I have initialized matrix:
M = np.array(((),()))
M has the shape of (2,0) now. I wish to fill M by steps: firstly to add 1 number like
M[0] = np.append(M[0],55)
By this operation I want to get such a matrix
((55),())
How can I do this? I can make this with standard pythons arrays [] by "append" operation like
arr = [[],[]]
arr[0].append(55)
But after that, I need this array to be a numpy array and there is one extra type transform operation which I wish to avoid.
The array you've written is no matrix because its axis has different dimensions. You could do it like this
import numpy as np
x = np.zeros((2,1))
x[0][0] = 55
Then if you want to append to it you can do something like:
x = np.append(x, [[42], [0]], axis=1)
Note that in order to append to a martrix all the dimensions except for the concatenation axis must match exactly
x = np.array([[0,1,2,3], [2,3,4]]) numpy will transform it into an 1d array of list-objects. This is not the intended use of numpy thoughI can start with a 2 element object dtype array:
In [351]: M = np.array((None,None))
In [352]: M.shape
Out[352]: (2,)
In [353]: M
Out[353]: array([None, None], dtype=object)
In [354]: M[0]=(5,)
In [355]: M[1]=()
In [356]: M
Out[356]: array([(5,), ()], dtype=object)
In [357]: print(M)
[(5,) ()]
Or more directly (from a list of tuples) (beware, sometimes this produces a error rather than object array).
In [362]: np.array([(55,),()])
Out[362]: array([(55,), ()], dtype=object)
But I don't see what it's good for. It would easier to construct a list of tuples:
In [359]: [(5,), ()]
Out[359]: [(5,), ()]
Do not try to use np.append like the list append. It is just a clumsy front end to np.concatenate.
M as you create it is:
In [360]: M = np.array(((),()))
In [361]: M
Out[361]: array([], shape=(2, 0), dtype=float64)
It can't hold any elements. And you can't change the shape of the slots as you can with a list. In numpy shape and dtype are significant.
You can specify object dtype:
In [367]: M = np.array([(),()], object)
In [368]: M
Out[368]: array([], shape=(2, 0), dtype=object)
but it's still impossible to reference and change one of those 0 elements.
ndarray? Whatshapeanddtypeare you aiming for? What kinds of calculations are you going to do with this array?ndarrayis fixed. You can reshape it, but the total number of elements has to remain unchanged. To add elements you have to make a new array, and copy values. If you do need to create an array incrementally, it's best to create lists (or lists of lists) and do one type conversion, or preallocate the array, and fill in values incrementally.