Determine the type from error message.

[MattiasEppler]

I get the erros
"could not find to_json() method in T's namespace"
"could not find from_json() method in T's namespace"

but I do not know for which type the methods are not found. How do I get this information?

[theodelrieu]

I assume you're using Visual Studio, which doesn't show a stack trace after the static_assert. I have to add that for 3.0.

Meanwhile you can modify json.hpp yourself:

// Somewhere in nlohmann::detail
struct from_json_fn {
static void call(T arg) {
  arg.show_me_the_type();
}
};

Do the same inside to_json_fn and that should help you to debug.

[MattiasEppler]

Hi. Yes i use Visual studio. Ok thanks for the information.

Do I have to add this

static void call(T arg) {
arg.show_me_the_type();
}

inside "struct from_json_fn"

or just the arg.show_me_the_type(); to the existing call?

[theodelrieu]

Sorry I was on my phone, so hard to be explicit. You have to add it to the existing call, the one with the static_assert.

[MattiasEppler]

Sorry.

this one:

    template<typename BasicJsonType, typename T>
    void call(const BasicJsonType&, T&, priority_tag<0>) const noexcept
    {
        static_assert(sizeof(BasicJsonType) == 0,
                      "could not find from_json() method in T's namespace");
    }

where to find the arg?

[theodelrieu]

You have to add it:

void call(const BasicJsonType&, T& arg, priority_tag<0>) const noexcept

[MattiasEppler]

Oh... I missed this :-)
ok.. now I get an error, 'show_me_the_type' is not a member of 'MyTpeToSerialize'.
Is this the trick?

[theodelrieu]

Yes it is. I have this in my TODO list, it will be done in 3.0 by default.

[MattiasEppler]

Hm ok. But I do not understand why arg.show_me_the_type(); is only called if the static_assert is false.

[theodelrieu]

In fact, this call overload is only called if you forgot to add a from/to_json method.
That's why the static_assert in this overload does a sizeof(T) == 0 check, that will always fail, this can be seen as the "failure" overload.

[MattiasEppler]

Ok. Thanks for the Explanation.

[nlohmann]

@theodelrieu Thanks a lot for taking care of this issue!