Why not Ruby?

Paul Rubin http
Fri Jan 2 04:51:53 CET 2009


Marek Kubica <marek at xivilization.net> writes:
> > I guess if it is functional then every mutation must copy and return a
> > new data structure? 

Yes.

> > (Which will be much more efficient in Haskell than
> > in Python - Haskell can share most of the underlying data whereas Python
> > would have to create a new dict every time. At least it only stores
> > references.)

The structure sharing is essential, but you can do it in Python, just
not using Python dicts as far as I can tell.

> Who says that it must create a whole new one? I could imagine that with a 
> bit weakref code and some thought an immutable dictionary that shares 
> data would be possible in Python too.

I don't see a way to do that.  Suppose d and e are dicts that are supposed
to share structure except d['name']='bob' and e['name']='joe'.  How do
weakrefs help?

Functional dictionaries are usually implemented using red-black trees
or AVL trees or similar data structures, rather than hash tables.



More information about the Python-list mailing list