Error replacing numbers in an array with predefined values in python

Question

I'm trying to replace values in an array (0-99 repeating) with a new set based on the value of n. For eg. if n=0, values 0,10,20..90 should be replaced by 0,1,2..9 and the rest should be 10. The following code works okay for all values of n (0-8) except 9. For 9 it gives the message long() argument must be a string or a number, not 'NoneType' I've tried a lot debugging this but can't seem to find what the problem is.

import numpy as np
arr1=[[19][29][ 0][11][ 1][86][90][28][23][31][39][96][82][17][71][39][ 8][97]]
n = 9
d = {}
for i, j in zip(range(n, 100, 10), range(10)):
    d[i] = j
arr2 = np.vectorize(d.get)(arr1)
arr2[arr2 == None] = 10

arr1 is the original array and arr2 is the new array.

output should be

arr2=[[ 1] [ 2] [10] [10] [10] [10] [10] [10] [10] [10] [ 3] [10] [10] [10] [10] [ 3] [10] [10]]

@sacul if n=1 values 1,11,21...91 should be replaced with 0,1,2..9 and the rest by 10. I need to create 10 such arrays. The first 0-8 worked fine except 9. — Noel DSouza
– Noel DSouza, Commented Dec 4, 2018 at 18:04
check np.putmask. Can you put a sample of your original array and the output you want? — Tarifazo
– Tarifazo, Commented Dec 4, 2018 at 18:13
@Mstaino original [[19] [29] [ 0] [11] [ 1] [86] [90] [28] [23] [31] [39] [96] [82] [17] [71] [39] [ 8] [97]... output [[ 1] [ 2] [10] [10] [10] [10] [10] [10] [10] [10] [ 3] [10] [10] [10] [10] [ 3] [10] [10].... — Noel DSouza
– Noel DSouza, Commented Dec 4, 2018 at 18:19
@NoelDSouza please edit your question to so that it contains a Minimal Working Example. I.e. you need to import numpy, define arr1, so that anyone can run your code and reproduce your error. — MountainDrew
– MountainDrew, Commented Dec 4, 2018 at 18:36

Tarifazo · Accepted Answer · 2018-12-04 19:06:31Z

2

You can use np.putmask (see here) to replace specific values with a formula based on those values (see here).

As for your case, you can modulus: it's easier and faster than using a dictionary. Does this represent your desired input/output?

import numpy as np
n = 9

arr1=np.random.randint(0, 100, size=20)
arr2 = arr1.copy()
np.putmask(arr2, (arr1-n)%10 == 0, arr1 % 10)

print(arr1)
print(arr2)

[69 70 63 52 27 96 0 40 2 90 36 24 17 90 67 58 74 50 11 58]

[ 9 70 63 52 27 96 0 40 2 90 36 24 17 90 67 58 74 50 11 58]

Edited for your desired output:

n = 9 
arr1=np.random.randint(0, 100, size=20)
arr2 = arr1.copy()
mask = (arr1-n)%10 == 0
np.putmask(arr2, mask , arr1 // 10)
np.putmask(arr2, ~mask , 10)
print(arr1)
print(arr2)

[28 72 87 31 87 3 34 96 61 14 25 79 74 25 38 87 38 8 6 8] [10 10 10 10 10 10 10 10 10 10 10 7 10 10 10 10 10 10 10 10]

If you want to use a dictionary, set the default value in the .get method

arr2 = np.vectorize(lambda x: d.get(x,10))(arr1)

edited Dec 4, 2018 at 19:06

answered Dec 4, 2018 at 18:33

Tarifazo

4,3531 gold badge11 silver badges24 bronze badges

Sign up to request clarification or add additional context in comments.

3 Comments

Tarifazo Over a year ago

edited: please put a sample of your array and what you expect of output in your post!

Noel DSouza Over a year ago

this works, thank you. any idea why my code gave an error for 9?

Tarifazo Over a year ago

Not sure, but my guess is that when you are creating your array using np.vectorize, when the first element yields a number (n=9), numpy interprets the array is an int type it throws an error when the next is None. This does not happen in the other cases as None is the first element (so the numpy knows it is a generic data type). You can bypass this error, if you want to use a dictionary, using the default for d.get (edit in answer)

hpaulj · Accepted Answer · 2018-12-04 19:25:52Z

Correction:

arr1=[[19],[29],[ 0],[11],[ 1],[86],[90],[28],[23],[31],[39],[96],[82],[17],[71],[39],[ 8],[97]]

d is:

{9: 0, 19: 1, 29: 2, 39: 3, 49: 4, 59: 5, 69: 6, 79: 7, 89: 8, 99: 9}

The error, with full traceback, is:

Traceback (most recent call last):
  File "stack53618793.py", line 8, in <module>
    arr2 = np.vectorize(d.get)(arr1)
  File "/usr/local/lib/python3.6/dist-packages/numpy/lib/function_base.py", line 1972, in __call__
    return self._vectorize_call(func=func, args=vargs)
  File "/usr/local/lib/python3.6/dist-packages/numpy/lib/function_base.py", line 2051, in _vectorize_call
    res = array(outputs, copy=False, subok=True, dtype=otypes[0])
TypeError: int() argument must be a string, a bytes-like object or a number, not 'NoneType'

With n=8 d is {8: 0, 18: 1, 28: 2, 38: 3, 48: 4, 58: 5, 68: 6, 78: 7, 88: 8, 98: 9}. arr2 has a lot of None because that's the default for get.

vectorize performs a test calculation with the first element of arr1, and uses the result to set the return dtype.

With n=8, get(19) returns None, so the return dtype is set to object.

With n=9, get(19) returns integer 1 (it's in d), so the return dtype is int. That produces an error when another get returns None.

One fix is to set the otypes.

arr2 = np.vectorize(d.get, otypes=[object])(arr1)

Another possibility is to replace get with a `get(

arr2 = np.vectorize(lambda x: d.get(x,10))(arr1)

Then you don't need the None replacement step.

This vectorized get is probably not the fastest way of doing this replacement. But if you do use vectorize you need to watch out for traps like this automatic otypes.

When you ask about an error, you should include the full traceback, or at least enough so we know exactly where the error occurs. It wasn't obvious to me until I ran the test case.

Collectives™ on Stack Overflow

Error replacing numbers in an array with predefined values in python

2 Answers 2

3 Comments

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

3 Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Related