MessagePack serializer implementation for Python msgpack.org[Python]
Python C C++ Other

README.rst

MessagePack for Python

Build Status Documentation Status

What's this

MessagePack is an efficient binary serialization format. It lets you exchange data among multiple languages like JSON. But it's faster and smaller. This package provides CPython bindings for reading and writing MessagePack data.

Very important notes for existing users

PyPI package name

TL;DR: When upgrading from msgpack-0.4 or earlier, don't do pip install -U msgpack-python. Do pip uninstall msgpack-python; pip install msgpack instead.

Package name on PyPI was changed to msgpack from 0.5. I upload transitional package (msgpack-python 0.5 which depending on msgpack) for smooth transition from msgpack-python to msgpack.

Sadly, this doesn't work for upgrade install. After pip install -U msgpack-python, msgpack is removed and import msgpack fail.

Deprecating encoding option

encoding and unicode_errors options are deprecated.

In case of packer, use UTF-8 always. Storing other than UTF-8 is not recommended.

For backward compatibility, you can use use_bin_type=False and pack bytes object into msgpack raw type.

In case of unpacker, there is new raw option. It is True by default for backward compatibility, but it is changed to False in near future. You can use raw=False instead of encoding='utf-8'.

Planned backward incompatible changes

When msgpack 1.0, I planning these breaking changes:

  • packer and unpacker: Remove encoding and unicode_errors option.
  • packer: Change default of use_bin_type option from False to True.
  • unpacker: Change default of raw option from True to False.
  • unpacker: Reduce all max_xxx_len options for typical usage.
  • unpacker: Remove write_bytes option from all methods.

To avoid these breaking changes breaks your application, please:

  • Don't use deprecated options.
  • Pass use_bin_type and raw options explicitly.
  • If your application handle large (>1MB) data, specify max_xxx_len options too.

Install

$ pip install msgpack

PyPy

msgpack provides a pure Python implementation. PyPy can use this.

Windows

When you can't use a binary distribution, you need to install Visual Studio or Windows SDK on Windows. Without extension, using pure Python implementation on CPython runs slowly.

For Python 2.7, Microsoft Visual C++ Compiler for Python 2.7 is recommended solution.

For Python 3.5, Microsoft Visual Studio 2015 Community Edition or Express Edition can be used to build extension module.

How to use

One-shot pack & unpack

Use packb for packing and unpackb for unpacking. msgpack provides dumps and loads as an alias for compatibility with json and pickle.

pack and dump packs to a file-like object. unpack and load unpacks from a file-like object.

>>> import msgpack
>>> msgpack.packb([1, 2, 3], use_bin_type=True)
'\x93\x01\x02\x03'
>>> msgpack.unpackb(_, raw=False)
[1, 2, 3]

unpack unpacks msgpack's array to Python's list, but can also unpack to tuple:

>>> msgpack.unpackb(b'\x93\x01\x02\x03', use_list=False, raw=False)
(1, 2, 3)

You should always specify the use_list keyword argument for backward compatibility. See performance issues relating to use_list option below.

Read the docstring for other options.

Streaming unpacking

Unpacker is a "streaming unpacker". It unpacks multiple objects from one stream (or from bytes provided through its feed method).

import msgpack
from io import BytesIO

buf = BytesIO()
for i in range(100):
   buf.write(msgpack.packb(range(i), use_bin_type=True))

buf.seek(0)

unpacker = msgpack.Unpacker(buf, raw=False)
for unpacked in unpacker:
    print(unpacked)

Packing/unpacking of custom data type

It is also possible to pack/unpack custom data types. Here is an example for datetime.datetime.

import datetime
import msgpack

useful_dict = {
    "id": 1,
    "created": datetime.datetime.now(),
}

def decode_datetime(obj):
    if b'__datetime__' in obj:
        obj = datetime.datetime.strptime(obj["as_str"], "%Y%m%dT%H:%M:%S.%f")
    return obj

def encode_datetime(obj):
    if isinstance(obj, datetime.datetime):
        return {'__datetime__': True, 'as_str': obj.strftime("%Y%m%dT%H:%M:%S.%f")}
    return obj


packed_dict = msgpack.packb(useful_dict, default=encode_datetime, use_bin_type=True)
this_dict_again = msgpack.unpackb(packed_dict, object_hook=decode_datetime, raw=False)

Unpacker's object_hook callback receives a dict; the object_pairs_hook callback may instead be used to receive a list of key-value pairs.

