Data Representation?

[Peter Otten]

Kris Caselden wrote:

> Say I have some data:
> 
>>>> a=[1]
>>>> b=[2]
>>>> link=[a,b]
> 
> The simplest why to write this to a file represents it as
> 
>>>> print str(link)
> [[1], [2]]
> 
> Unfortunately, if this is read back in via execfile(), the whole
> dynamic nature of changing 'link' by changing 'a' and 'b' is lost. Is
> there any way to write data so that the list name is written instead
> of the list's values? Essentially, '[[a], [b]]' instead of '[[1],
> [2]]'? Any help most appreciated.

I'm not sure what you mean by [[a], [b]]. If it is sufficcient that 
    
    link[0] is a and link[1] is b

i. e. they refer to the same list when restored, I suggest that you
introduce a namespace class and make a, b and link attributes. You can then
use the standard pickle procedure:

import pickle

class Namespace:
    def __str__(self):
        return str(self.__dict__)

ns = Namespace()
ns.a = [1]
ns.b = [2]
ns.link = [ns.a, ns.b]


pickle.dump(ns, file("tmp", "w"))

ns2 = pickle.load(file("tmp", "r"))

assert ns2.a is ns2.link[0]
assert ns2.b is ns2.link[1]

print ns2

ns2.b.append(3)
print ns2

Peter