CODEWITHSUNDEEP
pythonpython-2.7numpy
gutelfuldead
gutelfuldead

Reputation: 602

optimize iteration checking for maximum difference between consecutive values

I have a binary file of int_16's. They are sorted into an nx2 array where each column contains samples for a sinusoid.

I am checking each consecutive value to see if there is a magnitude difference between the two greater than some value. Here is some code that is performing the task. What is a better way of doing this? I am using Python 2.7

import numpy as np

DIFF_MAX = 100
errcnt = 0

f = open("samp.bin","rb")
a = np.fromfile(f, dtype=np.int16)
n = np.reshape(a, (-1,2))

for i in range(0,len(a)/2 - 2):
    if(abs(abs(n[i,0])-abs(n[i+1,0])) > DIFF_MAX):
            errcnt = errcnt + 1
            print "ERROR!!!"
    if(abs(abs(n[i,1])-abs(n[i+1,1])) > DIFF_MAX):
            errcnt = errcnt + 1
            print "ERROR!!!"

print str(errcnt) + " errors found out of " + str(len(a)/2)

Doing this I imagine will be inherently slow I was just curious if there was a better way. Thanks.

Runtime,

time python test.py 
0 errors found out of 4329472

real    0m21.025s
user    0m20.950s
sys 0m0.075s

Upvotes: 1

Views: 38

Answers (1)

Florian Drawitsch
Florian Drawitsch

Reputation: 715

You could do this by shifting the arrays by one position relative to each other and subtracting them.

abs(abs(n[1::,:]) - abs(n[:-1,:])) > DIFF_MAX

The resulting boolean array would indicate the positions with TRUE wherever the difference between two consecutive values along the first dimension is greater then DIFF_MAX and FALSE elsewhere.

Upvotes: 1

Related Questions