Your standard library for metaprogramming

This is not an official Boost library. However, a formal review will be asked for shortly. The library is fairly stable, but no effort will be made to stay backward compatible.

#include <boost/hana.hpp>
#include <boost/hana/ext/std/type_traits.hpp>
#include <cassert>
#include <string>
using namespace boost::hana;
using namespace boost::hana::literals;

struct Fish { std::string name; };
struct Cat  { std::string name; };
struct Dog  { std::string name; };

int main() {
  // Sequences capable of holding heterogeneous objects, and algorithms
  // to manipulate them.
  auto animals = make_tuple(Fish{"Nemo"}, Cat{"Garfield"}, Dog{"Snoopy"});
  auto names = transform(animals, [](auto a) {
    return a.name;
  });
  assert(reverse(names) == make_tuple("Snoopy", "Garfield", "Nemo"));

  // No compile-time information is lost: even if `animals` can't be a
  // constant expression because it contains strings, its length is constexpr.
  static_assert(length(animals) == 3u, "");

  // Computations on types can be performed with the same syntax as that of
  // normal C++. Believe it or not, everything is done at compile-time.
  auto animal_types = make_tuple(type<Fish*>, type<Cat&>, type<Dog*>);
  auto animal_ptrs = filter(animal_types, [](auto a) {
    return traits::is_pointer(a);
  });
  static_assert(animal_ptrs == make_tuple(type<Fish*>, type<Dog*>), "");

  // And many other goodies to make your life easier, including:
  // 1. Access to elements in a tuple with a sane syntax.
  static_assert(animal_ptrs[0_c] == type<Fish*>, "");
  static_assert(animal_ptrs[1_c] == type<Dog*>, "");

  // 2. Unroll loops at compile-time without hassle.
  std::string s;
  int_<10>.times([&]{ s += "x"; });
  // equivalent to s += "x"; s += "x"; ... s += "x";

  // 3. Easily check whether an expression is valid.
  //    This is usually achieved with complex SFINAE-based tricks.
  auto has_name = is_valid([](auto&& x) -> decltype(x.name) { });
  static_assert(!has_name(type<int>), "");
  static_assert(has_name(type<Cat>), "");
}

Boost.Hana is a header only library. To use it in your own project, just add the include directory to your compiler's header search path and you are done.

The library relies on a full-featured C++14 compiler and standard library, but nothing else is required. Here is a table of the current C++14 compilers with comments regarding support for Hana:

You can browse the documentation online at http://ldionne.github.io/hana. You can also get an offline version of the documentation by checking out the gh-pages branch. To avoid overwriting the current directory, you can clone the gh-pages branch into a subdirectory like doc/gh-pages:

git clone http://github.com/ldionne/hana --branch=gh-pages doc/gh-pages

After issuing this, doc/gh-pages will contain exactly the same static website that's available online. Note that doc/gh-pages is automatically ignored by git so updating the documentation won't pollute your index.

Setting yourself up to work on Boost.Hana is easy. First, you will need an installation of CMake. Once this is done, you can cd to the root of the project and setup the build directory:

mkdir build
cd build
cmake ..

Usually, you'll want to specify a custom compiler because the system's compiler is too old:

cmake .. -DCMAKE_CXX_COMPILER=/path/to/compiler

On Linux, you'll also need to specify a custom standard library (libc++). You can set the location of your custom libc++ installation as follows:

cmake .. -DLIBCXX_ROOT=/path/to/libc++

You may also want to use a recent version of Boost if you want to use all the adapters for Boost.MPL and Boost.Fusion. You can specify a custom path for your Boost installation if you don't want the system-wide Boost to be used:

cmake .. -DBOOST_ROOT=/path/to/boost

You can now build and run the unit tests and the examples. Assuming you used the Makefile generator with CMake, you can issue

make check

Tip

There is a Makefile at the root of the project which forwards everything to the build directory. This way, you can also issue those commands from the root of the project instead of the build directory.

You should be aware that compiling the unit tests is pretty time and RAM consuming, especially the tests for external adapters. This is due to the fact that Hana's unit tests are very thorough, and also that heterogeneous sequences in other libraries tend to have horrible compile-time performance.

There are also optional targets which are enabled only when the required software is available on your computer. For example, generating the documentation requires Doxygen to be installed. An informative message will be printed during the CMake generation step whenever an optional target is disabled. You can install any missing software and then re-run the CMake generation to update the list of available targets.

Tip

You can use the help target to get a list of all the available targets.

If you want to add unit tests or examples, just add a source file in test/ or example/ and then re-run the CMake generation step so the new source file is known to the build system. Let's suppose the relative path from the root of the project to the new source file is path/to/file.cpp. When you re-run the CMake generation step, a new target named path.to.file will be created, and a test of the same name will also be created. Hence,

make path.to.file # Compiles the unit test associated to path/to/file.cpp
cd build && ctest path.to.file # Runs the corresponding unit test

Tip for Sublime Text users

If you use the provided hana.sublime-project file, you can select the "[Hana] Build current file" build system. When viewing a file to which a target is associated (like a test or an example), you can then compile it by pressing ⌘B, or compile and then run it using ⇧⌘B.

The project is organized in a couple of subdirectories.

• The benchmark directory contains compile-time and runtime benchmarks to make sure the library is as fast as advertised. The benchmark code is written mostly in the form of eRuby templates. The templates are used to generate C++ files which are then compiled while gathering compilation and execution statistics.
• The cmake directory contains various CMake modules needed by the build system.
• The doc directory contains configuration files needed to generate the documentation. The doc/gh-pages subdirectory is automatically ignored by git; you can conveniently store a local copy of the documentation by cloning the gh-pages branch into that directory, as explained above.
• The example directory contains the source code for all the examples of both the tutorial and the reference documentation.
• The include directory contains the library itself, which is header only.
• The test directory contains the source code for all the unit tests.

• Talk on metaprogramming at C++Now 2015
• Talk on Hana at CppCon 2014 (slides)
• The MPL11 library, which is how Hana started out
• Talk on the MPL11 at C++Now 2014 (slides)

I also wrote my bachelor's thesis on a formalization of C++ metaprogramming through category theory and Hana. It is available here, but unfortunately it is in french only. I also presented a talk on this formalization for my bachelor's thesis (french only).

Please see CONTRIBUTING.md.

Please see LICENSE.md.