Serialization

When serializing ObjectProperty, a set of bit flags may be employed to change serialization behavior.

Bit	Name	Description
0	UseFlags	Indicates that the set of flags should be used during data processing.
1	CompactPrefix	Enables the use of smaller data types for length prefixes whenever possible.
2	StringEnum	Indicates that enums should be represented as strings.
3	ZLibCompress	Enables data compression using the Zlibrary. See Compression.
4	Encode	Indicates that all properties with bitflag of 8 must be included.

Figure A.I - A table of all possible bit flags for an object serializer.

Serialization Modes

ObjectProperty serializers should anticipate two separate types of serialization: verbose, and compact.

In either case, a PropertyClass buffer will always prefix itself with its property hash.

Compact

In passage through the DML, a compact serializer is used. In compact mode, property hashes and their length are skipped during serialization. Properties are listed in order as they are written in reflection.

WARNING

If the field is another PropertyClass and it is null, an empty uint32_t will be serialized in place of the missing object.

It should also be noted that any property with the deprecated flag is still serialized in this mode.

Verbose

In all other cases, a verbose serializer is employed. This is an exhaustive serialization method that allows for out of order properties.

In this mode, property hashes and their exact binary size will be serialized.

Compression

A serialized buffer may be compressed in one of two ways. The code that reads the buffer decides which one applies, not the buffer itself, so the two are not interchangeable.

Serializer Compression

When the ZLibCompress flag is set, the serializer reads and writes a single bit ahead of the object, stating whether the buffer is compressed. Everything after that bit starts on the next byte boundary.

Field	Type	Description
Flags	uint32_t	Only present when UseFlags is set. Replaces the flags the serializer was created with.
Compressed	1 bit	Set when the rest of the buffer is compressed.
Size	uint32_t	The uncompressed size. Only present when the buffer is compressed.
Data	bytes	A zlib stream when the buffer is compressed. Otherwise, the object itself.

Figure A.II - The layout of a buffer serialized with ZLibCompress.

The mail list in MSG_MAIL_DATA is read this way.

Standalone Compression

Some messages compress a buffer outside of the serializer. The receiver inflates the field first, then deserializes the result without the ZLibCompress flag. There is no leading bit.

Field	Type	Description
Size	uint32_t	The uncompressed size. When the highest bit is set, the data is stored uncompressed and the size is the remaining 31 bits.
Data	bytes	A zlib stream, or the raw buffer when the highest bit of the size is set.

Figure A.III - The layout of a standalone compressed buffer.

The character object in MSG_LOGINCOMPLETE is read this way, as are badge data and the character editing messages.

WARNING

Sending one layout where the other is expected does not fail cleanly. A standalone buffer read with ZLibCompress can report an unknown type hash of 0x78000000: the reader takes the lowest bit of the size as the compressed bit, then reads the hash one byte in, across the rest of the size and the zlib header.

In Imcodec, serializer compression is SerializerFlags.Compress. Standalone compression is Compression.CompressWithLength and Compression.DecompressWithLength, applied to a buffer that was serialized without the flag.