The document of sizes() says tha "Note that this function does not get the same result for integral or floating scalar types". What does it mean here? Does it mean that for integral type like AlignedBox1i, the sizes() would be max() - min() + 1 instead of max() - min()? But the result returned is still max() - min().

Thanks.

Good question, I've no idea. As you said, it is a shortcut for max() - min() regardless of the scalar type.