Source: dijitso Section: python Priority: optional Maintainer: Debian Science Team Uploaders: Johannes Ring , Drew Parsons Build-Depends: debhelper (>= 12), dh-python, python3-all, python3-setuptools, python3-numpy Standards-Version: 4.3.0 Homepage: https://fenicsproject.org Vcs-Git: https://salsa.debian.org/science-team/fenics/dijitso.git Vcs-Browser: https://salsa.debian.org/science-team/fenics/dijitso Package: python3-dijitso Architecture: all Depends: python3-numpy, python3-mpi4py, python3-pkg-resources, ${python3:Depends}, ${misc:Depends} Suggests: python-dijitso-doc Conflicts: python-dijitso (<< 2018.1~) Replaces: python-dijitso Description: distributed just-in-time building of shared libraries (Python 3) Dijitso was written to improve a core component of the FEniCS framework, namely the just in time compilation of C++ code that is generated from Python modules, but is only called from within a C++ library, and thus do not need wrapping in a nice Python interface. . The main approach of dijitso is to use ctypes to import the dynamic shared library directly with no attempt at wrapping it in a Python interface. . As long as the compiled code can provide a simple factory function to a class implementing a predefined C++ interface, there is no limit to the complexity of that interface as long as it is only called from C++ code, If you want a Python interface to your generated code, dijitso is probably not the answer. . Although dijitso serves a very specific role within the FEniCS project, it does not depend on other FEniCS components. . The parallel support depends on the mpi4py interface, although mpi4py is not actually imported within the dijitso module so it would be possible to mock the communicator object with a similar interface. . This package installs the library for Python 3. Package: python-dijitso Architecture: all Depends: python3-dijitso (>= 2018.1), ${misc:Depends} Description: distributed just-in-time building of shared libraries Dijitso was written to improve a core component of the FEniCS framework, namely the just in time compilation of C++ code that is generated from Python modules, but is only called from within a C++ library, and thus do not need wrapping in a nice Python interface. . The main approach of dijitso is to use ctypes to import the dynamic shared library directly with no attempt at wrapping it in a Python interface. . As long as the compiled code can provide a simple factory function to a class implementing a predefined C++ interface, there is no limit to the complexity of that interface as long as it is only called from C++ code, If you want a Python interface to your generated code, dijitso is probably not the answer. . Although dijitso serves a very specific role within the FEniCS project, it does not depend on other FEniCS components. . The parallel support depends on the mpi4py interface, although mpi4py is not actually imported within the dijitso module so it would be possible to mock the communicator object with a similar interface. . This is a dummy package that depends on python3-dijitso. (Dijitso is no longer available for Python 2).