Streaming API implementation of MessagePack binary serialization format - msgpack.

The basics

Packer and Unpacker classes provide Streaming API for serializing and deserializing data.
Unpacker also provide automatic unpacking Map and List by implicitly unpacking internal items.

Streaming API Example - Simple

For the simplest data packaging to bytes packet like this:

-------------------------------------------   
| version | userId | broadcast? | message |
-------------------------------------------

Packer part

import 'package:messagepack/messagepack.dart';

final p = Packer();
p.packInt(1); // packet format version
p.packInt(222); // user id
p.packBool(false); // broadcast message to others
p.packString('hi'); // user message text
final bytes = p.takeBytes(); //Uint8List
yourFunctionSendToServer(bytes); //sends [1, 204, 222, 195, 162, 104, 105]

1 encodes to 1 byte with value 1.
222 encodes to 2 bytes with values [204, 222] because 222 > 127.
true encodes to 1 byte with value 195.
'hi' encodes to 3 bytes with first byte containing str length info and other 2 bytes hold symbols values.

For more information, refer to the msgpack documentation.

Streaming packing process not like the way we usually pack data in json, by specifying keys.
The current example is more like how data is packed in TCP/IP frame structure.

Unpacker part

final List<int> rawBytes = yourFunctionReceiveFromServer(); // receive List<int> bytes from server
final u = Unpacker.fromList(rawBytes);
final version = u.unpackInt();
final userId = u.unpackInt();
final broadcast = u.unpackBool();
final message = u.unpackString();

// check values in test
expect(version, equals(1));
expect(userId, equals(222));
expect(broadcast, equals(true));
expect(message, equals('hi'));

Streaming packing and automatic/implicit unpacking example - complex

final p = Packer()
  ..packListLength(10)              //pack 10 different types items to list
  ..packInt(99)
  ..packBool(true)
  ..packString('hi')
  ..packNull()                      // explicitly pack null
  ..packString(null)                // implicitly any type can accept null 
  ..packBinary(<int>[104, 105])     // hi codes
  ..packListLength(2)               // pack 2 elements list ['elem1',3.14]
  ..packString('elem1')             // list[0]
  ..packDouble(3.14)                // list[1]
  ..packString('continue to pack other elements')
  ..packMapLength(1)                // map {'key1':false}
  ..packString('key1')              // map key
  ..packBool(false)                 // map value 
  ..packInt(9223372036854775807);   // next root list item (map ended)

final bytes = p.takeBytes();
final u = Unpacker(bytes);
// Unpack by the same sequential/streaming way 
// or implicitly/automatically
final l = u.unpackList(); //List<Object>
print(l);
// [99, true, hi, null, null, [104, 105], [elem1, 3.14], continue to pack other elements, {key1: false}, 9223372036854775807]

List<Object> items explicitly casted to corresponding types when using.

NOTES

Packing Maps and Lists

• firstly, pack Map or List header length
• secondly, manually pack all items - that's all

Only need put length header before packing items

After packing all items no need to stop or finish or end this map / list

final list = ['i1','i2'];
final map = {'k1': 11, 'k2': 22};
final p = Packer();
p.packListLength(list.length);
list.forEach(p.packString);
p.packMapLength(map.length);
map.forEach((key, v) {
  p.packString(key);
  p.packInt(v);
});
final bytes = p.takeBytes();

More examples

More examples can be found in:

• test/messagepack_test.dart
• example/example.dart

Don't use Packer after calling .takeBytes()

Internally it cleans up underlying _builder, so further behaviour will be useless.
Call .takeBytes() only once. It returns Uint8List of packed bytes.
After that don't continue to use Packer instance (don't call .packXXX() methods and .takeBytes() anymore).
Instead create new Packer instance.

Motivation for creating this package

No other packages available that give streaming API for processing data

Contributing

If you have advice how to improve library code or making it lighter or blazingly faster - don't hesitate to open an issue or pull request!