Avro SerDe for PHP 7.3+ and 8.0
Motivation
When serializing and deserializing messages using the Avro serialization format, especially when integrating with the Confluent Platform, you want to make sure that schemas are evolved in a way that downstream consumers are not affected.
Hence Confluent developed the Schema Registry which has the responsibility to validate a given schema evolution against a configurable compatibility policy.
Unfortunately Confluent is not providing an official Avro SerDe package for PHP. This library aims to provide an Avro SerDe library for PHP that implements the Confluent wire format and integrates FlixTech's Schema Registry Client.
Installation
This library is using the composer package manager for PHP.
composer require 'flix-tech/avro-serde-php:^1.6'Quickstart
NOTE
You should always use a cached schema registry client, since otherwise you'd make an HTTP request for every message serialized or deserialized.
1. Create a cached Schema Registry client
See the Schema Registry client documentation on caching for more detailed information.
<?php
use FlixTech\SchemaRegistryApi\Registry\Cache\AvroObjectCacheAdapter;
use FlixTech\SchemaRegistryApi\Registry\CachedRegistry;
use FlixTech\SchemaRegistryApi\Registry\PromisingRegistry;
use GuzzleHttp\Client;
$schemaRegistryClient = new CachedRegistry(
new PromisingRegistry(
new Client(['base_uri' => 'registry.example.com'])
),
new AvroObjectCacheAdapter()
);2. Build the RecordSerializer instance
The RecordSerializer is the main way you interact with this library. It provides the encodeRecord and
decodeMessage methods for SerDe operations.
<?php
use FlixTech\AvroSerializer\Objects\RecordSerializer;
/** @var \FlixTech\SchemaRegistryApi\Registry $schemaRegistry */
$recordSerializer = new RecordSerializer(
$schemaRegistry,
[
// If you want to auto-register missing schemas set this to true
RecordSerializer::OPTION_REGISTER_MISSING_SCHEMAS => false,
// If you want to auto-register missing subjects set this to true
RecordSerializer::OPTION_REGISTER_MISSING_SUBJECTS => false,
]
);3. Encoding records
This is a simple example on how you can use the RecordSerializer to encode messages in the Confluent Avro wire format.
<?php
/** @var \FlixTech\AvroSerializer\Objects\RecordSerializer $recordSerializer */
$subject = 'my-topic-value';
$avroSchema = AvroSchema::parse('{"type": "string"}');
$record = 'Test message';
$encodedBinaryAvro = $recordSerializer->encodeRecord($subject, $avroSchema, $record);
// Send this over the wire...4. Decoding messages
This is a simple example on how you can use the RecordSerializer to decode messages.
<?php
/** @var \FlixTech\AvroSerializer\Objects\RecordSerializer $recordSerializer */
/** @var string $encodedBinaryAvro */
$record = $recordSerializer->decodeMessage($encodedBinaryAvro);
echo $record; // 'Test message'Schema Resolvers
Schema Resolvers are responsible to know which Avro schema belongs to which type of record. This is especially useful if you want to manage your Avro schemas in separate files. Schema Resolvers enable you to integrate with whatever schema management concept you may have outside of the scope of this library.
Schema Resolvers take a $record of any type and try to resolve a matching AvroSchema instance for it.
FileResolver
In even moderately complicated applications you want to manage your schemas within the VCS, most probably as .avsc
files. These files contain JSON that is describing the Avro schema.
The resolver takes a $baseDir in which you want to manage the files and an inflector callable, which is a simple
function that takes the record as first parameter, and a second boolean $isKey parameter indicating if the inflection
is targeting a key schema.
<?php
namespace MyNamespace;
use FlixTech\AvroSerializer\Objects\SchemaResolvers\FileResolver;
use function get_class;use function is_object;
use function str_replace;
class MyRecord {}
$record = new MyRecord();
$baseDir = __DIR__ . '/files';
$inflector = static function ($record, bool $isKey) {
$ext = $isKey ? '.key.avsc' : '.avsc';
$fileName = is_object($record)
? str_replace('\\', '.', get_class($record))
: 'default';
return $fileName . $ext;
};
echo $inflector($record, false); // MyNamespace.MyRecord.avsc
echo $inflector($record, true); // MyNamespace.MyRecord.key.avsc
$resolver = new FileResolver($baseDir, $inflector);
$resolver->valueSchemaFor($record); // This will load from $baseDir . '/' . MyNamespace.MyRecord.avsc
$resolver->keySchemaFor($record); // This will load from $baseDir . '/' . MyNamespace.MyRecord.key.avscCallableResolver
This is the simplest but also most flexible resolver. It just takes two callables that are responsible to fetch either
value- or key-schemas respectively. A key schema resolver is optional.
