Awesome
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.
<img src="the_earth_gshhg.png" width="100%" />(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:
pip install roaring-landmask
To build from source, you can use pip:
pip install .
or maturin:
-
Install maturin.
-
Build and install
maturin build --release
pip install target/wheels/... # choose your whl