Introduction

You can have an IMP module depend on some external library, for example to take advantage of a method implemented in that library, or to link IMP with another piece of software at runtime.

Considerations

Think very carefully before introducing a new external dependency. Each additional dependency makes it harder for people to use your code. If you do need to add a dependency, it needs to be open source and available under a suitably permissive license (for example if it is available under the GPL then you cannot license your module as LGPL, but will need to also make it GPL).

Generally if you need to add a new dependency you should probably also put your code in a new module, rather than adding it to an existing module. That way, people that elect not to install that dependency will only be deprived of your code, not of the existing module.

Try not to require the very latest version of a dependency, as again that may make it harder for people to use. (If necessary, you can use preprocessor macros so that your code works with both old and new versions of a dependency; for examples, look in the C++ code for BOOST_VERSION.) We try to make IMP work with the versions of packages available in the oldest supported versions of RedHat Enterprise Linux (RHEL), Ubuntu LTS, and macOS. For example, RHEL 7 ships with Boost 1.66 and Python 3.6, so IMP works with both of those.

Simple dependencies

The simplest way to add a C/C++ dependency foo is to create a file foo.description in the dependency subdirectory of your module. This is a simple text file containing several variables:

• headers: a colon-separated list of any C/C++ headers that need to be included to use the dependency.
• libraries: a colon-separated list of any libraries that need to be linked against to use the dependency. (There is a similar extra_libraries variable, if needed, for libraries that aren't part of the dependency but that need to also be linked in order for it to work.)
• body: a fragment of C++ code that uses the dependency.

When CMake is run, it will use the variables in foo.description to build a small test program in order to make sure the dependency is available and that it works. See modules/kernel/dependency/GPerfTools.description for an example.

More complex dependencies

For more complex dependencies, you can also create a foo.cmake file containing arbitrary CMake instructions to configure the dependency. See for example modules/rmf/dependency/RMF.cmake.

Using the dependency

The next step in adding a foo dependency is to list foo in the module's dependencies.py file. foo can be listed either in that file's required_dependencies variable or in optional_dependencies, both of which are colon-separated lists of external dependencies. See for example modules/kernel/dependencies.py.

If placed in required_dependencies, the module cannot be built unless your dependency is found. The module will be automatically linked against the dependency.

If placed in optional_dependencies, the module can be built either with or without the dependency. If the dependency is available, a preprocessor macro will be set when the module is built; protect your code with that macro. For example, IMP::score_functor can be built with or without the HDF5 library. Any code that requires HDF5 is conditional on the IMP_SCORE_FUNCTOR_USE_HDF5 preprocessor macro.