RE: new minor getAttr feature

Lenard Lindstrom wrote:

> In Python the getattr function has matching setattr and 
> delattr functions for setting and deleting attributes respectively.

Yes, but I thought attrs_ would suffice for setAttr() and delAttr()
unless you want to handle the prototype chain cases.  Then you should be
explicit and use getAttr() first.

> The non-super version of getAttr is simply obj.attrs_[name] .
> 
> getAttr should through a NameError if the attribute is not 
> found since None is a valid attribute value.

Yes, I will change that.  That was stupid of me.

> And I guess getAttr, like attrs_, will be a truly special 
> method, uneffected by wild-card property methods. I had not 
> noticed it before but the following does not cause infinite recursion:
> 
> O>> object Wild:
> ...    def get_(attr):
> ...       print("Getting", attr)
> ...       return self.attrs_[attr]
> ...    d = 8
> ...    e = 9
> ...
> 9
> O>> Wild.d
> Getting $d
> 8
> O>> Wild.e
> Getting $e
> 9
> 
> I guess there is no way around these magic attributes since 
> there are no classes, which bypass the property mechanism. 
> Alternatively, one could have functions.
> 
> getAttr(obj, name)
> getAttrs(obj)        # Return object attribute list; replaces attrs_
> setAttr(obj, name, value)
> delAttr(obj, name)
> 
> Nothing magical about these. But I would have these do normal 
> attribute lookup, which means honouring properties.
> 
> To bypass properties I would have special getAttrIgnoreProp, 
> setAttrIgnoreProp and delAttrIgnoreProp functions. They have 
> verbose names but they are for special use only.

Do you really need all these functions?  How about:

     value = getAttr(name)
     (value, cont) = getAttr(name, cont=True)

	tuple[1].name = newValue # replaces setAttr()
      del tuple[1].name        # replaces delAttr()

Then we would have two functions total:

	getAttr(name, cont=False, super=None)	
      getAttrNoProp(name, cont=False, super=None)