The purpose of this Python code is to parse the .osm data, which are downloadable from various links, for example here, and then parse those data to produce a real road graph, for example, the car graph.

An .osm file, for example, greece-latest.osm, or europe-latest.osm. And, a distance in meters, for example 250 which will be the highest distance that we want to have to any edge in our graph. If an edge is more than 250 when we split it and add more nodes, so that it remains short. This is optional. The reasoning behind why we do this is explained here in detail.

2 .txt files, named nodes.txt and ways.txt.

The nodes.txt file contains all the nodes, and follows the following format:

id
latitude
longitude

0 37.5042194 23.4567436
1 37.5039174 23.456488
2 37.503815 23.4567841
3 37.5036181 23.4573291
...

The ways.txt file contains all edges, and follows the following format:

A
B
forward_mode
backward_mode
forward_duration
backward_duration
duration
is_startpoint
is_access_restricted
road_name

292 293 DRIVING INACCESSIBLE 1.66359093354 1.66359093354 None True False Ρήγα Φεραίου
293 294 DRIVING INACCESSIBLE 1.42879660177 1.42879660177 None True False Ρήγα Φεραίου
294 295 DRIVING INACCESSIBLE 1.92864344384 1.92864344384 None True False Ρήγα Φεραίου
295 296 DRIVING INACCESSIBLE 1.56874919292 1.56874919292 None True False Ρήγα Φεραίου

The parsing process can be tricky, and there is more than one correct ways to do it, meaning that someone could choose to include parking or private roads in his graph and someone might opt to not include them. Another example is that our graph's edge weights could slightly differ from someone else's edge weights.

The process I followed when implementing this graph-extractor is very similar to the script of Project-Osrm as seen here.

When parsing the .osm data, we also want to associate the osm_node_ids to our own, smaller ids, named my_ids, which take values from 0 to N. The reason behind that is that we need those ids as indexes in our c++ application, which solves the shortest path problem. To do that, I am using a temporary Sqlite database (which is built-in python) to help me with the data parsing. Otherwise, an in-memory approach would require way too much memory. The Sqlite database file which contains the osm_node_id, my_id, latitude, longitude takes up 80GBs of space for the Europe car graph.

• Road's type. This is indicated in the highway tag, which takes values such as motorway, residential, living_street, etc.
• Road's texture. For example, asphalt, dirt, etc.
• Road's accessibility tags. For example, accessible by the general public, private parking lots, roads that are accessible only in certain hours, etc.
• Max-speed road signs