GeoDjango is an included contrib module for Django that turns it into a world-class geographic Web framework. GeoDjango strives to make it as simple as possible to create geographic Web applications, like location-based services. Its features include:
This tutorial assumes familiarity with Django; thus, if you’re brand new to Django, please read through the regular tutorial to familiarize yourself with Django first.
Note
GeoDjango has additional requirements beyond what Django requires – please consult the installation documentation for more details.
This tutorial will guide you through the creation of a geographic web application for viewing the world borders. [1] Some of the code used in this tutorial is taken from and/or inspired by the GeoDjango basic apps project. [2]
Note
Proceed through the tutorial sections sequentially for step-by-step instructions.
Typically no special setup is required, so you can create a database as you would for any other project. We provide some tips for selected databases:
Use the standard django-admin script to create a project called
geodjango:
$ django-admin startproject geodjango
...\> django-admin startproject geodjango
This will initialize a new project. Now, create a world Django application
within the geodjango project:
$ cd geodjango
$ python manage.py startapp world
...\> cd geodjango
...\> py manage.py startapp world
settings.py¶The geodjango project settings are stored in the geodjango/settings.py
file. Edit the database connection settings to match your setup:
DATABASES = {
'default': {
'ENGINE': 'django.contrib.gis.db.backends.postgis',
'NAME': 'geodjango',
'USER': 'geo',
},
}
In addition, modify the INSTALLED_APPS setting to include
django.contrib.admin, django.contrib.gis,
and world (your newly created application):
INSTALLED_APPS = [
'django.contrib.admin',
'django.contrib.auth',
'django.contrib.contenttypes',
'django.contrib.sessions',
'django.contrib.messages',
'django.contrib.staticfiles',
'django.contrib.gis',
'world',
]
The world borders data is available in this zip file. Create a data
directory in the world application, download the world borders data, and
unzip. On GNU/Linux platforms, use the following commands:
$ mkdir world/data
$ cd world/data
$ wget https://thematicmapping.org/downloads/TM_WORLD_BORDERS-0.3.zip
$ unzip TM_WORLD_BORDERS-0.3.zip
$ cd ../..
...\> mkdir world\data
...\> cd world\data
...\> wget https://thematicmapping.org/downloads/TM_WORLD_BORDERS-0.3.zip
...\> unzip TM_WORLD_BORDERS-0.3.zip
...\> cd ..\..
The world borders ZIP file contains a set of data files collectively known as an ESRI Shapefile, one of the most popular geospatial data formats. When unzipped, the world borders dataset includes files with the following extensions:
.shp: Holds the vector data for the world borders geometries..shx: Spatial index file for geometries stored in the .shp..dbf: Database file for holding non-geometric attribute data
(e.g., integer and character fields)..prj: Contains the spatial reference information for the geographic
data stored in the shapefile.ogrinfo to examine spatial data¶The GDAL ogrinfo utility allows examining the metadata of shapefiles or
other vector data sources:
$ ogrinfo world/data/TM_WORLD_BORDERS-0.3.shp
INFO: Open of `world/data/TM_WORLD_BORDERS-0.3.shp'
using driver `ESRI Shapefile' successful.
1: TM_WORLD_BORDERS-0.3 (Polygon)
...\> ogrinfo world\data\TM_WORLD_BORDERS-0.3.shp
INFO: Open of `world/data/TM_WORLD_BORDERS-0.3.shp'
using driver `ESRI Shapefile' successful.
1: TM_WORLD_BORDERS-0.3 (Polygon)
ogrinfo tells us that the shapefile has one layer, and that this
layer contains polygon data. To find out more, we’ll specify the layer name
and use the -so option to get only the important summary information:
$ ogrinfo -so world/data/TM_WORLD_BORDERS-0.3.shp TM_WORLD_BORDERS-0.3
INFO: Open of `world/data/TM_WORLD_BORDERS-0.3.shp'
using driver `ESRI Shapefile' successful.
