python's OOP question
neoedmund
"expected aaa". how to make it work?
[code]
class C1(object):
def v(self, o):
return "expected "+o
class C2(object):
def v(self, o):
return "unexpected "+o
def m(self):
print self.v("aaa")
class C3(object):
def nothing(self):
pass
def test1():
o = C3()
setattr(o,"m",C2().m)
setattr(o,"v",C1().v)
o.m()
test1()
[/code]
Kay Schluehr
class C3(C1, C2):pass
>>> C3.mro() # shows method resolution order
[<class '__main__.C3'>, <class '__main__.C1'>, <class '__main__.C2'>,
<type 'object'>]
>>> o = C3()
>>> o.m()
expected aaa
Ben Finney
Setting attributes on an object externally isn't the same thing as
making bound methods of that object.
In this case, 'o.m' is a bound method of a C2 instance, and has no
knowledge of C1. 'o.v' is a bound method of a C1 instance, and has no
knowledge of C2. Neither of them has any knowledge of C3.
What is it you're trying to achieve?
--
\ "Unix is an operating system, OS/2 is half an operating system, |
`\ Windows is a shell, and DOS is a boot partition virus." -- |
_o__) Peter H. Coffin |
Ben Finney
neoedmund
But i need a "dynamic" way. Say i have a existing class, and add some
method from other class into it.
neoedmund
be done in python? Can anybody help me？
Ben Finney
"neoedmund" <neoe...@gmail.com> writes:
> I'm trying to achieve a higher level of "reusability". Maybe it
> cannot be done in python? Can anybody help me？
What, specifically, are you trying to achieve? What problem needs
solving?
--
\ "If you're a horse, and someone gets on you, and falls off, and |
`\ then gets right back on you, I think you should buck him off |
_o__) right away." -- Jack Handey |
Ben Finney
neoedmund
but "inheritance" means you must inherite all methods from super type.
now i just need "some" methods from one type and "some" methods from
other types,
to build the new type.
Do you think this way is more flexible than tranditional inheritance?
Gregor Horvath
> python use multiple inheritance.
> but "inheritance" means you must inherite all methods from super type.
> now i just need "some" methods from one type and "some" methods from
> other types,
> to build the new type.
> Do you think this way is more flexible than tranditional inheritance?
>
Probably your problem is better solved with delegation instead of
inheritance.
--
Servus, Gregor
George Sakkis
> python use multiple inheritance.
> but "inheritance" means you must inherite all methods from super type.
> now i just need "some" methods from one type and "some" methods from
> other types,
> to build the new type.
> Do you think this way is more flexible than tranditional inheritance?
The following does the trick:
from types import MethodType
def addMethod(meth, obj):
f = meth.im_func
setattr(obj, f.__name__, MethodType(f,obj))
def test1():
addMethod(C2.m, C3)
addMethod(C1.v, C3)
o = C3()
o.m()
The same works as is on modifying individual instances, rather than
their class:
def test2():
o = C3()
addMethod(C2.m, o)
addMethod(C1.v, o)
o.m()
# raises AttributeError
# C3().m()
George
neoedmund
def MixIn(pyClass, mixInClass):
if mixInClass not in pyClass.__bases__:
pyClass.__bases__ += (mixInClass,)
def test1():
o = C3()
MixIn(C3,C1)
MixIn(C3,C2)
o.m()
"expected aaa"
neoedmund
Thank you George.
Bruno Desthuilliers
(*PLEASE* stop top-posting - corrected)
>
> Ben Finney wrote:
>> [Please don't top-post above the text to which you're replying.]
>>
>> "neoedmund" <neoe...@gmail.com> writes:
>>
>>> I'm trying to achieve a higher level of "reusability". Maybe it
>>> cannot be done in python? Can anybody help me？
>> What, specifically, are you trying to achieve? What problem needs
>> solving?
> python use multiple inheritance.
> but "inheritance" means you must inherite all methods from super type.
> now i just need "some" methods from one type and "some" methods from
> other types, to build the new type.
> Do you think this way is more flexible than tranditional inheritance?
While dynamically adding attributes (and methods - which are attributes
too) is not a problem, I'd second Gregor's anwser : it might be better
to use composition/delegation here.
--
bruno desthuilliers
python -c "print '@'.join(['.'.join([w[::-1] for w in p.split('.')]) for
p in 'on...@xiludom.gro'.split('@')])"
neoedmund
Bruno Desthuilliers wrote:
> neoedmund wrote:
> (*PLEASE* stop top-posting - corrected)
> >
> > Ben Finney wrote:
> >> [Please don't top-post above the text to which you're replying.]
