Snowflake-Labs/jts-udf

Snowflake-Labs/jts-udf

Description: Advanced geospatial functions for Snowflake SQL from the JTS Java library

Language: Java

License: Apache-2.0

Stars: 0

Forks: 0

Open issues: 0

Created: 2026-03-02T06:06:38Z

Pushed: 2026-03-09T22:13:42Z

Default branch: main

Fork: no

Archived: no

README:

User Defined JTS Functions for Snowflake

The Snowflake geography and geometry types have a few dozen functions defined, covering the most frequent use cases. The JTS computational geometry library has many more functions, providing unique geospatial capabilities.

This repository contains the code and instructions to enable several of the unique functions in JTS as Java user-defined functions, bringing the capabilities of the open source JTS library into the Snowflake SQL environment.

[Learn more...](docs/about.md)

Available Functions

• jts.buffer(geom GEOMETRY, radius DOUBLE, quadsegs INT DEFAULT 8) -> GEOMETRY

Build buffer of geometry. Negative radius shrinks output, positive expands. More quad segments for more detail on curves.

• jts.buffersinglesided(geom GEOMETRY, radius DOUBLE) -> GEOMETRY

Build single sided buffer. Negative radius builds to right, positive to left.

• jts.concavehull(geom GEOMETRY, lengthRatio DOUBLE, isHolesAllowed BOOLEAN DEFAULT false) -> GEOMETRY

Returns concave hull. Length ratio from 0 (convex) to 1 (highly concave).

• jts.pointonsurface(geom GEOMETRY) -> GEOMETRY

Returns point guaranteed to be within geometry.

• jts.convexhull(geom GEOMETRY) -> GEOMETRY

Construct convex hull of geometry.

• jts.reduceprecision(geom GEOMETRY, gridSize DOUBLE) -> GEOMETRY

Apply precision grid to geometry, returns valid output.

• jts.isvalid(geom GEOMETRY) -> BOOLEAN

Return true for valid geometry, false for invalid.

• jts.makevalid(geom GEOMETRY) -> GEOMETRY

Repair invalidities in geometry, if there are any.

• jts.makevalid(geom GEOGRAPHY) -> GEOGRAPHY

Repair invalidities in geography, if there are any, using gnomonic projection.

• jts.intersection(geom0 GEOMETRY, geom1 GEOMETRY, gridSize DOUBLE DEFAULT 0) -> GEOMETRY

Calculate intersection of geometries. Optional grid size to force precision grid on output.

• jts.gunion(geom0 GEOMETRY, geom1 GEOMETRY, gridSize DOUBLE DEFAULT 0) -> GEOMETRY

Calculate union of geometries. Optional grid size to force precision grid on output.

• jts.difference(geom0 GEOMETRY, geom1 GEOMETRY, gridSize DOUBLE DEFAULT 0) -> GEOMETRY

Calculate difference of second geometry from first. Optional grid size to force precision grid on output.

• jts.symdifference(geom0 GEOMETRY, geom1 GEOMETRY, gridSize DOUBLE DEFAULT 0) -> GEOMETRY

Calculate symmetric difference of geometries. Optional grid size to force precision grid on output.

• jts.relate(geom0 GEOMETRY, geom1 GEOMETRY) -> GEOMETRY

Calculate full DE9IM relate matrix, flattened to string.

Build / Install

For now this library is available to build and install in your own organization, and is not built out to the marketplace.

Connect to Snowflake

The directions below make use of the snow commandline tool, so install the tool and setup your local login configuration using the Configuring Snowflake CLI and connecting to Snowflake docs.

Install Build Dependencies

To build the Java portions of the project, you will need:

• Java Development Kit (JDK) installation
• You may already have one, type which java at the command-line
• Maven build tool

To build the SQL portions of the project, you will need:

• Python
• Make

Build and Install the UDF Native App

The JTS UDF is structured as a Snowflake Native App. For this project that means:

• a Java JAR file that contains the JTS code and shims to connect it to Snowflake,
• a SQL file that defines the function names and connects them to the Java code; and,
• some YML manifest files to wrap the whole thing up as an "application package" that the snow commandline program can understand.

To build all the artefacts, run:

make app.prepare

This will generate all the necessary Java and SQL code.

To install the application package, run:

# check you have a working connection
snow connection list
snow connection test

# make sure the app passes basic tests
snow app validate

# copy the app package components into your account
# when complete you should be able to see the package
# in your account browser
snow app deploy

# enable the application
# when complete you should be able to run JTS functions
# in SQL
snow app run

Try the JTS Native App

A basic test is if a very simple geometry is valid (points are always valid).

SELECT jts.isvalid(to_geometry('POINT(0 0)'));

From there, do something that only JTS can do, like properly clean a figure-8 geometry polygon.

SELECT jts.makevalid(to_geometry('POLYGON((0 0, 0 1, 2 1, 2 2, 1 2, 1 0, 0 0))'));

Or clean an invalid geography polygon (the lower edge crosses the upper edge in geographics).

SELECT jts.makevalid(to_geography('POLYGON((0 60, 0 61, 5 60.05, 10 61, 10 60, 0 60))', True));

Development Installation

If you are planning to add new functions or work actively on the application, a manual installation might be more effective.

Set your role to SYSADMIN.

-- Create a database for our UDF work
CREATE OR REPLACE DATABASE java_udf_db;

-- Create a schema within that database
CREATE OR REPLACE SCHEMA java_udf_schema;

For working on the app package, installing it piecemeal is more convenient than installing it as an app package.

First, ensure all the components are built.

make dev.prepare

Then create a stage, copy the JAR onto that stage, and then create all the UDFs with reference to that JAR.

# make sure you have a working connection
snow connection test

# put a stage into the schema you created earlier
snow stage create \
--database java_udf_db \
--schema java_udf_schema \
--role sysadmin \
java_udf_stage

# copy the jar file to that stage
snow stage copy --overwrite ./target/snowflake-jts-1.0.0.jar @java_udf_stage

#…