Layer name: TM_WORLD_BORDERS-0.3
Geometry: Polygon
Feature Count: 246
Extent: (-180.000000, -90.000000) - (180.000000, 83.623596)
Layer SRS WKT:
GEOGCS["GCS_WGS_1984",
DATUM["WGS_1984",
SPHEROID["WGS_1984",6378137.0,298.257223563]],
PRIMEM["Greenwich",0.0],
UNIT["Degree",0.0174532925199433]]
FIPS: String (2.0)
ISO2: String (2.0)
ISO3: String (3.0)
UN: Integer (3.0)
NAME: String (50.0)
AREA: Integer (7.0)
POP2005: Integer (10.0)
REGION: Integer (3.0)
SUBREGION: Integer (3.0)
LON: Real (8.3)
LAT: Real (7.3)
...\> ogrinfo -so world\data\TM_WORLD_BORDERS-0.3.shp TM_WORLD_BORDERS-0.3
INFO: Open of `world/data/TM_WORLD_BORDERS-0.3.shp'
using driver `ESRI Shapefile' successful.
Layer name: TM_WORLD_BORDERS-0.3
Geometry: Polygon
Feature Count: 246
Extent: (-180.000000, -90.000000) - (180.000000, 83.623596)
Layer SRS WKT:
GEOGCS["GCS_WGS_1984",
DATUM["WGS_1984",
SPHEROID["WGS_1984",6378137.0,298.257223563]],
PRIMEM["Greenwich",0.0],
UNIT["Degree",0.0174532925199433]]
FIPS: String (2.0)
ISO2: String (2.0)
ISO3: String (3.0)
UN: Integer (3.0)
NAME: String (50.0)
AREA: Integer (7.0)
POP2005: Integer (10.0)
REGION: Integer (3.0)
SUBREGION: Integer (3.0)
LON: Real (8.3)
LAT: Real (7.3)
This detailed summary information tells us the number of features in the layer
(246), the geographic bounds of the data, the spatial reference system
(“SRS WKT”), as well as type information for each attribute field. For example,
FIPS: String (2.0) indicates that the FIPS character field has
a maximum length of 2. Similarly, LON: Real (8.3) is a floating-point
field that holds a maximum of 8 digits up to three decimal places.
Now that you’ve examined your dataset using ogrinfo, create a GeoDjango
model to represent this data:
from django.contrib.gis.db import models
class WorldBorder(models.Model):
# Regular Django fields corresponding to the attributes in the
# world borders shapefile.
name = models.CharField(max_length=50)
area = models.IntegerField()
pop2005 = models.IntegerField('Population 2005')
fips = models.CharField('FIPS Code', max_length=2)
iso2 = models.CharField('2 Digit ISO', max_length=2)
iso3 = models.CharField('3 Digit ISO', max_length=3)
un = models.IntegerField('United Nations Code')
region = models.IntegerField('Region Code')
subregion = models.IntegerField('Sub-Region Code')
lon = models.FloatField()
lat = models.FloatField()
# GeoDjango-specific: a geometry field (MultiPolygonField)
mpoly = models.MultiPolygonField()
# Returns the string representation of the model.
def __str__(self):
return self.name
Note that the models module is imported from django.contrib.gis.db.
The default spatial reference system for geometry fields is WGS84 (meaning
the SRID is 4326) – in other words, the field coordinates are in
longitude, latitude pairs in units of degrees. To use a different
coordinate system, set the SRID of the geometry field with the srid
argument. Use an integer representing the coordinate system’s EPSG code.
