Roads from OpenStreetMap¶
No road data of your own? Download a place from OpenStreetMap with osmnx and draw it with four lines of code.
osmnx downloads the road network of an area from OpenStreetMap. roadstyle takes its edges as they are.
import osmnx as ox # pip install osmnx
import roadstyle as rs
G = ox.graph_from_place("Tartu, Estonia", network_type="drive", simplify=False)
G = ox.simplify_graph(G, edge_attrs_differ=["bridge", "tunnel"]) # see "Bridges and tunnels" below
rs.render_edges(ox.graph_to_gdfs(G, nodes=False)).save("tartu.html")
Choose the area and the roads¶
| You have | Call |
|---|---|
| a place name | ox.graph_from_place("Södermalm, Stockholm", ...) |
| a point and a distance | ox.graph_from_point((59.315, 18.07), dist=1000, ...): (lat, lon), metres |
| a box | ox.graph_from_bbox((west, south, east, north), ...) |
| a polygon | ox.graph_from_polygon(polygon, ...): shapely, in EPSG:4326 |
Give simplify=False to each call, then simplify as in the example above. The download of a district takes one or two minutes
(Södermalm: about 90 seconds). A county or a country can take much longer.
network_type sets which roads osmnx downloads. Always give it, because the default of osmnx is "all":
network_type |
Roads |
|---|---|
"drive" |
public roads for cars |
"drive_service" |
the same, and service roads (driveways, parking aisles) |
"walk" |
all that a pedestrian can use, also footways and steps |
"bike" |
all that a cyclist can use |
"all", "all_public" |
all ways, or all public ways |
roadstyle styles each edge by its highway class, so footways and cycleways have their own look.
Two things about osmnx data¶
The graph has directions. A two-way street is two edges, one for each direction, and a one-way street is one edge. roadstyle
expects this: it draws the two directions side by side and puts arrows on the one-way streets. Do not change the graph with
ox.convert.to_undirected.
Bridges and tunnels¶
By default osmnx joins the OpenStreetMap ways that follow each other into one edge, when no junction is between them. Then a tunnel that
goes straight onto a bridge becomes one edge with both tunnel and bridge, and roadstyle draws all of it as a bridge. In Stockholm,
Söderledstunneln and Centralbron became one edge of 1.5 km.
To prevent this, download with no simplification, then simplify with no join at the start or the end of a bridge or a tunnel:
G = ox.graph_from_place(place, network_type="drive", simplify=False)
G = ox.simplify_graph(G, edge_attrs_differ=["bridge", "tunnel"])
This gives some more edges (1,470 in place of 1,361 for Södermalm). If you forget it, roadstyle gives a warning: "6 edges are tagged both bridge and tunnel and will be drawn as bridges".
A joined edge can also have lists, for example name = ['Götgatan', 'Ringvägen']. roadstyle keeps the first value of each list, and
changes the (u, v, key) index of osmnx into normal columns.
Style by an OSM tag¶
osmnx keeps the OpenStreetMap tags as columns (name, maxspeed, lanes, oneway, ...), so all of
Colour by your data works on them. maxspeed is text in OSM ("30", sometimes "SE:urban"), so make it a number
first:
import pandas as pd
edges = ox.graph_to_gdfs(G, nodes=False)
edges["speed_kmh"] = pd.to_numeric(edges["maxspeed"], errors="coerce")
rs.render_edges(edges, color_by="speed_kmh", cmap="plasma").save("speed.html")
When the download fails¶
osmnx keeps each download in a cache/ folder in the folder where you run it, so it downloads an area only once. To get new data,
delete that folder.
The downloads come from the public Overpass servers, which are sometimes busy or off. Then osmnx stops with Connection refused or a
timeout. Wait and try again, or use another server:
This is the setting of osmnx 2.x. Older examples use ox.settings.overpass_endpoint, which osmnx 2.x does not read.
More: notebooks/10_osmnx.ipynb shows all of this on
Södermalm, with the maps. An AI agent can do the same with no code: the tool render_place of the MCP server downloads and draws a
place in one call (With AI agents).