Ufdjrw

I have a numpy array whose rows is ordered and a vector. I need to insert this vector into the array if the value of the vector is smaller than some values of each array row. Furthermore, I need to get the index of the position. Can I use a vector like operation rather than two for loops?

For example, the i-th row of the array is [1,3,5,6], and the i-th value of vector is 4, the result is ([1,3,4,5], 2).
Here is the code example.

import random
import numpy as np

random.seed(0)
num_query = 2
topk = 3

distance = np.Inf * np.ones((num_query, topk))

print(distance)

for iter in range(4):
 print('------------loop: :d------------'.format(iter))
 dist = np.random.randn(num_query) / (0.3 * (iter+1))
 dist = dist * np.sign(dist)
 rows, columns = distance.shape
 for row in range(rows):
 d = dist[row]
 if d > distance[row, columns - 1]:
 continue
 min_col = 0
 for column in range(columns-1, -1, -1):
 if distance[row, column] < d:
 min_col = column + 1
 break
 distance[row, min_col + 1:] = distance[row, min_col:-1]
 distance[row, min_col] = d
 print(dist)
 print(distance)
 print('+++++++++++++++++++++++++++++++')
print(distance)

The output is:

[[inf inf inf]
 [inf inf inf]]
------------loop: 0------------
[3.60155727 0.0991411 ]
[[3.60155727 inf inf]
 [0.0991411 inf inf]]
+++++++++++++++++++++++++++++++
------------loop: 1------------
[1.4839764 1.99641269]
[[1.4839764 3.60155727 inf]
 [0.0991411 1.99641269 inf]]
+++++++++++++++++++++++++++++++
------------loop: 2------------
[2.44132044 0.19221564]
[[1.4839764 2.44132044 3.60155727]
 [0.0991411 0.19221564 1.99641269]]
+++++++++++++++++++++++++++++++
------------loop: 3------------
[0.29206705 1.17013653]
[[0.29206705 1.4839764 2.44132044]
 [0.0991411 0.19221564 1.17013653]]
+++++++++++++++++++++++++++++++
[[0.29206705 1.4839764 2.44132044]
 [0.0991411 0.19221564 1.17013653]]