migrate¶After defining your model, you need to sync it with the database. First, create a database migration:
$ python manage.py makemigrations
Migrations for 'world':
world/migrations/0001_initial.py:
- Create model WorldBorder
...\> py manage.py makemigrations
Migrations for 'world':
world/migrations/0001_initial.py:
- Create model WorldBorder
Let’s look at the SQL that will generate the table for the WorldBorder
model:
$ python manage.py sqlmigrate world 0001
...\> py manage.py sqlmigrate world 0001
This command should produce the following output:
BEGIN;
--
-- Create model WorldBorder
--
CREATE TABLE "world_worldborder" (
"id" serial NOT NULL PRIMARY KEY,
"name" varchar(50) NOT NULL,
"area" integer NOT NULL,
"pop2005" integer NOT NULL,
"fips" varchar(2) NOT NULL,
"iso2" varchar(2) NOT NULL,
"iso3" varchar(3) NOT NULL,
"un" integer NOT NULL,
"region" integer NOT NULL,
"subregion" integer NOT NULL,
"lon" double precision NOT NULL,
"lat" double precision NOT NULL
"mpoly" geometry(MULTIPOLYGON,4326) NOT NULL
)
;
CREATE INDEX "world_worldborder_mpoly_id" ON "world_worldborder" USING GIST ( "mpoly" );
COMMIT;
If this looks correct, run migrate to create this table in the
database:
$ python manage.py migrate
Operations to perform:
Apply all migrations: admin, auth, contenttypes, sessions, world
Running migrations:
...
Applying world.0001_initial... OK
...\> py manage.py migrate
Operations to perform:
Apply all migrations: admin, auth, contenttypes, sessions, world
Running migrations:
...
Applying world.0001_initial... OK
This section will show you how to import the world borders shapefile into the database via GeoDjango models using the LayerMapping data import utility.
There are many different ways to import data into a spatial database – besides the tools included within GeoDjango, you may also use the following:
Earlier, you used ogrinfo to examine the contents of the world borders
shapefile. GeoDjango also includes a Pythonic interface to GDAL’s powerful OGR
library that can work with all the vector data sources that OGR supports.
First, invoke the Django shell:
$ python manage.py shell
...\> py manage.py shell
If you downloaded the World Borders data earlier in the
tutorial, then you can determine its path using Python’s built-in
os module:
>>> import os
>>> import world
>>> world_shp = os.path.abspath(os.path.join(os.path.dirname(world.__file__),
... 'data', 'TM_WORLD_BORDERS-0.3.shp'))
Now, open the world borders shapefile using GeoDjango’s
DataSource interface:
>>> from django.contrib.gis.gdal import DataSource
>>> ds = DataSource(world_shp)
>>> print(ds)
/ ... /geodjango/world/data/TM_WORLD_BORDERS-0.3.shp (ESRI Shapefile)
Data source objects can have different layers of geospatial features; however, shapefiles are only allowed to have one layer:
>>> print(len(ds))
1
>>> lyr = ds[0]
>>> print(lyr)
TM_WORLD_BORDERS-0.3
You can see the layer’s geometry type and how many features it contains:
>>> print(lyr.geom_type)
Polygon
>>> print(len(lyr))
246
Note
Unfortunately, the shapefile data format does not allow for greater
specificity with regards to geometry types. This shapefile, like
many others, actually includes MultiPolygon geometries, not Polygons.
It’s important to use a more general field type in models: a
GeoDjango MultiPolygonField will accept a Polygon geometry, but a
PolygonField will not accept a MultiPolygon type geometry. This
is why the WorldBorder model defined above uses a MultiPolygonField.
The Layer may also have a spatial reference
system associated with it. If it does, the srs attribute will return a
SpatialReference object:
>>> srs = lyr.srs
>>> print(srs)
GEOGCS["GCS_WGS_1984",
DATUM["WGS_1984",
SPHEROID["WGS_1984",6378137.0,298.257223563]],
PRIMEM["Greenwich",0.0],
UNIT["Degree",0.0174532925199433]]
>>> srs.proj4 # PROJ.4 representation
'+proj=longlat +ellps=WGS84 +datum=WGS84 +no_defs '
This shapefile is in the popular WGS84 spatial reference system – in other words, the data uses longitude, latitude pairs in units of degrees.
