`Double.MAX_VALUE-1< Double.MAX_VALUE` is wron...

# datascience

[holgerbrandl]

Double.MAX_VALUE-1< Double.MAX_VALUE is wrong on the JVM (because of limited precision) although it is conceptually true. But what's the largest Double for which the assumption holds also on the JVM?

[altavir]

It is a bit off topic for the channel. The relative precision of floating point number is one bit. At maximum precision it should be like a half of the value. So I have not tested it, but it should be like MAX_VALUE/2 or somewhere around.