weird dict problem, how can this even happen?

Joel Hedlund joel.hedlund at gmail.com
Tue Dec 16 14:32:39 CET 2008


Duncan Booth wrote:
> I think you probably are correct. The only thing I can think that might 
> help is if you can catch all the situations where changes to the dependent 
> values might change the hash and wrap them up: before changing the hash pop 
> the item out of the dict, then reinsert it after the change.

That would probably require a lot of uncomfortable signal handling, 
especially for a piece of functionality I'd like to be as unobtrusive as 
possible in the application.

> Alternatively give up on defining hash and __eq__ for FragmentInfo and rely 
> on object identity instead.

Object identity wouldn't work so well for caching. Objects would always 
be drawn as they appeared for the first time. No updates would be shown 
until the objects were flushed from the cache.

I've been experimenting with a list cache now and I can't say I'm 
noticing any change in performance for a cache of 100 items. I'm still 
using the hash to "freeze" a sort of object tag in order to detect 
changes, and I require both hash and object equality for cache hits, 
like so:

> def index(self, key):
>     h = hash(key)
>     for i, item in enumerate(self.items):
>         if item.hash == h and item.key == key:
>             return i
>     raise KeyError(key)

This seems to do what I want and does OK performance wise.

Thanks again!

/Joel



More information about the Python-list mailing list