This is ctypes wrapper for BMI models.

The BMI describes a low level interface for numerical models.

This module is based on code from:

• CSMDS: https://csdms.colorado.edu/svn/bmi/trunk/python/bmi/BMI.py
• OpenEarth: http://svn.oss.deltares.nl/repos/openearthtools/trunk/python/OpenEarthTools
• 3Di: https://github.com/nens/python-subgrid

Several models implement the BMI interface.

• Subgrid: https://github.com/nens/python-subgrid
• Swan: https://github.com/SiggyF/chenopis
• XBeach: http://svn.oss.deltares.nl/repos/xbeach
• DFlow-FM: https://repos.deltares.nl/repos/ds/trunk/additional/unstruc/python/dflowfm

We need a compiled BMI library (dll, so, dylib). There are a couple of common locations where we look for it.:

.
~/local/lib
~/.local/lib
/opt/modelname/lib
/usr/local/lib
/usr/lib

A convention on linux is to install the library into /opt/modelname/. If you are using one of the models above, the modelname will be 3di, dflowfm, xbeach or swan

In case you have an alternative location, you can set the LD_LIBRARY_PATH, (DYLD_LIBRARY_PATH in OSX, PATH in windows) environment variable, for example for 3Di:

$ export LD_LIBRARY_PATH=/home/user/svn/3di/trunk/subgridf90/src/.libs

(On windows the command is set instead of export).

The virtualenv way (assumes virtualenvwrapper and virtualenv are installed):

mkvirtualenv main
workon main
# get the version from pypi
pip install bmi
# or if you want to add your source directory to the path
pip install -e .

Combination Windows and Anaconda:

- Download + Install Anaconda
- Download and install the :faulthandler: package, from http://www.lfd.uci.edu/~gohlke/pythonlibs/#faulthandler
- Download and install the :NetCDF4: package, from http://www.lfd.uci.edu/~gohlke/pythonlibs/#netcdf4
- Open an (Anaconda) Python-terminal (Press :[Ctrl]:+:[Alt]+:A:).
- pip.bat install -e ``<path to your bmi-python GIT working copy>``

There are two ways to use the wrapper. A handy way is as a context manager, so with a with statement:

with BMIWrapper(engine="model", configfile='/full/path/model.ini') as model:
    # model is the actual library.
    model.something()

The second way is by calling :meth:`start` and :meth:`stop` yourself and using the :attr:`library` attribute to access the Fortran library:

wrapper = BMIWrapper(engine="model", configfile='/full/path/model.mdu')
wrapper.start()
wrapper.library.something()
...
wrapper.stop()

Note: Without the mdu argument, no model is loaded and you're free to use the library as you want.

The python bmi library contains a script that can be used as a command line runner for your model:

bmi-runner <engine> <configfile>