Question with floating point addition

Question

float totalAmount = 0;
.
.
.//totalAmount assigned value 1.05 correctly
.
totalAmount  += float.Parse(dataRow["Amt"].ToString()); //where dataRow["Amt"] has value 4.93

The answer I get for totalAmount is 5.97999954 instead of 5.98

Why is this happening?

Yeah, why the heck *doesn't* `4.93 + 1.05` equal `4.98` ?? ;) — Roatin Marth, Dec 15 '09 at 19:21
You could get away with doing this if the internal object is a float. (float)dataRow["Amt"] — ChaosPandion, Dec 15 '09 at 19:26
Use "decimal" instead of "float" if you want exact representations of decimal quantities. "float" gives exact representations of binary quantities. — Eric Lippert, Dec 15 '09 at 19:29
This question is so common it should be added to the FAQ. Read http://stackoverflow.com/questions/56947/ for a good overview. — Dour High Arch, Dec 15 '09 at 19:29

jdmichal · Accepted Answer · 2009-12-15T19:29:02.837

9

You state that totalAmount is 1.05, before the accumulation. That would give expected results of:

1.05
4.93 +
------
5.98

You are getting 5.97999954, which is basically the answer as best represented by IEEE floating point, which is a binary format that cannot exactly express every decimal number. For instance, the rather common 0.1₁₀ has an infinite binary floating point representation .0001100110011...₂.

And a Wikipedia link, for good measure: http://en.wikipedia.org/wiki/Floating_point#Accuracy_problems ;)

edited Dec 15 '09 at 19:29

answered Dec 15 '09 at 19:23

jdmichal

10,984
4
43
42

Hey, I even learned something. Neither .1 nor .01 are representable in floating point. Which makes squaring .1 an interesting case. – jdmichal Dec 15 '09 at 19:30
I didn't even know we were having a war. I was just editing as requested. – jdmichal Dec 15 '09 at 19:33
Opimistic concurrency at it's worst :) Luckily there's a rollback feature. – Richard Szalay Dec 15 '09 at 19:49

score 1 · Answer 2 · answered Dec 15 '09 at 19:19

1

This is due to representing base-10 numbers in a base-2 system.

Floating point math on computers always does this. If you know the level of precision you'll need you should use the appropriate decimal representation.

answered Dec 15 '09 at 19:19

Ben S

68,394
30
171
212

score 0 · Answer 3 · answered Dec 15 '09 at 19:30

0

jdmichal is correct, but I'll add that, if you actually want this to add up correctly, you could use the Decimal type (including decimal literals).

answered Dec 15 '09 at 19:30

Steven Sudit

19,391
1
51
53

There are many options available to fix this, depending on the context of the problem. For instance, fixed precision (such as currency) allows separation of the decimal portion into an integral type. – jdmichal Dec 15 '09 at 19:37

Question with floating point addition

3 Answers3