Spatial SQL app

The page outlines the steps to manage spatial data using SedonaSQL.

Note

Since v1.5.0, Sedona assumes geographic coordinates to be in longitude/latitude order. If your data is lat/lon order, please use ST_FlipCoordinates to swap X and Y.

SedonaSQL supports SQL/MM Part3 Spatial SQL Standard. It includes four kinds of SQL operators as follows. All these operators can be directly called through:

ScalaJavaPython

var myDataFrame = sedona.sql("YOUR_SQL")
myDataFrame.createOrReplaceTempView("spatialDf")

Dataset<Row> myDataFrame = sedona.sql("YOUR_SQL")
myDataFrame.createOrReplaceTempView("spatialDf")

myDataFrame = sedona.sql("YOUR_SQL")
myDataFrame.createOrReplaceTempView("spatialDf")

Detailed SedonaSQL APIs are available here: SedonaSQL API. You can find example county data (i.e., county_small.tsv) in Sedona GitHub repo.

Set up dependencies¶

Scala/JavaPython

Read Sedona Maven Central coordinates and add Sedona dependencies in build.sbt or pom.xml.
Add Apache Spark core, Apache SparkSQL in build.sbt or pom.xml.
Please see SQL example project

Please read Quick start to install Sedona Python.
This tutorial is based on Sedona SQL Jupyter Notebook example. You can interact with Sedona Python Jupyter notebook immediately on Binder. Click to interact with Sedona Python Jupyter notebook immediately on Binder.

Create Sedona config¶

Use the following code to create your Sedona config at the beginning. If you already have a SparkSession (usually named spark) created by Wherobots/AWS EMR/Databricks, please skip this step and can use spark directly.

Sedona >= 1.4.1

You can add additional Spark runtime config to the config builder. For example, SedonaContext.builder().config("spark.sql.autoBroadcastJoinThreshold", "10485760")

ScalaJavaPython

import org.apache.sedona.spark.SedonaContext

val config = SedonaContext.builder()
.master("local[*]") // Delete this if run in cluster mode
.appName("readTestScala") // Change this to a proper name
.getOrCreate()

If you use SedonaViz together with SedonaSQL, please add the following line after SedonaContext.builder() to enable Sedona Kryo serializer:

.config("spark.kryo.registrator", classOf[SedonaVizKryoRegistrator].getName) // org.apache.sedona.viz.core.Serde.SedonaVizKryoRegistrator

import org.apache.sedona.spark.SedonaContext;

SparkSession config = SedonaContext.builder()
.master("local[*]") // Delete this if run in cluster mode
.appName("readTestScala") // Change this to a proper name
.getOrCreate()

If you use SedonaViz together with SedonaSQL, please add the following line after SedonaContext.builder() to enable Sedona Kryo serializer:

.config("spark.kryo.registrator", SedonaVizKryoRegistrator.class.getName) // org.apache.sedona.viz.core.Serde.SedonaVizKryoRegistrator

from sedona.spark import *

config = SedonaContext.builder() .\
    config('spark.jars.packages',
           'org.apache.sedona:sedona-spark-shaded-3.0_2.12:1.5.1,'
           'org.datasyslab:geotools-wrapper:1.5.1-28.2'). \
    getOrCreate()

If you are using Spark versions >= 3.4, please replace the 3.0 in package name of sedona-spark-shaded with the corresponding major.minor version of Spark, such as sedona-spark-shaded-3.4_2.12:1.5.1.

Sedona < 1.4.1

The following method has been deprecated since Sedona 1.4.1. Please use the method above to create your Sedona config.

ScalaJavaPython

var sparkSession = SparkSession.builder()
.master("local[*]") // Delete this if run in cluster mode
.appName("readTestScala") // Change this to a proper name
// Enable Sedona custom Kryo serializer
.config("spark.serializer", classOf[KryoSerializer].getName) // org.apache.spark.serializer.KryoSerializer
.config("spark.kryo.registrator", classOf[SedonaKryoRegistrator].getName)
.getOrCreate() // org.apache.sedona.core.serde.SedonaKryoRegistrator

If you use SedonaViz together with SedonaSQL, please use the following two lines to enable Sedona Kryo serializer instead:

.config("spark.serializer", classOf[KryoSerializer].getName) // org.apache.spark.serializer.KryoSerializer
.config("spark.kryo.registrator", classOf[SedonaVizKryoRegistrator].getName) // org.apache.sedona.viz.core.Serde.SedonaVizKryoRegistrator

SparkSession sparkSession = SparkSession.builder()
.master("local[*]") // Delete this if run in cluster mode
.appName("readTestScala") // Change this to a proper name
// Enable Sedona custom Kryo serializer
.config("spark.serializer", KryoSerializer.class.getName) // org.apache.spark.serializer.KryoSerializer
.config("spark.kryo.registrator", SedonaKryoRegistrator.class.getName)
.getOrCreate() // org.apache.sedona.core.serde.SedonaKryoRegistrator

If you use SedonaViz together with SedonaSQL, please use the following two lines to enable Sedona Kryo serializer instead:

