XML Data Source for Apache Spark

• A library for parsing and querying XML data with Apache Spark, for Spark SQL and DataFrames. The structure and test tools are mostly copied from CSV Data Source for Spark.

• This package supports to process format-free XML files in a distributed way, unlike JSON datasource in Spark restricts in-line JSON format.

• Compatible with Spark 3.0 and later with Scala 2.12, and also Spark 3.2 and later with Scala 2.12 or 2.13. Scala 2.11 and Spark 2 support ended with version 0.13.0.

Linking

You can link against this library in your program at the following coordinates:

groupId: com.databricks
artifactId: spark-xml_2.12
version: 0.15.0

Using with Spark shell

This package can be added to Spark using the --packages command line option. For example, to include it when starting the spark shell:

$SPARK_HOME/bin/spark-shell --packages com.databricks:spark-xml_2.12:0.15.0

Features

This package allows reading XML files in local or distributed filesystem as Spark DataFrames.

When reading files the API accepts several options:

• path: Location of files. Similar to Spark can accept standard Hadoop globbing expressions.
• rowTag: The row tag of your xml files to treat as a row. For example, in this xml <books> <book><book> ...</books>, the appropriate value would be book. Default is ROW.
• samplingRatio: Sampling ratio for inferring schema (0.0 ~ 1). Default is 1. Possible types are StructType, ArrayType, StringType, LongType, DoubleType, BooleanType, TimestampType and NullType, unless user provides a schema for this.
• excludeAttribute : Whether you want to exclude attributes in elements or not. Default is false.
• treatEmptyValuesAsNulls : (DEPRECATED: use nullValue set to "") Whether you want to treat whitespaces as a null value. Default is false
• mode: The mode for dealing with corrupt records during parsing. Default is PERMISSIVE.
  • PERMISSIVE :
    • When it encounters a corrupted record, it sets all fields to null and puts the malformed string into a new field configured by columnNameOfCorruptRecord.
    • When it encounters a field of the wrong datatype, it sets the offending field to null.
  • DROPMALFORMED : ignores the whole corrupted records.
  • FAILFAST : throws an exception when it meets corrupted records.
• inferSchema: if true, attempts to infer an appropriate type for each resulting DataFrame column, like a boolean, numeric or date type. If false, all resulting columns are of string type. Default is true.
• columnNameOfCorruptRecord: The name of new field where malformed strings are stored. Default is _corrupt_record.
• attributePrefix: The prefix for attributes so that we can differentiate attributes and elements. This will be the prefix for field names. Default is _. Can be empty, but only for reading XML.
• valueTag: The tag used for the value when there are attributes in the element having no child. Default is _VALUE.
• charset: Defaults to 'UTF-8' but can be set to other valid charset names
• ignoreSurroundingSpaces: Defines whether or not surrounding whitespaces from values being read should be skipped. Default is false.
• wildcardColName: Name of a column existing in the provided schema which is interpreted as a 'wildcard'. It must have type string or array of strings. It will match any XML child element that is not otherwise matched by the schema. The XML of the child becomes the string value of the column. If an array, then all unmatched elements will be returned as an array of strings. As its name implies, it is meant to emulate XSD's xs:any type. Default is xs_any. New in 0.11.0.
• rowValidationXSDPath: Path to an XSD file that is used to validate the XML for each row individually. Rows that fail to validate are treated like parse errors as above. The XSD does not otherwise affect the schema provided, or inferred. Note that if the same local path is not already also visible on the executors in the cluster, then the XSD and any others it depends on should be added to the Spark executors with SparkContext.addFile. In this case, to use local XSD /foo/bar.xsd, call addFile("/foo/bar.xsd") and pass just "bar.xsd" as rowValidationXSDPath.
• ignoreNamespace: If true, namespaces prefixes on XML elements and attributes are ignored. Tags <abc:author> and <def:author> would, for example, be treated as if both are just <author>. Note that, at the moment, namespaces cannot be ignored on the rowTag element, only its children. Note that XML parsing is in general not namespace-aware even if false. Defaults to false. New in 0.11.0.
• timestampFormat: Specifies an additional timestamp format that will be tried when parsing values as TimestampType columns. The format is specified as described in DateTimeFormatter. Defaults to try several formats, including ISO_INSTANT, including variations with offset timezones or no timezone (defaults to UTC). New in 0.12.0.
• dateFormat: Specifies an additional timestamp format that will be tried when parsing values as DateType columns. The format is specified as described in DateTimeFormatter. Defaults to ISO_DATE. New in 0.12.0.

