CODEWITHSUNDEEP
pythonfloating-point
Dave
Dave

Reputation: 8099

Check if floating point number is normal in Python

Are there any python library functions that discriminate between regular floating point numbers, and all of the atypical ones, NaN, +/-Inf, denormal etc.?

Essentially, I'm asking the Python version of this question: check if floating point variable has 'normal' values

Upvotes: 4

Views: 2377

Answers (2)

aka.nice
aka.nice

Reputation: 9572

To check if a float x is finite, just use math.isfinite(x).

To check if a float x is denormal (subnormal), here is a possible snippet:

import numpy
x = 1.0e-320
fmin_normalized = numpy.finfo(type(x)).tiny
x_is_subnormal =  numpy.isfinite(x) and x!= 0 and abs(x) < fmin_normalized
print(x_is_subnormal)

I let you finish the work (write the appropriate lambda and filter the collection)

Upvotes: 7

William Herewini
William Herewini

Reputation: 199

EDIT: As pointed out, my previous reply didn't quite work and I'm unsure of a clean method of doing this. So I guess you would have to pick out the types you would like to remove like so:

test = [4, 4e99, 0b0101, 3.2, np.nan, 6, np.nan, np.inf]

result = []
for value in test:
    if not np.isnan(value) and np.isfinite(value):
        result.append(value)

print(result)

OUTPUT:
[4, 4e+99, 5, 3.2, 6]

Upvotes: 0

Related Questions