Python Dict to numpy multidimensional array

Question

I have a python dictionary with keys (0,1,2....n) with each key holding the location co-ordinates in a tuple such as

{0:(x1,y1), 1:(x2,y2), 2:(x3,y3), ....., n-1:(xn,yn)}

I want to create a multidimensional array like

coordinates= np.array([
       [x1, y1],
       [x2, y2],
        ...
       [xn, yn]
       ])

I have tried using array = numpy.array(vehPositions.values()) but could not get the expected result.

Why didn't numpy.array(vehPositions.values()) work? Are the values in different order or wrong values? — Divakar, Jun 29 at 13:43
I only get a array of type (object) like array(dict_values([(x1,y1),(x2,y2),...])). If i use array(x), it returns me 'numpy.ndarray' object is not callable. If I do array[x], it gives me too many indices for array — Mechanic, Jun 29 at 13:55

Moses Koledoye · Accepted Answer · 2016-06-29 13:52:41Z

If you're using python 3.x you should cast explicitly to list as the values() method of dict will return a dict_values object.

You should do:

array = numpy.array(list(vehPositions.values()))

To create your list of lists from the tuple values in your dictionary and build your array, you will not need an explicit cast to list:

array = numpy.array([list(v) for v in vehPositions.values()])

Be careful to not use a generator expression as this will return an array containing a generator object:

>>> numpy.array(list(v) for v in vehPositions.values())
array(<generator object <genexpr> at 0x03B60AA8>, dtype=object)

The trial below demonstrates the procedure:

>>> d = {0: (1,2), 1: (3,4), 2: (5,6)}
>>> numpy.array([list(v) for v in d.values()])
array([[1, 2],
       [3, 4],
       [5, 6]])

1 Answer 1