mhm 5.13.3

Python distribution of mHM with bindings.

mHM - Python bindings

[TOC]

Python bindings to control mHM.

The wrapper (mhm/wrapper.f90) is just a small layer on top of the interfaces provided by mHM to be compatible with f2py.

Installation

There is a PyPI package to install the latest release:

pip install mhm

Installing the mHM Python package will provide the mhm command to execute mHM the traditional way.

In order to compile the Python bindings from scratch you need:

1. Python with version at least v3.8 and pip
2. a Fortran, a C and a C++ compiler (set the environment variables FC (and F77), CC and CXX accordingly). In case of gcc, this could look like:

   export FC="gfortran"
   export F77="gfortran"
   export CC="gcc"
   export CXX="g++"
   

3. NetCDF-C installed in your system path

See the Compilation instructions for these dependencies.

You can also use a conda environment (set up with miniforge for example) to get everything:

conda install -y pip libnetcdf fortran-compiler c-compiler cxx-compiler

To compile everything after cloning/downloading, you can use pip:

pip install -v .

To install it directly from the git repository you can type:

pip install -v git+https://git.ufz.de/mhm/mhm.git

Environment variables

The following environment variables can be used to control the compilation and installation of the python bindings for mHM:

• SKBUILD_CMAKE_BUILD_TYPE=[Release|Debug]: build type for the mhm library (default: Release)
• MHM_BUILD_FORCES_PATH=<path>: custom path to forces source dir (default: None)
• MHM_BUILD_PARALLEL=[0|1]: default value for CMAKE_WITH_OpenMP when building mHM (default: 0)
• CMAKE_ARGS="<args>": additional CMake defines, for example -DCMAKE_BUILD_TYPE=RelWithDebInfo -DCMAKE_WITH_OpenMP=ON -DMHM_ENABLE_NATIVE=ON

Test domain download tool

Together with the Python bindings comes a command line script to download the test domains:

mhm-download --verbose --branch develop --domain 1 --path mhm_domain/

You can then run mHM on this test domain with:

mhm mhm_domain/

You can get help on how to use this script with mhm-download -h:

$ mhm-download -h
usage: mhm-download [-h] [-V] [-v] [-b BRANCH] [-d {1,2}] [-p PATH]

Download tool to retrieve the test domains for mHM.

optional arguments:
  -h, --help            show this help message and exit
  -V, --version         display version information
  -v, --verbose         be verbose (default: False)
  -b BRANCH, --branch BRANCH
                        branch, tag, or commit of the mHM repository to take the test domain from,
                        by default tag determined from the mHM version (default: None)
  -d {1,2}, --domain {1,2}
                        test domain '1' or '2' (default: 1)
  -p PATH, --path PATH  destination path for the downloaded folder,
                        by default the original folder name in the current directory (default: None)

Within python scripts, you can use this tool with mhm.download_test. See below for examples.

Documentation

See mhm.tools and wrapper.f90 for further information on the provided routines.

Examples

If you have cloned the repository, you can do the following to simply run mhm without optimization:

import mhm

# download test domain 1
mhm.download_test(path="example_domain")
# run the downloaded example
mhm.model.init(cwd="example_domain")
mhm.model.run()
mhm.model.finalize()

Or you can do the following to control each timestep:

import mhm

# assuming to run from the mhm repo root
mhm.model.init()
mhm.run.prepare()
ndomians = mhm.run.get_ndomains()
for i in range(1, ndomians + 1):
    mhm.run.prepare_domain(domain=i) # 0 by default
    while not mhm.run.finished():
        mhm.run.do_time_step()
        mhm.run.write_output()
    mhm.run.finalize_domain()
mhm.run.finalize()
mhm.model.finalize()

See also the examples folder.

License

LGPLv3 (c) 2005-2023 mHM-Developers