Read and Write XML files in PySpark

This article shows you how to read and write XML files in Spark.

Sample XML file

Create a sample XML file named test.xml with the following content:

<?xml version="1.0"?>
<data>
    <record id="1">
        <rid>1</rid>
        <name>Record 1</name>
    </record>
    <record id="2">
        <rid>2</rid>
        <name>Record 2</name>
    </record>
    <record id="3">
        <rid>3</rid>
        <name>Record 3</name>
    </record>
</data>

Dependent library

For more information, refer to the following repo in GitHub. 

spark-xml

You can download this package directly from Maven repository: https://mvnrepository.com/artifact/com.databricks/spark-xml.

Make sure this package exists in your Spark environment.  Alternatively you can pass in this package as parameter when running Spark job using spark-submit or pyspark command.  For example:

spark-submit --jars spark-xml_2.12-0.6.0.jar ...

Error debug

You may encounter the following error:

py4j.protocol.Py4JJavaError: An error occurred while calling o44.load.
: java.lang.BootstrapMethodError: java.lang.NoClassDefFoundError: scala/runtime/java8/JFunction0$mcD$sp
        at com.databricks.spark.xml.XmlOptions.<init>(XmlOptions.scala:36)
        at com.databricks.spark.xml.XmlOptions$.apply(XmlOptions.scala:65)
        at com.databricks.spark.xml.DefaultSource.createRelation(DefaultSource.scala:66)
        at com.databricks.spark.xml.DefaultSource.createRelation(DefaultSource.scala:29)

This occurred because Scala version is not matching with spark-xml dependency version.

For example, spark-xml_2.12-0.6.0.jar depends on Scala version 2.12.8. For example, you can change to a different version of Spark XML package. 

spark-submit --jars spark-xml_2.11-0.4.1.jar ...

Read XML file

Remember to change your file location accordingly. 

from pyspark.sql import SparkSession
from pyspark.sql.types import StructType, StructField, StringType, IntegerType
from decimal import Decimal
appName = "Python Example - PySpark Read XML"
master = "local"

# Create Spark session
spark = SparkSession.builder \
    .appName(appName) \
    .master(master) \
    .getOrCreate()

schema = StructType([
    StructField('_id', IntegerType(), False),
    StructField('rid', IntegerType(), False),
    StructField('name', StringType(), False)
])

df = spark.read.format("com.databricks.spark.xml") \
    .option("rowTag","record").load("file:///home/tangr/python-examples/test.xml", schema=schema)

df.show()

Output

The attribute is converted to column _${AttributeName} (with prefix _) while the child element is converted to column.

+---+---+--------+
|_id|rid|    name|
+---+---+--------+
|  1|  1|Record 1|
|  2|  2|Record 2|
|  3|  3|Record 3|
+---+---+--------+

Write XML file

df.select("rid","name").write.format("com.databricks.spark.xml").option("rootTag", "data").option("rowTag", "record").mode(
    "overwrite").save('file:///home/tangr/python-examples/test2.xml')

Files are saved as partition files based on your parallelism setup in Spark session.

Output

<data>
    <record>
        <rid>1</rid>
        <name>Record 1</name>
    </record>
    <record>
        <rid>2</rid>
        <name>Record 2</name>
    </record>
    <record>
        <rid>3</rid>
        <name>Record 3</name>
    </record>
</data>

References

If you want to read single local file using Python, refer to the following article:

Read and Write XML Files with Python