CODEWITHSUNDEEP
python
Jason Coon
Jason Coon

Reputation: 18441

Python dictionary: are keys() and values() always the same order?

It looks like the lists returned by keys() and values() methods of a dictionary are always a 1-to-1 mapping (assuming the dictionary is not altered between calling the 2 methods).

For example:

>>> d = {'one':1, 'two': 2, 'three': 3}
>>> k, v = d.keys(), d.values()
>>> for i in range(len(k)):
    print d[k[i]] == v[i]

True
True
True

If you do not alter the dictionary between calling keys() and calling values(), is it wrong to assume the above for-loop will always print True? I could not find any documentation confirming this.

Upvotes: 436

Views: 143784

Answers (9)

Gregory Sky
Gregory Sky

Reputation: 130

I would agree with others that in python 3.6+ it should stay the same if unaffected by user.

an example from my code few days ago:

ips = { '001' : '199.250.178.14', '002' : '199.18.2.89', '003' : '109.251.63.21' }
def run(self):
    for x, y in self.ips.items():
                try:
                    subprocess.check_call(
                    ['ping', '-n', '1', y],
                    stdout=DEVNULL,  # suppress output
                    stderr=DEVNULL
                    )
                except subprocess.CalledProcessError:
                    nextServer = ('HUB ' + x + ' is OFFLINE   ' + " IP:   " + y)

and the output is always of the same order, exactly as i inputted it i.e. only offline servers will show up in order: HUB 003 is OFFLINE IP: 109.251.63.21 becasue first two are reachable

Hope that clears it out

Upvotes: 2

Alex Martelli
Alex Martelli

Reputation: 882781

Yes, what you observed is indeed a guaranteed property -- keys(), values() and items() return lists in congruent order if the dict is not altered. iterkeys() &c also iterate in the same order as the corresponding lists.

Upvotes: 107

user3064538
user3064538

Reputation:

Yes. Starting with CPython 3.6, dictionaries return items in the order you inserted them.

Ignore the part that says this is an implementation detail. This behaviour is guaranteed in CPython 3.6 and is required for all other Python implementations starting with Python 3.7.

Upvotes: 22

Sassa NF
Sassa NF

Reputation: 5406

Good references to the docs. Here's how you can guarantee the order regardless of the documentation / implementation:

k, v = zip(*d.iteritems())

Upvotes: 7

jlansey
jlansey

Reputation: 4259

I wasn't satisfied with these answers since I wanted to ensure the exported values had the same ordering even when using different dicts.

Here you specify the key order upfront, the returned values will always have the same order even if the dict changes, or you use a different dict.

keys = dict1.keys()
ordered_keys1 = [dict1[cur_key] for cur_key in keys]
ordered_keys2 = [dict2[cur_key] for cur_key in keys]

Upvotes: -1

nosklo
nosklo

Reputation: 223172

Found this:

If items(), keys(), values(), iteritems(), iterkeys(), and itervalues() are called with no intervening modifications to the dictionary, the lists will directly correspond.

On 2.x documentation and 3.x documentation.

Upvotes: 478

Bjorn
Bjorn

Reputation: 71970

Yes it is guaranteed in python 2.x:

If keys, values and items views are iterated over with no intervening modifications to the dictionary, the order of items will directly correspond.

Upvotes: 51

Koen Bok
Koen Bok

Reputation: 3284

For what it's worth, some heavy used production code I have written is based on this assumption and I never had a problem with it. I know that doesn't make it true though :-)

If you don't want to take the risk I would use iteritems() if you can.

for key, value in myDictionary.iteritems():
    print key, value

Upvotes: 4

sykloid
sykloid

Reputation: 101416

According to http://docs.python.org/dev/py3k/library/stdtypes.html#dictionary-view-objects , the keys(), values() and items() methods of a dict will return corresponding iterators whose orders correspond. However, I am unable to find a reference to the official documentation for python 2.x for the same thing.

So as far as I can tell, the answer is yes, but only in python 3.0+

Upvotes: 5

Related Questions