[Python-Dev] Non-string keys in namespace dicts

[Jim Jewett]

PJE wrote:
> Isn't the simplest way to cache attribute
> lookups to just have a cache dictionary in the type,
> and update that  dictionary whenever a change is
> made to a superclass?  That's  essentially how
> __slotted__ attribute changes on base classes
> work now, isn't it?

Neil Toronto wrote:

> The nice thing about caching pointers to dict
> entries is that they don't change as often as
> values do.

Is this really true for namespaces?

I was thinking that the typical namespace usage is a bunch of inserts
(possibly with lookups mixed in), followed by never changing it again
until it is deallocated.

> There are fewer ways to invalidate an
> entry pointer: inserting set, resize, clear, and delete.

I'm not sure how to resize without an inserting set.

I'm not sure I've ever seen clear on a namespace.  (I have seen it on
regular dicts being used as a namespace, such as tcl config options.)

I have seen deletes (deleting a temp name) and non-inserting sets ...
but they're both rare enough that letting them force the slow path
might be a good trade, if the optimization is otherwise simpler.

> Rare updating also means it's okay to invalidate the
> entire cache rather than single entries

Changing __bases__ seems to do that already.

(See http://svn.python.org/view/python/trunk/Objects/typeobject.c?rev=59106&view=markup
functions like update_subclasses.)

So I think an alternate version PJE's question would be:

Why not just extend that existing mechanism to work on non-slot,
non-method attributes?

-jJ