When writing files the API accepts several options:

• path: Location to write files.
• rowTag: The row tag of your xml files to treat as a row. For example, in <books> <book><book> ...</books>, the appropriate value would be book. Default is ROW.
• rootTag: The root tag of your xml files to treat as the root. For example, in <books> <book><book> ...</books>, the appropriate value would be books. It can include basic attributes by specifying a value like books foo="bar" (as of 0.11.0). Default is ROWS.
• declaration: Content of XML declaration to write at the start of every output XML file, before the rootTag. For example, a value of foo causes <?xml foo?> to be written. Set to empty string to suppress. Defaults to version="1.0" encoding="UTF-8" standalone="yes". New in 0.14.0.
• nullValue: The value to write null value. Default is string null. When this is null, it does not write attributes and elements for fields.
• attributePrefix: The prefix for attributes so that we can differentiating attributes and elements. This will be the prefix for field names. Default is _. Cannot be empty for writing XML.
• valueTag: The tag used for the value when there are attributes in the element having no child. Default is _VALUE.
• compression: compression codec to use when saving to file. Should be the fully qualified name of a class implementing org.apache.hadoop.io.compress.CompressionCodec or one of case-insensitive shorten names (bzip2, gzip, lz4, and snappy). Defaults to no compression when a codec is not specified.
• timestampFormat: Controls the format used to write TimestampType format columns. The format is specified as described in DateTimeFormatter. Defaults to ISO_INSTANT. New in 0.12.0.
• dateFormat: Controls the format used to write DateType format columns. The format is specified as described in DateTimeFormatter. Defaults to ISO_DATE. New in 0.12.0.

Currently it supports the shortened name usage. You can use just xml instead of com.databricks.spark.xml.

XSD Support

Per above, the XML for individual rows can be validated against an XSD using rowValidationXSDPath.

The utility com.databricks.spark.xml.util.XSDToSchema can be used to extract a Spark DataFrame schema from some XSD files. It supports only simple, complex and sequence types, only basic XSD functionality, and is experimental.

import com.databricks.spark.xml.util.XSDToSchema
import java.nio.file.Paths

val schema = XSDToSchema.read(Paths.get("/path/to/your.xsd"))
val df = spark.read.schema(schema)....xml(...)

Parsing Nested XML

Although primarily used to convert (portions of) large XML documents into a DataFrame, spark-xml can also parse XML in a string-valued column in an existing DataFrame with from_xml, in order to add it as a new column with parsed results as a struct.

import com.databricks.spark.xml.functions.from_xml
import com.databricks.spark.xml.schema_of_xml
import spark.implicits._
val df = ... /// DataFrame with XML in column 'payload' 
val payloadSchema = schema_of_xml(df.select("payload").as[String])
val parsed = df.withColumn("parsed", from_xml($"payload", payloadSchema))

• This can convert arrays of strings containing XML to arrays of parsed structs. Use schema_of_xml_array instead
• com.databricks.spark.xml.from_xml_string is an alternative that operates on a String directly instead of a column, for use in UDFs
• If you use DROPMALFORMED mode with from_xml, then XML values that do not parse correctly will result in a null value for the column. No rows will be dropped.
• If you use PERMISSIVE mode with from_xml et al, which is the default, then the parse mode will actually instead default to DROPMALFORMED. If however you include a column in the schema for from_xml that matches the columnNameOfCorruptRecord, then PERMISSIVE mode will still output malformed records to that column in the resulting struct.

Pyspark notes

The functions above are exposed in the Scala API only, at the moment, as there is no separate Python package for spark-xml. They can be accessed from Pyspark by manually declaring some helper functions that call into the JVM-based API from Python. Example:

from pyspark.sql.column import Column, _to_java_column
from pyspark.sql.types import _parse_datatype_json_string

def ext_from_xml(xml_column, schema, options={}):
    java_column = _to_java_column(xml_column.cast('string'))
    java_schema = spark._jsparkSession.parseDataType(schema.json())
    scala_map = spark._jvm.org.apache.spark.api.python.PythonUtils.toScalaMap(options)
    jc = spark._jvm.com.databricks.spark.xml.functions.from_xml(
        java_column, java_schema, scala_map)
    return Column(jc)

def ext_schema_of_xml_df(df, options={}):
    assert len(df.columns) == 1

    scala_options = spark._jvm.PythonUtils.toScalaMap(options)
    java_xml_module = getattr(getattr(
        spark._jvm.com.databricks.spark.xml, "package$"), "MODULE$")
    java_schema = java_xml_module.schema_of_xml_df(df._jdf, scala_options)
    return _parse_datatype_json_string(java_schema.json())

Structure Conversion

Due to the structure differences between DataFrame and XML, there are some conversion rules from XML data to DataFrame and from DataFrame to XML data. Note that handling attributes can be disabled with the option excludeAttribute.

