Constructing sparse from list of lists of tuples

I have a Python list of row information for a sparse matrix. Each row is represented as a list of (column, value) tuples. Call it alist:

alist = [[(1,10), (3,-3)],
         [(2,12)]]

How can I efficiently construct a scipy sparse matrix from this list of lists, resulting in a matrix like this:

0  10   0  -3
0   0  12   0

The obvious approach is to make a scipy.sparse.lil_matrix, which internally has this "list of lists" structure. But from the scipy.sparse.lil_matrix — SciPy v0.19.0 Reference Guide I see just three ways of constructing them:

• starting from a dense array
• starting from another sparse array
• just constructing an empty array

So the only way to get fresh data in is either to solve this problem with some other sparse matrix representation, or to start with a dense array, neither of which address the initial problem, and both of which seem likely to be less efficient representations than lil_matrix itself for this data.

I guess I can make an empty one, and use a loop to add values, but surely I'm missing something.

The scipy documentation is really frustrating when it comes to sparse matrices.

from Constructing sparse from list of lists of tuples