object data member dumper?

Bruno Desthuilliers onurb at xiludom.gro
Wed Nov 8 12:24:32 CET 2006

tom arnall wrote:
> Bruno Desthuilliers wrote:
>> tom arnall a écrit :
>>> does anyone know of a utility to do a recursive dump of object data
>>> members?
>> What are "object data members" ? (hint: in Python, everything is an
>> object - even functions and methods).
>> What is your real use case ?
> something like:
>    class A:

Make it :
     class A(object):

>       def __init__(self, p1):
>          self.p1 = p1
>    class B: 
>       def __init__(self,p1, p2):
>          self.a = A(p1)
>          self.p2 = p2
>          self.v1 = '3'
>    class C:
>       def __init__(self):
>          self.b = B(3,4)
>          self.p3 = 5
>    class D:
>       def __init__(self):
>          self.v2=2
>          self.o1 = C()
>          self.o2 = B(11,12)
>    d = D()
>    objectDataDumper(d)
> would produce something like:
>    object of class D with:
>    o1(C)->b(B)->a(A)->p1=3
>    o1(C)->b(B)->p2=4
>    o1(C)->b(B)->v1=3
>    o1(C)->p3=5
>    o2(B)->a(A)->p1=11
>    o2(B)->p2=12
>    o2(B)->v1=3
>    v2=2

Ok, so this is for debugging purpose ?

A small question : how should this behave for:
- _implementation attributes
- __magic__ attributes
- class attributes
- callable attributes
- properties and other descriptors  ?

In fact, the problem is that in Python, everything's an object, and an
object is mostly a composed namespace (ie a set of name-object mappings)
subject to some lookup rules. FWIW, the class of an object is an
attribute (__class__) referencing the class objet; superclasses are
class objects referenced by the __bases__ and __mro__ attributes of the
class object; method objects are usually attributes of the class object,
but it's possible to add/replace methods on a per-instance basis;
properties are class attributes that most of the time works on some
instance attribute. etc, etc, etc... Even the code of a function is an
attribute of the function object. And as an icing on top of the cake,
there's the __getattr__ magic method...

With all this in mind, writing a generic python object inspector is not
that trivial : either you'll end up with way too much informations or
you'll arbitrary skip informations that would be useful in a given
context or you'll need to pass so much args to the inspector that it'll
become too heavy for a quick, intuitive use... Well, that's at least my
own experience.

OTOH, "live" inspection of an object (either in the Python shell or in a
pdb session) works fine. The pprint and inspect modules may help too.

My 2 cents...
bruno desthuilliers
python -c "print '@'.join(['.'.join([w[::-1] for w in p.split('.')]) for
p in 'onurb at xiludom.gro'.split('@')])"

More information about the Python-list mailing list