69 lines
2.3 KiB
Plaintext
69 lines
2.3 KiB
Plaintext
== Why is this an issue?
|
|
|
|
Floating point numbers in C# (and in most other programming languages) are not precise. They are a binary approximation of the actual value. This means that even if two floating point numbers appear to be equal, they might not be due to the tiny differences in their binary representation.
|
|
|
|
Even simple floating point assignments are not simple:
|
|
|
|
[source,csharp]
|
|
----
|
|
float f = 0.100000001f; // 0.1
|
|
double d = 0.10000000000000001; // 0.1
|
|
----
|
|
(Note: Results may vary based on the compiler and its settings)
|
|
|
|
This issue is further compounded by the https://en.wikipedia.org/wiki/Associative_property[non-associative] nature of floating point arithmetic.
|
|
The order in which operations are performed can affect the outcome due to the rounding that occurs at each step. Consequently, the outcome of a series of mathematical operations can vary based on the order of operations.
|
|
|
|
As a result, using the equality (`==`) or inequality (`!=`) operators with `float` or `double` values is typically a mistake, as it can lead to unexpected behavior.
|
|
|
|
== How to fix it
|
|
|
|
Consider using a small tolerance value to check if the numbers are "close enough" to be considered equal. This tolerance value, often called `epsilon`, should be chosen based on the specifics of your program.
|
|
|
|
=== Code examples
|
|
|
|
==== Noncompliant code example
|
|
|
|
[source,csharp,diff-id=1,diff-type=noncompliant]
|
|
----
|
|
float myNumber = 3.146f;
|
|
|
|
if (myNumber == 3.146f) // Noncompliant: due to floating point imprecision, this will likely be false
|
|
{
|
|
// ...
|
|
}
|
|
|
|
if (myNumber < 4 || myNumber > 4) // Noncompliant: indirect inequality test
|
|
{
|
|
// ...
|
|
}
|
|
----
|
|
|
|
==== Compliant solution
|
|
|
|
[source,csharp,diff-id=1,diff-type=compliant]
|
|
----
|
|
float myNumber = 3.146f;
|
|
float epsilon = 0.0001f; // or some other small value
|
|
|
|
if (Math.Abs(myNumber - 3.146f) < epsilon)
|
|
{
|
|
// ...
|
|
}
|
|
|
|
if (myNumber <= 4 - epsilon || myNumber >= 4 + epsilon)
|
|
{
|
|
// ...
|
|
}
|
|
----
|
|
|
|
== Resources
|
|
|
|
=== Documentation
|
|
|
|
* https://docs.oracle.com/cd/E19957-01/806-3568/ncg_goldberg.html[Floating-Point Arithmetic Complexities] by David Goldberg
|
|
* Microsoft Learn - https://learn.microsoft.com/en-us/dotnet/csharp/language-reference/builtin-types/floating-point-numeric-types#comparing-floating-point-numbers[Floating-point numeric types]
|
|
* Wikipedia - https://en.wikipedia.org/wiki/Associative_property[Associative property]
|
|
|
|
include::../rspecator.adoc[]
|