[Tutor] Re: Unique Items in Lists

[Brian van den Broek]

Wolfram Kraus said unto the world upon 2005-01-27 03:24:
> Brian van den Broek wrote:

<SNIP>

Thanks Wolfram,

I knew someone would improve what I posted. (It can always be done ;-)

>>     for i in a_list:
>>         if i in items_dict:
>>             items_dict[i] = items_dict[i] + 1
>>         else:
>>             items_dict[i] = 1
> 
> get(key, default) is your friend here:
>       for i in a_list:
>           items_dict[i] = items_dict.get(i, 0) + 1
> 
> get() (and his "brother" setdefault()) are mighty dictionary-methods.

What a brain fart I had! I even got as far as reading then entry in 
the Library Ref Mapping Types table for setdefault, 'cause I knew 
there was a way to do this, but couldn't recall it. How they expect me 
to find it when they hide it on the previous line I'll never know ;-) 
/ :-[

>>     for key in items_dict.copy():   # Try it without the .copy()
>>         if items_dict[key] == 1:    # and see what happens.
>>             del items_dict[key]
>>
>>     dict_keys = items_dict.keys()
>>     dict_keys.sort()
> 
>  >
> 
>>     for key in dict_keys:
>>         print '%s occurred %s times' %(key, items_dict[key])
> 
> This whole part can be rewritten (without sorting, but in Py2.4 you can 
> use sorted() for this) with a list comprehension (Old Python2.1 style, 
> with a newer version the keys() aren't needed):
>       for k,v in [(k, items_dict[k]) \
>       for k in items_dict.keys() if items_dict[k] > 1]:
>           print '%s occurred %s times' %(key, items_dict[key])

I knew sorted() only from the 'What's New in 2.4'. Thanks for 
reminding me to look closer.

I get that the for loops in mine are more costly than the list comps 
in yours. But, I can't help but feel that the list comp isn't as clear 
or readable. Of course, that may say more about me (and the fact that 
while I can read list comps, they aren't really in my working 
vocabulary yet) than it does about the code. I think I like the way I 
wrote it in that it breaks the discrete steps apart. But, for long 
lists, that will be a performance bite indeed.

(I went from my first posts being too concerned about speed to not 
giving a toss. Perhaps the pendulum needs to swing back a bit.)

Anyway, thanks for showing me (and the OP) these alternatives!

Best,

Brian vdB