Range-based for over a struct object

01 Sep 2018 in English on C++

One of the things that I miss the most in the C++ is the ability to iterate over every single field in the POD (plain old data) structure. A basic example from the top of my head is when you need to preprocess the data, received with the different byte endianness (big vs little endian).

What would I want to see

C++11 has introduced an extremely useful construction called range-based for loop. It is great if you only care to perform some actions on every element of the container:

std::vector<int> vec { 5, 4, 3, 2, 1, };

for(auto &el: vec)
    Process(el);

The case is why aren’t we allowed to do the following?

struct {
    std::uint32_t fw_version = 0;
    std::uint16_t sector_0_version = 0;
    std::string id = "";
    std::array<std::uint8_t, 6> options{};
} data;

for (auto &el: data)
    Process(el);

Current language status (C++17)

We’re encountering a problem here: latest standard (C++17) doesn’t allow this. Well, if your class is some sort of a container with the values of the same type, you can implement begin() and end(), but it is impossible for the heterogeneous structures like the one above.

Unfortunately, the only way to iterate over every element of the structure is to write it yourself:

struct {
    std::uint32_t fw_version = 0;
    std::uint16_t sector_0_version = 0;
    std::string id = "";
    std::array<std::uint8_t, 6> options{};

    void Process() {
        Process(fw_version);
        Process(sector_0_version);
        Process(id);
        Process(options);
    }
} data;

data.Process();

I think you’ll agree with me that it’s awful. Every time you change the structure you have to remember to update the Process() method.

There is, however, a solution: static reflection. This isn’t (yet) part of the language, but we already have some libraries to perform similar kind of tasks. In my opinion, most promising are magic_get by Antony Polukhin (may be included in the boost as the boost::pfr library) and tinyrefl by Manu Sánchez. Since I wasn’t able to build tinyrefl for the Visual Studio, I’ll focus on the magic_get.

Using the compile-time static reflection

With the magic_get library our code transforms into this:

struct {
    std::uint32_t fw_version = 0;
    std::uint16_t sector_0_version = 0;
    std::string id = "";
    std::array<std::uint8_t, 6> options{};
} data;

boost::pfr::for_each_field(std::forward<decltype(data)>(data), [](auto&& val) { 
    Process(val); 
});

I agree it’s not as pretty as the code I wanted to work but imagine my happiness when I was able to get rid of all the boilerplate code I had to write earlier. And since everything is resolved at the compile-time there is no performance penalty.

If you ever had any similar task, check out this library, it may help you a lot.