Conversion from XML to DataFrame

• Attributes: Attributes are converted as fields with the heading prefix, attributePrefix.

  <one myOneAttrib="AAAA">
      <two>two</two>
      <three>three</three>
  </one>

  produces a schema below:

  root
   |-- _myOneAttrib: string (nullable = true)
   |-- two: string (nullable = true)
   |-- three: string (nullable = true)
  

• Value in an element that has no child elements but attributes: The value is put in a separate field, valueTag.

  <one>
      <two myTwoAttrib="BBBBB">two</two>
      <three>three</three>
  </one>

  produces a schema below:

  root
   |-- two: struct (nullable = true)
   |    |-- _VALUE: string (nullable = true)
   |    |-- _myTwoAttrib: string (nullable = true)
   |-- three: string (nullable = true)
  

Conversion from DataFrame to XML

• Element as an array in an array: Writing a XML file from DataFrame having a field ArrayType with its element as ArrayType would have an additional nested field for the element. This would not happen in reading and writing XML data but writing a DataFrame read from other sources. Therefore, roundtrip in reading and writing XML files has the same structure but writing a DataFrame read from other sources is possible to have a different structure.

  DataFrame with a schema below:

   |-- a: array (nullable = true)
   |    |-- element: array (containsNull = true)
   |    |    |-- element: string (containsNull = true)
  

  with data below:

  +------------------------------------+
  |                                   a|
  +------------------------------------+
  |[WrappedArray(aa), WrappedArray(bb)]|
  +------------------------------------+
  

  produces a XML file below:

  <a>
      <item>aa</item>
  </a>
  <a>
      <item>bb</item>
  </a>

Examples

These examples use a XML file available for download here:

$ wget https://github.com/databricks/spark-xml/raw/master/src/test/resources/books.xml

SQL API

XML data source for Spark can infer data types:

CREATE TABLE books
USING com.databricks.spark.xml
OPTIONS (path "books.xml", rowTag "book")

You can also specify column names and types in DDL. In this case, we do not infer schema.

CREATE TABLE books (author string, description string, genre string, _id string, price double, publish_date string, title string)
USING com.databricks.spark.xml
OPTIONS (path "books.xml", rowTag "book")

Scala API

Import com.databricks.spark.xml._ to get implicits that add the .xml(...) method to DataFrame. You can also use .format("xml") and .load(...).

import org.apache.spark.sql.SparkSession
import com.databricks.spark.xml._

val spark = SparkSession.builder().getOrCreate()
val df = spark.read
  .option("rowTag", "book")
  .xml("books.xml")

val selectedData = df.select("author", "_id")
selectedData.write
  .option("rootTag", "books")
  .option("rowTag", "book")
  .xml("newbooks.xml")

You can manually specify the schema when reading data:

import org.apache.spark.sql.SparkSession
import org.apache.spark.sql.types.{StructType, StructField, StringType, DoubleType}
import com.databricks.spark.xml._

val spark = SparkSession.builder().getOrCreate()
val customSchema = StructType(Array(
  StructField("_id", StringType, nullable = true),
  StructField("author", StringType, nullable = true),
  StructField("description", StringType, nullable = true),
  StructField("genre", StringType, nullable = true),
  StructField("price", DoubleType, nullable = true),
  StructField("publish_date", StringType, nullable = true),
  StructField("title", StringType, nullable = true)))


val df = spark.read
  .option("rowTag", "book")
  .schema(customSchema)
  .xml("books.xml")

val selectedData = df.select("author", "_id")
selectedData.write
  .option("rootTag", "books")
  .option("rowTag", "book")
  .xml("newbooks.xml")

Java API

import org.apache.spark.sql.SparkSession;

