What is the equivalent of map<int, vector<int> > in Python?

Question

In C++ often do something like this:

typedef map<int, vector<int> > MyIndexType;

Where I then use it like this:

MyIndexType myIndex;
for( ... some loop ...)
{
  myIndex[someId].push_back(someVal);
}

If there was no entry in the map the code will insert a new empty vector and then append to it.

In Python it would look like this:

myIndex = {}

for (someId,someVal) in collection:
   try:
      myIndex[someId].append(someVal)
   except KeyError:
      myIndex[someId] = [someVal]

The try except is a bit ugly here. Is there a way to tell the dictionary an object type to insert when a KeyError is encountered at dictionary declaration time?

Answer 1

Something like this perhaps:

myIndex = {}
for (someId,someVal) in collection:
    myIndex.setdefault(someId, []).append(someVal)

Answer 2

You want to use:

from collections import defaultdict
myIndex = defaultdict(list)
myIndex[someId].append(someVal)

Standard Library defaultdict objects.

Example usage from the Python documentation:

>>> s = [('yellow', 1), ('blue', 2), ('yellow', 3), ('blue', 4), ('red', 1)]
>>> d = defaultdict(list)
>>> for k, v in s:
        d[k].append(v)

>>> d.items()
[('blue', [2, 4]), ('red', [1]), ('yellow', [1, 3])]

Answer 3

How about this? It may not be performance-optimal, but I think it's the "simplest thing that could possibly work".

myIndex = {}

for (someId,someVal) in collection:
   if someId not in myIndex:
       myIndex[someId] = []
   myIndex[someId].append(someVal)

Answer 4

Just to complete the answer by Alastair: There is also the get equivalent of setdefault, which is called get (and not getdefault, as one might think):

myIndex = {}
someId = None
myList = myIndex.get(someId, []) # myList is [] now

Answer 5

From Python 2.5 and on you can get the behavior of setdefault or using defaultdict by implementing

__missing__(k)

as in note 10 here.