.config("spark.serializer", KryoSerializer.class.getName) // org.apache.spark.serializer.KryoSerializer
.config("spark.kryo.registrator", SedonaVizKryoRegistrator.class.getName) // org.apache.sedona.viz.core.Serde.SedonaVizKryoRegistrator

sparkSession = SparkSession. \
    builder. \
    appName('appName'). \
    config("spark.serializer", KryoSerializer.getName). \
    config("spark.kryo.registrator", SedonaKryoRegistrator.getName). \
    config('spark.jars.packages',
           'org.apache.sedona:sedona-spark-shaded-3.0_2.12:1.5.1,'
           'org.datasyslab:geotools-wrapper:1.5.1-28.2'). \
    getOrCreate()

If you are using Spark versions >= 3.4, please replace the 3.0 in package name of sedona-spark-shaded with the corresponding major.minor version of Spark, such as sedona-spark-shaded-3.4_2.12:1.5.1.

Initiate SedonaContext¶

Add the following line after creating Sedona config. If you already have a SparkSession (usually named spark) created by Wherobots/AWS EMR/Databricks, please call SedonaContext.create(spark) instead.

Sedona >= 1.4.1

ScalaJavaPython

import org.apache.sedona.spark.SedonaContext

val sedona = SedonaContext.create(config)

import org.apache.sedona.spark.SedonaContext;

SparkSession sedona = SedonaContext.create(config)

from sedona.spark import *

sedona = SedonaContext.create(config)

Sedona < 1.4.1

The following method has been deprecated since Sedona 1.4.1. Please use the method above to create your SedonaContext.

ScalaJavaPython

SedonaSQLRegistrator.registerAll(sparkSession)

SedonaSQLRegistrator.registerAll(sparkSession)

from sedona.register import SedonaRegistrator

SedonaRegistrator.registerAll(spark)

You can also register everything by passing --conf spark.sql.extensions=org.apache.sedona.sql.SedonaSqlExtensions to spark-submit or spark-shell.

Load data from files¶

Assume we have a WKT file, namely usa-county.tsv, at Path /Download/usa-county.tsv as follows:

POLYGON (..., ...)  Cuming County
POLYGON (..., ...)  Wahkiakum County
POLYGON (..., ...)  De Baca County
POLYGON (..., ...)  Lancaster County

The file may have many other columns.

Use the following code to load the data and create a raw DataFrame:

ScalaJavaPython

var rawDf = sedona.read.format("csv").option("delimiter", "\t").option("header", "false").load("/Download/usa-county.tsv")
rawDf.createOrReplaceTempView("rawdf")
rawDf.show()

Dataset<Row> rawDf = sedona.read.format("csv").option("delimiter", "\t").option("header", "false").load("/Download/usa-county.tsv")
rawDf.createOrReplaceTempView("rawdf")
rawDf.show()

rawDf = sedona.read.format("csv").option("delimiter", "\t").option("header", "false").load("/Download/usa-county.tsv")
rawDf.createOrReplaceTempView("rawdf")
rawDf.show()

The output will be like this:

|                 _c0|_c1|_c2|     _c3|  _c4|        _c5|                 _c6|_c7|_c8|  _c9|_c10| _c11|_c12|_c13|      _c14|    _c15|       _c16|        _c17|
+--------------------+---+---+--------+-----+-----------+--------------------+---+---+-----+----+-----+----+----+----------+--------+-----------+------------+
|POLYGON ((-97.019...| 31|039|00835841|31039|     Cuming|       Cuming County| 06| H1|G4020|null| null|null|   A|1477895811|10447360|+41.9158651|-096.7885168|
|POLYGON ((-123.43...| 53|069|01513275|53069|  Wahkiakum|    Wahkiakum County| 06| H1|G4020|null| null|null|   A| 682138871|61658258|+46.2946377|-123.4244583|
|POLYGON ((-104.56...| 35|011|00933054|35011|    De Baca|      De Baca County| 06| H1|G4020|null| null|null|   A|6015539696|29159492|+34.3592729|-104.3686961|
|POLYGON ((-96.910...| 31|109|00835876|31109|  Lancaster|    Lancaster County| 06| H1|G4020| 339|30700|null|   A|2169240202|22877180|+40.7835474|-096.6886584|

Create a Geometry type column¶

All geometrical operations in SedonaSQL are on Geometry type objects. Therefore, before any kind of queries, you need to create a Geometry type column on a DataFrame.

