# Installation¶

## Requirements¶

pyNRC requires Python 2.7+ or 3.5+ along with the following packages:

## Installing with conda¶

Todo

pyNRC has been placed on conda-forge so you can manage the package through your Conda installation. Simply add conda-forge to your .condarc file, which appends the appropriate URL to Conda’s channel search path:

$conda config --add channels conda-forge  With the conda-forge channel added, it’s a simple matter to run: $ conda install pynrc


## Installing with pip¶

You can install the pynrc package through pip:

$pip install pynrc  Note that the pip command only installs the program code. You still must download and install the data files, as described below. ## Installing from source¶ To get the most up to date version of pynrc, install directly from source, though stability is not guarenteed. The development version can be found on GitHub. In this case, you will need to clone the git repository: $ git clone https://github.com/JarronL/pynrc


Then install the package with:

$cd pynrc$ pip install .


For development purposes:

$cd pynrc$ pip install -e .


in order to create editable installations. This is great for helping to develop the code, create bug reports, pull requests to GitHub, etc.

## Installing the data files¶

Files containing such information as the instrument throughputs, SCA biases and darks, stellar models, and exoplanet models are distributed separately. To run pynrc, you must download these files and define the PYNRC_PATH environment variable. This is also the location that PSF coefficients will be saved to during normal operations of pynrc.

2. Untar into a directory of your choosing.

3. Set the environment variable PYNRC_PATH to point to that directory. For bash, for example:

$export PYNRC_PATH=$HOME/data/pynrc_data


You will probably want to add this to your .bashrc.

You should now be able to successfully import pynrc in a Python session.

## Testing¶

Todo

If you want to check that all the tests are running correctly with your Python configuration, you can also run:

\$ python setup.py test


in the source directory. If there are no errors, you are good to go!