[ Python 常見問題 ] Numpy - From ND to 1D arrays

Source From Here
Question
Say I have an array a:

view plaincopy to clipboardprint?

1. >>> import numpy as np  
2. >>> ary = np.array([[1,2,3],[4,5,6]])  
3. >>> ary  
4. array([[1, 2, 3],  
5.        [4, 5, 6]])  

I would like to convert it to a 1D array (i.e. a column vector) !

HowTo
Use np.ravel (for a 1D view) or np.ndarray.flatten (for a 1D copy) or np.ndarray.flat (for an 1D iterator):

view plaincopy to clipboardprint?

1. >>> ex1 = ary.reshape(1, -1)  
2. >>> ex1  
3. array([[1, 2, 3, 4, 5, 6]])  
4. >>> ex1.shape  
5. (1, 6)  
6.   
7. >>> ex2 = ary.ravel()  
8. >>> ex2  
9. array([1, 2, 3, 4, 5, 6])  
10. >>> ex2.shape  
11. (6,)  

Note that ravel() returns a view of a when possible. So modifying ary also modifies ex2. ]ravel() returns a view when the 1D elements are contiguous in memory, but would return a copy if, for example, a were made from slicing another array using a non-unit step size (e.g. ary = x[::2]). If you want a copy rather than a view, use:

view plaincopy to clipboardprint?

1. >>> ex3 = ary.flatten()  
2. >>> ex3  
3. array([1, 2, 3, 4, 5, 6])  
4. >>> ex3.shape  
5. (6,)  
6. >>> ary[0,0]=10  
7. >>> ary  
8. array([[10,  2,  3],  
9.        [ 4,  5,  6]])  
10. >>> ex3  
11. array([1, 2, 3, 4, 5, 6])  

If you just want an iterator, use np.ndarray.flat:

view plaincopy to clipboardprint?

1. >>> ex4 = ary.flat  
2. >>> ex4  
3. <numpy.flatiter object at 0x2cbfce0>  
4. >>> list(ex4)  
5. [10, 2, 3, 4, 5, 6]