Outputting JSON with reduced floating point precision#

datasette-leaflet-geojson outputs GeoJSON geometries in HTML pages in a way that can be picked up by JavaScript and used to plot a Leaflet map.

These geometries often look something like this:

1
{
2
  "type":"MultiPolygon",
3
  "coordinates":[[[[-122.4457678900319,37.77292891669105],[-122.441075063058,37.77352490695095]...

Decimal_degrees: Precision on Wikipedia says that 0.00001 should be accurate to within around a meter.

Shortening these floating point representations can shave 100KB+ off an HTML page with a lot of GeoJSON shapes on it!

There’s a lengthy Stack Overflow about this but it’s difficult to follow because ways of doing this changed between Python 2 and Python 3. Here’s what worked for me:

1
def round_floats(o):
2
    if isinstance(o, float):
3
        return round(o, 5)
4
    if isinstance(o, dict):
5
        return {k: round_floats(v) for k, v in o.items()}
6
    if isinstance(o, (list, tuple)):
7
        return [round_floats(x) for x in o]
8
    return o
9

10
data = json.dumps(round_floats(data))

See issue 11 for details.