With the help of an outstanding community effort named conda-forge, wradlib can now be more conveniently installed on linux, windows and osx.
Until now, installing wradlib and its dependencies could be tricky, with each OS having its own issues. On Windows, we so far recommended to satisfy all depencies via Python(x,y). This was convenient; however, it limited users to Python 2.7, and, more importantly, to 32-bit Python. This was a serious drawback particularly for memory-intensive applications.
With this post, we present a new installation approach that is harmonised across platforms. Using conda-forge/wradlib-feedstock, we provide installable wradlib packages for all major OS (accounting for different python and numpy versions, and offering 32-bit builts for Windows, if desired). All builts are tested and uploaded to the conda-forge channel.
The default-conda channel provides many wradlib dependencies out of the box, but not all. Hence, we also contributed to the conda-forge/gdal-feedstock making it the first feedstock serving two different package versions (gdal 1.11.4 and 2.0.2).
As a result, wradlib can now be conveniently installed using the conda package manager. Windows users should be aware, though, that this approach is not compatible with Python(x,y). So you need to make a decision.
So this is the basic walk through:
Install the Anaconda environment of your choice
Clone the root environment or create one from scratch
$ conda create --name wradlib --clone root or $ conda create --name wradlib python=2.7
Add the conda-forge channel
$ conda config --add channels conda-forge
Activate wradlib environment
$ source activate wradlib
> activate wradlib
Install wradlib (and dependencies)
(wradlib) $ conda install wradlib
Set GDAL_DATA environment variable (needed for georeferencing)
(wradlib) $ export GDAL_DATA=/path/to/anaconda/envs/wradlib/share/gdal
[wradlib] > setx GDAL_DATA C:\path\to\anaconda\envs\wradlib\Library\share\gdal
Optional dependencies can be installed OS independent with
(wradlib) $ pip install xmltodict
We hope that this new approach will make the installation of wradlib more convenient, and, as a result, enhance its usability on all major platforms - thanks to Anaconda from Continuum(R) and the conda-forge community effort.