ABRiS - Avro Bridge for Spark

• Pain free Spark/Avro integration.

• Seamlessly integrate with Confluent platform, including Schema Registry with all available naming strategies and schema evolution.

• Seamlessly convert your Avro records from anywhere (e.g. Kafka, Parquet, HDFS, etc) into Spark Rows.

• Convert your Dataframes into Avro records without even specifying a schema.

• Go back-and-forth Spark Avro (since Spark 2.4).

Coordinates for Maven POM dependency

Scala	Abris
2.11
2.12
2.13

Supported versions

Abris	Spark	Scala
6.2.0 - 6.x.x	3.2.1 - 3.5.x	2.12 / 2.13
6.0.0 - 6.1.1	3.2.0	2.12 / 2.13
5.0.0 - 5.x.x	3.0.x / 3.1.x	2.12
5.0.0 - 5.x.x	2.4.x	2.11 / 2.12

From version 6.0.0, ABRiS only supports Spark 3.2.x.

ABRiS 5.0.x is still supported for older versions of Spark (see branch-5)

Older Versions

This is documentation for Abris version 6. Documentation for older versions is located in corresponding branches: branch-5, branch-4, branch-3.2.

Confluent Schema Registry Version

Abris by default uses Confluent client version 6.2.0.

Installation

Abris needs spark-avro to run, make sure you include the spark-avro dependency when using Abris. The version of spark-avro and Spark should be identical.

Example: submitting a Spark job:

./bin/spark-submit \
    --packages org.apache.spark:spark-avro_2.12:3.5.0,za.co.absa:abris_2.12:6.4.0 \
    ...rest of submit params...

Example: using Abris in maven project:

<dependency>
    <groupId>org.apache.spark</groupId>
    <artifactId>spark-core_2.12</artifactId>
    <version>3.5.0</version>
    <scope>provided</scope>
</dependency>
<dependency>
    <groupId>org.apache.spark</groupId>
    <artifactId>spark-avro_2.12</artifactId>
    <version>3.5.0</version> <!-- version must be the same as Spark -->
</dependency>
<dependency>
    <groupId>za.co.absa</groupId>
    <artifactId>abris_2.12</artifactId>
    <version>6.4.0</version>
</dependency>

Example: using Abris in SBT project:

libraryDependencies ++= Seq(
  "org.apache.spark" %% "spark-core" % "3.5.0" % Provided,
  "org.apache.spark" %% "spark-avro" % "3.5.0",
  "za.co.absa" %% "abris" % "6.4.0"
)

Usage

ABRiS API is in it's most basic form almost identical to Spark built-in support for Avro, but it provides additional functionality. Mainly it's support of schema registry and also seamless integration with confluent Avro data format.

The API consists of two Spark SQL expressions (to_avro and from_avro) and fluent configurator (AbrisConfig)

Using the configurator you can choose from four basic config types:

• toSimpleAvro, toConfluentAvro, fromSimpleAvro and fromConfluentAvro

And configure what you want to do, mainly how to get the avro schema.

Example of usage:

val abrisConfig = AbrisConfig
  .fromConfluentAvro
  .downloadReaderSchemaByLatestVersion
  .andTopicNameStrategy("topic123")
  .usingSchemaRegistry("http://localhost:8081")

import za.co.absa.abris.avro.functions.from_avro
val deserialized = dataFrame.select(from_avro(col("value"), abrisConfig) as 'data)

Detailed instructions for many use cases are in separated documents:

• How to use Abris with vanilla avro (with examples)
• How to use Abris with Confluent avro (with examples)
• How to use Abris in Python (with examples)

Full runnable examples can be found in the za.co.absa.abris.examples package. You can also take a look at unit tests in package za.co.absa.abris.avro.sql.

IMPORTANT: Spark dependencies have provided scope in the pom.xml, so when running the examples, please make sure that you either, instruct your IDE to include dependencies with provided scope, or change the scope directly.

Confluent Avro format

The format of Avro binary data is defined in Avro specification. Confluent format extends it and prepends the schema id before the actual record. The Confluent expressions in this library expect this format and add the id after the Avro data are generated or remove it before they are parsed.

You can find more about Confluent and Schema Registry in Confluent documentation.

Schema Registry security and other additional settings

Only Schema registry client setting that is mandatory is the url, but if you need to provide more the configurer allows you to provide a whole map.

