Maps of Imaginary Lands

Introduction

Supporting code and slides for a talk introducing customised usage of Django's GIS components, Mapnik and OpenLayers. Originally presented at DjangoCon-US, September 2010 (Portland, Oregon, USA) in a short form and then at Kiwi Pycon (Bay of Islands, New Zealand) in November, 2010 and PyCon-APAC in June, 2012.

Short Description

The GIS features of Django aren't restricted to being applied to real world maps and planets. This talk will show how to display and interact with maps of imaginary lands, such as game maps or lands in science fiction novels. We'll uncover a bit of how Django GIS works in the process, separating the map display from the modelling.

Abstract

Whilst django.contrib.gis isn't particularly difficult to get started with, particularly if you follow the tutorials, it can sometimes seem a bit overwhelming in the way it goes from zero to fancy maps in only a couple of steps. I'd like to demystify some of the pieces of the stack, pulling apart the modelling support — specifying the data are we trying to work with — from the display and client-side portion.

To make this more than a dry technical dive, I'll show how to add extras to an imaginary map, rather than something pulled from Google Maps or Open Street Map. We'll take on the task of plotting features on a landscape from a potential role-playing game and show how the GIS data manipulation features, such as calculating region intersections, nearby points, and Javascript client-side display work the in a familiar way against this slightly unusual background.

Some basic familiarity with Django's GIS features would be useful for this talk, although it might also serve as a motivating introduction to trying things out.

Aside: Creating map graphics

Over the years, a number of people have asked how I created the island map image for this presentation.

The main graphic was done using the Gimp, following tutorials at The Cartographer's Guild website. I have spent hours playing with their map-making tutorials and trying out different things. The World Building subreddit is also worth regularly dipping into for inspiration and critiques.

The shapefiles describing the island boundaries and paths and points of interest were creating in QGis, a professional level tool that is highly accessible to amateurs and with a very supportive online community of tutorials and tips.

Setting up

This code is intended as a small self-contained, executable example, written with reasonably professional coding standards in mind.

All the code here has been verified to run against Django 1.4 (as of June, 2012).

That being said, a few prerequisites need to be installed in order to run the example. I am assuming you have worked through the django.contrib.gis tutorial and thus have all the necessary prerequisites installed. If you can view any Django admin page that involves GIS information, you have met the requirements here.

Secondly, you'll need to have Mapnik installed (and Mapnik's Python bindings). Major Linux distributions will ship these as packages that won't require anything more than a yum install ... or aptitude install.... I've been led to believe it isn't amazingly difficult to get the necessary pieces installed on a Mac OS X and Windows as well¹. If you can run the following at a Python prompt and no exception is raised, you have the necessary components installed:

>>> from mapnik2 import ogcserver

This assumes you are running mapnik-python version 2.0. If you are using an earlier version (0.7) of mapnik and mapnik-python, you will need to import mapnik, rather than mapnik2². There are obviously other differences between the earlier and later versions of mapnik, however, for the most part, things are forwards compatible and the code here will work without change once the import is correct.

The included settings file is set up to use Postgis as the database. I don't believe anything Postgis-specific is used in the code, however, so any spatial database backend should work (again, being able to work through the Django GIS tutorial is probably both necessary and sufficient).

To do a setup from scratch (assuming Postgis for the first step), I ran the following steps, in order:

1. Create the GIS-aware database: createdb -T template_postgis imaginary_lands. Assumes you already have template_postgis created as per Geodjango setup instructions.
2. python manage.py syncdb --noinput to do the basic model creation. This will load an initial fixtures file to create an admin user. The username and password for this user are both "admin" (without the quotes).
3. python manage.py import_lands and python manage.py import_adventures to load initial shape data into the GIS models.

4. Create the GeoTiff version of the base map (only the PNG version is checked in):

   bin/create_tiff.sh

   This is a shell script that uses some GDAL utilities to make the conversion (Windows users may need to find an equivalent alternative).

You then need to start both the Django development webserver and a local Mapnik server. Start the Mapnik server by:

cd map_server
python wms.py

That will launch the server on port 8001. Start Django's development server with the usual:

python manage.py runserver

You can now browse to http://localhost:8000/ to see the island map with a few features enabled (try zooming in or panning around), or http://localhost:8000/admin/ to view the same data in the admin interface, with the imaginary island base map, instead of the normal world map.

Navigating the code

Hopefully the code is written and commented clearly enough that somebody familiar with Django can follow along. I have tried to write in a non-throwaway style, so any code you see here is how I would do something in production.

The standalone Mapnik server configuration is all in the map_server/ directory. The only piece that is particular to a local development installation is wms.py. Otherwise one would do a normal mod_wsgi installation (for example) to call map_factory.py in a production environment.

The lands/ and adventure/ directories are two small Django apps that purely manage the GIS data. Have a look at the models there, as well as the data import scripts in lands/management/commands/import_land.py and similarly in the adventure/ directory. Fairly standard GeoDjango import techniques being used there. Also note the admin.py files in both directories and how the overrides to use my imaginary map is set up via utils/admin_helper.py. Each directory contains a data/ subdirectory that contains the raw shape files that are imported into GeoDjango. You can inspect those with tools like ogrinfo, as described in the GeoDjango tutorial.

The interface/ directory contains the main HTML-generating view, as well as the views that are called by OpenLayers to populate the data (it's the web interface for the data). These would be a fair bit more fleshed out in a "real world" application, but they are correct for the small-scale operation here. The javascript code in interface/templates/interface/simple.html is also a key part of this functionality.

Best of luck!

Malcolm Tredinnick (Sydney, Australia)

---

1. I have no direct experience with either platform. However, a credible source wrote to say that installing Mapnik and Python bindings on Windows XP, SP2 was "a breeze."↩

2. In Mapnik 2.1, the Python module will again be called mapnik. Then mapnik2 name was to allow 0.7 and 2.0 to be run in parallel for a while.↩