undesired interraction between property and __setattr__

[Alex Martelli]

dman wrote:
        ...
>     def __setattr__( self , name , value ) :
>         if name == "_s" :
>             self.__dict__[name] = value
>         elif name in self.__class__.__dict__ :
>             self.__dict__[ name ] = self.__class__.__dict__[ name ]
>             self.__dict__[ name ].__set__( self , value )
>         else :
>             print name , "=>" , value
>             setattr( self._s , name , value )
        ...
> Is this the proper/ideal solution, or is there something I'm missing,
> or is this a bug in python?

I think a somewhat more natural way to code this __setattr__ might be:

    def __setattr__( self , name , value ) :
        if name == "_s" or hasattr(
            getattr(self.__class__, name, None), '__set__'):
                object.__setattr__(self, name, value)
        else :
            print name , "=>" , value
            setattr( self._s , name , value )

i.e.: if the name is _s, OR is the name of a data attribute defined in the
current class, THEN use the normal way to set the attribute as defined by
class object (which includes property setting or stashing in self.__dict__),
ELSE print something and set the name/value attribute in self._s instead.

I find the nested combination of hasattr and setattr here a bit hard to
follow, though.  My personal preference would be something like:

    def __setattr__( self , name , value ) :
        try: setter = self.__class__.name.__set__
        except AttributeError:
            if name == '_s':
                self.__dict__['_s'] = value
            else:
                print name , "=>" , value
                setattr( self._s , name , value )
        else: setter(self, value)

but it's an issue of personal stylistic preference.


Alex