model-manager 0.8

A python package to manage your (scientific) model runs, input and output.

modelmanager

Version number: 0.8

Author: Michel Wortmann, Potsdam Institute of Climate Impact Research, Germany

Overview

A python package to create flexible APIs for (scientific) models.

Dependencies

Modelmanager itself has no dependencies, but some of its plugins do. They will direct you two install them through pip. Python development dependencies are listed in requirements_dev.txt. Before installing the any package (see below) consider setting up a virtual python environment (virtualenv mydevenv).

Installation

To install use pip:

$ pip install modelmanager

Or clone the repo:

$ git clone https://github.com/mwort/modelmanager.git
$ python setup.py install

Concept

The model The modelmanager links with your model like illustrated in this file structure:

modeldir/               # your main model directory
    mm/                 # the modelmanager resource directory
        settings.py     # define or import here all variables, functions and plugins
	      browser/        # this is a plugin directory, e.g. for the browser app

    modelexec           # all your model resources
    input/
    output/

With this setup you can either use your model interface through the commandline or through the python API (see Usage below).

Usage

Initialise project where in your model root directory:

cd home/mymodel
modelmanager init

Add some variables, functions or plugins for your model in mm/settings.py and call them on the commandline like this:

modelmanager example_function --example_argument=2

Or use your new model api in a Python script like this:

import modelmanager as mm

project = mm.Project()
result = project.example_function()

Use the browser app by adding this line to your settings:

from modelmanager.plugins.browser import *

Then start the application on the commandline:

modelmanager startbrowser

Navigate to localhost/admin in your browser.

Contributing

Bug reports, ideas and feature requests welcome on Github.

Testing

Run test in tests/ like this:

make                                # runs all tests
python test_project.py              # just runs test_projects with call stats
python -m unittest test_project.Settings  # just run Settings tests
make clean                          # clean any leftover test output

make should pass before submitting a pull/merge request.

Releasing

• add entry in CHANGELOG.md
• change version number in __init__.py and README.md, tag and commit (make version)
• build sdist and push to git and pypi (make release)