Installation¶
Conda Installation (recommended)¶
There are conda distributions available for cis_interface from conda-forge. You can install cis_interface from conda-forge by calling:
$ conda install -c conda-forge cis_interface
from your terminal prompt (or Anaconda prompt on Windows). This will
install cis_interface and all of its dependencies in your active
conda environment from the conda-forge
channel.
Manual Installation¶
Note
Before installing cis_interface from pip
or the cloned repository, you
should install the non-Python dependencies, particularly the
ZeroMQ C and C++ libraries (see below).
If you do not want to use conda, cis_interface can also be installed
from either PyPI
using pip
:
$ pip install cis_interface
or by cloning the Git repository on Github:
$ git clone --recurse-submodules https://github.com/cropsinsilico/cis_interface.git
and then building the distribution.:
$ cd cis_interface
$ python setup.py install
If the --recurse-submodules
option was not included when cloning the repo,
you will need to run the following from within the repository before calling
python setup.py install
to ensure that
rapidjson is cloned as a submodule:
$ git submodule init
$ git submodule update
If you do not have admin privileges on the target machine, --user
can be
added to the end of either the pip
or setup.py
installation commands.
When using the --user
, you may need to add the directory containing the
entry point scripts to your PATH
environment variable in order to use
cis_interface command line tools (e.g. cisrun
) without specifying
their full path. Usually, this directory can be found using the following
Python commands:
>>> import os
>>> from distutils.sysconfig import get_python_lib
>>> os.path.realpath(os.path.join(get_python_lib(), '../../../bin/'))
The displayed path can then be added either on the command link or in a startup
script (e.g. .bashrc
or .bash_profile
), using one of the following:
$ export PATH=$PATH:<scripts_dir> # (linux/osx, bash)
$ setenv PATH $PATH:<scripts_dir> # (linux/osx, tcsh)
$ set PATH "%PATH%:<scripts_dir> # (windows)
These commands will only add the directory to your path for the current
session. For the change to be permanent on Linux/MacOS, the appropriate command
from above can be added to your .bashrc
or .bash_profile
. On
Windows (>=7), the following command will permanently modify your path:
$ setx PATH "%PATH%:<scripts_dir>
The changes will take affect the next time you open the terminal.
User Defined rapidjson¶
If you would like to use an existing installation of the
rapidjson
header-only library, you can pass the flag
--rapidjson-include-dir=<user_defined_dir>
to either the pip
or setup.py
installation commands from above with the location of the
existing rapidjson include directory.
Additional Steps on Windows¶
As local communication on Windows is handled by ZeroMQ, running models written in C or C++ will require installing the ZeroMQ libraries for C and C++. If you install cis_interface using conda, these will be installed automatically as depencies. If you are not using conda, you will need to install them yourself.
Instructions for installing the ZeroMQ C and C++ libraries can be found
here
At install, cis_interface will attempt to search for those libraries.
To speed up the search you can (temporarily) add the directories
containing the libraries to your PATH environment variable prior to
running one of the above install commands. If cis_interface complains
that it cannot find these libraries, you can manually set them in your
.cis_interface.cfg
file (See Configuration Options).
If you install these libraries after installing cis_interface you can re-configure
cis_interface and have it search for the libraries again by calling cisconfig
from the command line.
Note
Although not required, the ZeroMQ libraries are also recommended for message passing on Linux and MacOS operating systems as the IPC V message queues have default upper limits of 2048 bytes on some operating systems and will have to send larger messages piecemeal, adding to the message passing overhead.
Additional Steps for Matlab Models¶
To run Matlab models, you will also need to install the Matlab engine for Python. This requires that you have an existing Matlab installation and license. cis_interface will attempt to install the Matlab engine for Python at install, but should it fail or if you want to use a non-default version of Matlab, you can also do it manually.
Instructions for installing the Matlab engine as a python package can be found on the
Mathworks website. Once you have installed the Matlab engine as a python
package, you can re-configure cis_interface by calling cisconfig
. from the comamnd
line.
Note
The version of Matlab that you are using will determine the versions of Python that you can use with cis_interface. The chart below shows the versions of Python that are compatible with several versions of Matlab. If you are using an incompatible version, the instructions above for manually installing the Matlab engine as a Python package will fail with an error message indicating which versions of Python you can use.
Matlab Version | Max Python Version |
---|---|
R2015b | 2.7, 3.3, 3.4 |
R2017a | 2.7, 3.3, 3.4, 3.5 |
R2017b | 2.7, 3.3, 3.4, 3.5, 3.6 |
Additional Steps for RabbitMQ Message Passing¶
RabbitMQ connections allow messages to be passed between models when the
models are not running on the same machine. To use these connections
(those with the prefix ‘RMQ’), the framework must have access to a
RabbitMQ server. If you have access to an existing RabbitMQ server,
the information for that server can either be provided to the
RabbitMQ connection driver
(See cis_interface.drivers.RMQDriver.RMQDriver
) or added
to the cis_interface config file (See
Configuration Options for information on setting
config options).
Starting a local RabbitMQ Server is also relatively easy. Details on downloading, installing, and starting a RabbitMQ server can be found here. The default values for RabbitMQ related properties in the config file are set to the defaults for starting a RabbitMQ server.