rspec/rules/S2688/java/rule.adoc

When working with ``++float++`` or ``++double++`` primitive types, it may be required to deal with ``++NaN++`` (Not a Number) values. When tested against itself, ``++NaN++`` will always answer ``++false++`` as long as the primitive wrapper type is not used. When the wrapper is used, it will always be ``++true++``. This property is illustrated in the code snipped below.


----
double d = getValue();
if (d == d) { // false for primitive 'double' when NaN, and true for any non-NaN values
  doSomething();
}

Double bigD = getValue();
if (bigD == bigD) { // always true for wrapper type 'Double' when NaN, AND with any other Double value
  doSomething(); 
}
----

In order to remove any ambiguity, this rule raises an issue every time an equality test is used with ``++double++``, ``++Double++``, ``++float++`` or ``++Float++``, when both sides of the test are the same variable. The ``++isNaN(...)++`` methods from ``++Double++`` and ``++Float++`` should be preferred.

== Noncompliant Code Example

----
double x = getValue();
if (x == x) { // Noncompliant
  doSomething();
}
if (x == Double.NaN) { // Noncompliant
  doSomething();
}
----

== Compliant Solution

----
double x = getValue();
if (Double.isNaN(x)) { // compliant
  doSomething();
}
----

include::../see.adoc[]