For example, you may want to provide basic.auth.user.info and basic.auth.credentials.source required for user authentication. You can do it this way:

val registryConfig = Map(
  AbrisConfig.SCHEMA_REGISTRY_URL -> "http://localhost:8081",
  "basic.auth.credentials.source" -> "USER_INFO",
  "basic.auth.user.info" -> "srkey:srvalue"
)

val abrisConfig = AbrisConfig
  .fromConfluentAvro
  .downloadReaderSchemaByLatestVersion
  .andTopicNameStrategy("topic123")
  .usingSchemaRegistry(registryConfig) // use the map instead of just url

Other Features

Generating Avro schema from Spark data frame column

There is a helper method that allows you to generate schema automatically from spark column. Assuming you have a data frame containing column "input". You can generate schema for data in that column like this:

val schema = AvroSchemaUtils.toAvroSchema(dataFrame, "input")

Using schema manager to directly download or register schema

You can use SchemaManager directly to do operations with schema registry. The configuration is identical to Schema Registry Client. The SchemaManager is just a wrapper around the client providing helpful methods and abstractions.

val schemaRegistryClientConfig = Map( ...configuration... )
val schemaManager = SchemaManagerFactory.create(schemaRegistryClientConfig)

// Downloading schema:
val schema = schemaManager.getSchemaById(42)

// Registering schema:
val schemaString = "{...avro schema json...}"
val subject = SchemaSubject.usingTopicNameStrategy("fooTopic")
val schemaId = schemaManager.register(subject, schemaString)

// and more, check SchemaManager's methods

De-serialisation Error Handling

There are 2 ways ABRiS handles de-serialisation errors:

FailFast (Default)

Given no provided de-serialisation handler, a failure will result in a spark exception being thrown and with the error being outputted. This is the default procedure.

SpecificRecordHandler

The second option requires providing a default record that will be outputted in the event of a failure. This should be used as a flag to be deleted outside ABRiS that should mean the spark job will not stop. Beware however, a null or empty record will also result in an error so a record with a different input should be chosen.

This can be provided as such:

val abrisConfig = AbrisConfig
  .fromConfluentAvro
  .downloadReaderSchemaByLatestVersion
  .andTopicNameStrategy("topic123")
  .usingSchemaRegistry(registryConfig)
  .withSchemaConverter("custom")
  .withExceptionHandler(new SpecificRecordExceptionHandler(providedDefaultRecord))

This is only for confluent-based configuration, not for standard avro.

PermissiveRecordExceptionHandler

The third option is to use the PermissiveRecordExceptionHandler. In case of a deserialization failure, this handler replaces the problematic record with a fully null record, instead of throwing an exception. This allows the data processing pipeline to continue without interruption.

The main use case for this option is when you want to prioritize continuity of processing over individual record integrity. It's especially useful when dealing with large datasets where occasional malformed records could be tolerated.

Here's how to use it:

val abrisConfig = AbrisConfig
  .fromConfluentAvro
  .downloadReaderSchemaByLatestVersion
  .andTopicNameStrategy("topic123")
  .usingSchemaRegistry(registryConfig)
  .withSchemaConverter("custom")
  .withExceptionHandler(new PermissiveRecordExceptionHandler())

With this configuration, in the event of a deserialization error, the PermissiveRecordExceptionHandler will log a warning, substitute the malformed record with a fully null one, and allow the data processing pipeline to continue.

Data Conversions

This library also provides convenient methods to convert between Avro and Spark schemas.

If you have an Avro schema which you want to convert into a Spark SQL one - to generate your Dataframes, for instance - you can do as follows:

val avroSchema: Schema = AvroSchemaUtils.load("path_to_avro_schema")
val sqlSchema: StructType = SparkAvroConversions.toSqlType(avroSchema) 

You can also do the inverse operation by running:

val sqlSchema = new StructType(new StructField ....
val avroSchema = SparkAvroConversions.toAvroSchema(sqlSchema, avro_schema_name, avro_schema_namespace)

Custom data conversions

If you would like to use custom logic to convert from Avro to Spark, you can implement the SchemaConverter trait. The custom class is loaded in ABRiS using the service provider interface (SPI), so you need to register your class in your META-INF/services resource directory. You can then configure the custom class with its short name or the fully qualified name.

Example

Custom schema converter implementation

package za.co.absa.abris.avro.sql
import org.apache.avro.Schema
import org.apache.spark.sql.types.DataType

class CustomSchemaConverter extends SchemaConverter {
  override val shortName: String = "custom"
  override def toSqlType(avroSchema: Schema): DataType = ???
}

Provider configuration file META-INF/services/za.co.absa.abris.avro.sql.SchemaConverter:

za.co.absa.abris.avro.sql.CustomSchemaConverter

Abris configuration

val abrisConfig = AbrisConfig
  .fromConfluentAvro
  .downloadReaderSchemaByLatestVersion
  .andTopicNameStrategy("topic123")
  .usingSchemaRegistry(registryConfig)
  .withSchemaConverter("custom")

Multiple schemas in one topic

The naming strategies RecordName and TopicRecordName allow for a one topic to receive different payloads, i.e. payloads containing different schemas that do not have to be compatible, as explained here.

When you read such data from Kafka they will be stored as binary column in a dataframe, but once you convert them to Spark types they cannot be in one dataframe, because all rows in dataframe must have the same schema.

So if you have multiple incompatible types of avro data in a dataframe you must first sort them out to several dataframes. One for each schema. Then you can use Abris and convert the avro data.

How to measure code coverage

./mvn clean verify -Pcode-coverage,scala-2.12
or
./mvn clean verify -Pcode-coverage,scala-2.13

Code coverage reports will be generated on paths:

{local-path}\ABRiS\target\jacoco

---

Copyright 2018 ABSA Group Limited

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at

    http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.