Are the following initializations supposed to be different?

[LinZhihao-723]

I'm initializing a json object using the following two different alternatives in c++17:

nlohmann::json A = nlohmann::json::parse(metadata_buffer.begin(), metadata_buffer.end()); // Object A
nlohmann::json B{nlohmann::json::parse(metadata_buffer.begin(), metadata_buffer.end())}; //Object B

They are initialized with the same data; however, A differs from B.
Basically, B is equivalent to [A], where everything parsed is considered the first element in the list. Is this behavior expected?

[gregmarr]

Supposed to be, not really, but they are. Unfortunately, it can't easily be fixed. In fact, the behavior varies depending on the compiler, so avoiding B is recommended.

[LinZhihao-723]

So suppose I have a vector of bytes, the correct way to convert them into JSON object is using A? Or is there any more proper way such as directly calling a constructor?

[gregmarr]

Yes, A is the proper form. You can even simplify it a bit:

auto A = nlohmann::json::parse(metadata_buffer.begin(), metadata_buffer.end());