Package Jenks with `mapboxgl` library about mapboxgl-jupyter HOT 3 CLOSED

What specific problems are users having installing jenks? I'd anticipate two main issues:

1. they don't have git installed
2. they don't have a compiler toolchain to build C extension modules

Publishing an official release of the jenks module to pypi will solve (1) - Just pip install jenks. Honestly something I should do anyways. I'd much prefer to do this than fork the code and bundle it with an unrelated repo.

But (2) will still be an issue - the jenks module is implemented in Cython for speed reasons and needs a C compiler to work. This will typically affect Windows users more severly as the compiler toolchain is relatively easy to install on OSX and Linux. A few options to solve this

• Build binary wheels for every target platform and python version
• Rewrite the module setup to fall back onto a (slow but working) pure python implementation in the absence of a compiler.
• Rely on another third party classification library that has these issues solved:
  • PySAL has a very robust module for handling classification with Jenks and other methods: https://pysal.readthedocs.io/en/latest/library/esda/mapclassify.html
  • There is a jenkspy module on pypi which is implemented in C but distributes windows wheels with each release. https://pypi.python.org/pypi/jenkspy

from mapboxgl-jupyter.

@perrygeo I think relying on a third party classification lib makes sense here - jenks is just the beginning of classifications users may want to use to calculate a domain of breaks.

PySAL seems like a great option.

from mapboxgl-jupyter.

Let's keep data transformation outside of the scope of this library, and instead build examples using PySAL for how to prepare data to use with mapboxgl.

I updated the current example notebook to use pySAL to calculate natural breaks. Here's the code, and the live example at www.mapbox.com/labs/jupyter

from mapboxgl-jupyter.