Well, here is another attempt at a formal definition.
If f(x) is an expanding transformation, then for each x <= y , f(x) <= f(y) . In other words, expanding conversions preserve a partial order of values.
That which has not lost information about the total value may mean.
The loss of accuracy in this structure means that for some x < y , f(x) = f(y) .
The only problem with this definition is the (float)Double.MAX_VALUE resulting from Float.POSITIVE_INFINITY , which technically fulfills the above criteria, but should not really be considered not to lose information about the total value.
source share