Serialisation Support (pytyp.s11n.base)

Pytyp can encode from Python classes to dicts. It can also decode dicts back into Python classes.

The functions and classes here form the basis for pytyp.s11n.json and pytyp.s11n.yaml - you may find the examples here useful in understanding how the process works, but probably want to call the routines in those packages.

Encoding Support

To encode data, pytyp looks at the constructor arguments. For each argument it assumes that the class has an attribute or property that provides a value.

So, for example, this class can be encoded:

>>> class EncExample():
...     def __init__(self, a, b=None):
...         self.a = a
...         self.b = b
>>> encode = Encoder()
>>> encode(EncExample(1, 2))
{'a': 1, 'b': 2}

but this class cannot:

>>> class BadEncExample():
...     def __init__(self, q):
...         self.z = q
>>> encode(BadEncExample(1))
Traceback (most recent call last):
AttributeError: 'BadEncExample' object has no attribute 'q'

If you do not want your objects to be mutable you can expose the same information through read-only properties:

>>> class ReadOnly():
...     def __init__(self, value):
...         self._value = value
...     @property
...     def value(self):
...         return self._value
>>> encode(ReadOnly(1))
{'value': 1}
class pytyp.s11n.base.Encoder(recurse=True, strict=True, check_circular=True)[source]

An instance of this class can be called to encode data:

>>> class MyClass:
...    def __init__(self, arg1, arg2):
...        self.arg1 = arg1
...        self.arg2 = arg2
>>> myInstance = MyClass(42, 'foo')
>>> encode = Encoder()
>>> # use pprint to sort dicts for consistency
>>> pprint(encode([1,myInstance,{'a':2}]))
[1, {'arg1': 42, 'arg2': 'foo'}, {'a': 2}]
  • recurse – Should included values also be encoded? This depends on the requirements of the calling code (JSON and YAML differ).
  • strict – If true, raise an error if “special” attributes (corresponding to *args etc) are missing.
  • check_circular – If true, detect and abort on encoding circular data structures.

Decoding Support

To decode data, pytyp looks at the type specification and constructs the class by calling the constructor. The specification can contain lists, tuples and dictionaries, but must have the same form as the input.

For example, here decode() is called with a type specification for a list of DecExample() instances:

>>> class DecExample():
...     def __init__(self, a):
...         self.a = a
...     def __repr__(self):
...         return '<DecExample({0})>'.format(self.a)
>>> decode([{'a': 1}, {'a': 2}], [DecExample])
[<DecExample(1)>, <DecExample(2)>]

To handle nested types the constructor of the container class must have a type declaration (another type specification):

>>> class Container():
...     def __init__(self, ex:DecExample):
...         self.ex = ex
...     def __repr__(self):
...         return '<Container({0})>'.format(self.ex)
>>> decode({'ex': {'a': 1}}, Container)

Note the type declaration in the constructor above. Without that declaration pytyp will incorrectly interpret the data:

>>> class BadContainer():
...     def __init__(self, ex):
...         self.ex = ex
...     def __repr__(self):
...         return '<BadContainer({0})>'.format(self.ex)
>>> decode({'ex': {'a': 1}}, BadContainer)
<BadContainer({'a': 1})>

In type specifications, lists must be of a single type, but tuples and dicts have a specific type for each member:

>>> decode(({'ex': {'a': 1}}, {'a': 2}), (Container, DecExample))
(<Container(<DecExample(1)>)>, <DecExample(2)>)

A value of None can be matched by an optional type:

>>> from pytyp.spec.abcs import Opt
>>> decode((None, {'a': 2}), (Opt(Container), DecExample))
(None, <DecExample(2)>)
pytyp.s11n.base.decode(data, spec)[source]

Rewrite the given data so that it conforms to the type specification.

Table Of Contents

Previous topic

Serialisation (pytyp.s11n)

Next topic

JSON Serialisation (pytyp.s11n.json)

This Page