SELECT ST_GeomFromWKT(_c0) AS countyshape, _c1, _c2

You can select many other attributes to compose this spatialdDf. The output will be something like this:

|                 countyshape|_c1|_c2|     _c3|  _c4|        _c5|                 _c6|_c7|_c8|  _c9|_c10| _c11|_c12|_c13|      _c14|    _c15|       _c16|        _c17|
+--------------------+---+---+--------+-----+-----------+--------------------+---+---+-----+----+-----+----+----+----------+--------+-----------+------------+
|POLYGON ((-97.019...| 31|039|00835841|31039|     Cuming|       Cuming County| 06| H1|G4020|null| null|null|   A|1477895811|10447360|+41.9158651|-096.7885168|
|POLYGON ((-123.43...| 53|069|01513275|53069|  Wahkiakum|    Wahkiakum County| 06| H1|G4020|null| null|null|   A| 682138871|61658258|+46.2946377|-123.4244583|
|POLYGON ((-104.56...| 35|011|00933054|35011|    De Baca|      De Baca County| 06| H1|G4020|null| null|null|   A|6015539696|29159492|+34.3592729|-104.3686961|
|POLYGON ((-96.910...| 31|109|00835876|31109|  Lancaster|    Lancaster County| 06| H1|G4020| 339|30700|null|   A|2169240202|22877180|+40.7835474|-096.6886584|

Although it looks same with the input, but actually the type of column countyshape has been changed to Geometry type.

To verify this, use the following code to print the schema of the DataFrame:

spatialDf.printSchema()

The output will be like this:

root
 |-- countyshape: geometry (nullable = false)
 |-- _c1: string (nullable = true)
 |-- _c2: string (nullable = true)
 |-- _c3: string (nullable = true)
 |-- _c4: string (nullable = true)
 |-- _c5: string (nullable = true)
 |-- _c6: string (nullable = true)
 |-- _c7: string (nullable = true)

Note

SedonaSQL provides lots of functions to create a Geometry column, please read SedonaSQL constructor API.

Load GeoJSON using Spark JSON Data Source¶

Spark SQL's built-in JSON data source supports reading GeoJSON data. To ensure proper parsing of the geometry property, we can define a schema with the geometry property set to type 'string'. This prevents Spark from interpreting the property and allows us to use the ST_GeomFromGeoJSON function for accurate geometry parsing.

ScalaJavaPython

val schema = "type string, crs string, totalFeatures long, features array<struct<type string, geometry string, properties map<string, string>>>"
sedona.read.schema(schema).json(geojson_path)
    .selectExpr("explode(features) as features") // Explode the envelope to get one feature per row.
    .select("features.*") // Unpack the features struct.
    .withColumn("geometry", expr("ST_GeomFromGeoJSON(geometry)")) // Convert the geometry string.
    .printSchema()

String schema = "type string, crs string, totalFeatures long, features array<struct<type string, geometry string, properties map<string, string>>>";
sedona.read.schema(schema).json(geojson_path)
    .selectExpr("explode(features) as features") // Explode the envelope to get one feature per row.
    .select("features.*") // Unpack the features struct.
    .withColumn("geometry", expr("ST_GeomFromGeoJSON(geometry)")) // Convert the geometry string.
    .printSchema();

schema = "type string, crs string, totalFeatures long, features array<struct<type string, geometry string, properties map<string, string>>>";
(sedona.read.json(geojson_path, schema=schema)
    .selectExpr("explode(features) as features") # Explode the envelope to get one feature per row.
    .select("features.*") # Unpack the features struct.
    .withColumn("geometry", f.expr("ST_GeomFromGeoJSON(geometry)")) # Convert the geometry string.
    .printSchema())

Load Shapefile and GeoJSON using SpatialRDD¶

Shapefile and GeoJSON can be loaded by SpatialRDD and converted to DataFrame using Adapter. Please read Load SpatialRDD and DataFrame <-> RDD.

Load GeoParquet¶

Since v1.3.0, Sedona natively supports loading GeoParquet file. Sedona will infer geometry fields using the "geo" metadata in GeoParquet files.

Scala/JavaJavaPython

val df = sedona.read.format("geoparquet").load(geoparquetdatalocation1)
df.printSchema()

Dataset<Row> df = sedona.read.format("geoparquet").load(geoparquetdatalocation1)
df.printSchema()

df = sedona.read.format("geoparquet").load(geoparquetdatalocation1)
df.printSchema()

The output will be as follows:

root
 |-- pop_est: long (nullable = true)
 |-- continent: string (nullable = true)
 |-- name: string (nullable = true)
 |-- iso_a3: string (nullable = true)
 |-- gdp_md_est: double (nullable = true)
 |-- geometry: geometry (nullable = true)

Sedona supports spatial predicate push-down for GeoParquet files, please refer to the SedonaSQL query optimizer documentation for details.

GeoParquet file reader can also be used to read legacy Parquet files written by Apache Sedona 1.3.1-incubating or earlier. Please refer to Reading Legacy Parquet Files for details.

Warning

GeoParquet file reader does not work on Databricks runtime when Photon is enabled. Please disable Photon when using GeoParquet file reader on Databricks runtime.

Inspect GeoParquet metadata¶

Since v1.5.1, Sedona provides a Spark SQL data source "geoparquet.metadata" for inspecting GeoParquet metadata. The resulting dataframe contains the "geo" metadata for each input file.

