Consider a dataframe like:
df = pd.DataFrame({'r': [1, 1, 2, 2], 'c': [0, 2, 1, 2], 'v': [2, 4, 3, 5],})
I want to extract a numpy array or tensor considering 'r' and 'c' as row and column index of the matrix. so the corresponding matrix will be like:
arr = array([[0, 0, 0],
[2, 0, 4],
[0, 3, 5]])
So is there a decent way to do that or I have to loop through each df row? what about extracting dataframe df from a matrix like arr?
To create a numpy array from the dataframe considering r and c as the row and column index:
r, c, v = df.T.values
arr = np.zeros((r.max() + 1, c.max() + 1))
arr[r, c] = v
>>> arr
array([[0., 0., 0.],
[2., 0., 4.],
[0., 3., 5.]])
To recreate the dataframe from the numpy array like above you can use np.nonzero to get the indices of the elements that are non-zero:
r, c = np.nonzero(arr)
df = pd.DataFrame({'r': r, 'c': c, 'v': arr[r, c]})
>>> df
r c v
0 1 0 2.0
1 1 2 4.0
2 2 1 3.0
3 2 2 5.0
