Stream Codecs

Stream codecs are a serialization tool used to describe how an object should be stored and read from a stream, such as buffers. Stream codecs are primarly used by Vanilla's networking system to sync data.

정보

As stream codecs are roughly analagous to codecs, this page has been formatted in the same way to show the similarities.

Using Stream Codecs

Stream codecs encode and decode objects into some stream using StreamCodec#encode and StreamCodec#decode, respectively. encode takes in the stream and the object to encode into the stream. decode takes in the stream and returns the decoded object. Typically, the stream is either a ByteBuf, FriendlyByteBuf, or RegistryFriendlyByteBuf.

// Let exampleStreamCodec represent a StreamCodec<ExampleJavaObject>
// Let exampleObject be a ExampleJavaObject
// Let buffer be a RegistryFriendlyByteBuf

// Encode Java object into the buffer stream
exampleStreamCodec.encode(buffer, exampleObject);

// Read Java object from buffer stream
ExampleJavaObject obj = exampleStreamCodec.decode(buffer);

노트

Unless you are manually handling the buffer object, you will generally never call encode and decode.

Existing Stream Codecs

ByteBufCodecs

ByteBufCodecs contains static instances of codecs for certain primitives and objects.

Stream Codec	Java Type
BOOL	Boolean
BYTE	Byte
SHORT	Short
INT	Integer
FLOAT	Float
DOUBLE	Double
BYTE_ARRAY	byte[]*
STRING_UTF8	String**
TAG	Tag
COMPOUND_TAG	CompoundTag
VECTOR3F	Vector3f
QUATERNIONF	Quaternionf
GAME_PROFILE	GameProfile

* byte[] can be limited to a certain number of values via ByteBufCodecs#byteArray.

* String can be limited to a certain number of characters via ByteBufCodecs#stringUtf8.

Additionally, there are some static instances that encode and decode primivites and objects using a different method.

Unsigned Shorts

UNSIGNED_SHORT is an alternative of SHORT that is meant to be treated as an unsigned number. As numbers are signed in Java, unsigned shorts are sent and received as Integers with the upper two bytes masked out.

Variable-Sized Number

VAR_INT and VAR_LONG are stream codecs where the value is encoded to be as small as possible. This is done by encoding seven bits at a time, using the upper bit as a marker of whether there is more data for this number. Numbers between 0 and 2^28-1 for integers or 0 and 2^56-1 for longs will be sent shorter or equal to the number of bytes in a integer or long, respectively. If the values of your numbers are normally in this range and generally at the lower end of it, then these variable stream codecs should be used.

노트

VAR_INT is an alternative for INT.

Trusted Tags

TRUSTED_TAG and TRUSTED_COMPOUND_TAG are variants of TAG and COMPOUND_TAG, respectively, that have an unlimited heap to decode the tag to, compared to the 2MiB limit of TAG and COMPOUND_TAG. Trusted tag stream codecs should ideally only be used in clientbound packets, such as what Vanilla does for block entity data packet and entity data serializers.

If a different limit should be used, then a NbtAccounter can be supplied with the given size using ByteBufCodecs#tagCodec or #compoundTagCodec.

Vanilla and NeoForge

Minecraft and NeoForge define many stream codecs for objects that are frequently encoded and decoded. Some examples include ResourceLocation#STREAM_CODEC for ResourceLocations or NeoForgeStreamCodecs#CHUNK_POS for ChunkPoss.

Most of the stream codecs can be found within the object class itself or within StreamCodec, ByteBufCodecs, or NeoForgeStreamCodecs.

Creating Stream Codecs

Stream codecs can be created for reading or writing any object to a stream. This documentation will focus on the stream as a buffer as that is its primary purpose.

Stream codecs have two generics: B representing the buffer and V representing the object value. B is generally one of three types: ByteBuf, FriendlyByteBuf, RegistryFriendlyByteBuf, each extending one another. FriendlyByteBuf adds Minecraft-specific read and write methods while RegistryFriendlyByteBuf provides access to the list of registries and its objects.