Scala/JavaJavaPython

val df = sedona.read.format("geoparquet.metadata").load(geoparquetdatalocation1)
df.printSchema()

Dataset<Row> df = sedona.read.format("geoparquet.metadata").load(geoparquetdatalocation1)
df.printSchema()

df = sedona.read.format("geoparquet.metadata").load(geoparquetdatalocation1)
df.printSchema()

The output will be as follows:

root
 |-- path: string (nullable = true)
 |-- version: string (nullable = true)
 |-- primary_column: string (nullable = true)
 |-- columns: map (nullable = true)
 |    |-- key: string
 |    |-- value: struct (valueContainsNull = true)
 |    |    |-- encoding: string (nullable = true)
 |    |    |-- geometry_types: array (nullable = true)
 |    |    |    |-- element: string (containsNull = true)
 |    |    |-- bbox: array (nullable = true)
 |    |    |    |-- element: double (containsNull = true)
 |    |    |-- crs: string (nullable = true)

If the input Parquet file does not have GeoParquet metadata, the values of version, primary_column and columns fields of the resulting dataframe will be null.

Note

geoparquet.metadata only supports reading GeoParquet specific metadata. Users can use G-Research/spark-extension to read comprehensive metadata of generic Parquet files.

Load data from JDBC data sources¶

The 'query' option in Spark SQL's JDBC data source can be used to convert geometry columns to a format that Sedona can interpret. This should work for most spatial JDBC data sources. For Postgis there is no need to add a query to convert geometry types since it's already using EWKB as it's wire format.

ScalaJavaPython

// For any JDBC data source, including Postgis.
val df = sedona.read.format("jdbc")
    // Other options.
    .option("query", "SELECT id, ST_AsBinary(geom) as geom FROM my_table")
    .load()
    .withColumn("geom", expr("ST_GeomFromWKB(geom)"))

// This is a simplified version that works for Postgis.
val df = sedona.read.format("jdbc")
    // Other options.
    .option("dbtable", "my_table")
    .load()
    .withColumn("geom", expr("ST_GeomFromWKB(geom)"))

// For any JDBC data source, including Postgis.
Dataset<Row> df = sedona.read().format("jdbc")
    // Other options.
    .option("query", "SELECT id, ST_AsBinary(geom) as geom FROM my_table")
    .load()
    .withColumn("geom", expr("ST_GeomFromWKB(geom)"))

// This is a simplified version that works for Postgis.
Dataset<Row> df = sedona.read().format("jdbc")
    // Other options.
    .option("dbtable", "my_table")
    .load()
    .withColumn("geom", expr("ST_GeomFromWKB(geom)"))

# For any JDBC data source, including Postgis.
df = (sedona.read.format("jdbc")
    # Other options.
    .option("query", "SELECT id, ST_AsBinary(geom) as geom FROM my_table")
    .load()
    .withColumn("geom", f.expr("ST_GeomFromWKB(geom)")))

# This is a simplified version that works for Postgis.
df = (sedona.read.format("jdbc")
    # Other options.
    .option("dbtable", "my_table")
    .load()
    .withColumn("geom", f.expr("ST_GeomFromWKB(geom)")))

Transform the Coordinate Reference System¶

Sedona doesn't control the coordinate unit (degree-based or meter-based) of all geometries in a Geometry column. The unit of all related distances in SedonaSQL is same as the unit of all geometries in a Geometry column.

By default, this function uses lon/lat order since v1.5.0. Before, it used lat/lon order. You can use ST_FlipCoordinates to swap X and Y.

For more details, please read the ST_Transform section in Sedona API References.

To convert Coordinate Reference System of the Geometry column created before, use the following code:

SELECT ST_Transform(countyshape, "epsg:4326", "epsg:3857") AS newcountyshape, _c1, _c2, _c3, _c4, _c5, _c6, _c7
FROM spatialdf

The first EPSG code EPSG:4326 in ST_Transform is the source CRS of the geometries. It is WGS84, the most common degree-based CRS.

The second EPSG code EPSG:3857 in ST_Transform is the target CRS of the geometries. It is the most common meter-based CRS.

This ST_Transform transform the CRS of these geometries from EPSG:4326 to EPSG:3857. The details CRS information can be found on EPSG.io

The coordinates of polygons have been changed. The output will be like this:

+--------------------+---+---+--------+-----+-----------+--------------------+---+
|      newcountyshape|_c1|_c2|     _c3|  _c4|        _c5|                 _c6|_c7|
+--------------------+---+---+--------+-----+-----------+--------------------+---+
|POLYGON ((-108001...| 31|039|00835841|31039|     Cuming|       Cuming County| 06|
|POLYGON ((-137408...| 53|069|01513275|53069|  Wahkiakum|    Wahkiakum County| 06|
|POLYGON ((-116403...| 35|011|00933054|35011|    De Baca|      De Baca County| 06|
|POLYGON ((-107880...| 31|109|00835876|31109|  Lancaster|    Lancaster County| 06|

Run spatial queries¶

After creating a Geometry type column, you are able to run spatial queries.

Range query¶

Use ST_Contains, ST_Intersects, ST_Within to run a range query over a single column.

