I'm following the book 'Data Science from Scratch' and this is a piece of code in it:
dd_pair = defaultdict(lambda: [0, 0])
dd_pair[2][1] = 1 # now dd_pair contains {2: [0, 1]}
Can someone please help me understand why and how it works?
defaultdict takes a data-type as an initializer. Let's consider we have a dictionary called "users" with "ID" being key and a list as value.
We have to check if a "ID" exists in the dictionary, if yes we append something to the list, else we put an empty list in that place.
So with a regular dictionary, we do something like:
users = {}
if "id1" not in users:
users["id1"] = []
users["id1"].append("log")
Now with defaultdict, all we have to do is to set an initialiser as:
from collections import defaultdict
users = defaultdict(list) # Any key not existing in the dictionary will get assigned a `list()` object, which is an empty list
users["id1"].append("log")
So coming to your code,
dd_pair = defaultdict(lambda: [0, 0])
This says, any key which doesn't exist in dd_pair will get a list of two elements initialised to 0 as their initial value. So if you just do print(dd_pair["somerandomkey"]) it should print [0,0].
Therefore, dd_pair[2][1] translates roughly to look like this:
dd_pair[2] = [0,0] # dd_pair looks like: {2:[0,0]}
dd_pair[2][1] = 1 # dd_pair looks like: {2:[0,1]}
Why the need for
lambda, why not just use [0,0] ?
The defaultdict constructor expects a callable (The constructor actually expects a default_factory, check out Python docs). In extremely simple terms, if we do defaultdict(somevar), somevar() should be valid.
So, if you just pass [0,0] to defaultdict it'll be wrong since [0,0]() is not valid at all.
So what you need is a function which returns [0,0], which can be simply implemented using lambda:[0,0]. (To verify, just do (lambda:[0,0])() , it will return [0,0]).
One more way is to create a class for your specific type, which is better explained in this answer: https://stackoverflow.com/a/36320098/
