deepcopy alternative?

[none]

I have a very complex data structure which is basically a class object
containing (sometimes many) other class objects, function references,
ints, floats, etc. The man for the copy module states pretty clearly
that it will not copy methods or functions. I've looked around for a
while (prob just using the wrong keywords) and haven't found a good
solution. As a workaround I've been using cPickle, loads(dumps(obj ))
which is incredibly slow (~8 sec for a 1000 elem list).

I believe the only thing stopping me from doing a deepcopy is the
function references, but I'm not sure. If so is there any way to
transform a string into a function reference(w/o eval or exec)?

Example
from copy import deepcopy
a = ClassObj([ClassObj([3.2, 'str', funcRef]), 4, ClassObj[funcRef])
b = deepcopy(a)
TypeError: function() takes at least 2 arguments (0 given)

All I want is a deepcopy of the list. Any suggestions would be appreciated.

Jack Trades

PS: If the answer to this is in __getstate__() or __setstate__() is
there some reference to how these should be implemented besides the
library reference?

[Szabolcs Nagy]

Re: deepcopy alternative?

I believe the only thing stopping me from doing a deepcopy is the

function references, but I'm not sure. If so is there any way to
transform a string into a function reference(w/o eval or exec)?

what's your python version?
for me deepcopy(lambda :1) does not work in py2.4 but it works in py2.5
(in py2.4 i tried to override __deepcopy__ but it had no effect)

[Gabriel Genellina]

Re: deepcopy alternative?

En Mon, 29 Jan 2007 20:17:11 -0300, <"none <"@bag.python.o rgescribió:

I have a very complex data structure which is basically a class object
containing (sometimes many) other class objects,

What are "class objects"? instances of a given class, or a class itself?

function references,
ints, floats, etc. The man for the copy module states pretty clearly
that it will not copy methods or functions. I've looked around for a
while (prob just using the wrong keywords) and haven't found a good
solution. As a workaround I've been using cPickle, loads(dumps(obj ))
which is incredibly slow (~8 sec for a 1000 elem list).
>
I believe the only thing stopping me from doing a deepcopy is the
function references, but I'm not sure. If so is there any way to
transform a string into a function reference(w/o eval or exec)?

What do you mean by "function reference"? A module-level function works
fine both with pickle and copy. Methods do not; but it's not common to
take methods out of a class and put them inside a tuple, by example...

>
Example
from copy import deepcopy
a = ClassObj([ClassObj([3.2, 'str', funcRef]), 4, ClassObj[funcRef])
b = deepcopy(a)
TypeError: function() takes at least 2 arguments (0 given)

Please provide a *running* example. I've added some definitions, corrected
some syntax errors, and this full example works for me:

--- cut ---
from copy import deepcopy

class ClassObj(object ):
def __init__(self, args):
self.data = args

def funcRef(x):
print x

a = ClassObj([ClassObj([3.2, 'str', funcRef]), 4, ClassObj([funcRef])])
b = deepcopy(a)
print a is b
a.data.append(' hello')
print len(a.data), len(b.data)
--- cut ---

PS: If the answer to this is in __getstate__() or __setstate__() is
there some reference to how these should be implemented besides the
library reference?

Until you provide more info on your problem it's hard to tell.


--
Gabriel Genellina

[none]

Re: deepcopy alternative?

Szabolcs Nagy wrote:

>I believe the only thing stopping me from doing a deepcopy is the
>function references, but I'm not sure. If so is there any way to
>transform a string into a function reference(w/o eval or exec)?

>
what's your python version?
for me deepcopy(lambda :1) does not work in py2.4 but it works in py2.5
(in py2.4 i tried to override __deepcopy__ but it had no effect)
>

Thanks that fixed the problem real quick :)

Jack Trades