python's OOP question

[neoedmund]

Re: python's OOP question



On Oct 16, 9:01 pm, Bruno Desthuilliers <o...@xiludom.g rowrote:

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...@gmai l.comwrites:

>

>>>I'm trying to achieve a higher level of "reusabilit y". 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 "inheritanc e" 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

the main idea is to use the magic '__getattr__(se lf, 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 "inheritanc e" 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 ?
>
--
bruno desthuilliers
python -c "print '@'.join(['.'.join([w[::-1] for w in p.split('.')]) for
p in 'on...@xiludom. gro'.split('@')])"

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 "inheritanc e", 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]

Re: python's OOP question

neoedmund wrote:
(snip)

So I can reuse a method freely only if it's worth reusing.
For the word "inheritanc e", 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...).

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

[neoedmund]

Re: python's OOP question


Bruno Desthuilliers wrote:

neoedmund wrote:
(snip)

So I can reuse a method freely only if it's worth reusing.
For the word "inheritanc e", 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...).
>
--
bruno desthuilliers
python -c "print '@'.join(['.'.join([w[::-1] for w in p.split('.')]) for
p in 'onurb@xiludom. gro'.split('@')])"

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]

Re: python's OOP question


Bruno Desthuilliers wrote:

neoedmund wrote:
(snip)

So I can reuse a method freely only if it's worth reusing.
For the word "inheritanc e", 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...).
>
--
bruno desthuilliers
python -c "print '@'.join(['.'.join([w[::-1] for w in p.split('.')]) for
p in 'onurb@xiludom. gro'.split('@')])"

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]

Re: python's OOP question

neoedmund 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"

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]

Re: python's OOP question

"neoedmund" <neoedmund@gmai l.comwrites:

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 "preventati ve 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]

Re: python's OOP question

On 2006-10-18, neoedmund <neoedmund@gmai l.comwrote:

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]

Re: python's OOP question

Neil Cerutti wrote:

On 2006-10-18, neoedmund <neoedmund@gmai l.comwrote:

>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]

Re: python's OOP question

Neil Cerutti wrote:

On 2006-10-18, neoedmund <neoedmund@gmai l.comwrote:

>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