I had a rule in my head for implicit type conversions of primitives: if one operand is "wider" than the other, "promote" the smaller one. For example: given
int narrow = 1;
long wide = Long.MAX_VALUE;
then the expression
wide - narrow is equivalent to
wide - (long)narrow and the result is
long.
WRONG! Well, partly. The example above is correct, but consider this:
byte narrow = (byte)1;
short wide = Short.MAX_VALUE;
The expression
wide - narrow is equivalent to
(int)wide - (int)narrow and the result is
int. But I didn't discover that until I'd had a lot of fun with "possible loss of precision" errors from the compiler.
There's
rules for binary operators and
rules for unary operators.