<?php
use FlixTech\AvroSerializer\Objects\SchemaResolvers\CallableResolver;
use PHPUnit\Framework\Assert;
use function Widmogrod\Functional\constt;
$valueSchemaJson = '
{
"type": "record",
"name": "user",
"fields": [
{"name": "name", "type": "string"},
{"name": "age", "type": "int"}
]
}
';
$valueSchema = AvroSchema::parse($valueSchemaJson);
$resolver = new CallableResolver(
constt(
AvroSchema::parse($valueSchemaJson)
)
);
$record = [ 'foo' => 'bar' ];
$schema = $resolver->valueSchemaFor($record);
Assert::assertEquals($schema, $valueSchema);DefinitionInterfaceResolver
This library also provides a HasSchemaDefinitionInterface
that exposes two static methods:
HasSchemaDefinitionInterface::valueSchemaJsonreturns the schema definition for the value as JSON stringHasSchemaDefinitionInterface::keySchemaJsonreturns eitherNULLor the schema definition for the key as JSON string.
The DefinitionInterfaceResolver checks if a given record implements that interface (if not it will throw an
InvalidArgumentException) and resolves the schemas via the static methods.
<?php
namespace MyNamespace;
use FlixTech\AvroSerializer\Objects\HasSchemaDefinitionInterface;
use FlixTech\AvroSerializer\Objects\SchemaResolvers\DefinitionInterfaceResolver;
class MyRecord implements HasSchemaDefinitionInterface {
public static function valueSchemaJson() : string
{
return '
{
"type": "record",
"name": "user",
"fields": [
{"name": "name", "type": "string"},
{"name": "age", "type": "int"}
]
}
';
}
public static function keySchemaJson() : ?string
{
return '{"type": "string"}';
}
}
$record = new MyRecord();
$resolver = new DefinitionInterfaceResolver();
$resolver->valueSchemaFor($record); // Will resolve from $record::valueSchemaJson();
$resolver->keySchemaFor($record); // Will resolve from $record::keySchemaJson();ChainResolver
The chain resolver is a useful tool for composing multiple resolvers. The first resolver to be able to resolve a schema
will win. If none of the resolvers in the chain is able to determine a schema, an InvalidArgumentException is thrown.
<?php
namespace MyNamespace;
use FlixTech\AvroSerializer\Objects\SchemaResolvers\ChainResolver;
$record = ['foo' => 'bar'];
/** @var \FlixTech\AvroSerializer\Objects\SchemaResolvers\FileResolver $fileResolver */
/** @var \FlixTech\AvroSerializer\Objects\SchemaResolvers\CallableResolver $callableResolver */
$resolver = new ChainResolver($fileResolver, $callableResolver);
// or new ChainResolver(...[$fileResolver, $callableResolver]);
$resolver->valueSchemaFor($record); // Will resolve $fileResolver, then $callableResolver
$resolver->keySchemaFor($record); // Will resolve $fileResolver, then $callableResolverSymfony Serializer Integration
This library provides integrations with the Symfony Serializer component.
