Working with Spatial Data

On this page Carat arrow pointing down
Warning:
As of November 18, 2022, CockroachDB v21.1 is no longer supported. For more details, refer to the Release Support Policy.

This page provides information about working with spatial data in CockroachDB.

Supported data types

Supported spatial data types include:

  • Geometric objects such as points, lines, and polygons in 2-dimensional space. These are projected onto the flat surface of a plane and are represented in SQL by the GEOMETRY data type.

  • Geographic objects, which are also made up of points, lines, polygons, etc., in 2-dimensional space. They are projected onto the surface of a sphere and are represented in SQL by the GEOGRAPHY data type. (Technically, they are projected onto a spheroid: "a sphere with a bulge"). The spheroid projection means that:

    • The X and Y coordinates of 2-dimensional points are actually Longitude and Latitude values.
    • The paths between geographic objects are not straight lines; they are curves, and so the distances between objects are calculated using great circle math.

Compatibility

Just as CockroachDB strives for Postgres compatibility, our spatial data support is designed to be as compatible as possible with the functionality provided by the PostGIS extension.

However, we do not yet implement the full list of PostGIS built-in functions and operators. Also, our spatial indexing works differently (see the Performance section below). For a list of the spatial functions we support, see Geospatial functions.

If your application needs support for functions that are not yet implemented, please check out our meta-issue for built-in function support on GitHub, which describes how to find an issue for the built-in function(s) you need.

For a list of other known limitations, see Known Limitations.

ORM compatibility

The following ORM spatial libraries are fully compatible with CockroachDB's spatial features:

Note:

Most PostGIS-compatible client libraries are incompatible with CockroachDB's spatial features without an adapter.

Troubleshooting

For general CockroachDB troubleshooting information, see this troubleshooting overview.

If you need help troubleshooting an issue with our spatial support, please get in touch using our Support resources.

Performance

In order to avoid full table scans, make sure to add indexes to any columns that are accessed as part of a predicate in the WHERE clause. For geospatial columns, the index will only be used if the column is accessed using an index-accelerated geospatial function from the list below (all of these functions work on GEOMETRY data types; a * means that a function also works on GEOGRAPHY data types):

To use a version of a function from the list above that will explicitly not use the index, add an underscore (_) to the beginning of the function name, e.g., _ST_Covers.

You can check which queries are using which indexes using the EXPLAIN statement. For more information about general query tuning (including index usage), see Optimize Statement Performance.

The syntax for adding an index to a geometry column is CREATE INDEX index_name ON table_name USING GIST (column_name).

For example, to add an index to the geom column of the sample tornadoes table:

icon/buttons/copy
CREATE INDEX tornado_geom_idx ON tornadoes USING GIST (geom);

This creates a (spatial) GIN index on the geom column.

Because CockroachDB is a scale-out, multi-node database, our spatial indexing strategy is based on a space-filling curve/quad-tree design (also known as "divide the space"), rather than the R-Tree data structure used by some other spatial databases (also known as "divide the objects"). Other databases that use a "divide the space" strategy include Microsoft SQL Server and MongoDB.

For more detailed information about how CockroachDB's spatial indexes work, see Spatial indexes.

If you encounter behavior that you think is due to a performance issue, please get in touch using our Support resources.

Example: Load NYC data for the PostGIS tutorial

Follow the steps below to load the SQL for the NYC data used in the Introduction to PostGIS tutorial.

Note:

CockroachDB can work with the tutorial up to Chapter 22, with the following exceptions:

  • Do not try to load Shapefiles via the GUI as shown in the tutorial. Instead, follow the steps below to load the SQL data directly into CockroachDB. (We have already converted the tutorial Shapefiles to SQL for you.)
  • We do not support GML or KML data.
  • We do not support SVG.

Before you begin

  • Install a build of CockroachDB with support for spatial data by following the instructions at Install CockroachDB.

  • Start a local insecure cluster and connect to that cluster from a SQL client:

    icon/buttons/copy
    cockroach sql --insecure --host=localhost --port=26257
    

    Leave this shell open for use in the examples below.

Step 1. Load the NYC data

Clone the data set:

icon/buttons/copy
git clone https://github.com/otan-cockroach/otan-scripts

Load the SQL files into your CockroachDB cluster:

icon/buttons/copy
cat otan-scripts/geospatial_sql/*.sql | cockroach sql --insecure --host=localhost --port=26257

The command above will take a few minutes to run.

Step 2. Follow the PostGIS tutorial

When the cluster is finished loading the data, open a SQL shell and start working through the Introduction to PostGIS tutorial:

icon/buttons/copy
cockroach sql --insecure --host=localhost --port=26257

See also


Yes No
On this page

Yes No