How to index and slice a NumPy array

Rows and columns in numerical data often need to be isolated before a calculation, plot, or quality check. NumPy expresses those selections inside square brackets, so a scalar or subarray follows the same coordinate order as its source ndarray.

A two-dimensional index addresses the row axis first and the column axis second. An integer removes the selected axis, while a colon retains every position on that axis; positions start at zero, and negative positions count backward from the end.

Slice bounds use start:stop:step and exclude the stop position. Basic slices ordinarily share the original data buffer, which avoids copying but also means that changing a slice can change its source; call .copy() when the selected data must be independent.

1. Create the input array in array-index-slice.py.

   array-index-slice.py

   import numpy as np
    
   readings = np.arange(1, 21).reshape(4, 5)
    
   print("readings:")
   print(readings)

2. Append the integer-indexing selections to array-index-slice.py.

   array-index-slice.py

   single = readings[2, 3]
   row = readings[2, :]
   column = readings[:, 3]
    
   print("single readings[2, 3]:", single)
   print("row readings[2, :]:", row)
   print("column readings[:, 3]:", column)

   readings[2, 3] returns one scalar. readings[2, :] keeps every column in row 2, while readings[:, 3] keeps column 3 from every row.

3. Append the basic-slice selections to array-index-slice.py.

   array-index-slice.py

   window = readings[1:3, 2:5]
   reversed_rows = readings[::2, ::-1]
    
   print("window readings[1:3, 2:5]:")
   print(window)
   print("reversed rows readings[::2, ::-1]:")
   print(reversed_rows)

   The rectangular slice keeps rows 1 and 2 with columns 2 through 4. The stepped slice keeps every second row and reverses its column order.

4. Append fail-capable selection checks to array-index-slice.py.

   array-index-slice.py

   assert single == 14
   assert row.tolist() == [11, 12, 13, 14, 15]
   assert column.tolist() == [4, 9, 14, 19]
   assert window.tolist() == [[8, 9, 10], [13, 14, 15]]
   assert reversed_rows.tolist() == [[5, 4, 3, 2, 1], [15, 14, 13, 12, 11]]
   assert np.shares_memory(readings, window)
    
   print("window shares memory:", np.shares_memory(readings, window))
   print("indexing and slicing checks passed")

   Python stops with an AssertionError if any selection differs from the expected values or the basic rectangular slice does not share memory with readings.

5. Run the completed indexing and slicing script.

   $ python3 array-index-slice.py
   readings:
   [[ 1  2  3  4  5]
    [ 6  7  8  9 10]
    [11 12 13 14 15]
    [16 17 18 19 20]]
   single readings[2, 3]: 14
   row readings[2, :]: [11 12 13 14 15]
   column readings[:, 3]: [ 4  9 14 19]
   window readings[1:3, 2:5]:
   [[ 8  9 10]
    [13 14 15]]
   reversed rows readings[::2, ::-1]:
   [[ 5  4  3  2  1]
    [15 14 13 12 11]]
   window shares memory: True
   indexing and slicing checks passed

   The final line appears only after every scalar, row, column, rectangular-window, stepped-slice, and memory-sharing assertion succeeds.