Installation

C++ library

Gemmi was a header-only library until ver. 0.6.0. Some parts, such as CIF syntax parsing, remain header-only. If you happen to use only these parts, just ensure that Gemmi’s include directory is in your project’s include path. For example:

git clone https://github.com/project-gemmi/gemmi.git
c++ -Igemmi/include -O2 my_program.cpp

However, in most cases, you need to build the gemmi_cpp library and link your project against it.

If you use CMake, you may:

  • first install gemmi and then use find_package:

    find_package(gemmi 0.7.0 CONFIG REQUIRED)

  • or add gemmi as a git submodule and use add_subdirectory:

    add_subdirectory(gemmi EXCLUDE_FROM_ALL)

  • or use FetchContent:

    include(FetchContent)
FetchContent_Declare(
  gemmi
  GIT_REPOSITORY https://github.com/project-gemmi/gemmi.git
  GIT_TAG        ...
)
FetchContent_GetProperties(gemmi)
if (NOT gemmi_POPULATED)
  FetchContent_Populate(gemmi)
  add_subdirectory(${gemmi_SOURCE_DIR} ${gemmi_BINARY_DIR} EXCLUDE_FROM_ALL)
endif()

Then link your target with the library (this also takes care of includes):

target_link_libraries(example PRIVATE gemmi::gemmi_cpp)

If a target only needs gemmi headers, do this instead:

target_link_libraries(example PRIVATE gemmi::headers)

The gemmi::headers interface, which is also included in gemmi::gemmi_cpp, adds two things: the include directory and the compile feature cxx_std_14 (a minimal requirement for compilation).

Gemmi can be compiled with either zlib or zlib-ng. The only difference is that zlib-ng is faster. Here are the relevant CMake options:

  • FETCH_ZLIB_NG – download, build statically, and use zlib-ng.

  • USE_ZLIB_NG – find zlib-ng installed on the system.

  • INTERNAL_ZLIB – compile third_party/zlib (a subset of zlib distributed with gemmi).

  • None of the above – find zlib installed on the system; if not found, use third_party/zlib.

On Windows, when a program or library is linked with a zlib(-ng) DLL, it may require the DLL to be in the same directory. It is simpler to build zlib-ng statically or use -D FETCH_ZLIB_NG=ON.

Note on Unicode: if a file name is passed to Gemmi (through std::string), it is assumed to be in ASCII or UTF-8.

Python module

From PyPI

To install the gemmi module, run:

pip install gemmi

We have binary wheels for several Python versions (for all supported CPython versions and one PyPy version), so the command usually downloads binaries. If a matching wheel is not available, the module is compiled from source – it takes a few minutes and requires a C++ compiler that supports C++17.

Gemmi 0.7+ supports only Python 3.8+.

Other binaries

You can find gemmi:

If you use the CCP4 suite, you can find gemmi there.

If you use conda, the gemmi package, which includes also a command-line program and C++ dev files, can be installed from conda-forge:

conda install -c conda-forge gemmi

These distribution channels may have an older version of gemmi.

From git

The latest version can be installed directly from the repository. Either use:

pip install git+https://github.com/project-gemmi/gemmi.git

or clone the project (or download a zip file) and from the top-level directory run:

pip install .

Building with pip uses scikit-build-core and CMake underneath. You can pass options to CMake either using the --config-settings option in recent pip versions:

pip install . --config-settings="cmake.args=-DFETCH_ZLIB_NG=ON"

or by using environment variables such as CMAKE_ARGS. See scikit-build-core docs for details.

If gemmi is already installed, uninstall the old version first (pip uninstall) or add the --upgrade option.

Alternatively, you can manually install nanobind and cmake (using pip) and build a cloned project directly with CMake:

cmake -D USE_PYTHON=1 .
make -j4 gemmi_py

Fortran and C bindings

The Fortran bindings are in an early stage and are not documented yet. They use the ISO_C_BINDING module introduced in Fortran 2003 and shroud. You can check the fortran/ directory to see what to expect. This directory contains a Makefile – run make to build the bindings. (They are currently not integrated with the CMake build.)

The C bindings are used only for making Fortran bindings, but they should be usable on their own.

WebAssembly

The Gemmi library can be compiled with Emscripten to WebAssembly. Since compiling the entire library is unlikely to be necessary, we’ll show how to compile a subset needed for a particular project, adding bindings for JavaScript. We present two approaches.

With Embind

The wasm/ subdirectory contains bindings that use Embind to expose C++ classes to JavaScript. Currently, they consist of two parts:

  • Minimal bindings to the macromolecular Structure that allow reading a PDB or mmCIF file and iterating over models, chains, residues and atoms (see mol.test.js). This serves as an example and a starting point for further work (which can be carried on either as part of gemmi or in the user’s own project). Feel free to reach out if you have questions.

  • Bindings to class Mtz that enable map calculation (via FFT) from map coefficients. This part was previously provided in the separate mtz module, the first library to enable the use of MTZ files in molecular graphics apps.

The files from the Gemmi library used for building the wasm module are listed as GEMMI_OBJS in the Makefile.

With C API

As part of the Gemmi project, we maintain a set of web tools (mostly file converters), which are single-page applications powered by Gemmi functions in WASM. The source code of these tools is in the wasm repository (not to be confused with the wasm subdirectory of the gemmi repo – one of them should be renamed).

These tools don’t use the bindings described above. They demonstrate an alternative approach. For each page we wrote a dedicated C++ function, with a C API, that performs the bulk of the work. The bindings to these functions are generated using Emscripten (without Embind). This approach – writing part of the web app in C++ – is more performant, as it keeps all computations on the WebAssembly side and minimizes the number of calls across the JS/WASM boundary.

Check the Makefiles in subdirectories to see how the wasm modules are built.

Program

The library comes with a command-line program also named gemmi.

Binaries

Binaries are distributed with the CCP4 suite and with Global Phasing software. They are also in PyPI (pip install gemmi-program), conda-forge packages, and a few Linux (and FreeBSD) distros.

The very latest builds (as well as a little older ones) can be downloaded from CI jobs:

  • For Windows – click the first (green) job in AppVeyor CI and find gemmi.exe in the Artifacts tab (if there is also a dll file there, it’s a dynamically linked build and both files are needed).

  • For Linux and Mac – sign in to GitHub (no special permissions are needed, but GitHub requires sign-in for artifacts), go to gemmi’s gemmi’s CI workflow, click the latest job with ✅, scroll to the bottom of the page, and download one of the zip files from the Artifacts section.

From source

To build it from source, first make sure you have git, cmake and C++ compiler installed (on Ubuntu: sudo apt install git cmake make g++), then:

git clone https://github.com/project-gemmi/gemmi.git
cd gemmi
cmake .
make

Alternatively, you can use pip install git+https://..., which installs both the Python module and the program. If you are not using the Python module, you can use pip to build only the program:

pip install git+https://github.com/project-gemmi/gemmi.git --config-settings=cmake.args=-DONLY_PROGRAM=ON

Testing

The main automated tests are in Python:

python3 -m unittest discover -v tests/

We also have Python doctest tests in the documentation, and a few other test routines. All the commands used for testing are listed in the run-tests.sh script in the repository.