Apache Avro Data Source Guide

Since Spark 2.4 release, Spark SQL provides built-in support for reading and writing Apache Avro data.

Deploying

The spark-avro module is external and not included in spark-submit or spark-shell by default.

As with any Spark applications, spark-submit is used to launch your application. spark-avro_2.12 and its dependencies can be directly added to spark-submit using --packages, such as,

./bin/spark-submit --packages org.apache.spark:spark-avro_2.12:3.0.3 ...

For experimenting on spark-shell, you can also use --packages to add org.apache.spark:spark-avro_2.12 and its dependencies directly,

./bin/spark-shell --packages org.apache.spark:spark-avro_2.12:3.0.3 ...

See Application Submission Guide for more details about submitting applications with external dependencies.

Load and Save Functions

Since spark-avro module is external, there is no .avro API in DataFrameReader or DataFrameWriter.

To load/save data in Avro format, you need to specify the data source option format as avro(or org.apache.spark.sql.avro).

val usersDF = spark.read.format("avro").load("examples/src/main/resources/users.avro")
usersDF.select("name", "favorite_color").write.format("avro").save("namesAndFavColors.avro")
Dataset<Row> usersDF = spark.read().format("avro").load("examples/src/main/resources/users.avro");
usersDF.select("name", "favorite_color").write().format("avro").save("namesAndFavColors.avro");
df = spark.read.format("avro").load("examples/src/main/resources/users.avro")
df.select("name", "favorite_color").write.format("avro").save("namesAndFavColors.avro")
df <- read.df("examples/src/main/resources/users.avro", "avro")
write.df(select(df, "name", "favorite_color"), "namesAndFavColors.avro", "avro")

to_avro() and from_avro()

The Avro package provides function to_avro to encode a column as binary in Avro format, and from_avro() to decode Avro binary data into a column. Both functions transform one column to another column, and the input/output SQL data type can be a complex type or a primitive type.

Using Avro record as columns is useful when reading from or writing to a streaming source like Kafka. Each Kafka key-value record will be augmented with some metadata, such as the ingestion timestamp into Kafka, the offset in Kafka, etc.

Both functions are currently only available in Scala, Java, and Python.

import org.apache.spark.sql.avro.functions._

// `from_avro` requires Avro schema in JSON string format.
val jsonFormatSchema = new String(Files.readAllBytes(Paths.get("./examples/src/main/resources/user.avsc")))

val df = spark
  .readStream
  .format("kafka")
  .option("kafka.bootstrap.servers", "host1:port1,host2:port2")
  .option("subscribe", "topic1")
  .load()

// 1. Decode the Avro data into a struct;
// 2. Filter by column `favorite_color`;
// 3. Encode the column `name` in Avro format.
val output = df
  .select(from_avro('value, jsonFormatSchema) as 'user)
  .where("user.favorite_color == \"red\"")
  .select(to_avro($"user.name") as 'value)

val query = output
  .writeStream
  .format("kafka")
  .option("kafka.bootstrap.servers", "host1:port1,host2:port2")
  .option("topic", "topic2")
  .start()
import static org.apache.spark.sql.functions.col;
import static org.apache.spark.sql.avro.functions.*;

// `from_avro` requires Avro schema in JSON string format.
String jsonFormatSchema = new String(Files.readAllBytes(Paths.get("./examples/src/main/resources/user.avsc")));

Dataset<Row> df = spark
  .readStream()
  .format("kafka")
  .option("kafka.bootstrap.servers", "host1:port1,host2:port2")
  .option("subscribe", "topic1")
  .load();

// 1. Decode the Avro data into a struct;
// 2. Filter by column `favorite_color`;
// 3. Encode the column `name` in Avro format.
Dataset<Row> output = df
  .select(from_avro(col("value"), jsonFormatSchema).as("user"))
  .where("user.favorite_color == \"red\"")
  .select(to_avro(col("user.name")).as("value"));

StreamingQuery query = output
  .writeStream()
  .format("kafka")
  .option("kafka.bootstrap.servers", "host1:port1,host2:port2")
  .option("topic", "topic2")
  .start();
from pyspark.sql.avro.functions import from_avro, to_avro

# `from_avro` requires Avro schema in JSON string format.
jsonFormatSchema = open("examples/src/main/resources/user.avsc", "r").read()

df = spark\
  .readStream\
  .format("kafka")\
  .option("kafka.bootstrap.servers", "host1:port1,host2:port2")\
  .option("subscribe", "topic1")\
  .load()

# 1. Decode the Avro data into a struct;
# 2. Filter by column `favorite_color`;
# 3. Encode the column `name` in Avro format.
output = df\
  .select(from_avro("value", jsonFormatSchema).alias("user"))\
  .where('user.favorite_color == "red"')\
  .select(to_avro("user.name").alias("value"))

query = output\
  .writeStream\
  .format("kafka")\
  .option("kafka.bootstrap.servers", "host1:port1,host2:port2")\
  .option("topic", "topic2")\
  .start()

Data Source Option

Data source options of Avro can be set via:

