It turns out it is possible, although that surprised me. Section 3.10.2 of the JLS gives the structure of floating point literals, including HexadecimalFloatingPointLiteral.
public class Test {
public static void main(String[] args) {
double d1 = 0xAAAAAAAAAAAAAAAAAAp0d;
double d2 = 0x1.8p1d;
System.out.println(d1); // A very big number
System.out.println(d2); // 24 = 1.5 * 2^1
}
}
The p
is required as part of the binary exponent – the value after the p
is the number of bits to shift the value left. Other examples:
0x1.4p0d => 1.25 (binary 0.01 shifted 0 bits)
0x8p-4d => 0.5 (binary 1000 shifted *right* 4 bits)