cerial 0.0.1

Python3 serializer with memoryview support

Fast, simple, and more pythonic binary serializer library.

Cerial combines Python 3 features, Cython, and memory views to create a fast and easy library for packing and unpacking binary structures.

Removes a lot of the boilerplate that sometimes ends up surrounding struct when used for serialization.

Example

Here’s a quick example of how to use cerial to handle a UDP packet. First we define the structure:

>>> UDP = Struct('UDP',
...    ('sport', 'uint16'),
...    ('dport', 'uint16'),
...    ('len', 'uint16'),
...    ('csum', 'uint16'),
...    endianness='!'
...)

Unlike the builtin struct library, cerial needs the fields to be bound to a name. This is because we expose the binary structure with dictionary interface instead of as a tuple of values.

Now that we have a structure defined, we can use to parse data:

>>> data = bytearray(b'\x08R\x02\x02\x00w\xa9k')
>>> hdr = UDP.parse(data)
>>> hdr
<UDP b'\x08R\x02\x02\x00w\xa9k'>

We can now inspect the binary structure:

>>> hdr['csum']
43371
>>> dict(hdr)
{'csum': 43371, 'len': 119, 'sport': 2130, 'dport': 514}
>>> bytes(hdr)
b'\x08R\x02\x02\x00w\xa9k'

And we can also update individual fields, which will directly write to the underlying memory view, no rendering step needed.

>>> hdr['csum'] = 0xffff
>>> data
bytearray(b'\x08R\x02\x02\x00w\xff\xff')
>>> bytes(hdr)
b'\x08R\x02\x02\x00w\xff\xff'

TODO

• Feature parity with the various primitives the struct library supports

• First class support for bit fields

• Improve performance

• Support readonly memoryviews (e.g. support bytes). An upstream Cython issue.

Performance

Currently slightly slower than the struct module, but as far as I can tell, significantly faster than other similar libraries like pypacker or scrapy at parsing packets.