Skip to main content

Fast linestring simplification using RDP or Visvalingam-Whyatt and a Rust binary

Project description

Build Status Coverage Status DownloadsDOI


Simplify a LineString using the Ramer–Douglas–Peucker or Visvalingam-Whyatt algorithms



pip install simplification
Please use a recent (>= 8.1.2) version of pip.

Supported Python Versions (Linux x86_64, macOS x86_64 + aarch64, Windows amd64)

  • Python 3.7
  • Python 3.8
  • Python 3.9
  • Python 3.10

Supported Platforms

  • Linux (manylinux-compatible) x86_64
  • macOS Darwin x86_64 and aarch64
  • Windows 64-bit


from simplification.cutil import (

# Using Ramer–Douglas–Peucker
coords = [
    [0.0, 0.0],
    [5.0, 4.0],
    [11.0, 5.5],
    [17.3, 3.2],
    [27.8, 0.1]

# For RDP, Try an epsilon of 1.0 to start with. Other sensible values include 0.01, 0.001
simplified = simplify_coords(coords, 1.0)

# simplified is [[0.0, 0.0], [5.0, 4.0], [11.0, 5.5], [27.8, 0.1]]

# Using Visvalingam-Whyatt
# You can also pass numpy arrays, in which case you'll get numpy arrays back
import numpy as np
coords_vw = np.array([
    [5.0, 2.0],
    [3.0, 8.0],
    [6.0, 20.0],
    [7.0, 25.0],
    [10.0, 10.0]
simplified_vw = simplify_coords_vw(coords_vw, 30.0)

# simplified_vw is [[5.0, 2.0], [7.0, 25.0], [10.0, 10.0]]

Passing empty and/or 1-element lists will return them unaltered.

But I only want the simplified Indices

simplification now has:

  • cutil.simplify_coords_idx
  • cutil.simplify_coords_vw_idx

The values returned by these functions are the retained indices. In order to use them as e.g. a masked array in Numpy, something like the following will work:

import numpy as np
from simplification.cutil import simplify_coords_idx

# assume an array of coordinates: orig
simplified = simplify_coords_idx(orig, 1.0)
# build new geometry using only retained coordinates
orig_simplified = orig[simplified]

But I need to ensure that the resulting geometries are valid

You can use the topology-preserving variant of VW for this: simplify_coords_vwp. It's slower, but has a far greater likelihood of producing a valid geometry.

But I Want to Simplify Polylines

No problem; Decode them to LineStrings first.

# pip install pypolyline before you do this
from pypolyline.cutil import decode_polyline
# an iterable of Google-encoded Polylines, so precision is 5. For OSRM &c., it's 6
decoded = (decode_polyline(line, 5) for line in polylines)
simplified = [simplify_coords(line, 1.0) for line in decoded]

How it Works

FFI and a Rust binary

Is It Fast

I should think so.

What does that mean

Using numpy arrays for input and output, the library can be reasonably expected to process around 2500 1000-point LineStrings per second on a Core i7 or equivalent, for a 98%+ reduction in size.
A larger LineString, containing 200k+ points can be reduced to around 3k points (98.5%+) in around 50ms using RDP.

This is based on a test harness available here.


All benchmarks are subjective, and pathological input will greatly increase processing time. Error-checking is non-existent at this point.



Citing Simplification

If Simplification has been significant in your research, and you would like to acknowledge the project in your academic publication, we suggest citing it as follows (example in APA style, 7th edition):

Hügel, S. (2021). Simplification (Version X.Y.Z) [Computer software].

In Bibtex format:

author = {Hügel, Stephan},
doi = {10.5281/zenodo.5774852},
license = {MIT},
month = {12},
title = {{Simplification}},
url = {},
version = {X.Y.Z},
year = {2021}

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Built Distributions

simplification-0.6.1-cp310-cp310-win_amd64.whl (287.2 kB view hashes)

Uploaded cp310

simplification-0.6.1-cp39-cp39-win_amd64.whl (287.2 kB view hashes)

Uploaded cp39

simplification-0.6.1-cp38-cp38-win_amd64.whl (287.2 kB view hashes)

Uploaded cp38

simplification-0.6.1-cp37-cp37m-win_amd64.whl (286.6 kB view hashes)

Uploaded cp37

Supported by

AWS AWS Cloud computing Datadog Datadog Monitoring Facebook / Instagram Facebook / Instagram PSF Sponsor Fastly Fastly CDN Google Google Object Storage and Download Analytics Huawei Huawei PSF Sponsor Microsoft Microsoft PSF Sponsor NVIDIA NVIDIA PSF Sponsor Pingdom Pingdom Monitoring Salesforce Salesforce PSF Sponsor Sentry Sentry Error logging StatusPage StatusPage Status page