README.md revision 12037:d28054ac6ec9
1 2 3# pybind11 — Seamless operability between C++11 and Python 4 5[](http://pybind11.readthedocs.org/en/master/?badge=master) 6[](http://pybind11.readthedocs.org/en/stable/?badge=stable) 7[](https://gitter.im/pybind/Lobby) 8[](https://travis-ci.org/pybind/pybind11) 9[](https://ci.appveyor.com/project/wjakob/pybind11) 10 11**pybind11** is a lightweight header-only library that exposes C++ types in Python 12and vice versa, mainly to create Python bindings of existing C++ code. Its 13goals and syntax are similar to the excellent 14[Boost.Python](http://www.boost.org/doc/libs/1_58_0/libs/python/doc/) library 15by David Abrahams: to minimize boilerplate code in traditional extension 16modules by inferring type information using compile-time introspection. 17 18The main issue with Boost.Python—and the reason for creating such a similar 19project—is Boost. Boost is an enormously large and complex suite of utility 20libraries that works with almost every C++ compiler in existence. This 21compatibility has its cost: arcane template tricks and workarounds are 22necessary to support the oldest and buggiest of compiler specimens. Now that 23C++11-compatible compilers are widely available, this heavy machinery has 24become an excessively large and unnecessary dependency. 25 26Think of this library as a tiny self-contained version of Boost.Python with 27everything stripped away that isn't relevant for binding generation. Without 28comments, the core header files only require ~4K lines of code and depend on 29Python (2.7 or 3.x, or PyPy2.7 >= 5.7) and the C++ standard library. This 30compact implementation was possible thanks to some of the new C++11 language 31features (specifically: tuples, lambda functions and variadic templates). Since 32its creation, this library has grown beyond Boost.Python in many ways, leading 33to dramatically simpler binding code in many common situations. 34 35Tutorial and reference documentation is provided at 36[http://pybind11.readthedocs.org/en/master](http://pybind11.readthedocs.org/en/master). 37A PDF version of the manual is available 38[here](https://media.readthedocs.org/pdf/pybind11/master/pybind11.pdf). 39 40## Core features 41pybind11 can map the following core C++ features to Python 42 43- Functions accepting and returning custom data structures per value, reference, or pointer 44- Instance methods and static methods 45- Overloaded functions 46- Instance attributes and static attributes 47- Arbitrary exception types 48- Enumerations 49- Callbacks 50- Iterators and ranges 51- Custom operators 52- Single and multiple inheritance 53- STL data structures 54- Iterators and ranges 55- Smart pointers with reference counting like ``std::shared_ptr`` 56- Internal references with correct reference counting 57- C++ classes with virtual (and pure virtual) methods can be extended in Python 58 59## Goodies 60In addition to the core functionality, pybind11 provides some extra goodies: 61 62- Python 2.7, 3.x, and PyPy (PyPy2.7 >= 5.7) are supported with an 63 implementation-agnostic interface. 64 65- It is possible to bind C++11 lambda functions with captured variables. The 66 lambda capture data is stored inside the resulting Python function object. 67 68- pybind11 uses C++11 move constructors and move assignment operators whenever 69 possible to efficiently transfer custom data types. 70 71- It's easy to expose the internal storage of custom data types through 72 Pythons' buffer protocols. This is handy e.g. for fast conversion between 73 C++ matrix classes like Eigen and NumPy without expensive copy operations. 74 75- pybind11 can automatically vectorize functions so that they are transparently 76 applied to all entries of one or more NumPy array arguments. 77 78- Python's slice-based access and assignment operations can be supported with 79 just a few lines of code. 80 81- Everything is contained in just a few header files; there is no need to link 82 against any additional libraries. 83 84- Binaries are generally smaller by a factor of at least 2 compared to 85 equivalent bindings generated by Boost.Python. A recent pybind11 conversion 86 of PyRosetta, an enormous Boost.Python binding project, 87 [reported](http://graylab.jhu.edu/RosettaCon2016/PyRosetta-4.pdf) a binary 88 size reduction of **5.4x** and compile time reduction by **5.8x**. 89 90- When supported by the compiler, two new C++14 features (relaxed constexpr and 91 return value deduction) are used to precompute function signatures at compile 92 time, leading to smaller binaries. 93 94- With little extra effort, C++ types can be pickled and unpickled similar to 95 regular Python objects. 96 97## Supported compilers 98 991. Clang/LLVM 3.3 or newer (for Apple Xcode's clang, this is 5.0.0 or newer) 1002. GCC 4.8 or newer 1013. Microsoft Visual Studio 2015 Update 3 or newer 1024. Intel C++ compiler 16 or newer (15 with a [workaround](https://github.com/pybind/pybind11/issues/276)) 1035. Cygwin/GCC (tested on 2.5.1) 104 105## About 106 107This project was created by [Wenzel Jakob](http://rgl.epfl.ch/people/wjakob). 108Significant features and/or improvements to the code were contributed by 109Jonas Adler, 110Sylvain Corlay, 111Trent Houliston, 112Axel Huebl, 113@hulucc, 114Sergey Lyskov 115Johan Mabille, 116Tomasz Miąsko, 117Dean Moldovan, 118Ben Pritchard, 119Jason Rhinelander, 120Boris Schäling, 121Pim Schellart, 122Ivan Smirnov, and 123Patrick Stewart. 124 125### License 126 127pybind11 is provided under a BSD-style license that can be found in the 128``LICENSE`` file. By using, distributing, or contributing to this project, 129you agree to the terms and conditions of this license. 130
