osmnxr

osmnxr is OSMnx for R: a tidyverse-friendly toolkit, inspired by the OSMnx Python library, to download, model, simplify, analyze and visualize street networks and other geospatial features from OpenStreetMap.

It is a reimplementation — not a Python wrapper. Heavy graph computation (routing, simplification, metrics) runs in the bundled Rust crate osmnxr-core via extendr, so there is no Python runtime to manage. Everything you get back is tidy sf.

Installation

Install the released version from CRAN (macOS and Windows binaries need no Rust; building from source needs the Rust toolchain, rustup, version 1.81 or later — the crates it depends on come with the package, nothing is downloaded):

install.packages("osmnxr")

Or the development version from GitHub (always builds from source):

# install.packages("remotes")
remotes::install_github("StrategicProjects/osmnxr")

Quick start

library(osmnxr)

# Download a drivable street network for a place
g <- ox_graph_from_place("Olinda, Brazil", network_type = "drive")
g
plot(g)

# Route between two points (Rust Dijkstra)
from <- ox_nearest_nodes(g, x = -34.85, y = -8.01)
to   <- ox_nearest_nodes(g, x = -34.84, y = -8.00)
ox_shortest_path(g, from, to)

# Urban metrics
ox_basic_stats(g)
ox_orientation_entropy(g)   # street-grid order/disorder

No network access? Explore the whole API offline with a synthetic grid:

g <- example_osm_graph()
ox_basic_stats(g)
ox_shortest_path(g, ox_nearest_nodes(g, 0, 0), ox_nearest_nodes(g, 300, 300))

How it works

osmnxr is split into a tidy R API over a Rust compute core:

The pipeline, from a place name to an analyzable network:

Related work

osmnxr complements the R geospatial stack — it can hand graphs to sfnetworks, tidygraph and dodgr, and downloads via the same Overpass API used by osmdata.

Acknowledgements

osmnxr owes its design to Geoff Boeing's OSMnx (Boeing, G. 2025. "Modeling and Analyzing Urban Networks and Amenities with OSMnx." Geographical Analysis, 57(4): 567–577. doi:10.1111/gean.70009). It follows OSMnx's API and modelling choices; the code is a reimplementation and shares nothing with it. Its data come from the OpenStreetMap contributors and the public Overpass and Nominatim services.

Thanks to everyone who files issues — user reports shape this package. Srikanth Komala Sheshachala (@talegari) proposed ox_union() for multimodal routing (#1), ox_route_tsp() for travelling-salesman routes (#4), constrained / Pareto-optimal routing with road preferences (#5) and ox_blend() for snapping points and whole graphs into a network (#6); his careful reports with reproducible benchmarks also led to automatic retries on transient Overpass failures (#2), a ~250× faster graph-building phase on metro-sized extracts (#3), and the session edge-index cache plus the ox_route_shortest() interface (#5 feedback).

License

MIT © André Leite and contributors / StrategicProjects.