Avro schema generator for PHP

Installation

composer require php-kafka/php-avro-schema-generator "^3.0"

Note: For PHP7.x please use the 2.x version.

Description

This library enables you to:

• Manage your embedded schema as separate files
• The library is able to merge those files
• The library is able to generate avsc schema templates from PHP classes

Merging subschemas / schemas

Schema template directories: directories containing avsc template files (with subschema)
Output directory: output directory for the merged schema files

Console example

./vendor/bin/avro-cli avro:subschema:merge ./example/schemaTemplates ./example/schema

PHP example

<?php
use PhpKafka\PhpAvroSchemaGenerator\Registry\SchemaRegistry;
use PhpKafka\PhpAvroSchemaGenerator\Merger\SchemaMerger;
$registry = (new SchemaRegistry())
    ->addSchemaTemplateDirectory('./schemaTemplates')
    ->load();
$merger = new SchemaMerger('./schema');
$merger->setSchemaRegistry($registry);
$merger->merge();

Merge optimizers

There are optimizers that you can enable for merging schema:

• FullNameOptimizer: removes unneeded namespaces
• FieldOrderOptimizer: the first fields of a record schema will be: type, name, namespace (if present)
• PrimitiveSchemaOptimizer: Optimizes primitive schema e.g. {"type": "string"} to "string"

How to enable optimizer:

Console example

./bin/avro-cli  --optimizeFullNames --optimizeFieldOrder --optimizePrimitiveSchemas avro:subschema:merge ./example/schemaTemplates ./example/schema

PHP Example

<?php
use PhpKafka\PhpAvroSchemaGenerator\Registry\SchemaRegistry;
use PhpKafka\PhpAvroSchemaGenerator\Merger\SchemaMerger;
use PhpKafka\PhpAvroSchemaGenerator\Optimizer\FieldOrderOptimizer;
use PhpKafka\PhpAvroSchemaGenerator\Optimizer\FullNameOptimizer;
use PhpKafka\PhpAvroSchemaGenerator\Optimizer\PrimitiveSchemaOptimizer;
$registry = (new SchemaRegistry())
    ->addSchemaTemplateDirectory('./schemaTemplates')
    ->load();
$merger = new SchemaMerger('./schema');
$merger->setSchemaRegistry($registry);
$merger->addOptimizer(new FieldOrderOptimizer());
$merger->addOptimizer(new FullNameOptimizer());
$merger->addOptimizer(new PrimitiveSchemaOptimizer());
$merger->merge();

Generating schemas from classes

You will need to adjust the generated templates, but it gives you a good starting point to work with.
Class directories: Directories containing the classes you want to generate schemas from
Output directory: output directory for your generated schema templates
After you have reviewed and adjusted your templates you will need to merge them (see above)

Console example

./bin/avro-cli  avro:schema:generate ./example/classes ./example/schemaTemplates

PHP Example

<?php
use PhpKafka\PhpAvroSchemaGenerator\Converter\PhpClassConverter;
use PhpKafka\PhpAvroSchemaGenerator\Parser\ClassParser;
use PhpKafka\PhpAvroSchemaGenerator\Parser\DocCommentParser;
use PhpKafka\PhpAvroSchemaGenerator\Registry\ClassRegistry;
use PhpKafka\PhpAvroSchemaGenerator\Parser\ClassPropertyParser;
use PhpKafka\PhpAvroSchemaGenerator\Generator\SchemaGenerator;
use PhpParser\ParserFactory;
$parser = (new ParserFactory())->create(ParserFactory::PREFER_PHP7);
$classPropertyParser = new ClassPropertyParser(new DocCommentParser());
$classParser = new ClassParser($parser, $classPropertyParser);
$converter = new PhpClassConverter($classParser);
$registry = (new ClassRegistry($converter))->addClassDirectory('./classes')->load();
$generator = new SchemaGenerator('./schema');
$generator->setClassRegistry($registry);
$schemas = $generator->generate();
$generator->exportSchemas($schemas);

The generator is able to detect types from:

• doc comments
• property types
• doc annotations
  • @avro-type to set a fixed type instead of calculating one
  • @avro-default set a default for this property in your schema
  • @avro-doc to set schema doc comment
  • @avro-logical-type set logical type for your property (decimal is not yet supported, since it has additional parameters)

Disclaimer

In v1.3.0 the option --optimizeSubSchemaNamespaces was added. It was not working fully
in the 1.x version and we had some discussions (#13) about it.
Ultimately the decision was to adapt this behaviour fully in v2.0.0 so you might want to
upgrade if you rely on that behaviour.