Maybe it's a C++17 thing? The "self view" sounds a bit like the new string view.

[thomasdullien]

Hey there, so I downgraded to Clang 3.5, and it all works. I think we should not worry about it for now - it may be a transitory issue caused by me using trunk clang. Cheers, Thomas 2016-12-16 9:59 GMT+01:00 Jared Grubb <notifications@github.com>:

…

Maybe it's a C++17 thing? The "self view" sounds a bit like the new string view. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#144 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AEYBvBlDVClNhTaBZ6FhWtHvVLBm-9RAks5rIlLmgaJpZM4GZ3Rq> .