When is it safe to compare (==) two floats?

[Radical]

I am writing a program that will automatically check & pay off the amount I owe on my credit card. Since it takes a few days to process payment, I am saving the value I owed to a .txt file on my PC, then when the script goes off the following day, it will check the text file and see if the amount I currently owe is the same as the amount I owed yesterday. If it is the same it will not pay again.

Since I am not performing mathematical operations on the floats, it may be fine to compare. But I am just concerned because everyone says comparing floats is unreliable.

The exact steps I'm performing:
- Copy the amount owing from the text on the website (ex: 237.55)
- Read the contents of my text file, and assign it to a variable (ex: 237.55)
- Compare the two floats (==) and if they are equal, I exit the program.

Could this potentially cause issues?

[buran]

have a look at decimal module from standard library. you are working with money, so always 2 decimal places

[Radical]

have a look at decimal module from standard library. you are working with money, so always 2 decimal places

I just tried it out and it seems to be doing the exact thing I was afraid float would do.

previousOwing = decimal.Decimal(234.55)
print(previousOwing)

Output:
234.55000000000001136868377216160297393798828125

which is ironic, because float seems to work perfectly:

previousOwing = float(234.55)
print(previousOwing)

Output:
234.55

[deanhystad]

Read before using. You cannot create a decimal object using a float.

https://docs.python.org/3/library/decimal.html

import decimal

a = decimal.Decimal("234.55")
b = decimal.Decimal(234.55)
print(a, b)

Output:
234.55 234.55000000000001136868377216160297393798828125

Well, you CAN create a decimal object using a float, but it defeats the whole purpose of using decimal. The float number is already inaccurate, even if that doesn't show up when you print it. Using the inaccurate float to initialize the Decimal object doesn't magically create more accuracy.

[PyDan]

For CAD stuff, I use a tolerance, Python has a math.isclose
https://docs.python.org/3/whatsnew/3.5.h...e-equality