> >>
> >> "neoedmund" <neoe...@gmail.com> writes:
> >>
> >>> I'm trying to achieve a higher level of "reusability". Maybe it
> >>> cannot be done in python? Can anybody help me？
> >> What, specifically, are you trying to achieve? What problem needs
> >> solving?
>
> > python use multiple inheritance.
> > but "inheritance" means you must inherite all methods from super type.
> > now i just need "some" methods from one type and "some" methods from
> > other types, to build the new type.
> > Do you think this way is more flexible than tranditional inheritance?
>
> While dynamically adding attributes (and methods - which are attributes
> too) is not a problem, I'd second Gregor's anwser : it might be better
> to use composition/delegation here.
>
Could you show some code to help me know how composition/delegation can
be done here? Thanks.
Kay Schluehr
> Could you show some code to help me know how composition/delegation can
> be done here? Thanks.
Starting with your example C2 might just derive from C1 and perform a
supercall:
class C1(object):
def v(self, o):
return "expected "+o
class C2(C1):
def v(self, o):
return "unexpected "+o
def m(self):
print super(C2,self).v("aaa")
>>> c2 = C2()
>>> c2.m()
expected aaa
But in general there is no single pattern to deal with object
composition.
Bruno Desthuilliers
About composition/delegation, there's no "one-size-fits-all" answer, but
the main idea is to use the magic '__getattr__(self, name)' method.
Now back to your specific case : after a better reading of your original
question, straight composition/delegation wouldn't work here - at least
not without modifications to both C1 and C2 (sorry, should have read
better the first time).
Given the context (ie : "create a new type with methods from type X and
methods from type Y"), a very simple solution could be:
class C3(object):
m = C2.m.im_func
v = C1.v.im_func
FWIW, if you have full control over C1, C2 and C3, you could also just
'externalize' the functions definitions:
def v1(self, o):
return "expected "+o
def v2(self, o):
return "unexpected "+o
def m2(self):
""" requires that 'self' has a v(self, somestring) method """
print self.v("aaa")
class C1(object):
v = v1
class C2(object):
v = v2
m = m2
class C3(object):
v = v1
m = m2
The problem (with the whole approach, whatever the choosen technical
solution) is that if one of theses methods depends on another one (or on
any other attribute) that is not defined in your new class, you're in
trouble. This is not such a big deal in the above example, but might
become much more brittle in real life.
Now we can look at the problem from a different perspective. You wrote:
"""
but "inheritance" means you must inherite all methods from super type.
now i just need "some" methods from one type and "some" methods from
other types, to build the new type.
"""
What is your problem with having the other extra methods too ?
neoedmund
On Oct 16, 9:01 pm, Bruno Desthuilliers <o...@xiludom.gro> wrote:
> neoedmund wrote:
> > Bruno Desthuilliers wrote:
> >> neoedmund wrote:
> >> (*PLEASE* stop top-posting - corrected)
> >>> Ben Finney wrote:
> >>>> [Please don't top-post above the text to which you're replying.]
>
> >>>> "neoedmund" <neoedm...@gmail.com> writes:
>
> >>>>> I'm trying to achieve a higher level of "reusability". Maybe it
> >>>>> cannot be done in python? Can anybody help me？
> >>>> What, specifically, are you trying to achieve? What problem needs
> >>>> solving?
> >>> python use multiple inheritance.
> >>> but "inheritance" means you must inherite all methods from super type.
> >>> now i just need "some" methods from one type and "some" methods from
> >>> other types, to build the new type.
> >>> Do you think this way is more flexible than tranditional inheritance?
>
> >> While dynamically adding attributes (and methods - which are attributes
> >> too) is not a problem, I'd second Gregor's anwser : it might be better
> >> to use composition/delegation here.
>
> > Could you show some code to help me know how composition/delegation can
> > be done here? Thanks.About composition/delegation, there's no "one-size-fits-all" answer, but
Bruno , your 2 great and clear samples revealed what method is in
python, which is also I really want to ask. thank you.
So I can reuse a method freely only if it's worth reusing.
For the word "inheritance", in some aspect, meanings reuse the super
class, with the condition: must reuse everything from super class.
It's lack of a option to select which methods are to be reused.
this is something should be improved for general OOP i think.
So to answer " What is your problem with having the other extra methods
too ?",
in real life, a class is not defined so well that any method is needed
by sub-class. so contains a method never to be used is meaningless and
i think something should forbidden.
also, any handy methods in a class can be grabbed out for our reuse. so
we can forgotting the bundering inheritance tree and order.