When constructing a stream codec, B should be the least-specific buffer type. For example, a ResourceLocation is sent as a string. As strings are supported by a regular ByteBuf, its type should be StreamCodec<ByteBuf, ResourceLocation>. FriendlyByteBuf contains methods for writing a ChunkPos, so its type should be StreamCodec<FriendlyByteBuf, ChunkPos>. An Item needs access to the registry, so its type should be StreamCodec<RegistryFriendlyByteBuf, Item>.

Most methods that take in a stream codec look for ? super B for the buffer type, meaning that all three of the above examples can be used if the buffer type is a RegistryFriendlyByteBuf.

Member Encoders

StreamMemberEncoder is an alternative to StreamEncoder where the encoding object comes first and the buffer second. This is typically used when the encoding object contains an instance method to write the object to the buffer. A StreamMemberEncoder can be used to create the StreamCodec by calling StreamCodec#ofMember.

// Some object to create a stream codec for
public class ExampleObject {
    
    // The normal constructor
    public ExampleObject(String arg1, int arg2, boolean arg3) { /* ... */ }

    // The stream decoder reference
    public ExampleObject(ByteBuf buffer) { /* ... */ }

    // The stream encoder reference
    public void encode(ByteBuf buffer) { /* ... */ }
}

// What the stream codec would look like
public static StreamCodec<ByteBuf, ExampleObject> =
    StreamCodec.ofMember(ExampleObject::encode, ExampleObject::new);

Composites

Stream codecs can read and write objects via StreamCodec#composite. Each composite stream codec defines a list of stream codecs and getters which are read/written in the order they are provided. composite has overloads up to six parameters.

Every two parameters in a composite represents the stream codec used to read/write the field and a getter to get the field to encode from the object. The final parameter is a function to create a new instance of the object when decoding.

// Objects to create a stream codec for
public record SimpleExample(String arg1, int arg2, boolean arg3) {}
public record RegistryExample(double arg1, Holder<Item> arg2) {}

// The stream codecs
public static final StreamCodec<ByteBuf, SimpleExample> SIMPLE_STREAM_CODEC =
    StreamCodec.composite(
        // Stream codec and getter pair
        ByteBufCodecs.STRING_UTF8, SimpleExample::arg1,
        ByteBufCodecs.VAR_INT, SimpleExample::arg2,
        ByteBufCodecs.BOOL, SimpleExample::arg3,
        SimpleExample::new
    );

// Since this has a holder, a RegistryFriendlyByteBuf is used
public static final StreamCodec<RegistryFriendlyByteBuf, RegistryExample> REGISTRY_STREAM_CODEC =
    StreamCodec.composite(
        // Note that ByteBuf stream codecs can be used here
        ByteBufCodecs.DOUBLE, RegistryExample::arg1,
        ByteBufCodecs.holderRegistry(Registries.ITEM), RegistryExample::arg2,
        RegistryExample::new
    );

Transformers

Stream codecs can be transformed into equivalent, or partially equivalent, representations using mapping methods. Two mapping methods apply to the value while one mapping method applies to the buffer.

The map method transforms the value using two functions: one to transform the current type into the new type, and one to transform the new type back into the current type. This is analagous to codec transformers.

public static final StreamCodec<ByteBuf, ResourceLocation> STREAM_CODEC = 
    ByteBufCodecs.STRING_UTF8.map(
        // String -> ResourceLocation
        ResourceLocation::new,
        // ResourceLocation -> String
        ResourceLocation::toString
    );

The apply method transforms the value using a StreamCodec.CodecOperation. A StreamCodec.CodecOperation takes in a stream codec of the current type and returns a stream codec of the new type. These typically wrap around map or take in helper methods.

public static final StreamCodec<ByteBuf, List<ResourceLocation>> STREAM_CODEC =
    ResourceLocation.STREAM_CODEC.apply(ByteBufCodecs.list());

The mapStream method transforms the buffer using a function that takes in the new buffer type and returns the current buffer type. This method should rarely be used as most methods with stream codecs do not need to change the type of the buffer.

public static final StreamCodec<RegistryFriendlyByteBuf, Integer> STREAM_CODEC =
    ByteBufCodecs.VAR_INT.mapStream(buffer -> (ByteBuf) buffer);

