随机，非重复的2D列表Python

Question

I have a program that generates a two dimensional list containing a set number of 2 dimensional co-ordinates within a certain range. (It doesn't matter if the nested are lists or tuples) eg:

[[5, 0], [4, 6], [9, 7], [2, 9], [2, 6]]

The problem is that I wish for no two sublists to be equal while keeping the length of the outer list the same.

i have tried using random.sample:

pointx = random.sample(range(10),5)
pointy = random.sample(range(20),5)
points = list(zip(pointx,pointy))

where 10 is the x range, 20 is the y range and 5 is the amount of points:

[(1,0),(9,19),(8,13),(3,5),(0,14)]

However this method's maximum amount of points is only 10+20 where it clearly should be 10*20 because of the separate sample lists.

While writing this, i have realised that:

point2D = []
point1D = random.sample(range(10*20),5)
[point2D.append(divmod(i,10)) for i in point1D]

This is a viable approach, but also clumsy, i was wondering whether there was a better solve possibly involving numpy.

Thanks in advance.

Answer 1

Imagine a 2D grid of shape (10,20) with points at each integer points. Now, let's say we want to select 5 unique points on this grid. That's all! We translate this to NumPy.

Choose 5 unique indices from all possible 200 indices with :

idx = np.random.choice(200, size=5, replace=0)

Convert these indices into x and y coordinates that is speaking in terms of matrices terminology, get the row and column indices :

row, col = np.unravel_index(idx,(10,20))

These would be the desired tuples.

Let's put these two steps together for a complete code -

idx = np.random.choice(200, size=5, replace=0)
out = np.column_stack((np.unravel_index(idx,(10,20))))

Let's use a sample case -

In [461]: idx = np.random.choice(200, size=5, replace=0)
     ...: out = np.column_stack((np.unravel_index(idx,(10,20))))
     ...: 

In [462]: out
Out[462]: 
array([[ 6, 18],
       [ 9, 17],
       [ 5,  7],
       [ 8,  2],
       [ 6, 12]])

Answer 2

Using numpy we can create all of the ordered pairs of (x,y) and then randomly select a number of pairs.

# first create an array of all possible pairs from (0,0) to (9,19)
pairs = np.dstack(np.meshgrid(np.arange(10), np.arange(20))).reshape(-1,2)

# now select a random set of 25 of those pairs, which are unique
pairs[np.random.choice(np.arange(pairs.shape[0]), size=25, replace=False)]