The following example finds all counties that are within the given polygon:

SELECT *
FROM spatialdf
WHERE ST_Contains (ST_PolygonFromEnvelope(1.0,100.0,1000.0,1100.0), newcountyshape)

Note

Read SedonaSQL constructor API to learn how to create a Geometry type query window

KNN query¶

Use ST_Distance to calculate the distance and rank the distance.

The following code returns the 5 nearest neighbor of the given polygon.

SELECT countyname, ST_Distance(ST_PolygonFromEnvelope(1.0,100.0,1000.0,1100.0), newcountyshape) AS distance
FROM spatialdf
ORDER BY distance DESC
LIMIT 5

Join query¶

The details of a join query is available here Join query.

Other queries¶

There are lots of other functions can be combined with these queries. Please read SedonaSQL functions and SedonaSQL aggregate functions.

Visualize query results¶

Sedona provides SedonaPyDeck and SedonaKepler wrappers, both of which expose APIs to create interactive map visualizations from SedonaDataFrames in a Jupyter environment.

Note

Both SedonaPyDeck and SedonaKepler expect the default geometry order to be lon-lat. If your dataframe has geometries in the lat-lon order, please check out ST_FlipCoordinates

Note

Both SedonaPyDeck and SedonaKepler are designed to work with SedonaDataFrames containing only 1 geometry column. Passing dataframes with multiple geometry columns will cause errors.

SedonaPyDeck¶

Spatial query results can be visualized in a Jupyter lab/notebook environment using SedonaPyDeck.

SedonaPyDeck exposes APIs to create interactive map visualizations using pydeck based on deck.gl

Note

To use SedonaPyDeck, install sedona with the pydeck-map extra:

pip install sedona[pydeck-map]

The following tutorial showcases the various maps that can be created using SedonaPyDeck, the datasets used to create these maps are publicly available.

Each API exposed by SedonaPyDeck offers customization via optional arguments, details on all possible arguments can be found in the API docs of SedonaPyDeck.

Creating a Choropleth map using SedonaPyDeck¶

SedonaPyDeck exposes a create_choropleth_map API which can be used to visualize a choropleth map out of the passed SedonaDataFrame containing polygons with an observation:

Example (referenced from example notebook available via binder):

SedonaPyDeck.create_choropleth_map(df=groupedresult, plot_col='AirportCount')

Note

plot_col is a required argument informing SedonaPyDeck of the column name used to render the choropleth effect.

The dataset used is available here and can also be found in the example notebook available here

Creating a Geometry map using SedonaPyDeck¶

SedonaPyDeck exposes a create_geometry_map API which can be used to visualize a passed SedonaDataFrame containing any type of geometries:

Example (referenced from overture notebook available via binder):

SedonaPyDeck.create_geometry_map(df_building, elevation_col='height')

Tip

elevation_col is an optional argument which can be used to render a 3D map. Pass the column with 'elevation' values for the geometries here.

Creating a Scatterplot map using SedonaPyDeck¶

SedonaPyDeck exposes a create_scatterplot_map API which can be used to visualize a scatterplot out of the passed SedonaDataFrame containing points:

Example:

SedonaPyDeck.create_scatterplot_map(df=crimes_df)

The dataset used here is the Chicago crimes dataset, available here

Creating a heatmap using SedonaPyDeck¶

SedonaPyDeck exposes a create_heatmap API which can be used to visualize a heatmap out of the passed SedonaDataFrame containing points:

Example:

SedonaPyDeck.create_heatmap(df=crimes_df)

The dataset used here is the Chicago crimes dataset, available here

SedonaKepler¶

Spatial query results can be visualized in a Jupyter lab/notebook environment using SedonaKepler.

SedonaKepler exposes APIs to create interactive and customizable map visualizations using KeplerGl.

Note

To use SedonaKepler, install sedona with the kepler-map extra:

pip install sedona[kepler-map]

This tutorial showcases how simple it is to instantly visualize geospatial data using SedonaKepler.

Example (referenced from an example notebook via the binder):

SedonaKepler.create_map(df=groupedresult, name="AirportCount")

The dataset used is available here and can also be found in the example notebook available here

Details on all the APIs available by SedonaKepler are listed in the SedonaKepler API docs

Create a User-Defined Function (UDF)¶

User-Defined Functions (UDFs) are user-created procedures that can perform operations on a single row of information. To cover almost all use cases, we will showcase 4 types of UDFs for a better understanding of how to use geometry with UDFs. Sedona's serializer deserializes the SQL geometry type to JTS Geometry (Scala/Java) or Shapely Geometry (Python). You can implement any custom logic using the rich ecosystem around these two libraries.