Unit

A stream codec which supplies an in-code value and encodes to nothing can be represented using StreamCodec#unit. This is useful if no information should be synced across the network.

warning

Unit stream codecs expect that any encoded object must match the unit specified; otherwise an error will be thrown. Therefore, all objects must have some equals implementation that returns true for the unit object, or that the instance provided to the stream codec is always provided when encoding.

public static final StreamCodec<ByteBuf, Item> UNIT_STREAM_CODEC =
    StreamCodec.unit(Items.AIR);

Lazy Initialized

Sometimes, a stream codec may rely on data that is not present when it is constructed. In these situations NeoForgeStreamCodecs#lazy can be used for a stream codec to construct itself on first read/write. The method takes in a supplied stream codec.

public static final StreamCodec<ByteBuf, Item> LAZY_STREAM_CODEC = 
    NeoForgeStreamCodecs.lazy(
        () -> StreamCodec.unit(Items.AIR)
    );

Collections

A stream codec for collections can be generated from a object stream codec via collection. collection takes in an IntFunction that constructs the empty collection, a stream codec of the object, and an optional maximum size.

public static final StreamCodec<ByteBuf, Set<BlockPos>> COLLECTION_STREAM_CODEC =
    ByteBufCodecs.collection(
        HashSet::new, // Constructs a set with the specified capacity
        BlockPos.STREAM_CODEC,
        256 // The set can only have up to 256 elements
    );

Another overload of collection can be specified with StreamCodec#apply.

public static final StreamCodec<ByteBuf, Set<BlockPos>> COLLECTION_STREAM_CODEC =
    BlockPos.STREAM_CODEC.apply(
        ByteBufCodecs.collection(HashSet::new)
    );

List-based collections also can be specified through StreamCodec#apply by calling ByteBufCodecs#list with an optional maximum size.

public static final StreamCodec<ByteBuf, List<BlockPos>> LIST_STREAM_CODEC =
    BlockPos.STREAM_CODEC.apply(
        // The list can only have up to 256 elements
        ByteBufCodecs.list(256)
    );

Map

A stream codec for a map of key and value objects can be generated using two stream codecs via ByteBufCodecs#map. The function also takes in an IntFunction that constructs the empty map and an optional maximum size.

public static final StreamCodec<ByteBuf, Map<String, BlockPos>> MAP_STREAM_CODEC =
    ByteBufCodecs.map(
        HashMap::new, // Constructs a map with the specified capacity
        ByteBufCodecs.STRING_UTF8,
        BlockPos.STREAM_CODEC,
        256 // The map can only have up to 256 elements
    );

Either

A stream codec for two different methods of reading/writing some object data can be generated from two steram codecs via ByteBufCodecs#either. This method first reads/writes a boolean indicating whether to read/write the first or second stream codec, respectively.

public static final StreamCodec<ByteBuf, Either<Integer, String>> EITHER_STREAM_CODEC = 
    ByteBufCodecs.either(
        ByteBufCodecs.VAR_INT,
        ByteBufCodecs.STRING_UTF8
    );

Id Mapper

In most cases, when sending information across the network where an object is present on both sides, an integer representing an id is sent. Ids representing an object reduce the amount of information that need to be synced across the network. Both enums and registries make use of this.

ByteBufCodecs#idMapper provides a convenient way to send ids for objects. It either takes in two functions which convert an object to int and vice versa, or an IdMap.

// For some enum
public enum ExampleIdObject {
    ;

    // Gets Id -> Enum
    public static final IntFunction<ExampleIdObject> BY_ID = 
        ByIdMap.continuous(
            ExampleIdObject::getId,
            ExampleIdObject.values(),
            ByIdMap.OutOfBoundsStrategy.ZERO
    );
    
    ExampleIdObject(int id) { /* ... */ }
}

// The stream codec would look like
public static final StreamCodec<ByteBuf, ExampleIdObject> ID_STREAM_CODEC =
    ByteBufCodecs.idMapper(ExampleIdObject.BY_ID, ExampleIdObject::getId);

노트

NeoForge provides an alternative for id mappers that does not cache the enum values on construction via IExtensibleEnum#createStreamCodecForExtensibleEnum. However, this rarely needs to be used outside of extensible enums.

