Why 0.1 + 0.05 is not equal to 0.15

General Discussion
21

Oh, come on ! :smiley:

22

It reminds me of how there are countable infinities and uncountable infinities. For example, there are more numbers between 0 and 1 then there are integers. When we try to stuff the infinite into the finite space of a computer you have to make compromises somewhere.

1 Like
23

The Donald Rumfeld paradox ?

24

Hello everyone, here is a preview of my PR:

It has a default tolerance of 0.0000001, even if you don’t use a mutator. It should solve the problem of 0.1+0.2 not being equal to 0.3

image

image
image1440×1064 151 KB

PR:

25

You are missing the test case 0.3 > 0.1 + 0.2

It's not equivalent to the 0.1+0.2<0.3 test.

26

@ABG both return false in my test

image
image842×476 38.5 KB

1 Like
27

That should cover all the cases.

1 Like
28

Thanks :grinning_face: I’ll include it in my PR

29

I forgot to include tests with negative numbers.

Do those fool the offsets?

Also, you're going to love this one:

image
image576×163 6.37 KB

Take the 4 * 3 cross comparisons of the 4 complex numbers at +/- 1 +/- i (n X at the origin of the complex plane.)

Then take an aspirin.
You'll probably need it.

image
image491×110 4.79 KB

1 Like
30

No, they work fine. This is the internal operation for different operators:

image
image1408×566 61.5 KB

and, about the complex number comparison, the comparison itself is invalid. It should throw an error as ordering operators are not defined for complex numbers. The behavior is weird though :face_without_mouth:

image
image1244×644 59 KB

31

Are the two signs exchanged?

eq01a
eq01a690×277 43.5 KB

32

No, my intention behind using opposite directions of tolerance in LT & LTE is to include near-equal values in LTE.

Example: arg1 = 10 and tolerance = 0.001

LTE: if arg0 is 10.0005 I want LTE to return true as 10.0005 is a near-equal value to 10, that’s why + tolerance is there (10.0005 <= 10.001)

LT uses - for the opposite reason, it should exclude near-equal values.

33

Any chance of getting this up on a test server, so we can hit it with corner cases?

34

You need to fetch my PR branch & compile it locally to run on your localhost. Then you can test it OR you can send me test values, I’ll post the result here