FastNSA

FastNSA is a high-performance toolkit for network-constrained spatial analysis, designed for large-scale road networks and massive point datasets. It provides a Python interface backed by a native implementation for efficient spatial indexing and network-based statistical computation.

Installation

FastNSA is distributed via PyPI with prebuilt native backends.

pip install fastnsa

Supported platforms:

• Linux (x86_64)

• Python ≥ 3.9

Quick Example

import fastnsa as nsa

# 1. Build a road network from OpenStreetMap
# (internally constructs spatial and network indexes)
net = nsa.Network.from_osm(place="Central Park, NY")

# 2. Load event points and project them onto the network
gdf = nsa.load_points("accidents.csv")
events = nsa.PointEvents(gdf, target_network=net)

# 3. Compute a network-based K-function
results = nsa.network_k_function(
    network=net,
    points=events,
    r_values=[10, 50, 100]
)

print(results)

Spatiotemporal Network K-function

import fastnsa as nsa
import numpy as np

net = nsa.Network.from_osm(place="Central Park, NY")
gdf = nsa.load_points("accidents_with_time.csv", time_col="event_time")
events = nsa.PointEvents(gdf, target_network=net, time_col="event_time")

st_result = nsa.spatiotemporal_network_k_function(
    network=net,
    points=events,
    s_values=np.array([50.0, 100.0, 150.0]),
    t_values=np.array([1.0, 3.0, 7.0]),
    method="EARTH",
    simulations=99,
    alpha=0.05,
)

print(st_result.k_obs.shape)  # (n_s, n_t)

Design Overview

The Python layer provides:

• Network construction and I/O

• Event projection to networks

• tatistical analysis interfaces

Performance-critical components such as spatial indexing, point-to-network mapping, and distance evaluation are implemented in a native backend and exposed through Python bindings.

License

FastNSA is licensed under the GNU Lesser General Public License v3.0 (LGPL-2.1).

You are free to use this library in proprietary and open-source software, but any modifications to the FastNSA library itself must be released under LGPL.