June 2, 2021


geospatial python geocode osm

Learned about a couple of handy python geospatial projects.

PyPostal

Python bindings for the libpostal library. It basically provides a way to standardize an input string into something approximately a geocodable address. Note: it does not actually do geocoding; it does string standardization to match a wide array of international addressing schemes.

Following their instructions on a Mac:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20

mkdir address
cd address
python3 -m venv venv
brew install curl autoconf automake libtool pkg-config\n
git clone https://github.com/openvenues/libpostal\n
./bootstrap.sh\n
cd libpostal
./bootstrap.sh

# I'm just sticking whatever files they need in the local direction.
# No harm for a demo.
./configure --datadir=/Users/thomas/tmp/address/postaldata
make
sudo make install
cd ..
ls
source venv/bin/activate
ls postaldata
ls postaldata/libpostal
pip install postal

And now we’re ready to use it in Python. I just tried in an interactive python3 shell:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18

from postal.expand import expand_address
# This import takes seconds...I guess it's referencing a huge set of rules/data
from postal.parser import parse_address

# I have a friend living in a house in a town in Yorkshire with
# a typically English off the wall address scheme. Trying something like this
>>> parse_address('shoreditch, the hill, yarm, yorkshire')

# Not too bad given the weird input
[('shoreditch the hill', 'road'), ('yarm', 'city'), ('yorkshire', 'state_district')]

# One of their examples
>>> parse_address('The Book Club 100-106 Leonard St, Shoreditch, London, Greater London, EC2A 4RH, United Kingdom')
[('the book club', 'house'), ('100-106', 'house_number'),
  ('leonard st', 'road'),
  ('shoreditch', 'suburb'), ('london', 'city'),
  ('greater london', 'state_district'), ('ec2a 4rh', 'postcode'),
  ('united kingdom', 'country')]

I can see a lot of utility here. Good stuff.

OSMnx

In wondering how to get an isochrone on a street network I came across OSMnx. It’s a utility library and set of algorithms for both accessing (downloading) Open Street Map (OSM) data, and a set of algorithms for working with that data in a Python-native graph data structure. For example…how to download data for a city, build a walking graph of the street network, then calculate how far you can get from a single location in $X$ minutes.

The examples repo shows exactly how to do this.

See Also