Add a get overload taking a parameter.

[theodelrieu]

Closes #1227.

Takes an lvalue reference, and returns the same reference.

This allows non-default constructible types to be used with get (i.e. without specializing adl_serializer).
This overload does not require CopyConstructible either.

// New value use-case
auto values = j.get<std::vector<std::string>>();

// already created values
std::string str;
j.get(str);

Implements #1227

This can be a step forward to resolve #958. Indeed, we can recommend users to replace implicit conversions from json values by using this new overload. I believe we could start deprecating operator T if we decide to merge this PR.

Note that this new overload returns its parameter, which allows chaining (i.e. a bit like C++17's vector::emplace_back).

---

Pull request checklist

Read the Contribution Guidelines for detailed information.

• Changes are described in the pull request, or an existing issue is referenced.
• The test suite compiles and runs without error.
• Code coverage is 100%. Test cases can be added by editing the test suite.
• The source code is amalgamated; that is, after making changes to the sources in the include/nlohmann directory, run make amalgamate to create the single-header file single_include/nlohmann/json.hpp. The whole process is described here.

[coveralls]

Coverage remained the same at 100.0% when pulling 521fe49 on theodelrieu:feature/get_with_parameter into 680a4ab on nlohmann:develop.

[nlohmann]

The comment seems to end in the middle of the sentence.

[theodelrieu]

Oops, will fix

[theodelrieu]

Should I add support for basic_json instances as well? It's supported for the first get overload.

[nlohmann]

What do you mean?

[theodelrieu]

// one can do the following
nlohmann::json j;
auto v = j.get<my_type>();
auto j2 = j.get<nlohmann::json>();
auto j3 = j.get<my_custom_json>();

Should we support the following:

nlohmann::json j;
nlohmann::json j2;
my_custom_json j3;
j.get(j2);
j.get(j3);

It would be consistent with the non-parameter overload.

[theodelrieu]

I've added support for basic_json types.

I also added tests in unit-conversions.cpp, note that converting to C-arrays is possible with this overload.

However, it will throw when using string_t::value_type []. It's the current behavior with std::array<string_t::value_type, N>. It would be nice to fix both at the same time.

But this is a bit tricky, it requires retrieving the current string, and throw(?) if the string length does not match the C-array/std::array size...

[theodelrieu]

@nlohmann I think it would be good to release a 3.2.1 with #1238 in it before merging this PR. What do you think?

[nlohmann]

@theodelrieu Why not both?

[theodelrieu]

I don't really mind, but since it's a new feature it would be more fit in a minor release.

We can make 3.3 directly though

[nlohmann]

Oh, I see. Maybe I need to better understand the feature. Could you add a snippet for the README?

[theodelrieu]

I knew I forgot something...

[nlohmann]

Sorry to be so late to the party, but now when I see this in the documentation: isn't the name get a bit misleading? It's rather a setter, but to the passed reference. But in particular the example below

j.at("name").get(p.name);

seems counter-intuitive to me.

[gregmarr]

Maybe get_to?

[theodelrieu]

get_to sounds good to me.

[nlohmann]

Yes, could you please change the name to get_to and also add an example to doc/examples?

[theodelrieu]

Done!

[theodelrieu]

Let me know if the example looks fine, I could not upload them to wandbox by running make -C doc

[nlohmann]

Looks good to me.

[nlohmann]

Thanks a lot! Great work as always.