How to create a graph from German OSM data with OSMnx efficiently?

Question

I am currently working on a dataset with German postal codes. I aim to calculate "real" distances between the centers of each postal code area. To do so, I first downloaded the germany-latest.osm.bz2 file from geofabrik. Next I would like to create a graph from the downloaded file for the network_type = "driving".

However, I noticed that using the osmnx.graph.graph_from_xml() method is very time consuming for a large OMS file such as for Germany.

My questions are:

1. How can I create the graph more efficiently?
2. Once the graph has been created: how can I compute the "real" distances for a large set of origin-destination-pairs?

Thank you!

I am attaching my current code for reference:

import os
from pyrosm import OSM, get_data
import osmnx as ox
import networkx as nx

os.chdir(r"myPath")

xml = "germany-latest.osm.bz2"

graph = ox.graph.graph_from_xml(xml) #how to restrict to network_type = "driving"?

Answer 1

My suggestion is to download each country separately with https://download.geofabrik.de/europe/germany.html and save them to a PostGIS Database with osm2psgl (https://osm2pgsql.org/).

Setting up a PostGIS DB is done fast and working with osm2psgl is easier than it looks as well.

You can upload a bz2 File to your PostGIS DB from geofabrik with:

osm2pgsql bayern.osm.bz2 -d geodb -U USERNAME -W -H localhost -P 5432 -S default.style --output-pgsql-schema=osm --prefix bayern

When it's uploaded to PostGIS you can use all the PostGIS SQL magic you need.