Bruno Desthuilliers
(snip)
> So I can reuse a method freely only if it's worth reusing.
> For the word "inheritance", in some aspect, meanings reuse the super
> class, with the condition: must reuse everything from super class.
Not really. In fact, inheritance *is* a special case of
composition/delegation. A 'child' class is a class that has references
to other classes - it's 'parents' -, and then attributes that are not
found in the instance or child class are looked up in the parents
(according to mro rules in case of multiple inheritance). And that's all
there is.
> It's lack of a option to select which methods are to be reused.
Methods not redefined in the 'child' class or it's instance are
'reusable'. Now they are only effectively 'reused' if and when called by
client code. So the 'option to select which methods are to be reused' is
mostly up to both the 'child' class and code using it.
> this is something should be improved for general OOP i think.
> So to answer " What is your problem with having the other extra methods
> too ?",
> in real life, a class is not defined so well that any method is needed
> by sub-class.
Then perhaps is it time to refactor. A class should be a highly cohesive
unit. If you find yourself needing only a specific subset of a class, it
may be time to extract this subset in it's own class. Given Python's
support for both multiple inheritance and composition/delegation, it's
usually a trivial task (unless you already mixed up too many orthogonal
concerns in your base class...).
neoedmund
ivestgating the web, i found something similiar with my approch:
http://en.wikipedia.org/wiki/Duck_typing
"Duck-typing avoids tests using type() or isinstance(). Instead, it
typically employs hasattr() tests"
neoedmund
I donnot agree with your "it's time to refactory" very much, man has
probly never has time to do such things. suppose a system is working
soundly, you maybe has no time or motivation to do refactory instead of
having a vocation to a island. it's easy to say, at lease myself has
not the experience to do such things, :-)
Fredrik Lundh
> ivestgating the web, i found something similiar with my approch:
> http://en.wikipedia.org/wiki/Duck_typing
> "Duck-typing avoids tests using type() or isinstance(). Instead, it
> typically employs hasattr() tests"
that's not entirely correct, though: in Python, duck-typing typically
uses "Easier to Ask Forgiveness than Permission" (EAFP), aka "Just Do
It", rather than "Look Before You Leap" (LBYL).
</F>
Ben Finney
> Bruno Desthuilliers wrote:
> > neoedmund wrote:
> > > in real life, a class is not defined so well that any method is
> > > needed by sub-class.
> >
> > Then perhaps is it time to refactor. A class should be a highly
> > cohesive unit. If you find yourself needing only a specific subset
> > of a class, it may be time to extract this subset in it's own
> > class.
>
> I donnot agree with your "it's time to refactory" very much, man has
> probly never has time to do such things.
I respectfully suggest that the *reason* you find yourself with little
time to refactor is probably related to the fact that you *need* to
refactor. If your code is crufty and poorly-designed, it is costing
you every time you need to maintain it.
Would it help if we called it "preventative maintenance"?
--
\ "I bought a dog the other day. I named him Stay. It's fun to |
`\ call him. 'Come here, Stay! Come here, Stay!' He went insane. |
_o__) Now he just ignores me and keeps typing." -- Steven Wright |
Ben Finney
Neil Cerutti
> ivestgating the web, i found something similiar with my approch:
> http://en.wikipedia.org/wiki/Duck_typing
> "Duck-typing avoids tests using type() or isinstance(). Instead, it
> typically employs hasattr() tests"
It's pity it didn't get called quack typing. One ckecks if
some unknown noun can quack, not if a duck can do something
unknown.
--
Neil Cerutti
Peter Otten
> On 2006-10-18, neoedmund <neoe...@gmail.com> wrote:
>> ivestgating the web, i found something similiar with my approch:
>> http://en.wikipedia.org/wiki/Duck_typing
>> "Duck-typing avoids tests using type() or isinstance(). Instead, it
>> typically employs hasattr() tests"
>
> It's pity it didn't get called quack typing.
That's because the quacks recommend static typing.
Peter
Peter Otten
> On 2006-10-18, neoedmund <neoe...@gmail.com> wrote:
>> ivestgating the web, i found something similiar with my approch:
>> http://en.wikipedia.org/wiki/Duck_typing
>> "Duck-typing avoids tests using type() or isinstance(). Instead, it
>> typically employs hasattr() tests"
>
> It's pity it didn't get called quack typing.
That's because the quacks prescribe static typing.
Peter