Pickling Python Expressions

My last post introduced the concept of X, a class which “absorbs” operations and behaves like a function.
As many people pointed out, this was merely a syntactic alternative to lambda. You may like it, you may not.
Now, after a rewrite, X can now be pickled. But let me explain first.

Python lambdas cannot be pickled. In fact, python code cannot be pickled.
Pickling an object, aka serializing, is converting the object’s state (that is, its data) to a string, which can at a later time be unpickled to re-create the object with that state. The unpickling process instanciates that class, assuming it has not changed, and updating the new instance’s state to the pickled one. Python code is never stored.

Trying to pickle a class or a function might appear to work, but it does not really pickle it; it simply pickles the reference to it (and its state). Unpickling the string in a new terminal would prove that (as would a quick analysis of resulting string).

Attempts to pickle methods, nested functions or lambdas fail on the spot. That is because a reference to them cannot be kept (Actually, it can be. But they are rather volatile, so it might not be wise). Eventually, python code, or even expressions, cannot be pickled.

This brings me back to X. X allows you to do just that:

>>> expr = 1 + (X + 3) * 4
>>> s = pickle.dumps(expr)

(destory objects, change objects, switch an interpreter, whatever you wish)

>>> expr2 = pickle.loads(s)
>>> expr2(5)
33

By using X, the programmer can blend dynamic code with his data, and still be able to pickle it.
I believe this removes a very big limitation.

Just to be fair, I will note that there is another way to achieve this: Keep your expressions in a string, and eval it when you need it run. I highly recommend not doing it.

X’s new source code (a bit cryptic, but it’s the best I could do. Suggestions for simplification are welcomed) :

"""
x.py

Author: Erez Sh.
Date  : 11/11/2008
"""

import operator

def identity(x):
	return x

class _Return(object):
	"Pickle-able!"
	def __init__(self, value):
		self._value = value

	def __call__(self, *args):
		return self._value

class _Partial(object):
	"Pickle-able!"
	def __init__(self, callable, *args):
		self._callable = callable
		self._args = args

	def __call__(self, *args, **kwargs):
		args = self._args + args
		return self._callable(*args)

class _X(object):
	def __init__(self, func, *args_to_run):
		self.__func = func
		self.__args_to_run = tuple(args_to_run)

	def __getstate__(self):
		return self.__func, self.__args_to_run
	def __setstate__(self, state):
		self.__func, self.__args_to_run = state
	def __reduce__(self):
		#raise Exception("Deprecated!")
		return object.__reduce__(self)

	def __apply_un_func(self, func ):
		return _X(func, _Partial(self))
	def __apply_bin_func(self, func, arg ):
		return _X(func, _Partial(self), _Return(arg))
	def __apply_rbin_func(self, func, arg ):
		return _X(func, _Return(arg), _Partial(self))
	def __apply_multargs_func(self, func, *args ):
		return _X(func, _Partial(self), *map(_Return,args))

	def __call__(self, arg):
		return self.__func(*[x(arg) for x in self.__args_to_run])

	def __getattr__(self, attr):
		return self.__apply_bin_func( getattr, attr )

	def call(self, *args, **kwargs):
		return self.__apply_multargs_func( apply, args, kwargs)

	# Containers
	def __getitem__(self, other):
		return self.__apply_bin_func( operator.getitem, other )
	def __getslice__(self, a,b=None,c=None):
		return self.__apply_bin_func( operator.getslice, other )
	def in_(self, other):
		return self.__apply_bin_func( operator.contains, other )

	# Arith
	def __add__(self, other):
		return self.__apply_bin_func( operator.add, other )
	def __sub__(self, other):
		return self.__apply_bin_func( operator.sub, other )
	def __mul__(self, other):
		return self.__apply_bin_func( operator.mul, other )
	def __div__(self, other):
		return self.__apply_bin_func( operator.div, other )
	def __floordiv__(self, other):
		return self.__apply_bin_func( operator.floordiv, other )
	def __truediv__(self, other):
		return self.__apply_bin_func( operator.truediv, other )
	def __mod__(self, other):
		return self.__apply_bin_func( operator.mod, other )
	def __pow__(self, other):
		return self.__apply_bin_func( operator.pow, other )

	def __radd__(self, other):
		return self.__apply_rbin_func( operator.add, other )
	def __rsub__(self, other):
		return self.__apply_rbin_func( operator.sub, other )
	def __rmul__(self, other):
		return self.__apply_rbin_func( operator.mul, other )
	def __rdiv__(self, other):
		return self.__apply_rbin_func( operator.div, other )
	def __rfloordiv__(self, other):
		return self.__apply_rbin_func( operator.floordiv, other )
	def __rtruediv__(self, other):
		return self.__apply_rbin_func( operator.truediv, other )
	def __rmod__(self, other):
		return self.__apply_rbin_func( operator.mod, other )
	def __rpow__(self, other):
		return self.__apply_rbin_func( operator.pow, other )

	# bitwise
	def __and__(self, other):
		return self.__apply_bin_func( operator.and_, other )
	def __or__(self, other):
		return self.__apply_bin_func( operator.or_, other )
	def __xor__(self, other):
		return self.__apply_bin_func( operator.xor, other )

	def __rand__(self, other):
		return self.__apply_rbin_func( operator.and_, other )
	def __ror__(self, other):
		return self.__apply_rbin_func( operator.or_, other )
	def __rxor__(self, other):
		return self.__apply_rbin_func( operator.xor, other )
	
	def __rshift__(self, other):
		return self.__apply_bin_func( operator.rshift, other )
	def __lshift__(self, other):
		return self.__apply_bin_func( operator.lshift, other )

	# Comparison
	def __lt__(self, other):
		return self.__apply_bin_func( operator.lt, other )
	def __le__(self, other):
		return self.__apply_bin_func( operator.le, other )
	def __eq__(self, other):
		return self.__apply_bin_func( operator.eq, other )
	def __ne__(self, other):
		return self.__apply_bin_func( operator.ne, other )
	def __ge__(self, other):
		return self.__apply_bin_func( operator.ge, other )
	def __gt__(self, other):
		return self.__apply_bin_func( operator.gt, other )

	def __abs__(self):
		return self.__apply_un_func( abs )
	def __neg__(self):
		return self.__apply_un_func( operator.neg )
	

X = _X(identity, identity)
Advertisements

2 Responses to Pickling Python Expressions

  1. lorg says:

    I think your X is advancing from being “just nice” to really useful.
    I’ll have it in mind when I code, I’ll tell you if I use it somewhere.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: