Path: csiph.com!v102.xanadu-bbs.net!xanadu-bbs.net!eternal-september.org!feeder.eternal-september.org!border1.nntp.ams1.giganews.com!nntp.giganews.com!bcyclone05.am1.xlned.com!bcyclone05.am1.xlned.com!newsfeed.xs4all.nl!newsfeed2a.news.xs4all.nl!xs4all!newsgate.cistron.nl!newsgate.news.xs4all.nl!post.news.xs4all.nl!not-for-mail
To: python-list@python.org
From: Peter Otten <__peter__@web.de>
Subject: Re: Passing new fields to an object
Date: Sat, 13 Jun 2015 09:49:34 +0200
Organization: None
References: <mlev91$4ji$1@speranza.aioe.org> <557b868c$0$11125$c3e8da3@news.astraweb.com>
Mime-Version: 1.0
Content-Type: text/plain; charset="ISO-8859-1"
Content-Transfer-Encoding: 7Bit
User-Agent: KNode/4.13.3
Precedence: list
Newsgroups: comp.lang.python
Message-ID: <mailman.453.1434181789.13271.python-list@python.org>
Lines: 84
NNTP-Posting-Host: 2001:888:2000:d::a6
Xref: csiph.com comp.lang.python:92585

Steven D'Aprano wrote:

> On Fri, 12 Jun 2015 16:53:08 +0100, Paulo da Silva wrote:
> 
>> I would like to do something like this:
>> 
>> class C:
>> def __init__(self,**parms):
>> ...
>> 
>> c=C(f1=1,f2=None)
>> 
>> I want to have, for the object
>> self.f1=1
>> self.f2=None
>> 
>> for an arbitrary number of parameters.
>> 
>> What is the best way to achieve this?
> 
> 
> Others have suggested that you update the instance dict with the keyword
> parameters:
> 
>     self.__dict__.update(parms)
> 
> 
> But I suggest that you should be very careful with this technique,
> because it can lead to surprising problems and bugs in your code. If your
> class has methods (and what sort of class doesn't have methods?), this
> will override them and lead to mysterious failures in your code:
> 
> instance = C(a=1, b=2, method=3)
> # much later
> result = instance.method(args)  # will raise exception
> 
> 
> You should use SimpleNamespace, as Peter suggests, but *not* subclass it.
> If you subclass it and add methods:
> 
> class C(SimpleNamespace):
>     def foo(self, arg):
>         print("called foo")
> 
> 
> then you risk overriding foo method, as above. If you don't add methods,
> there is no need to subclass.

I sometimes do it anyway if only to get a meaningful class name.

> Instead, use composition: your class should *contain* a SimpleNamespace,
> not *be* one:
> 
> class C:
>     def __init__(self, **param):
>         self.ns = SimpleNamespace(param)

**param

>     def __getattr__(self, attrname):
>         return getattr(self.ns, attrname)
>     def foo(self, arg):
>         print("called foo")
> 
> 
> instance = C(a=1, b=2, foo=3)
> # later
> instance.foo("x")  # prints "called foo"

>>> c = C(ns=42)
>>> assert c.ns == 42
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
AssertionError

Yes, it's less likely, but now that the OP has two convenient ways to 
produce a name clash I think it's time to repeat that there is one bullet-
proof approach: use a dict ;)

> The special method __getattr__ only runs if the attribute name is not
> found in the usual way, so the method foo will continue to be found and
> not be overridden by the param foo.