In addition, shapefiles also support attribute fields that may contain additional data. Here are the fields on the World Borders layer:
>>> print(lyr.fields)
['FIPS', 'ISO2', 'ISO3', 'UN', 'NAME', 'AREA', 'POP2005', 'REGION', 'SUBREGION', 'LON', 'LAT']
The following code will let you examine the OGR types (e.g. integer or string) associated with each of the fields:
>>> [fld.__name__ for fld in lyr.field_types]
['OFTString', 'OFTString', 'OFTString', 'OFTInteger', 'OFTString', 'OFTInteger', 'OFTInteger', 'OFTInteger', 'OFTInteger', 'OFTReal', 'OFTReal']
You can iterate over each feature in the layer and extract information from both
the feature’s geometry (accessed via the geom attribute) as well as the
feature’s attribute fields (whose values are accessed via get()
method):
>>> for feat in lyr:
... print(feat.get('NAME'), feat.geom.num_points)
...
Guernsey 18
Jersey 26
South Georgia South Sandwich Islands 338
Taiwan 363
Layer objects may be sliced:
>>> lyr[0:2]
[<django.contrib.gis.gdal.feature.Feature object at 0x2f47690>, <django.contrib.gis.gdal.feature.Feature object at 0x2f47650>]
And individual features may be retrieved by their feature ID:
>>> feat = lyr[234]
>>> print(feat.get('NAME'))
San Marino
Boundary geometries may be exported as WKT and GeoJSON:
>>> geom = feat.geom
>>> print(geom.wkt)
POLYGON ((12.415798 43.957954,12.450554 ...
>>> print(geom.json)
{ "type": "Polygon", "coordinates": [ [ [ 12.415798, 43.957954 ], [ 12.450554, 43.979721 ], ...
LayerMapping¶To import the data, use a LayerMapping in a Python script.
Create a file called load.py inside the world application,
with the following code:
import os
from django.contrib.gis.utils import LayerMapping
from .models import WorldBorder
world_mapping = {
'fips' : 'FIPS',
'iso2' : 'ISO2',
'iso3' : 'ISO3',
'un' : 'UN',
'name' : 'NAME',
'area' : 'AREA',
'pop2005' : 'POP2005',
'region' : 'REGION',
'subregion' : 'SUBREGION',
'lon' : 'LON',
'lat' : 'LAT',
'mpoly' : 'MULTIPOLYGON',
}
world_shp = os.path.abspath(
os.path.join(os.path.dirname(__file__), 'data', 'TM_WORLD_BORDERS-0.3.shp'),
)
def run(verbose=True):
lm = LayerMapping(WorldBorder, world_shp, world_mapping, transform=False)
lm.save(strict=True, verbose=verbose)
A few notes about what’s going on:
world_mapping dictionary corresponds to a field in the
WorldBorder model. The value is the name of the shapefile field
that data will be loaded from.mpoly for the geometry field is MULTIPOLYGON, the
geometry type GeoDjango will import the field as. Even simple polygons in
the shapefile will automatically be converted into collections prior to
insertion into the database.world application (with data subdirectory) to a different location,
the script will still work.transform keyword is set to False because the data in the
shapefile does not need to be converted – it’s already in WGS84 (SRID=4326).Afterwards, invoke the Django shell from the geodjango project directory:
$ python manage.py shell
...\> py manage.py shell
Next, import the load module, call the run routine, and watch
LayerMapping do the work:
>>> from world import load
>>> load.run()
ogrinspect¶Now that you’ve seen how to define geographic models and import data with the
LayerMapping data import utility, it’s possible to further automate this process with
use of the ogrinspect management command. The ogrinspect
command introspects a GDAL-supported vector data source (e.g., a shapefile)
and generates a model definition and LayerMapping dictionary automatically.
The general usage of the command goes as follows:
$ python manage.py ogrinspect [options] <data_source> <model_name> [options]
...\> py manage.py ogrinspect [options] <data_source> <model_name> [options]
data_source is the path to the GDAL-supported data source and
model_name is the name to use for the model. Command-line options may
be used to further define how the model is generated.