SparkSession spark = SparkSession.builder().getOrCreate();
DataFrame df = spark.read()
  .format("xml")
  .option("rowTag", "book")
  .load("books.xml");

df.select("author", "_id").write()
  .format("xml")
  .option("rootTag", "books")
  .option("rowTag", "book")
  .save("newbooks.xml");

You can manually specify schema:

import org.apache.spark.sql.SparkSession;
import org.apache.spark.sql.types.*;

SparkSession spark = SparkSession.builder().getOrCreate();
StructType customSchema = new StructType(new StructField[] {
  new StructField("_id", DataTypes.StringType, true, Metadata.empty()),
  new StructField("author", DataTypes.StringType, true, Metadata.empty()),
  new StructField("description", DataTypes.StringType, true, Metadata.empty()),
  new StructField("genre", DataTypes.StringType, true, Metadata.empty()),
  new StructField("price", DataTypes.DoubleType, true, Metadata.empty()),
  new StructField("publish_date", DataTypes.StringType, true, Metadata.empty()),
  new StructField("title", DataTypes.StringType, true, Metadata.empty())
});

DataFrame df = spark.read()
  .format("xml")
  .option("rowTag", "book")
  .schema(customSchema)
  .load("books.xml");

df.select("author", "_id").write()
  .format("xml")
  .option("rootTag", "books")
  .option("rowTag", "book")
  .save("newbooks.xml");

Python API

from pyspark.sql import SparkSession
spark = SparkSession.builder.getOrCreate()

df = spark.read.format('xml').options(rowTag='book').load('books.xml')
df.select("author", "_id").write \
    .format('xml') \
    .options(rowTag='book', rootTag='books') \
    .save('newbooks.xml')

You can manually specify schema:

from pyspark.sql import SparkSession
from pyspark.sql.types import *

spark = SparkSession.builder.getOrCreate()
customSchema = StructType([
    StructField("_id", StringType(), True),
    StructField("author", StringType(), True),
    StructField("description", StringType(), True),
    StructField("genre", StringType(), True),
    StructField("price", DoubleType(), True),
    StructField("publish_date", StringType(), True),
    StructField("title", StringType(), True)])

df = spark.read \
    .format('xml') \
    .options(rowTag='book') \
    .load('books.xml', schema = customSchema)

df.select("author", "_id").write \
    .format('xml') \
    .options(rowTag='book', rootTag='books') \
    .save('newbooks.xml')

R API

Automatically infer schema (data types)

library(SparkR)

sparkR.session("local[4]", sparkPackages = c("com.databricks:spark-xml_2.12:0.15.0"))

df <- read.df("books.xml", source = "xml", rowTag = "book")

# In this case, `rootTag` is set to "ROWS" and `rowTag` is set to "ROW".
write.df(df, "newbooks.csv", "xml", "overwrite")

You can manually specify schema:

library(SparkR)

sparkR.session("local[4]", sparkPackages = c("com.databricks:spark-xml_2.12:0.15.0"))
customSchema <- structType(
  structField("_id", "string"),
  structField("author", "string"),
  structField("description", "string"),
  structField("genre", "string"),
  structField("price", "double"),
  structField("publish_date", "string"),
  structField("title", "string"))

df <- read.df("books.xml", source = "xml", schema = customSchema, rowTag = "book")

# In this case, `rootTag` is set to "ROWS" and `rowTag` is set to "ROW".
write.df(df, "newbooks.csv", "xml", "overwrite")

Hadoop InputFormat

The library contains a Hadoop input format for reading XML files by a start tag and an end tag. This is similar with XmlInputFormat.java in Mahout but supports to read compressed files, different encodings and read elements including attributes, which you may make direct use of as follows:

import com.databricks.spark.xml.XmlInputFormat
import org.apache.spark.SparkContext
import org.apache.hadoop.io.{LongWritable, Text}

val sc: SparkContext = _

// This will detect the tags including attributes
sc.hadoopConfiguration.set(XmlInputFormat.START_TAG_KEY, "<book>")
sc.hadoopConfiguration.set(XmlInputFormat.END_TAG_KEY, "</book>")

val records = sc.newAPIHadoopFile(
  "path",
  classOf[XmlInputFormat],
  classOf[LongWritable],
  classOf[Text])

Building From Source

This library is built with SBT. To build a JAR file simply run sbt package from the project root.

Acknowledgements

This project was initially created by HyukjinKwon and donated to Databricks.