Is it possible to link to an existing mfem library? about pymfem HOT 7 CLOSED

Hi @jan-janssen, I didn't use PyMFEM with conda mfem. Is it install a pre-compiled binary or source code?
If so, setup.py takes --mfem-prefix --mfem-source as parameters. With pip, you need to use --install-option.
to set each of these parameters. If you want to compile it with MPI, you also give --with-parallel, --hypre-prefix
and --metis-prefix.

from pymfem.

I am a bit confused, I have the libmfem.so installed on my system. I am just looking for the pip command to link to this already installed version.

from pymfem.

Setup.py in PyMFEM generates a wrapper code from the MFEM source code. MFEM can be compiled with various options.
Depending the option you use, the wrapper code has to change. Therefore, if you have only libmfem.so, you can not do pip install.

from pymfem.

It was compiled with: https://github.com/conda-forge/mfem-feedstock/blob/main/recipe/build.sh

make config CXX=$CXX PREFIX=$PREFIX MFEM_SHARED=YES MFEM_STATIC=NO MFEM_USE_MPI=YES MFEM_USE_METIS_5=YES CXXFLAGS="$CXXFLAGS" LDFLAGS="$LDFLAGS" LIBS="$LIBS" HYPRE_OPT="-I$PREFIX/include" GHV_CXX=$CXX_FOR_BUILD
make lib
make install

from pymfem.

We don't have a plan to change setup.py for this specific installation pattern. Probably, better approach would be to modify the feedstock and build the wrapper together.

from pymfem.

We don't have a plan to change setup.py for this specific installation pattern.

I can patch the setup.py I just do not fully understand which option to change. I tried to add python -m pip install . -vv --mfem-prefix=${PREFIX} but that did not work. Can you just tell me how to disable the build process?

from pymfem.

In setup.py, we run SWIG to generate a wrapper code. SWIG needs to read all C++ header files.
The location of header files is set by using --mfem-source option. If Conda package does not contain
header file, SWIG will fail and thus we can not build PyMFEM.

from pymfem.