maxscheurer/cppe

C++ and Python library for Polarizable Embedding

cppe's Issues

Refactor PeOptions py-side

Don't expose PeOptions to py-side, only communicate via Python dictionary.

Make CPPE pip-installable

The compiled Python interface should be installable via pip, making integration in Python packages easier (e.g., pyscf).

• minimal setup.py to trigger build via cmake
• install the resulting cppe.cpython-37m*.so library
• run tests via python3 setup.py test

Expand test suite

• tests against downloaded reference data (Dalton?)
• consistency tests (pyscf)
• ...

release v0.3.2

I was wondering why v0.3.2 is not on PyPI yet and why it was not tagged as a release?

Wrong vector size when not all sites are polarizable

In the case that not all sites are polarizable, the induced moments vector is too large (number of elements = number of sites which have a static moment, so most of the times = all sites).

More options

• isotropic polarizabilities (iso_pol)
• disable polarization
• disable electrostatics
• tests

Build system problems

The build system is a mess at the moment...
Problems arise with ExternalProjects (pybind11 and armadillo), if these are not system-installed.
One option would be to build the main cppe as an external project, too (as done in PSI4).

Refactor CMake

• build only one library, currently libcppe and pycppe are built
• cleanup required

Make evaluation of static energies & fields lazy

As it says in the title...
Makes it easier not to forget calling calculate_static_energies_and_fields.

v0.1.1

On Linux conda the new v0.0.9 builds just fine (https://anaconda.org/psi4/pycppe/files and https://anaconda.org/psi4/cppe/files). Request that you splurge on the next version and make it v0.2 so it sorts after the dummy v0.1 tag I had used originally.

Atomic force

Here is a question about the nuclear gradients raised in PySCF pyscf/pyscf#1321

PElib Integration

In the branches for Q-Chem and Psi4, I have disabled the PElib integration
because it is not as easy to expose to host programs (matrix elements need to be ordered
as in gen1int, for example). However, it should still be possible to compile the library including PElib and gen1int for test purposes and experimentation.

Code cleanup

• use <algorithm> when possible...
• remove unused code
• make a header for each source file
• refactor libcppe.hh (Related to #3)
• more pythonic handling of PE Options (#16)
• multiply pre-factors with moments?
• more pythonic handling of PE Energies (dictionary)
• more pythonic exposure of properties (get_site_position(), get_values())
• clang-format on all files

Update README

The README file is rather sparse at the moment...

PeOptions type caster

For a more pythonic experience, it would be nice to expose the PeOption class
as a plain python dictionary. This could be done with a custom type caster in pybind11.