For example, the following command nearly reproduces the WorldBorder model
and mapping dictionary created above, automatically:
$ python manage.py ogrinspect world/data/TM_WORLD_BORDERS-0.3.shp WorldBorder \
--srid=4326 --mapping --multi
...\> py manage.py ogrinspect world\data\TM_WORLD_BORDERS-0.3.shp WorldBorder \
--srid=4326 --mapping --multi
A few notes about the command-line options given above:
--srid=4326 option sets the SRID for the geographic field.--mapping option tells ogrinspect to also generate a
mapping dictionary for use with
LayerMapping.--multi option is specified so that the geographic field is a
MultiPolygonField instead of just a
PolygonField.The command produces the following output, which may be copied
directly into the models.py of a GeoDjango application:
# This is an auto-generated Django model module created by ogrinspect.
from django.contrib.gis.db import models
class WorldBorder(models.Model):
fips = models.CharField(max_length=2)
iso2 = models.CharField(max_length=2)
iso3 = models.CharField(max_length=3)
un = models.IntegerField()
name = models.CharField(max_length=50)
area = models.IntegerField()
pop2005 = models.IntegerField()
region = models.IntegerField()
subregion = models.IntegerField()
lon = models.FloatField()
lat = models.FloatField()
geom = models.MultiPolygonField(srid=4326)
# Auto-generated `LayerMapping` dictionary for WorldBorder model
worldborders_mapping = {
'fips' : 'FIPS',
'iso2' : 'ISO2',
'iso3' : 'ISO3',
'un' : 'UN',
'name' : 'NAME',
'area' : 'AREA',
'pop2005' : 'POP2005',
'region' : 'REGION',
'subregion' : 'SUBREGION',
'lon' : 'LON',
'lat' : 'LAT',
'geom' : 'MULTIPOLYGON',
}
GeoDjango adds spatial lookups to the Django ORM. For example, you
can find the country in the WorldBorder table that contains
a particular point. First, fire up the management shell:
$ python manage.py shell
...\> py manage.py shell
Now, define a point of interest [3]:
>>> pnt_wkt = 'POINT(-95.3385 29.7245)'
The pnt_wkt string represents the point at -95.3385 degrees longitude,
29.7245 degrees latitude. The geometry is in a format known as
Well Known Text (WKT), a standard issued by the Open Geospatial
Consortium (OGC). [4] Import the WorldBorder model, and perform
a contains lookup using the pnt_wkt as the parameter:
>>> from world.models import WorldBorder
>>> WorldBorder.objects.filter(mpoly__contains=pnt_wkt)
<QuerySet [<WorldBorder: United States>]>
Here, you retrieved a QuerySet with only one model: the border of the
United States (exactly what you would expect).
Similarly, you may also use a GEOS geometry object.
Here, you can combine the intersects spatial lookup with the get
method to retrieve only the WorldBorder instance for San Marino instead
of a queryset:
>>> from django.contrib.gis.geos import Point
>>> pnt = Point(12.4604, 43.9420)
>>> WorldBorder.objects.get(mpoly__intersects=pnt)
<WorldBorder: San Marino>
The contains and intersects lookups are just a subset of the
available queries – the GeoDjango Database API documentation has more.
When doing spatial queries, GeoDjango automatically transforms geometries if they’re in a different coordinate system. In the following example, coordinates will be expressed in EPSG SRID 32140, a coordinate system specific to south Texas only and in units of meters, not degrees:
>>> from django.contrib.gis.geos import GEOSGeometry, Point
>>> pnt = Point(954158.1, 4215137.1, srid=32140)
Note that pnt may also be constructed with EWKT, an “extended” form of
WKT that includes the SRID:
>>> pnt = GEOSGeometry('SRID=32140;POINT(954158.1 4215137.1)')
GeoDjango’s ORM will automatically wrap geometry values in transformation SQL, allowing the developer to work at a higher level of abstraction:
>>> qs = WorldBorder.objects.filter(mpoly__intersects=pnt)
>>> print(qs.query) # Generating the SQL
SELECT "world_worldborder"."id", "world_worldborder"."name", "world_worldborder"."area",
"world_worldborder"."pop2005", "world_worldborder"."fips", "world_worldborder"."iso2",
"world_worldborder"."iso3", "world_worldborder"."un", "world_worldborder"."region",
"world_worldborder"."subregion", "world_worldborder"."lon", "world_worldborder"."lat",
"world_worldborder"."mpoly" FROM "world_worldborder"
WHERE ST_Intersects("world_worldborder"."mpoly", ST_Transform(%s, 4326))
>>> qs # printing evaluates the queryset
<QuerySet [<WorldBorder: United States>]>
Raw queries
When using raw queries, you must wrap your geometry fields so that the field value can be recognized by GEOS:
from django.db import connection
# or if you're querying a non-default database:
from django.db import connections
connection = connections['your_gis_db_alias']
City.objects.raw('SELECT id, name, %s as point from myapp_city' % (connection.ops.select % 'point'))
You should only use raw queries when you know exactly what you’re doing.