Geometry to primitive¶

This UDF example takes a geometry type input and returns a primitive type output:

ScalaJavaPython

import org.locationtech.jts.geom.Geometry
import org.apache.spark.sql.types._

def lengthPoly(geom: Geometry): Double = {
    geom.getLength
}

sedona.udf.register("udf_lengthPoly", lengthPoly _)

df.selectExpr("udf_lengthPoly(geom)").show()

import org.apache.spark.sql.api.java.UDF1;
import org.apache.spark.sql.types.DataTypes;

// using lambda function to register the UDF
sparkSession.udf().register(
        "udf_lengthPoly",
        (UDF1<Geometry, Double>) Geometry::getLength,
        DataTypes.DoubleType);

df.selectExpr("udf_lengthPoly(geom)").show()

from sedona.sql.types import GeometryType
from pyspark.sql.types import DoubleType

def lengthPoly(geom: GeometryType()):
    return geom.length

sedona.udf.register("udf_lengthPoly", lengthPoly, DoubleType())

df.selectExpr("udf_lengthPoly(geom)").show()

Output:

+--------------------+
|udf_lengthPoly(geom)|
+--------------------+
|   3.414213562373095|
+--------------------+

Geometry to Geometry¶

This UDF example takes a geometry type input and returns a geometry type output:

ScalaJavaPython

import org.locationtech.jts.geom.Geometry
import org.apache.spark.sql.types._

def bufferFixed(geom: Geometry): Geometry = {
    geom.buffer(5.5)
}

sedona.udf.register("udf_bufferFixed", bufferFixed _)

df.selectExpr("udf_bufferFixed(geom)").show()

import org.apache.spark.sql.api.java.UDF1;
import org.apache.spark.sql.types.DataTypes;

// using lambda function to register the UDF
sparkSession.udf().register(
        "udf_bufferFixed",
        (UDF1<Geometry, Geometry>) geom ->
            geom.buffer(5.5),
        new GeometryUDT());

df.selectExpr("udf_bufferFixed(geom)").show()

from sedona.sql.types import GeometryType
from pyspark.sql.types import DoubleType

def bufferFixed(geom: GeometryType()):
    return geom.buffer(5.5)

sedona.udf.register("udf_bufferFixed", bufferFixed, GeometryType())

df.selectExpr("udf_bufferFixed(geom)").show()

Output:

+--------------------------------------------------+
|                             udf_bufferFixed(geom)|
+--------------------------------------------------+
|POLYGON ((1 -4.5, -0.0729967710887076 -4.394319...|
+--------------------------------------------------+

Geometry, primitive to geometry¶

This UDF example takes a geometry type input and a primitive type input and returns a geometry type output:

ScalaJavaPython

import org.locationtech.jts.geom.Geometry
import org.apache.spark.sql.types._

def bufferIt(geom: Geometry, distance: Double): Geometry = {
    geom.buffer(distance)
}

sedona.udf.register("udf_buffer", bufferIt _)

df.selectExpr("udf_buffer(geom, distance)").show()

import org.apache.spark.sql.api.java.UDF2;
import org.apache.spark.sql.types.DataTypes;

// using lambda function to register the UDF
sparkSession.udf().register(
        "udf_buffer",
        (UDF2<Geometry, Double, Geometry>) Geometry::buffer,
        new GeometryUDT());

df.selectExpr("udf_buffer(geom, distance)").show()

from sedona.sql.types import GeometryType
from pyspark.sql.types import DoubleType

def bufferIt(geom: GeometryType(), distance: DoubleType()):
    return geom.buffer(distance)

sedona.udf.register("udf_buffer", bufferIt, GeometryType())

df.selectExpr("udf_buffer(geom, distance)").show()

Output:

+--------------------------------------------------+
|                        udf_buffer(geom, distance)|
+--------------------------------------------------+
|POLYGON ((1 -9, -0.9509032201612866 -8.80785280...|
+--------------------------------------------------+

Geometry, primitive to Geometry, primitive¶

This UDF example takes a geometry type input and a primitive type input and returns a geometry type and a primitive type output:

ScalaJavaPython

import org.locationtech.jts.geom.Geometry
import org.apache.spark.sql.types._
import org.apache.spark.sql.api.java.UDF2

val schemaUDF = StructType(Array(
    StructField("buffed", GeometryUDT),
    StructField("length", DoubleType)
))

val udf_bufferLength = udf(
    new UDF2[Geometry, Double, (Geometry, Double)] {
        def call(geom: Geometry, distance: Double): (Geometry, Double) = {
            val buffed = geom.buffer(distance)
            val length = geom.getLength
            (buffed, length)
        }
    }, schemaUDF)

sedona.udf.register("udf_bufferLength", udf_bufferLength)

data.withColumn("bufferLength", expr("udf_bufferLengths(geom, distance)"))
    .select("geom", "distance", "bufferLength.*")
    .show()

import org.apache.spark.sql.api.java.UDF2;
import org.apache.spark.sql.types.DataTypes;
import org.apache.spark.sql.types.StructType;
import scala.Tuple2;

StructType schemaUDF = new StructType()
            .add("buffedGeom", new GeometryUDT())
            .add("length", DataTypes.DoubleType);

