Restructure dictionary (Python 2.5)

[Tim Chase]

> I tried this code:
> old_dict = {'xyz':['a','b','c'],'baz':['c','d']}
> new_dict = {}
> for dkey, vallist in old_dict.iteritems():
> 	for val in vallist:
> 		if val in new_dict:
> 			theset = new_dict[val]
> 			theset.add(dkey)
> 			new_dict[val] = theset
> 		else:
> 			new_dict[val] = set([dkey])
> print new_dict
> 
> Which yields the output {'a': set(['xyz']), 'c': set(['xyz', 'baz']), 
> 'b': set(['xyz']), 'd': set(['baz'])}, so it seems to work but I have a 
> feeling this is a crappy solution that's underusing Python. Please 
> enlighten me of a better way!

In python2.5, collections.defaultdict was added which could clean 
this up a bit:

   import collections
   # using a set to remove duplication
   new_dict = collections.defaultdict(set)
   for dkey, vallist in old_dict.iteritems():
     for val in vallist:
       new_dict[val].add(dkey)

   # using a list to keep all keys
   new_dict = collections.defaultdict(list)
   for dkey, vallist in old_dict.iteritems():
     for val in vallist:
       new_dict[val].append(dkey)

If you're working in pre-2.5 python, you could use

   new_dict = {}
   ...
       # use a set()
       new_dict.setdefault(val, set()).add(dkey)
       # use a list()
       new_dict.setdefault(val, []).append(dkey)

-tkc