python precision how to deal with without numpy

Question

i have a written a function that returns that 2 numbers are equal if the difference between 2 numbers is <=1e-6

if abs(a - b) > eps:
return False
return True
# 4 4.000001
in python i checked using

print(f"{abs(a - b):.20f}")
the difference is indeed saves as 1.e-6 something thats why i get false

Answer 1

Computers sometimes store numbers with tiny rounding errors.

To compare numbers like 4 and 4.000001, which should be 'equal' for your purpose, use math.isclose(a, b, rel_tol=1e-6) instead of if abs(a - b) > 1e-6.

This avoids those errors causing unexpected results because math.isclose() is designed to ignore tiny differences.

PS: I had a float error in my Diablo 1 curses clone when calculating remaining HP of enemies (int/int = float). I didn't use this, but just rounded as a simple solution. In your case, you might want floats so this is a Pythonic solution.

Note: Need to import math library

Answer 2

Dareen Moughrabi ,

if we have to handle decimals numbers without any compromises, we should not use float, but the data format `decimal`.

therefore we need to import the decimal modul. see a short comparison and sample:

https://sololearn.com/compiler-playground/cKNTzaOn1NV0/?ref=app

Answer 3

Dareen Moughrabi ,

here are some key points to `float vers. decimal in python`

> performance:
*float*: fast, since it is implemented in c and hardware-supported, suitable for most cases.
*decimal*: implemented in python (software based), so lower performance, used for high precise calculations.

> precision:
*float*: 64-bit, with rounding errors.
*decimal*: high precision, accurate for financial or math calculations.

> memory consumption:
*float*: fixed at 8 bytes
*decimal*: i think this is more memory-consuming due to storage model and some overhead ???

> handling:
*float*: is default in python, mostly supported in modules / libraries.
*decimal*: a module has to be imported, needs type conversions.

Answer 4

PietPadda thank you

Answer 5

Lothar do we usually default to float for memory reasons so decimal is more memory heavy but other than that it does the same as float?