// using lambda function to register the UDF
sparkSession.udf().register("udf_bufferLength",
            (UDF2<Geometry, Double, Tuple2<Geometry, Double>>) (geom, distance) -> {
                Geometry buffed = geom.buffer(distance);
                Double length = buffed.getLength();
                return new Tuple2<>(buffed, length);
            },
            schemaUDF);

df.withColumn("bufferLength", functions.expr("udf_bufferLength(geom, distance)"))
            .select("geom", "distance", "bufferLength.*")
            .show();

from sedona.sql.types import GeometryType
from pyspark.sql.types import *

schemaUDF = StructType([
    StructField("buffed", GeometryType()),
    StructField("length", DoubleType())
])

def bufferAndLength(geom: GeometryType(), distance: DoubleType()):
    buffed = geom.buffer(distance)
    length = buffed.length
    return [buffed, length]

sedona.udf.register("udf_bufferLength", bufferAndLength, schemaUDF)

df.withColumn("bufferLength", expr("udf_bufferLength(geom, buffer)")) \
            .select("geom", "buffer", "bufferLength.*") \
            .show()

Output:

+------------------------------+--------+--------------------------------------------------+-----------------+
|                          geom|distance|                                        buffedGeom|           length|
+------------------------------+--------+--------------------------------------------------+-----------------+
|POLYGON ((1 1, 1 2, 2 1, 1 1))|    10.0|POLYGON ((1 -9, -0.9509032201612866 -8.80785280...|66.14518337329191|
+------------------------------+--------+--------------------------------------------------+-----------------+

Save to permanent storage¶

To save a Spatial DataFrame to some permanent storage such as Hive tables and HDFS, you can simply convert each geometry in the Geometry type column back to a plain String and save the plain DataFrame to wherever you want.

Use the following code to convert the Geometry column in a DataFrame back to a WKT string column:

SELECT ST_AsText(countyshape)
FROM polygondf

Note

ST_AsGeoJSON is also available. We would like to invite you to contribute more functions

Save GeoParquet¶

Since v1.3.0, Sedona natively supports writing GeoParquet file. GeoParquet can be saved as follows:

df.write.format("geoparquet").save(geoparquetoutputlocation + "/GeoParquet_File_Name.parquet")

Since v1.5.1, Sedona supports writing GeoParquet files with custom GeoParquet spec version and crs. The default GeoParquet spec version is 1.0.0 and the default crs is null. You can specify the GeoParquet spec version and crs as follows:

val projjson = "{...}" // PROJJSON string for all geometry columns
df.write.format("geoparquet")
        .option("geoparquet.version", "1.0.0")
        .option("geoparquet.crs", projjson)
        .save(geoparquetoutputlocation + "/GeoParquet_File_Name.parquet")

If you have multiple geometry columns written to the GeoParquet file, you can specify the CRS for each column. For example, g0 and g1 are two geometry columns in the DataFrame df, and you want to specify the CRS for each column as follows:

val projjson_g0 = "{...}" // PROJJSON string for g0
val projjson_g1 = "{...}" // PROJJSON string for g1
df.write.format("geoparquet")
        .option("geoparquet.version", "1.0.0")
        .option("geoparquet.crs.g0", projjson_g0)
        .option("geoparquet.crs.g1", projjson_g1)
        .save(geoparquetoutputlocation + "/GeoParquet_File_Name.parquet")

The value of geoparquet.crs and geoparquet.crs.<column_name> can be one of the following:

"null": Explicitly setting crs field to null. This is the default behavior.
"" (empty string): Omit the crs field. This implies that the CRS is OGC:CRS84 for CRS-aware implementations.
"{...}" (PROJJSON string): The crs field will be set as the PROJJSON object representing the Coordinate Reference System (CRS) of the geometry. You can find the PROJJSON string of a specific CRS from here: https://epsg.io/ (click the JSON option at the bottom of the page). You can also customize your PROJJSON string as needed.

Please note that Sedona currently cannot set/get a projjson string to/from a CRS. Its geoparquet reader will ignore the projjson metadata and you will have to set your CRS via ST_SetSRID after reading the file. Its geoparquet writer will not leverage the SRID field of a geometry so you will have to always set the geoparquet.crs option manually when writing the file, if you want to write a meaningful CRS field.

Due to the same reason, Sedona geoparquet reader and writer do NOT check the axis order (lon/lat or lat/lon) and assume they are handled by the users themselves when writing / reading the files. You can always use ST_FlipCoordinates to swap the axis order of your geometries.

Sort then Save GeoParquet¶

To maximize the performance of Sedona GeoParquet filter pushdown, we suggest that you sort the data by their geohash values (see ST_GeoHash) and then save as a GeoParquet file. An example is as follows:

SELECT col1, col2, geom, ST_GeoHash(geom, 5) as geohash
FROM spatialDf
ORDER BY geohash

Save to Postgis¶

Unfortunately, the Spark SQL JDBC data source doesn't support creating geometry types in PostGIS using the 'createTableColumnTypes' option. Only the Spark built-in types are recognized. This means that you'll need to manage your PostGIS schema separately from Spark. One way to do this is to create the table with the correct geometry column before writing data to it with Spark. Alternatively, you can write your data to the table using Spark and then manually alter the column to be a geometry type afterward.