<?php
use FlixTech\AvroSerializer\Integrations\Symfony\Serializer\AvroSerDeEncoder;
use FlixTech\AvroSerializer\Objects\DefaultRecordSerializerFactory;
use PHPUnit\Framework\Assert;
use Symfony\Component\Serializer\Normalizer\GetSetMethodNormalizer;
use Symfony\Component\Serializer\Serializer;
class User
{
/** @var string */
private $name;
/** @var int */
private $age;
public function __construct(string $name, int $age)
{
$this->name = $name;
$this->age = $age;
}
public function getName(): string
{
return $this->name;
}
public function setName(string $name): void
{
$this->name = $name;
}
public function getAge(): int
{
return $this->age;
}
public function setAge(int $age): void
{
$this->age = $age;
}
}
$recordSerializer = DefaultRecordSerializerFactory::get(
getenv('SCHEMA_REGISTRY_HOST')
);
$avroSchemaJson = '{
"type": "record",
"name": "user",
"fields": [
{"name": "name", "type": "string"},
{"name": "age", "type": "int"}
]
}';
$user = new User('Thomas', 38);
$normalizer = new GetSetMethodNormalizer();
$encoder = new AvroSerDeEncoder($recordSerializer);
$symfonySerializer = new Serializer([$normalizer], [$encoder]);
$serialized = $symfonySerializer->serialize(
$user,
AvroSerDeEncoder::FORMAT_AVRO,
[
AvroSerDeEncoder::CONTEXT_ENCODE_SUBJECT => 'users-value',
AvroSerDeEncoder::CONTEXT_ENCODE_WRITERS_SCHEMA => AvroSchema::parse($avroSchemaJson),
]
);
$deserializedUser = $symfonySerializer->deserialize(
$serialized,
User::class,
AvroSerDeEncoder::FORMAT_AVRO
);
Assert::assertEquals($deserializedUser, $user);Name converter
Sometimes your property names may differ from the names of the fields in your schema. One option to solve this is by using custom Serializer annotations. However, if you're using the annotations provided by this library, you may use our name converter that parses these annotations and maps between the schema field names and the property names.
<?php
use FlixTech\AvroSerializer\Integrations\Symfony\Serializer\AvroSerDeEncoder;
use FlixTech\AvroSerializer\Integrations\Symfony\Serializer\NameConverter\AvroNameConverter;
use FlixTech\AvroSerializer\Objects\DefaultRecordSerializerFactory;
use Symfony\Component\Serializer\Normalizer\GetSetMethodNormalizer;
use Symfony\Component\Serializer\Serializer;
use Doctrine\Common\Annotations\AnnotationReader as DoctrineAnnotationReader;
use Doctrine\Common\Annotations\AnnotationRegistry;
use FlixTech\AvroSerializer\Objects\Schema\Generation\AnnotationReader;
$recordSerializer = DefaultRecordSerializerFactory::get(
getenv('SCHEMA_REGISTRY_HOST')
);
AnnotationRegistry::registerLoader('class_exists');
$reader = new AnnotationReader(
new DoctrineAnnotationReader()
);
$nameConverter = new AvroNameConverter($reader);
$normalizer = new GetSetMethodNormalizer(null, $nameConverter);
$encoder = new AvroSerDeEncoder($recordSerializer);
$symfonySerializer = new Serializer([$normalizer], [$encoder]);Schema builder
This library also provides means of defining schemas using php, very similar to the SchemaBuilder API provided by the Java SDK:
<?php
use FlixTech\AvroSerializer\Objects\Schema;
use FlixTech\AvroSerializer\Objects\Schema\Record\FieldOption;
Schema::record()
->name('object')
->namespace('org.acme')
->doc('A test object')
->aliases(['stdClass', 'array'])
->field('name', Schema::string(), FieldOption::doc('Name of the object'), FieldOption::orderDesc())
->field('answer', Schema::int(), FieldOption::default(42), FieldOption::orderAsc(), FieldOption::aliases('wrong', 'correct'))
->field('ignore', Schema::boolean(), FieldOption::orderIgnore())
->parse();Schema generator
Besides providing a fluent api for defining schemas, we also provide means of generating schema from
class metadata (annotations). For this to work, you have to install the doctrine/annotations package.
<?php
use FlixTech\AvroSerializer\Objects\DefaultSchemaGeneratorFactory;
use FlixTech\AvroSerializer\Objects\Schema\Generation\Annotations as SerDe;
/**
* @SerDe\AvroType("record")
* @SerDe\AvroName("user")
*/
class User
{
/**
* @SerDe\AvroType("string")
* @var string
*/
private $firstName;
/**
* @SerDe\AvroType("string")
* @var string
*/
private $lastName;
/**
* @SerDe\AvroType("int")
* @var int
*/
private $age;
public function __construct(string $firstName, string $lastName, int $age)
{
$this->firstName = $firstName;
$this->lastName = $lastName;
$this->age = $age;
}
public function getFirstName(): string
{
return $this->firstName;
}
public function getLastName(): string
{
return $this->lastName;
}
public function getAge(): int
{
return $this->age;
}
}
$generator = DefaultSchemaGeneratorFactory::get();
$schema = $generator->generate(User::class);
$avroSchema = $schema->parse();Further examples on the possible annotations can be seen in the test case.
Examples
This library provides a few executable examples in the examples folder. You should have a look to get an understanding how this library works.