python dict setdefault，困惑

Question

I was looking for an algorithm, and I can't figure out why the dict d has values in it and curr does not. I think it does not seem like anything is being done to dict d .

>>> def what(*words):
...     d = {}
...     print d
...     for word in words:
...     print 'word: ' + word
...         curr = d
...         for letter in word:
...             curr = curr.setdefault(letter, {})
...         curr = curr.setdefault('.', '.')
...     print d
...     print '?'
...     print curr
...     return 1
... 
>>> what('foo') 
{}
word: foo
{'f': {'o': {'o': {'.': '.'}}}}
?
.
1

Answer 1

Read the documentation for dict.setdefault : it is like get but if the key wasn't present then it is also set:

>>> my_dict = {}
>>> my_dict.setdefault('some key', 'a value')
'a value'
>>> my_dict
{'some key': 'a value'}
>>> my_dict.get('some key2', 'a value2')
'a value2'
>>> my_dict
{'some key': 'a value'}

Modifying a little your example:

>>> def what(*words):
...     d = dict()
...     for word in words:
...             curr = d
...             for letter in word:
...                     curr = curr.setdefault(letter, {})
...             curr = curr.setdefault('.', '.')
...             print 'curr is now: %r while d is %r' % (curr, d)
... 
>>> what('foo')
curr is now: '.' while d is {'f': {'o': {'o': {'.': '.'}}}}

As you can see curr changes, because when calling setdefault it sometimes(in your example always) create a new dict and set it as value to curr , while d always refers to the original dict . As you can see it is modified after the loop, since it's value is {'f': {'o': {'o': {'.': '.'}}}} which is quite different from {} .

Probably your confusion is due to the fact that curr = curr.setdefault(letter, {}) always create a new and empty dict , which is then assigned to curr (and thus for every letter you add a nesting level to the original dict instead of overwriting the values).

See this:

>>> my_dict = {}
>>> curr = my_dict
>>> for letter in 'foo':
...     print 'my_dict is now %r. curr is now %r' % (my_dict, curr)
...     curr = curr.setdefault(letter, {})
... 
my_dict is now {}. curr is now {}
my_dict is now {'f': {}}. curr is now {}
my_dict is now {'f': {'o': {}}}. curr is now {}
>>> my_dict
{'f': {'o': {'o': {}}}}

As you can see for every level the my_dict has a new nesting level.

Maybe, but I'm just guessing, you wanted to obtain something like 'foo' -> {'f': {}, 'o': {}} , in which case you should do:

>>> my_dict = {}
>>> for letter in 'foo':
...     my_dict.setdefault(letter, {})
... 
>>> my_dict
{'o': {}, 'f': {}}

Answer 2

d = dict() --> initializes an empty dictionary and binds it to the name d ; so you have a dictionary object ( {} ) referenced by name d

Inside the outer for loop
curr = d --> binds another name curr to the same object. So, names ( d and curr refer to the same object)

Inside the inner for loop
During the first iteration letter = 'f'

curr = curr.setdefault(letter, {})

There are 2 things that are happening in the above statement,

A) curr.setdefault(letter, {}) --> As per documentation:

"If key is in the dictionary, return its value. If not, insert key with a value of default and return default. default defaults to None.".

Since, the letter 'f' is not in the initial dictionary object it mutates the initial object to {'f':{}} and returns the value {} , which is not the initial dictionary object, but a new one that was created because of the setdefault statement. At this time both curr and d refer to the initial dictionary object which has since mutated to {'f':{}} .

B) Reassignment of the name curr to the return value mentioned above. Now, the names curr and d refer to different objects. d refers to the object {'f':{}} , while curr refers to an empty dictionary object, which is actually the value of d['f'] . This is why the nesting happens in the original dictionary object, as we go through the loop.

Answer 3

setdefault(key[, default)

From the docs :

If key is in the dictionary, return its value. If not, insert key with a value of default and return default . default defaults to None .

Usage examples

>>> d = {'a': 1, 'b': 2, 'c': 3}
>>> d.setdefault('a') # returns the corresponding value for key 'a'
1
>>> d.setdefault('a', 10) # returns the corresponding value for key 'a'
1
>>> d.setdefault('b') # returns the corresponding value for key 'b'
2
>>> d.setdefault('c', 100) # returns the corresponding value for key 'c'
3
>>> type(d.setdefault('z')) # because 'z' is not a key of d, None is returned which is the default value of default 
<class 'NoneType'>
>>> d.setdefault('z', 666) # returns 666 since key 'z' is not in d
666

In your code

I think you are confused because curr = curr.setdefault(letter, {}) always creates a new and empty dict , which is then assigned to curr . This means that instead of overwriting the values, you are adding a nesting level to the original dict for every element in words .

I also think that what you want to achieve with your code is to create a dictionary with every element in words as key with {} as value, so you can achieve it using the following code that uses a dict-comprehension :

def what(*words):
    return {word: {} for word in set(words)}

Note: I have added the explanation of setdefault since your questions has been viewed particularly for this case, but I wanted to cover your specific question too.