Is there a faster way to achieve the same result?

Question

I have this python code:

    for i, num in enumerate(num_arr):
        if num > threshold:
            num_arr[i] = threshold

'num_arr' is a simple array filled with integers, 'threshold' may vary from 10 to 100,000. Is there any faster way to achieve the same result? bitwise operation or something of that sort?

HAve a look at either `np.clip` or `np.minimum` – Paul Panzer Nov 02 '19 at 13:23 — Paul Panzer, Nov 02 '19 at 13:23

score 1 · Accepted Answer · answered Nov 02 '19 at 13:23

1

You can use .clip(..) [numpy-doc] for that. For example:

num_arr = num_arr.clip(max=threshold)

For example:

>>> a
array([14, 25,  7, 12,  2])
>>> a.clip(max=10)
array([10, 10,  7, 10,  2])

answered Nov 02 '19 at 13:23

Willem Van Onsem

443,496
30
428
555

yep, it seems to cut off like 0.1 sec from execution time, which is good. – Dan Shorla Ki Nov 02 '19 at 13:45
@DanShorlaKi: for small arrays, the boost will not be that significant. Usually numpy/pandas pay off when you process large amounts of data in "bulk" (like an array of 1'000, 10'000+, etc. elements). – Willem Van Onsem Nov 02 '19 at 13:47
i found this has about same execution time as np.clip array[array > threshold] = threshold – Dan Shorla Ki Nov 02 '19 at 14:12

Is there a faster way to achieve the same result?

1 Answers1