Optional

A stream codec for sending an Optional wrapped value can be generated by supplying a stream codec to ByteBufCodecs#optional. This method first reads/writes a boolean indicating whether to read/write the object.

public static final StreamCodec<RegistryFriendlyByteBuf, Optional<DataComponentType<?>>> OPTIONAL_STREAM_CODEC =
    DataComponentType.STREAM_CODEC.apply(ByteBufCodecs::optional);

Registry Objects

Registry objects can be sent across the network using one of three methods: registry, holderRegistry, or holder. Each takes in a ResourceKey representing the registry the registry object is in.

warning

Custom registries must be syncable by calling RegistryBuilder#sync and setting the value to true. Otherwise, the encoder will throw an exception.

registry and holderRegistry returns the registry object or a holder wrapped registry object, respectively. These methods send over an id representing the registry object.

// Registry object
public static final StreamCodec<RegistryFriendlyByteBuf, Item> VALUE_STREAM_CODEC =
    BytebufCodecs.registry(Registries.ITEM);

// Holder of registry object
public static final StreamCodec<RegistryFriendlyByteBuf, Holder<Item>> HOLDER_STREAM_CODEC =
    BytebufCodecs.holderRegistry(Registries.ITEM);

holder returns a holder wrapped registry object. This method sends over an id representing the registry object, or the registry object itself if the provided Holder is a direct reference. To do so, holder also takes in the stream codec of the registry object.

public static final StreamCodec<RegistryFriendlyByteBuf, Holder<SoundEvent>> STREAM_CODEC =
    ByteBufCodecs.holder(
        Registries.SOUND_EVENT, SoundEvent.DIRECT_STREAM_CODEC
    );

노트

holder will only throw an exception for a non-synced custom registry if the holder is not direct.

Holder Sets

Tags or sets of holder wrapped registry objects can be sent using holderSet. This takes in a ResourceKey representing the registry the registry objects are in.

public static final StreamCodec<RegistryFriendlyByteBuf, HolderSet<Item>> HOLDER_SET_STREAM_CODEC =
    BytebufCodecs.holderSet(Registries.ITEM);

Recursive

Sometimes, an object may reference an object of the same type as a field. For example, MobEffectInstance takes in an optional MobEffectInstance if there is a hidden effect. In this case, StreamCodec#recursive can be used to supply the stream codec as part of a function to create the stream codec.

// Define our recursive object
public record RecursiveObject(Optional<RecursiveObject> inner) { /* ... */ }

public static final StreamCodec<ByteBuf, RecursiveObject> RECURSIVE_CODEC = StreamCodec.recursive(
    recursedStreamCodec -> StreamCodec.composite(
        recursedStreamCodec.apply(ByteBufCodecs::optional),
        RecursiveObject::inner,
        RecursiveObject::new
    )
);

Dispatch

Stream codecs can have sub-stream codecs that can decode a particular object based on some specified type via StreamCodec#dispatch. This is typically used with registry objects that represent a type, like ParticleType for ParticleOptions or StatType for Stats.

A dispatch stream codec first attempts to read/write the type object. From there, the current object is read/written using one of the functions provided in the method. The first Function takes in the current object and gets the type to write the value. The second Function takes in the type object and gets the StreamCodec for the current object to read the value.

// Define our object(s)
public abstract class ExampleObject {

    // Define the method used to specify the object type for encoding
    public abstract StreamCodec<? super RegistryFriendlyByteBuf, ? extends ExampleObject> streamCodec();
}

// Assume there is a ResourceKey<StreamCodec< super RegistryFriendlyByteBuf, ? extends ExampleObject>> DISPATCH
public static final StreamCodec<RegistryFriendlyByteBuf, ExampleObject> DISPATCH_STREAM_CODEC =
    ByteBufCodecs.registry(DISPATCH).dispatch(
        // Get the stream codec from the specific object
        ExampleObject::streamCodec,
        // Get the stream codec from the registry object
        Function.identity()
    )