Postgis uses EWKB to serialize geometries. If you convert your geometries to EWKB format in Sedona you don't have to do any additional conversion in Postgis.

my_postgis_db# create table my_table (id int8, geom geometry);

df.withColumn("geom", expr("ST_AsEWKB(geom)")
    .write.format("jdbc")
    .option("truncate","true") // Don't let Spark recreate the table.
    // Other options.
    .save()

// If you didn't create the table before writing you can change the type afterward.
my_postgis_db# alter table my_table alter column geom type geometry;

Convert between DataFrame and SpatialRDD¶

DataFrame to SpatialRDD¶

Use SedonaSQL DataFrame-RDD Adapter to convert a DataFrame to an SpatialRDD. Please read Adapter Scaladoc

ScalaJavaPython

var spatialRDD = Adapter.toSpatialRdd(spatialDf, "usacounty")

SpatialRDD spatialRDD = Adapter.toSpatialRdd(spatialDf, "usacounty")

from sedona.utils.adapter import Adapter

spatialRDD = Adapter.toSpatialRdd(spatialDf, "usacounty")

"usacounty" is the name of the geometry column

Warning

Only one Geometry type column is allowed per DataFrame.

SpatialRDD to DataFrame¶

Use SedonaSQL DataFrame-RDD Adapter to convert a DataFrame to an SpatialRDD. Please read Adapter Scaladoc

ScalaJavaPython

var spatialDf = Adapter.toDf(spatialRDD, sedona)

Dataset<Row> spatialDf = Adapter.toDf(spatialRDD, sedona)

from sedona.utils.adapter import Adapter

spatialDf = Adapter.toDf(spatialRDD, sedona)

All other attributes such as price and age will be also brought to the DataFrame as long as you specify carryOtherAttributes (see Read other attributes in an SpatialRDD).

You may also manually specify a schema for the resulting DataFrame in case you require different column names or data types. Note that string schemas and not all data types are supported—please check the Adapter Scaladoc to confirm what is supported for your use case. At least one column for the user data must be provided.

Scala

val schema = StructType(Array(
  StructField("county", GeometryUDT, nullable = true),
  StructField("name", StringType, nullable = true),
  StructField("price", DoubleType, nullable = true),
  StructField("age", IntegerType, nullable = true)
))
val spatialDf = Adapter.toDf(spatialRDD, schema, sedona)

SpatialPairRDD to DataFrame¶

PairRDD is the result of a spatial join query or distance join query. SedonaSQL DataFrame-RDD Adapter can convert the result to a DataFrame. But you need to provide the name of other attributes.

ScalaJavaPython

var joinResultDf = Adapter.toDf(joinResultPairRDD, Seq("left_attribute1", "left_attribute2"), Seq("right_attribute1", "right_attribute2"), sedona)

import scala.collection.JavaConverters;

List leftFields = new ArrayList<>(Arrays.asList("c1", "c2", "c3"));
List rightFields = new ArrayList<>(Arrays.asList("c4", "c5", "c6"));
Dataset joinResultDf = Adapter.toDf(joinResultPairRDD, JavaConverters.asScalaBuffer(leftFields).toSeq(), JavaConverters.asScalaBuffer(rightFields).toSeq(), sedona);

from sedona.utils.adapter import Adapter

joinResultDf = Adapter.toDf(jvm_sedona_rdd, ["poi_from_id", "poi_from_name"], ["poi_to_id", "poi_to_name"], spark))

or you can use the attribute names directly from the input RDD

ScalaJavaPython

import scala.collection.JavaConversions._
var joinResultDf = Adapter.toDf(joinResultPairRDD, leftRdd.fieldNames, rightRdd.fieldNames, sedona)

import scala.collection.JavaConverters;
Dataset joinResultDf = Adapter.toDf(joinResultPairRDD, JavaConverters.asScalaBuffer(leftRdd.fieldNames).toSeq(), JavaConverters.asScalaBuffer(rightRdd.fieldNames).toSeq(), sedona);

from sedona.utils.adapter import Adapter

joinResultDf = Adapter.toDf(result_pair_rdd, leftRdd.fieldNames, rightRdd.fieldNames, spark)

All other attributes such as price and age will be also brought to the DataFrame as long as you specify carryOtherAttributes (see Read other attributes in an SpatialRDD).

You may also manually specify a schema for the resulting DataFrame in case you require different column names or data types. Note that string schemas and not all data types are supported—please check the Adapter Scaladoc to confirm what is supported for your use case. Columns for the left and right user data must be provided.

Scala

val schema = StructType(Array(
  StructField("leftGeometry", GeometryUDT, nullable = true),
  StructField("name", StringType, nullable = true),
  StructField("price", DoubleType, nullable = true),
  StructField("age", IntegerType, nullable = true),
  StructField("rightGeometry", GeometryUDT, nullable = true),
  StructField("category", StringType, nullable = true)
))
val joinResultDf = Adapter.toDf(joinResultPairRDD, schema, sedona)