Java Double NaN and signed zero: primitive and wrapper comparisons differ

Know which equality a collection uses

At the primitive level NaN is unequal to itself, while Double.equals treats canonical NaN values as equal. A HashSet of boxed doubles therefore holds one NaN key. Signed positive and negative zero compare equal with ==, but Double.equals distinguishes them. This matters for keyed metrics and deduplication.

The fixture prints both cases. Finite-value validation can reject NaN before it reaches a map or arithmetic pipeline when the domain has no meaning for it.

Define a metric-key policy

If zero signs should collapse, normalize them before using Double as a key. If NaN is invalid input, reject it rather than relying on wrapper equality. A comparator and equals policy must also agree with the collection being used.

Working program

import java.util.HashSet;
import java.util.Set;

public class SensorReadingIdentity {
    public static void main(String[] args) {
        double missingReading = Double.NaN;
        System.out.println("primitive NaN=" + (missingReading == missingReading));
        Set<Double> unique = new HashSet<>();
        unique.add(Double.NaN);
        unique.add(Double.NaN);
        System.out.println("boxed NaN count=" + unique.size());
        System.out.println("primitive zero=" + (-0.0 == 0.0));
        System.out.println("boxed zero=" + Double.valueOf(-0.0).equals(0.0));
    }
}

Output

primitive NaN=false
boxed NaN count=1
primitive zero=true
boxed zero=false

Cost and ownership

Each comparison is constant work. HashSet adds storage for boxed keys, and its identity semantics follow Double.equals rather than primitive ==. The correct policy depends on whether signed zero and NaN carry domain meaning.

Common Mistakes

• Do not assume boxed Double equality matches primitive == for NaN or signed zero.
• Do not store NaN as a normal measurement if the domain requires finite values.
• Do not use floating-point keys without defining normalization.

Read next

double finite validation, Java HashSet: deduplication and equality, Java comparators: tie-breakers, overflow and sorted-key identity, floating sum order.