Property NameDefaultMeaningScope
avroSchema None Optional schema provided by a user in JSON format.
  • When reading Avro, this option can be set to an evolved schema, which is compatible but different with the actual Avro schema. The deserialization schema will be consistent with the evolved schema. For example, if we set an evolved schema containing one additional column with a default value, the reading result in Spark will contain the new column too.
  • When writing Avro, this option can be set if the expected output Avro schema doesn't match the schema converted by Spark. For example, the expected schema of one column is of "enum" type, instead of "string" type in the default converted schema.
read, write and function from_avro
recordName topLevelRecord Top level record name in write result, which is required in Avro spec. write
recordNamespace "" Record namespace in write result. write
ignoreExtension true The option controls ignoring of files without .avro extensions in read.
If the option is enabled, all files (with and without .avro extension) are loaded.
The option has been deprecated, and it will be removed in the future releases. Please use the general data source option pathGlobFilter for filtering file names.
read
compression snappy The compression option allows to specify a compression codec used in write.
Currently supported codecs are uncompressed, snappy, deflate, bzip2 and xz.
If the option is not set, the configuration spark.sql.avro.compression.codec config is taken into account.
write
mode FAILFAST The mode option allows to specify parse mode for function from_avro.
Currently supported modes are:
  • FAILFAST: Throws an exception on processing corrupted record.
  • PERMISSIVE: Corrupt records are processed as null result. Therefore, the data schema is forced to be fully nullable, which might be different from the one user provided.
function from_avro

Configuration

Configuration of Avro can be done using the setConf method on SparkSession or by running SET key=value commands using SQL.

Property NameDefaultMeaningSince Version
spark.sql.legacy.replaceDatabricksSparkAvro.enabled true If it is set to true, the data source provider com.databricks.spark.avro is mapped to the built-in but external Avro data source module for backward compatibility. 2.4.0
spark.sql.avro.compression.codec snappy Compression codec used in writing of AVRO files. Supported codecs: uncompressed, deflate, snappy, bzip2 and xz. Default codec is snappy. 2.4.0
spark.sql.avro.deflate.level -1 Compression level for the deflate codec used in writing of AVRO files. Valid value must be in the range of from 1 to 9 inclusive or -1. The default value is -1 which corresponds to 6 level in the current implementation. 2.4.0

Compatibility with Databricks spark-avro

This Avro data source module is originally from and compatible with Databricks’s open source repository spark-avro.

By default with the SQL configuration spark.sql.legacy.replaceDatabricksSparkAvro.enabled enabled, the data source provider com.databricks.spark.avro is mapped to this built-in Avro module. For the Spark tables created with Provider property as com.databricks.spark.avro in catalog meta store, the mapping is essential to load these tables if you are using this built-in Avro module.

Note in Databricks’s spark-avro, implicit classes AvroDataFrameWriter and AvroDataFrameReader were created for shortcut function .avro(). In this built-in but external module, both implicit classes are removed. Please use .format("avro") in DataFrameWriter or DataFrameReader instead, which should be clean and good enough.

If you prefer using your own build of spark-avro jar file, you can simply disable the configuration spark.sql.legacy.replaceDatabricksSparkAvro.enabled, and use the option --jars on deploying your applications. Read the Advanced Dependency Management section in Application Submission Guide for more details.

Supported types for Avro -> Spark SQL conversion

Currently Spark supports reading all primitive types and complex types under records of Avro.

Avro typeSpark SQL type
boolean BooleanType
int IntegerType
long LongType
float FloatType
double DoubleType
string StringType
enum StringType
fixed BinaryType
bytes BinaryType
record StructType
array ArrayType
map MapType
union See below

In addition to the types listed above, it supports reading union types. The following three types are considered basic union types:

  1. union(int, long) will be mapped to LongType.
  2. union(float, double) will be mapped to DoubleType.
  3. union(something, null), where something is any supported Avro type. This will be mapped to the same Spark SQL type as that of something, with nullable set to true. All other union types are considered complex. They will be mapped to StructType where field names are member0, member1, etc., in accordance with members of the union. This is consistent with the behavior when converting between Avro and Parquet.

It also supports reading the following Avro logical types:

Avro logical typeAvro typeSpark SQL type
date int DateType
timestamp-millis long TimestampType
timestamp-micros long TimestampType
decimal fixed DecimalType
decimal bytes DecimalType

At the moment, it ignores docs, aliases and other properties present in the Avro file.

Supported types for Spark SQL -> Avro conversion

Spark supports writing of all Spark SQL types into Avro. For most types, the mapping from Spark types to Avro types is straightforward (e.g. IntegerType gets converted to int); however, there are a few special cases which are listed below:

Spark SQL typeAvro typeAvro logical type
ByteType int
ShortType int
BinaryType bytes
DateType int date
TimestampType long timestamp-micros
DecimalType fixed decimal

You can also specify the whole output Avro schema with the option avroSchema, so that Spark SQL types can be converted into other Avro types. The following conversions are not applied by default and require user specified Avro schema:

Spark SQL typeAvro typeAvro logical type
BinaryType fixed
StringType enum
TimestampType long timestamp-millis
DecimalType bytes decimal