Memory consumption of a dictionary: set a value None vs. delete the item

Question

I understand that del d[key] will delete the key-value pair, whereas d[key]=None only de-references the value.

However, in terms of memory management, is there any difference? Does setting a value None trigger garbage collection immediately, assuming that there is no other variable referencing this value?

I ran a little experiment:

In [74]: import sys
In [75]: a = {'a': 'blah'}

In [76]: sys.getsizeof(a)
Out[76]: 280

In [77]: a['a'] = None

In [79]: sys.getsizeof(a)
Out[79]: 280

In [80]: del a['a']

In [81]: sys.getsizeof(a)
Out[81]: 280

Not sure if the approach is valid, but it seems no difference in terms of the size of the dictionary at all. I must miss something here.

Answer 1

sys.getsizeof measures the size of the dict itself; not the size of the values it contains.

None is an object. It requires some memory.

To find the size of a dict including the size of the values it contains, you could use pympler:

In [26]: import pympler.asizeof as asizeof

In [27]: asizeof.asizeof({'a': None})
Out[27]: 168

In [28]: asizeof.asizeof({})
Out[28]: 136

In [29]: import sys

In [30]: 
In [31]: sys.getsizeof({})
Out[31]: 136

In [34]: sys.getsizeof({'a':None})
Out[34]: 136

Answer 2

unutbu is correct. But also the python garbage collector can be a little slow on the uptake sometimes.

Calling del on an object and removing all references to that object means that the garbage collector can collect it whenever it feels ready. This will not usually result in an immediate decrease in the amount of memory used.

you can make use of the gc module to force collection and look at what the garbage collector actually sees.