twisted.spread.jelly : API documentation

S-expression-based persistence of python objects.

It does something very much like Pickle; however, pickle's main goal seems to be efficiency (both in space and time); jelly's main goals are security, human readability, and portability to other environments.

This is how Jelly converts various objects to s-expressions.

Boolean:

   True --> ['boolean', 'true']

Integer:

   1 --> 1

List:

   [1, 2] --> ['list', 1, 2]

String:

   "hello" --> "hello"

Float:

   2.3 --> 2.3

Dictionary:

   {'a': 1, 'b': 'c'} --> ['dictionary', ['b', 'c'], ['a', 1]]

Module:

   UserString --> ['module', 'UserString']

Class:

   UserString.UserString --> ['class', ['module', 'UserString'], 'UserString']

Function:

   string.join --> ['function', 'join', ['module', 'string']]

Instance: s is an instance of UserString.UserString, with a __dict__ {'data': 'hello'}:

   ["UserString.UserString", ['dictionary', ['data', 'hello']]]

Class Method: UserString.UserString.center:

   ['method', 'center', ['None'], ['class', ['module', 'UserString'],
    'UserString']]

Instance Method: s.center, where s is an instance of UserString.UserString:

   ['method', 'center', ['instance', ['reference', 1, ['class',
   ['module', 'UserString'], 'UserString']], ['dictionary', ['data', 'd']]],
   ['dereference', 1]]

The set builtin and the sets.Set class are serialized to the same thing, and unserialized to set if available, else to sets.Set. It means that there's a possibility of type switching in the serialization process. The solution is to always use set if possible, and only use sets.Set under Python 2.3; this can be accomplished by using twisted.python.compat.set.

The same rule applies for frozenset and sets.ImmutableSet.

Author Glyph Lefkowitz

Function	setUnjellyableForClass	Set which local class will represent a remote type.
Function	setUnjellyableFactoryForClass	Set the factory to construct a remote instance of a type:
Function	setUnjellyableForClassTree	Set all classes in a module derived from `baseClass` as copiers for a corresponding remote class.
Function	getInstanceState	Utility method to default to 'normal' state rules in serialization.
Function	setInstanceState	Utility method to default to 'normal' state rules in unserialization.
Class	Unpersistable	This is an instance of a class that comes back when something couldn't be unpersisted.
Class	Jellyable	Inherit from me to Jelly yourself directly with the `getStateFor' convenience method.
Class	Unjellyable	Inherit from me to Unjelly yourself directly with the `setStateFor` convenience method.
Class	InsecureJelly	This exception will be raised when a jelly is deemed `insecure'; e.g. it contains a type, class, or module disallowed by the specified `taster'
Class	DummySecurityOptions	DummySecurityOptions() -> insecure security options Dummy security options -- this class will allow anything.
Class	SecurityOptions	This will by default disallow everything, except for 'none'.
Function	jelly	Serialize to s-expression.
Function	unjelly	Unserialize from s-expression.
Function	_newInstance	Make a new instance of a class without calling its __init__ method. Supports both new- and old-style classes.
Function	_maybeClass	Undocumented
Class	_Jellier	(Internal) This class manages state for a call to jelly()
Class	_Unjellier	No class docstring; 5/27 methods documented
Class	_Dummy	(Internal) Dummy class, used for unserializing instances.
Class	_DummyNewStyle	(Internal) Dummy class, used for unserializing instances of new-style classes.
Function	_newDummyLike	Create a new instance like `instance`.

def setUnjellyableForClass(classname, unjellyable): (source)

Set which local class will represent a remote type.

If you have written a Copyable class that you expect your client to be receiving, write a local "copy" class to represent it, then call:

   jellier.setUnjellyableForClass('module.package.Class', MyCopier).

Call this at the module level immediately after its class definition. MyCopier should be a subclass of RemoteCopy.

The classname may be a special tag returned by 'Copyable.getTypeToCopyFor' rather than an actual classname.

This call is also for cached classes, since there will be no overlap. The rules are the same.

def setUnjellyableFactoryForClass(classname, copyFactory): (source)

Set the factory to construct a remote instance of a type:

 jellier.setUnjellyableFactoryForClass('module.package.Class', MyFactory)

Call this at the module level immediately after its class definition. copyFactory should return an instance or subclass of RemoteCopy.

Similar to setUnjellyableForClass except it uses a factory instead of creating an instance.

def setUnjellyableForClassTree(module, baseClass, prefix=None): (source)

Set all classes in a module derived from baseClass as copiers for a corresponding remote class.

When you have a heirarchy of Copyable (or Cacheable) classes on one side, and a mirror structure of Copied (or RemoteCache) classes on the other, use this to setUnjellyableForClass all your Copieds for the Copyables.

Each copyTag (the "classname" argument to getTypeToCopyFor, and what the Copyable's getTypeToCopyFor returns) is formed from adding a prefix to the Copied's class name. The prefix defaults to module.__name__. If you wish the copy tag to consist of solely the classname, pass the empty string ''.

Parameters	module	a module object from which to pull the Copied classes. (passing sys.modules[__name__] might be useful)
	baseClass	the base class from which all your Copied classes derive.
	prefix	the string prefixed to classnames to form the unjellyableRegistry.

Parameters	state	A `dict` used to update `inst.__dict__` or `_NO_STATE` to skip this part of initialization.
Returns	A new instance of `cls`.

t.s.jelly : module documentation