GeoDjango loads geometries in a standardized textual representation. When the
geometry field is first accessed, GeoDjango creates a
GEOSGeometry object, exposing powerful
functionality, such as serialization properties for popular geospatial
formats:
>>> sm = WorldBorder.objects.get(name='San Marino')
>>> sm.mpoly
<MultiPolygon object at 0x24c6798>
>>> sm.mpoly.wkt # WKT
MULTIPOLYGON (((12.4157980000000006 43.9579540000000009, 12.4505540000000003 43.9797209999999978, ...
>>> sm.mpoly.wkb # WKB (as Python binary buffer)
<read-only buffer for 0x1fe2c70, size -1, offset 0 at 0x2564c40>
>>> sm.mpoly.geojson # GeoJSON
'{ "type": "MultiPolygon", "coordinates": [ [ [ [ 12.415798, 43.957954 ], [ 12.450554, 43.979721 ], ...
This includes access to all of the advanced geometric operations provided by the GEOS library:
>>> pnt = Point(12.4604, 43.9420)
>>> sm.mpoly.contains(pnt)
True
>>> pnt.contains(sm.mpoly)
False
GeoDjango also offers a set of geographic annotations to compute distances and several other operations (intersection, difference, etc.). See the Geographic Database Functions documentation.
GeoDjango extends Django’s admin application with support for editing geometry fields.
GeoDjango also supplements the Django admin by allowing users to create and modify geometries on a JavaScript slippy map (powered by OpenLayers).
Let’s dive right in. Create a file called admin.py inside the
world application with the following code:
from django.contrib.gis import admin
from .models import WorldBorder
admin.site.register(WorldBorder, admin.GeoModelAdmin)
Next, edit your urls.py in the geodjango application folder as follows:
from django.contrib.gis import admin
from django.urls import include, path
urlpatterns = [
path('admin/', admin.site.urls),
]
Create an admin user:
$ python manage.py createsuperuser
...\> py manage.py createsuperuser
Next, start up the Django development server:
$ python manage.py runserver
...\> py manage.py runserver
Finally, browse to http://localhost:8000/admin/, and log in with the user
you just created. Browse to any of the WorldBorder entries – the borders
may be edited by clicking on a polygon and dragging the vertices to the desired
position.
OSMGeoAdmin¶With the OSMGeoAdmin, GeoDjango uses
a Open Street Map layer in the admin.
This provides more context (including street and thoroughfare details) than
available with the GeoModelAdmin
(which uses the Vector Map Level 0 WMS dataset hosted at OSGeo).
The PROJ.4 datum shifting files must be installed (see the PROJ.4 installation instructions for more details).
If you meet this requirement, then just substitute the OSMGeoAdmin
option class in your admin.py file:
admin.site.register(WorldBorder, admin.OSMGeoAdmin)
Footnotes
| [1] | Special thanks to Bjørn Sandvik of thematicmapping.org for providing and maintaining this dataset. |
| [2] | GeoDjango basic apps was written by Dane Springmeyer, Josh Livni, and Christopher Schmidt. |
| [3] | This point is the University of Houston Law Center. |
| [4] | Open Geospatial Consortium, Inc., OpenGIS Simple Feature Specification For SQL. |
Nov 02, 2020