Skip to content

Raster writer


Sedona writers are available in Scala, Java and Python and have the same APIs.

Write Array[Double] to GeoTiff files

Introduction: You can write a GeoTiff dataframe as GeoTiff images using the spark write feature with the format geotiff. The geotiff raster column needs to be an array of double type data.

Since: v1.2.1

Spark SQL example:

The schema of the GeoTiff dataframe to be written can be one of the following two schemas:

 |-- image: struct (nullable = true)
 |    |-- origin: string (nullable = true)
 |    |-- Geometry: geometry (nullable = true)
 |    |-- height: integer (nullable = true)
 |    |-- width: integer (nullable = true)
 |    |-- nBands: integer (nullable = true)
 |    |-- data: array (nullable = true)
 |    |    |-- element: double (containsNull = true)


 |-- origin: string (nullable = true)
 |-- Geometry: geometry (nullable = true)
 |-- height: integer (nullable = true)
 |-- width: integer (nullable = true)
 |-- nBands: integer (nullable = true)
 |-- data: array (nullable = true)
 |    |-- element: double (containsNull = true)

Field names can be renamed, but schema should exactly match with one of the above two schemas. The output path could be a path to a directory where GeoTiff images will be saved. If the directory already exists, write should be called in overwrite mode.

var dfToWrite ="geotiff").option("dropInvalid", true).option("readToCRS", "EPSG:4326").load("PATH_TO_INPUT_GEOTIFF_IMAGES")

You can override an existing path with the following approach:


You can also extract the columns nested within image column and write the dataframe as GeoTiff image.

dfToWrite = dfToWrite.selectExpr("image.origin as origin","image.geometry as geometry", "image.height as height", "image.width as width", " as data", "image.nBands as nBands")

If you want the saved GeoTiff images not to be distributed into multiple partitions, you can call coalesce to merge all files in a single partition.


In case, you rename the columns of GeoTiff dataframe, you can set the corresponding column names with the option parameter. All available optional parameters are listed below:

 |-- writeToCRS: (Default value "EPSG:4326") => Coordinate reference system of the geometry coordinates representing the location of the Geotiff.
 |-- fieldImage: (Default value "image") => Indicates the image column of GeoTiff DataFrame.
 |-- fieldOrigin: (Default value "origin") => Indicates the origin column of GeoTiff DataFrame.
 |-- fieldNBands: (Default value "nBands") => Indicates the nBands column of GeoTiff DataFrame.
 |-- fieldWidth: (Default value "width") => Indicates the width column of GeoTiff DataFrame.
 |-- fieldHeight: (Default value "height") => Indicates the height column of GeoTiff DataFrame.
 |-- fieldGeometry: (Default value "geometry") => Indicates the geometry column of GeoTiff DataFrame.
 |-- fieldData: (Default value "data") => Indicates the data column of GeoTiff DataFrame.

An example:

dfToWrite ="geotiff").option("dropInvalid", true).option("readToCRS", "EPSG:4326").load("PATH_TO_INPUT_GEOTIFF_IMAGES")
dfToWrite = dfToWrite.selectExpr("image.origin as source","ST_GeomFromWkt(image.geometry) as geom", "image.height as height", "image.width as width", " as data", "image.nBands as bands")
dfToWrite.write.mode("overwrite").format("geotiff").option("writeToCRS", "EPSG:4326").option("fieldOrigin", "source").option("fieldGeometry", "geom").option("fieldNBands", "bands").save("DESTINATION_PATH")

Write Array[Double] to other formats


Introduction: Return a Base64 String from a geotiff image

Format: RS_Base64 (height:Int, width:Int, redBand: Array[Double], greenBand: Array[Double], blackBand: Array[Double], optional: alphaBand: Array[Double])

Since: v1.1.0

Spark SQL example:

val BandDF = spark.sql("select RS_Base64(h, w, band1, band2, RS_Array(h*w, 0)) as baseString from dataframe")


|          baseString|


Although the 3 RGB bands are mandatory, you can use RS_Array(h*w, 0.0) to create an array (zeroed out, size = h * w) as input.


Introduction: Return a html img tag with the base64 string embedded

Format: RS_HTML(base64:String, optional: width_in_px:String)

Spark SQL example:

df.selectExpr("RS_HTML(encodedstring, '300') as htmlstring" ).show()


|          htmlstring|
|<img src="data:im...|
|<img src="data:im...|

Last update: March 16, 2023 00:00:53