GeoTable
Read and write spatial vectors in the following formats thanks to GDAL and pandas.
- GeoJSON
- KMZ
- SHP
- CSV
Install
sudo dnf -y install gdal-python3 libkml
# sudo apt-get -y install python3-gdal libkml
virtualenv -p $(which python3) --system-site-packages \
~/.virtualenvs/crosscompute
source ~/.virtualenvs/crosscompute/bin/activate
pip install -U geotable
Use
If you cloned or downloaded the repository, you can run these examples in the tests folder.
$ cd tests
Load URLs.
In [1]: import geotable
In [2]: t = geotable.load(
'https://data.cityofnewyork.us/api/geospatial/tqmj-j8zm'
'?method=export&format=Original')
Load KMZ files.
In [1]: import geotable
In [2]: t = geotable.load('xyz.kmz')
Load shapefiles.
In [1]: import geotable
In [2]: t = geotable.load('shp.zip')
In [3]: t.iloc[0]
Out[3]:
name b
quantity 2
cost 0.66
date 1990-01-01 00:00:00
geometry_object POINT (-91.5305465 14.8520705)
geometry_layer b
geometry_proj4 +proj=longlat +datum=WGS84 +no_defs
Name: 0, dtype: object
Load CSVs containing spatial information.
geotable.load('csv/wkt.csv') # Load single CSV
geotable.load('csv.zip') # Load archive of multiple CSVs
geotable.load('csv.zip', parse_dates=['date']) # Configure pandas.read_csv
Handle CSVs with different geometry columns.
$ cat csv/latitude_longitude.csv name,quantity,cost,date,latitude,longitude b,2,0.66,1990-01-01,14.8520705,-91.5305465 $ cat csv/lat_lon.csv name,quantity,cost,date,lat,lon c,3,0.99,2000-01-01,42.2808256,-83.7430378 $ cat csv/latitude_longitude_wkt.csv name,quantity,cost,date,latitude_longitude_wkt a,1,0.33,1980-01-01,POINT(42.3736158 -71.10973349999999) $ cat csv/longitude_latitude_wkt.csv name,quantity,cost,date,longitude_latitude_wkt a,1,0.33,1980-01-01,POINT(-71.10973349999999 42.3736158) $ cat csv/wkt.csv name,quantity,cost,date,wkt aaa,1,0.33,1980-01-01,"POINT(-71.10973349999999 42.3736158)" bbb,1,0.33,1980-01-01,"LINESTRING(-122.1374637 37.3796627,-92.5807231 37.1067189)" ccc,1,0.33,1980-01-01,"POLYGON ((-83.10973350093332 42.37361082304877, -103.5305394806998 14.85206885307358, -95.7430260175515 42.28082607112266, -83.10973350093332 42.37361082304877))"
Handle CSVs with different spatial references.
$ cat proj4_from_file.csv name,wkt aaa,"POLYGON((326299 4693415,-1980130 1771892,-716771 4787516,326299 4693415))" $ cat proj4_from_file.proj4 +proj=utm +zone=17 +ellps=WGS84 +datum=WGS84 +units=m +no_defs $ cat proj4_from_row.csv name,wkt,geometry_layer,geometry_proj4 aaa,"LINESTRING(-122.1374637 37.3796627,-92.5807231 37.1067189)",l1,+proj=longlat +datum=WGS84 +no_defs aaa,"POLYGON((326299 4693415,-1980130 1771892,-716771 4787516,326299 4693415))",l2,+proj=utm +zone=17 +ellps=WGS84 +datum=WGS84 +units=m +no_defs
Load and save in different spatial references.
from geotable.projections import SPHERICAL_MERCATOR_PROJ4
t = geotable.load('shp.zip', target_proj4=SPHERICAL_MERCATOR_PROJ4)
from geotable.projections import LONGITUDE_LATITUDE_PROJ4
t.save_shp('/tmp/shp.zip', target_proj4=LONGITUDE_LATITUDE_PROJ4)
Use LONGITUDE_LATITUDE_PROJ4 for compatibility with algorithms that use geodesic distance such as those found in geopy and pysal. Geodesic distance is also known as arc distance and is the distance between two points as measured using the curvature of the Earth. If your locations are spread over a large geographic extent, geodesic longitude and latitude coordinates provide greater accuracy than Euclidean XY coordinates.
from geotable.projections import LONGITUDE_LATITUDE_PROJ4
t = geotable.load('shp.zip', target_proj4=LONGITUDE_LATITUDE_PROJ4)
t.save_csv('/tmp/csv.zip', target_proj4=LONGITUDE_LATITUDE_PROJ4)
t.save_shp('/tmp/shp.zip', target_proj4=LONGITUDE_LATITUDE_PROJ4)
t.save_kmz('/tmp/xyz.kmz', target_proj4=LONGITUDE_LATITUDE_PROJ4)
Use the Universal Transverse Mercator (UTM) projection for compatibility with algorithms that use Euclidean distance on XY coordinates such as those found in scipy.spatial. If you know that your locations are confined to a small region, you can use the projected XY coordinates with standard Euclidean based algorithms, which tend to be significantly faster than their geodesic variants.
utm_proj4 = geotable.load_utm_proj4('shp.zip')
t = geotable.load('csv.zip', target_proj4=utm_proj4)
t.save_csv('/tmp/csv.zip', target_proj4=utm_proj4)
t.save_shp('/tmp/shp.zip', target_proj4=utm_proj4)
t.save_kmz('/tmp/xyz.kmz', target_proj4=utm_proj4)
Use the Spherical Mercator projection when visualization is more important than accuracy. Do not use this projection for algorithms where spatial accuracy is important.
from geotable.projections import SPHERICAL_MERCATOR_PROJ4
t = geotable.load('csv/wkt.csv', target_proj4=SPHERICAL_MERCATOR_PROJ4)
t.save_csv('/tmp/csv.zip', target_proj4=SPHERICAL_MERCATOR_PROJ4)
t.save_shp('/tmp/shp.zip', target_proj4=SPHERICAL_MERCATOR_PROJ4)
t.save_kmz('/tmp/xyz.kmz', target_proj4=SPHERICAL_MERCATOR_PROJ4)
You can render your spatial vectors in Jupyter Notebook with the draw function. Each geometry layer will appear in a different color.
t = geotable.load('csv/wkt.csv')
t.draw() # Render the geometries in Jupyter Notebook
You can also use ColorfulGeometryCollection in Jupyter Notebook directly.
from geotable import ColorfulGeometryCollection from shapely.geometry import Point ColorfulGeometryCollection([Point(0, 0), Point(1, 1)])
Here are some other convenience functions.
import geotable
# Show WKT for first geometry
geotable.load('xyz.kmz').geometries[0].wkt
# Load without z coordinates
geotable.load('xyz.kmz', drop_z=True).geometries[0].wkt
# Restrict geometries to bounding box
geotable.load('xyz.kmz', bounding_box=(-71.2, 42.37, -71.1, 42.38))
# Restrict geometries to bounding polygon
from shapely.geometry import Polygon
polygon = Polygon([
(-71.2, 42.37),
(-71.1, 42.37),
(-71.1, 42.38),
(-71.2, 42.38)])
geotable.load('xyz.kmz', bounding_polygon=polygon)
# Load files according to a reference file's UTM projection
reference_path = 'xyz.kmz'
load = geotable.define_load_with_utm_proj4(reference_path)
load('csv/wkt.csv')
Test
pip install pytest pytest-cov py.test --cov-report term-missing --cov=geotable tests
