p23serialize

Library to prepare objects to be easier to serialize/deserialize with standard formats e.g. json/msgpack.

The main goals are:

• Be able to serialize circular objects. (E.g. serialize a list that also contains itself nested somewhere)
• Be able to serialize arbitrary numpy objects so that their representation is the same on python 2 and python 3.
• Objects must serialize / deserialize to the same underlying structure in both python 2 and 3. (compatable format)

It's currently work in progress. Have a look at the tests to see what it can currently do.

License

MIT license - see LICENSE.txt

README TODO'S:

• Explain caveats with dictionary keys (what limitations json has)

• Simple examples of some typical objects after they are serialized and how to read the format

• Bytes/unicode pains. (e.g. in python 2 'aa' == u'aa' BUT in python 3 'aa' != b'aa'

• Deserialization: object creation vs finalization (or, "what is needed for circular deserialization")

• Other libraries out there - and why I decided to write my own instead. (hint: unicode vs str and python 2 vs 3 issues)

• Notes about json behaviour with python 2, 3, bytes and unicode (e.g. json.dumps(['aa', u'aa']) vs json.dumps(['aa', b'aa']) and associated json.loads stuff)

• Explain different language limitations when it comes to dictionary keys.

  (e.g. Python dictionary keys must be hashable. Javascript is even more strict - only strings are allowed as keys)

REPO TODO's

• Add requirements.txt for pip
• How to structure so that it packages nicely? (for pip installs / pypi listing)