Extended types

It is also possible to pack/unpack custom data types using the ext type.

>>> import msgpack
>>> import array
>>> def default(obj):
...     if isinstance(obj, array.array) and obj.typecode == 'd':
...         return msgpack.ExtType(42, obj.tostring())
...     raise TypeError("Unknown type: %r" % (obj,))
...
>>> def ext_hook(code, data):
...     if code == 42:
...         a = array.array('d')
...         a.fromstring(data)
...         return a
...     return ExtType(code, data)
...
>>> data = array.array('d', [1.2, 3.4])
>>> packed = msgpack.packb(data, default=default, use_bin_type=True)
>>> unpacked = msgpack.unpackb(packed, ext_hook=ext_hook, raw=False)
>>> data == unpacked
True

Advanced unpacking control

As an alternative to iteration, Unpacker objects provide unpack, skip, read_array_header and read_map_header methods. The former two read an entire message from the stream, respectively de-serialising and returning the result, or ignoring it. The latter two methods return the number of elements in the upcoming container, so that each element in an array, or key-value pair in a map, can be unpacked or skipped individually.

Each of these methods may optionally write the packed data it reads to a callback function:

from io import BytesIO

def distribute(unpacker, get_worker):
    nelems = unpacker.read_map_header()
    for i in range(nelems):
        # Select a worker for the given key
        key = unpacker.unpack()
        worker = get_worker(key)

        # Send the value as a packed message to worker
        bytestream = BytesIO()
        unpacker.skip(bytestream.write)
        worker.send(bytestream.getvalue())

Notes

string and binary type

Early versions of msgpack didn't distinguish string and binary types (like Python 1). The type for representing both string and binary types was named raw.

For backward compatibility reasons, msgpack-python will still default all strings to byte strings, unless you specify the use_bin_type=True option in the packer. If you do so, it will use a non-standard type called bin to serialize byte arrays, and raw becomes to mean str. If you want to distinguish bin and raw in the unpacker, specify raw=False.

Note that Python 2 defaults to byte-arrays over Unicode strings:

>>> import msgpack
>>> msgpack.unpackb(msgpack.packb([b'spam', u'eggs']))
['spam', 'eggs']
>>> msgpack.unpackb(msgpack.packb([b'spam', u'eggs'], use_bin_type=True),
                    raw=False)
['spam', u'eggs']

This is the same code in Python 3 (same behaviour, but Python 3 has a different default):

>>> import msgpack
>>> msgpack.unpackb(msgpack.packb([b'spam', u'eggs']))
[b'spam', b'eggs']
>>> msgpack.unpackb(msgpack.packb([b'spam', u'eggs'], use_bin_type=True),
                    raw=False)
[b'spam', 'eggs']

ext type

To use the ext type, pass msgpack.ExtType object to packer.

>>> import msgpack
>>> packed = msgpack.packb(msgpack.ExtType(42, b'xyzzy'))
>>> msgpack.unpackb(packed)
ExtType(code=42, data='xyzzy')

You can use it with default and ext_hook. See below.

Note about performance

GC

CPython's GC starts when growing allocated object. This means unpacking may cause useless GC. You can use gc.disable() when unpacking large message.

use_list option

List is the default sequence type of Python. But tuple is lighter than list. You can use use_list=False while unpacking when performance is important.

Python's dict can't use list as key and MessagePack allows array for key of mapping. use_list=False allows unpacking such message. Another way to unpacking such object is using object_pairs_hook.

Development

Test

MessagePack uses pytest for testing. Run test with following command:

$ make test