Ceres Serialization Library

Ceres is a relatively simple serialization library. The reason it exists is that there are some requirements to behaviour that I didn't really find a way of handling with other serialization logic such as Java's built-in serialization or Gson.

Features

Separation of object deconstruction and serialized data output. This means the same serializer logic can be used in combination with various serialization formats. The only built-in format is plain binary data using DataOutputStreams and DataInputStreams producing and consuming ByteBuffers, respectively. This is implemented and made available in the BinarySerialization class.
Deserialization into existing objects. This means it is possible to "load" some persisted state into an object that was re-created outside the deserialization logic. This is useful in cases where the deserializer would otherwise need mechanisms to access context for constructing certain objects.
Automatic serializer generation. Types can be annotated to have a serializer generated for them automatically. By default, these serializers will be realized as generated classes for maximum performance.

Limitations

Only supports creating objects in deserialization if the type has a default constructor.
Aliasing is not supported. Each occurrence of a reference to some object will be serialized separately. Consequently, cycles are not supported.
Polymorphism support requires an explicit serializer for the supertype to be serialized and deserialized.
Can only deserialize into existing values of final fields. I.e. deserialization will not replace the object assigned to a final field.

Design

The core principle of Ceres is to separate serialization into a deconstruction step, and a data storage step.

Serializers take care of the deconstruction step, i.e. they simply deconstruct complex data types into primitive data types, which can then be written and read through the SerializationVisitor and DeserializationVisitor interfaces. These visitors' implementations take care of writing and reading primitive data types to and from a backing storage.

Finally, I like to use the term serialization backend for the collection of utility methods that can be used for correctly invoking a registered serializer for serializing or deserializing a value using some a pair of visitors. An example for this is the BinarySerialization class.

One concession is that implementations may choose to ignore the names passed for serialized and deserialized values. As such, serializers should, where possible ensure writing and reading happens in the same order.

Usage

To serialize an object the recommended way is to go through a serialization backend. See the built-in BinarySerialization for reference. Alternatively the serializer for a type can be obtained and invoked directly.

public class Example {
    @Serialized
    public static class MySerializableType {
        public int a;
        public int b;
    }

    // The above is equivalent to marking all fields directly like so:
    public static class MySerializableType {
        @Serialized public int a;
        @Serialized public int b;
    }

    public static void main(String[] args) {
        // Through the BinarySerialization backend:

        ByteBuffer serialized = BinarySerialization.serialize(new MySerializableType());

        // Deserialize into a new object. This requires Example to have a default constructor.
        MySerializableType deserialized = BinarySerialization.deserialize(serialized, MySerializableType.class);

        // Deserialize into an existing object, if possible.
        // This will use the serializer for deserialized.getClass().
        // As such, deserialized must not be null.
        BinarySerialization.deserialize(serialized, deserialized);

        // Alternatively, specify the type explicitly. Here deserialized may be null.
        BinarySerialization.deserialize(serialized, MySerializableType.class, deserialized);

        // If there's a chance we cannot deserialize into the specified object,
        // e.g. for arrays if the length may differ, use the returned value.
        deserialized = BinarySerialization.deserialize(serialized, MySerializableType.class, deserialized);


        // Using serializers directly:

        // Create the part that will actually write the data to some output format (e.g. MySerializedDataType).
        MySerializer serializer = new MySerializer();
        Ceres.getSerializer(MySerializableType.class)
                .serialize(serializer, MySerializableType.class, new MySerializableType());

        MyDeserializer deserializer = new MyDeserializer(serializer.getSerializedData());
        deserialized = Ceres.getSerializer(MySerializableType.class)
                .deserialize(deserializer, MySerializableType.class, deserialized);
    }

    private static class MySerializedDataType {
        // Something that stores the serialized data, e.g. binary, json, ...
    }

    private static class MySerializer implements SerializationVisitor {
        public MySerializedDataType getSerializedData();

        // Implementation ...
    }

    private static class MyDeserializer implements DeserializationVisitor {
        public DeserializationVisitor(MySerializedDataType serializedData) {
            // Store access to serialized data for reading via the implemented interface.
        }

        // Implementation ...
    }
}

Maven

Ceres can be included into a project via the Github Package Repository. See the documentation for more information on how to set that up. In short, you'll want to add your username and a public access token into your ~/.gradle/gradle.properties and use those variables in your repository declaration. Note that the public access token will need read:packages permissions.

For example, using Gradle:

repositories {
  maven {
    url = uri("https://maven.pkg.github.com/fnuecke/ceres")
    credentials {
      username = project.findProperty("gpr.user") ?: System.getenv("USERNAME")
      password = project.findProperty("gpr.key") ?: System.getenv("TOKEN")
    }
  }
}

dependencies {
  implementation 'li.cil.ceres:ceres:0.0.4'
}