#bitmap #gis #openstreetmap #point #roaring-bitmaps #landmask #gshhg

bin+lib roaring-landmask

A fast and limited-memory structure with a landmask based on GSHHG for determing whether a point on Earth is on land or in the ocean

14 unstable releases (6 breaking)

0.9.0 Sep 11, 2024
0.7.1 Nov 7, 2022
0.5.1 Apr 1, 2022
0.5.0 Mar 24, 2022
0.2.1 Jun 14, 2021

#59 in Geospatial

Download history 133/week @ 2024-09-03 167/week @ 2024-09-10 22/week @ 2024-09-17 11/week @ 2024-09-24 21/week @ 2024-10-01

562 downloads per month

MIT license

555KB
910 lines

Crates.io Documentation PyPI Actions Status

The Roaring Landmask

Have you ever needed to know whether you are in the ocean or on land? And you need to know it fast? And you need to know it without using too much memory or too much disk? Then try the Roaring Landmask!

The roaring landmask is a Rust + Python package for quickly determining whether a point given in latitude and longitude is on land or not. A landmask is stored in a tree of Roaring Bitmaps. Points close to the shore might still be in the ocean, so a positive value is then checked against the vector shapes of the coastline.

(source)

The landmask is generated from the GSHHG shoreline database (Wessel, P., and W. H. F. Smith, A Global Self-consistent, Hierarchical, High-resolution Shoreline Database, J. Geophys. Res., 101, 8741-8743, 1996) and the OpenStreetMap (© OpenStreetMap).

An alternative is the opendrift-landmask-data, which is slightly faster, is pure Python, but requires more memory and disk space (memory-mapped 3.7Gb).

Performance

Microbenchmarks:

test tests::test_contains_in_ocean         ... bench:          24 ns/iter (+/- 0)
test tests::test_contains_on_land          ... bench:       3,795 ns/iter (+/- 214)

Many points, through Python:

------------------------------------------------------------------------------------------------------ benchmark: 5 tests -----------------------------------------------------------------------------------------------------
Name (time in us)                       Min                     Max                    Mean                StdDev                  Median                   IQR            Outliers           OPS            Rounds  Iterations
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
test_landmask_many_par          34,335.6220 (>1000.0)   39,922.9660 (>1000.0)   36,167.6438 (>1000.0)  1,602.6359 (>1000.0)   35,658.2270 (>1000.0)  1,722.6990 (>1000.0)       9;1       27.6490 (0.00)         30           1
test_landmask_many             130,760.1480 (>1000.0)  131,155.3400 (>1000.0)  130,863.7110 (>1000.0)    137.1064 (598.03)   130,809.7410 (>1000.0)    135.3770 (>1000.0)       1;1        7.6415 (0.00)          8           1
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

The parallel version is significantly faster, while the sequential version is slightly slower than the equivalent benchmark in opendrift-landmask-data, which uses about 120 ms.

Usage from Python

from roaring_landmask import RoaringLandmask

l = RoaringLandmask.new()
x = np.arange(-180, 180, .5)
y = np.arange(-90, 90, .5)

xx, yy = np.meshgrid(x,y)

print ("points:", len(xx.ravel()))
on_land = l.contains_many(xx.ravel(), yy.ravel())

Building & installing

Pre-built wheels are available on PyPI:

  1. pip install roaring-landmask

To build from source, you can use pip:

  1. pip install .

or maturin:

  1. Install maturin.

  2. Build and install

maturin build --release
pip install target/wheels/... # choose your whl

Dependencies

~11–22MB
~325K SLoC