If the value is None, I'd like to change it to "" (empty string).
I start off like this, but I forget:
for k, v in mydict.items():
if v is None:
... right?
Add a comment
|
3 Answers
for k, v in mydict.iteritems():
if v is None:
mydict[k] = ''
In a more general case, e.g. if you were adding or removing keys, it might not be safe to change the structure of the container you're looping on -- so using items to loop on an independent list copy thereof might be prudent -- but assigning a different value at a given existing index does not incur any problem, so, in Python 2.any, it's better to use iteritems.
In Python3 however the code gives AttributeError: 'dict' object has no attribute 'iteritems' error. Use items() instead of iteritems() here.
Refer to this post.
answered Feb 23, 2010 at 1:21
4 Comments
Alex Martelli
@John, it applies to any built-in container -- just as much to a list as to a dict -- and I think "index" is a more generic term than "key".
CivFan
mydict[k] = '' -- your example updates k index to an immutable object, ''. So after this, k would point to an entirely different object than v. Would this mutate the k index in some way that might cause problems with .iteritems()?Alex Martelli
@CivFan, nope, altering or re-assigning a value (as opposed to a key) does not give problems with
dict.iteritems.
mgilson
Hmm ... I wonder if the advice to use
.items to loop if you're adding/removing keys is prudent. I don't know the implementation details of dict_items objects, but it seems that you might still have issues iterating over them in python3.x if you're adding/deleting keys. In that case, it's probably safest to iterate over a list of the keys: for k in list(mydict): v = mydict[k]; ...You could create a dict comprehension of just the elements whose values are None, and then update back into the original:
tmp = dict((k,"") for k,v in mydict.iteritems() if v is None)
mydict.update(tmp)
Update - did some performance tests
Well, after trying dicts of from 100 to 10,000 items, with varying percentage of None values, the performance of Alex's solution is across-the-board about twice as fast as this solution.
2 Comments
ジョージ
Hmm, for Python 2.7 I see about the same performance ( though the Alex's method is still always faster ) ; I have tried dicts with 1000 .. 100000 values and None percentage from 2 to 50 %% ; for comparison, the method from buckley ( with s/items/iteritems/ ) was about 4 times slower. I think this is kind of interesting, since your form allows one to alter the dictionary on iteration.
Muhammad Khuzaima Umair
list comprehensions make a new dict, that's why it is slower as compared to the accepted answer.
Comprehensions are usually faster, and this has the advantage of not editing mydict during the iteration:
mydict = dict((k, v if v else '') for k, v in mydict.items())
3 Comments
John La Rooy
If you are using
.items() it doesn't matter (for python2) if you modify mydict as the list that .items() returns will not change even if you add/remove keys of mydict
John Machin
Changing only values in a dict is never a problem; grief is caused by adding/deleting KEYS while iterating over the dict. Comprehensions are faster than WHAT? Really fast: copying the whole dict when there's one or two Nones to change. Speaking of Nones, you should have
if v is not None instead of if v (re-read the question). Overall summary: -1
Statistic Dean
Your solution replaces 0, empty list and everything that is False in a boolean context by
''. I would suggest replacing if v by if v is None, this is also more PEP 8 compliant : python.org/dev/peps/pep-0008/#programming-recommendations