Petrock

A small, dependency-free geocoding and address-normalisation library for Python.

Petrock takes messy postal addresses and returns a normalised structure: street, house number, postcode, locality. It ships with rule sets for eleven European countries and does not call any remote service — everything runs locally against a compiled rule table of about 4 MB.

# install
pip install petrock

# normalise a single address
from petrock import parse
addr = parse("kralovska 12/4, praha 6", country="cz")
print(addr.street, addr.number, addr.postcode)

Why not an API?

Address normalisation is usually the first step of a longer pipeline, often inside a batch job with millions of rows. Round-tripping every row through a network service is slow, costs money and drags a privacy question into what should be a pure function. Petrock keeps it local and deterministic: same input, same output, same version — forever.

What it does not do

• No latitude/longitude lookup. Petrock parses and normalises, it does not geocode to coordinates.
• No fuzzy matching against a delivery-point database. If the street does not exist, Petrock will still happily normalise it.
• No validation of whether a building number is real.

Supported countries