[Python-Dev] Re: [PEP 224] Attribute Docstrings

[Alex Martelli]

"M.-A. Lemburg" <mal at lemburg.com> wrote in message
news:39AAA306.2CBD5383 at lemburg.com...
    [snip]
> This will give you:
>
> A.__doc__x__ == " name of the database; override in subclasses ! "
> A.__doc__y__ == " run in auto-commit ? "
> D.__doc__x__ == """ name of the attached database; note that this must
support
>         transactions
>     """
> D.__doc__y__ == " run in auto-commit ? "
>
> There's no way you are going to achieve this using dictionaries.

But there must be... after all, if I ask for D.y, I get the same as A.y
by inheritance, don't I?  Yet, it IS "achieved using dictionaries" --
plus, I guess, a tiny amount of "magic" (not particularly black) in
the lookup process (if y is not found in D.__dict__, we walk depth
first among the base classes, etc).

So, if it works for getattr (through the __dict__ and an appropriate
walk among bases), why shouldn't it work just as well for a
hypothetical getdoc...?  Indeed, a "dictionary with inheritance from
other dictionaries" would seem to be a pretty natural building block
(conceptually identical to the lookup process for attributes, though
the latter may use its own shortcuts for speed); D.__docs__ would
just be one such.

E.g., if read-only, something like:

class compositeMapping:
    def __init__(self, *bases, **ownvalues):
        self.bases=bases
        self.data=ownvalues
    def __getattr__(self,key):
        try:
            return self.data[key]
        except KeyError:
            for base in bases:
                try:
                    return base[key]
                except KeyError:
                    pass
            raise KeyError, key

Isn't this "achieved using dictionaries"?  And couldn't it be just
what's wanted for SomeClass.__docs__...?


Alex