Skip to content

Latest commit

 

History

History
91 lines (68 loc) · 3.22 KB

INSTALL.md

File metadata and controls

91 lines (68 loc) · 3.22 KB
Raw

Using conda for the required installations works like magic. Except when it doesn't.

If it's the latter case ... after about 2–3 hours of chasing answers and comments on StackOverflow, swearing, pouring through GitHub issues to test every elusive option mentioned, swearing some more, inadvertently frightening the cat, etc., you may get portions of geoplot to install correctly.

One difficulty is that geoplot builds atop several other packages. Some may trigger the infamous "conda: Solving environment" infinite spinner.

Nonetheless, conda is a favorite of many people for installing Python packages. It's especially brilliant to use for Jupyter, spaCy, etc. Anaconda has published notes about how to improve its performance, as a possible workaround:

https://www.anaconda.com/understanding-and-improving-condas-performance/

QuantStack released a conda-compatible package manager called mamba which is much faster, although no guarantees about resolving these installation issues:

https://medium.com/@wolfv/making-conda-fast-again-4da4debfb3b7

While developing this tutorial, conda installers failed on an up-to-date macOS laptop, then again on a pristine cloud server instance running Ubuntu 18.04 LTS. That indicates trouble ahead in Python installation land...

Another alternative is to use pip and PyPi. One of the differences between conda and pip is how the latter can also build wheels as binary distributions. That leads to faster installation for pure Python packages. However, when a library builds on top of low-level libraries in C or C++ (or FORTRAN) then it may be necessary to skip the binaries – ignore the pre-compiled wheels – and instead force compilation from source code within your environment.

We've tried to save people some time and agony by describing an installation recipe that works on a pristine Ubuntu 18.04 LTS server. First, install the libraries that may require admin privileges:

sudo apt-get update
sudo apt-get install virtualenv python-pip python-dev python3-dev

sudo apt-get install libproj-dev proj-data proj-bin
sudo apt-get install libgeos-dev
sudo apt-get install libspatialindex-dev

Then create a virtual environment for Python3 and activate it:

virtualenv -p python3 ~/venv
source ~/venv/bin/activate

Now install the Python dependencies into that virtual environment:

pip install -r requirements.txt

pip install shapely --no-binary shapely 
pip install git+https://github.com/SciTools/cartopy.git --no-binary cartopy
pip install pysal --no-binary pysal

pip install geopandas
pip install geoplot

Notice two subtle points about this approach:

  • Cython and descartes get installed early in the sequence
  • shapely, cartopy, and pysal require the --no-binary option

Okay, that approach – or something similar – should work on a recent release of Ubuntu/Debian Linux.

BTW, compared with what the installation of geospatial libraries required a decade ago, these steps are simple!

If you have Windows you'll probably have other issues to resolve. YMMV.

If you're using a Mac, well there's brew and many swear by it, while many others swear at it, since use of brew may cause other installed dependencies to get mangled.