This driver implements basic GIS (geodjango) support for Microsoft SQL Server, built on top of django-pyodbc-azure.

It should be considered very alpha-quality at this stage! Feedback, issues, and patches are all very welcome.

Supported and unsupported operations

Most possible operations are supported. The primary exceptions are those that include the boundary itself, and convenience operations such as left/right, overlaps_above, etc.

The following spatial lookups are not supported:

• bounding-box related: contains_properly, covered_by, covers
• specialist positional: left, right, overlaps_left, overlaps_right, overlaps_above, overlaps_below, strictly_above, strictly_below
• miscellaneous: dwithin, exact, relate, same_as

The following spatial aggregate operations are not supported:

• extent3d and make_line

In addition, for performance reasons not all geometry operations have a corresponding geography analogue. The following operations are not available on geography types:

• bbcontains, bboverlaps, contained, crosses, touches

Limitations of SQL Server

SQL Server is OGC compliant, but does fall short of the functionality provided by PostGIS and Oracle Spatial. In particular, all of the boundary inclusion operations are missing: for example, contains is supported, but not covers.

Type information is also slightly different in SQL Server. Instead of keeping the geometry type (Point, Polygon, etc) in the column's metadata, it is a property of the instance (and hence so is the dimensionality), and similarly for the SRID. This means you can in theory store geometries of different types and SRIDs in the same column; this driver creates a constraint to check the type, but nothing else. It also means that introspection is rather fragile.

Geometries cannot be transformed to a different SRID (such as with ST_Transform in PostGIS).

Admin Interface

The admin interface works. This is worth noting here simply because the interface has to be pretend to be MySQL in order to run! There are some hard-coded checks for MySQL in the framework, and the limitations (with respect to introspection) of SQL Server are actually similar enough that this works for SQL Server too.

Installation and Setup

The only direct dependency is django-pyodbc-azure. If you are on linux this will require installing freetds and odbcinst. You will also need to configure it (the most important is odbcinst.ini).

To use the driver, your Django database configuration section should look something like this:

DATABASES = {
    'default': {
        'NAME': 'dbname',
        'ENGINE': 'django_pyodbc_gis',
        'HOST': '127.0.0.1,1433',
        'USER': 'django',
        'PASSWORD': 'pwd123',
        'OPTIONS': {
            'host_is_server': True,
            # 'dsn': 'mssql',
            'extra_params': 'TDS_Version=8.0'
        }
    }
}

You have two options regarding specifying the host connection details; if you have configured a DSN you may omit the HOST key and use the dsn key in OPTIONS to specify it. If not, you will probably need to specify the TDS version in extra_params (if you get error messages about unicode you may well have gotten this wrong)

TODO

• extended operations (gml, geojson, etc. Further investigation: SQL Server only supports GML, but treats it as an instance method where-as geodjango assumes it is a function. This might remain on the back-burner for now)
• Check inspectdb support
• Test suite!
• Test against 2008, 2005 as well