Originally the final project for HIS 575 at Southern Connecticut State University, Spring 2016, Professor Troy Paddock.
Use the KyngChaos binary to install GDAL
Some people use Homebrew. Out of inertia, concern for running non-Homebrew stuff in parallel with Homebrew, and not knowing the consequences of installing Homebrew on a machine that's had a lot installed without it, I don't. But people have documented installing GDAL with Homebrew
clhenrick on GitHub has some nifty shell scripts for doing some things with GDAL quickly.
Get GeoJSON data or get data into GeoJSON
I made the choice to use modern boundaries and used a Polish government site for Ciechocinek, Łódź, Lowicz, and Sochaczew, then found Mapzen's metro extracts service and used it for Lisboa/Lisbon, Metz, Biarritz, Nancy, Zbaszn (changed to Zbąszynek after the war), and Saarbrücken.
The Polish government files were ESRI shapefiles, so I needed to convert them with GDAL, making sure to set the output to UTF-8 as they were previously in Windows-1250 encoding (understandable given the source, but really, folks, do everything in UTF-8 if you have any control at all over that):
export SHAPE_ENCODING="Windows-1250"
ogr2ogr -f GeoJSON -t_srs crs:84 gminy-municipalities.geojson gminy.shp -lco ENCODING=UTF-8
I found a complete set of French commune boundary files, but it was in MapInfo format, so I used ogr2ogr to transform them. Simple format change was not terribly complicated:
ogr2ogr -f GeoJSON [target datafile].tab [source datafile]
But then it turned out that they used RGF93 /Lambert-93 projection, which was initially beyond my ken. To the Internet! And I found a helpful blog post from Frank Donnelly of Baruch College describing what to do.
ogr2ogr -f GeoJSON -t_srs EPSG:4326 communes-00.geojson COMMUNE.tab
It's a good idea to validate your GeoJSON syntax. In this project, I tended to use JSONLint. There's probably a shell utility that will do that, if you're strictly a shell person.
Though it's not validation, strictly speaking, I also used JSON Pretty Print to neaten things up. Helps make it more readable.
Use this regex to cut the number of points in a complex polygon in half. I don't need the city boundaries to be more accurate than that. For Łódź I cut it down into 1/4 or maybe even 1/8 of the original number of points.
Find:
(\[ -?\d{1,2}\.\d{9,15}, \d{1,2}\.\d{9,15} \],) \[ -?\d{1,2}\.\d{9,15}, \d{1,2}\.\d{9,15} \],
Replace:
\1
Then strip the prefatory JSON until all you have are the coordinates for the polygon and the brackets and use this regex to format the pairs:
Find:
\[ (-?\d{1,2}\.\d{9,15}), (\d{1,2}\.\d{9,15}) \],
Replace:
\1 \2,
Do what you need to do to end up with a string that looks like
"POLYGON((LON LAT,LON LAT,etc.))"
Use it with Python to get the centroid of a city's polygon with a variation of Jared Erickson's create centroids recipe in his Python+GDAL/OGR cookbook.
tl;dr:
import ogr
poly_wkt= "[string from above text manipulations]"
geom_poly = ogr.CreateGeometryFromWkt(poly_wkt)
centroid = geom_poly.Centroid()
print centroid # get the value pair from this output