The "+ -" operation does not produce a compile time error. Effectively it, negates the second operand, to produce x - y. This makes some sense because we often say . Still, I can't find this explicitly stated in the spec. Does anyone know where this behavior is given (either directly or more likely indirectly)? (We came across this when a developer made a typo and got the wrong numerical result. We were surprised, at first, because we would have figured it would cause a compile time error.)
Hi, z=x + - y; Since the precedence of + & - are same therefore compiler will check their associativity which is left to right & take - as a unary operator.You can say it, z = x + (- y); Regards, Hassan.
Always Belive On Logic!!
Jim Yingst
Wanderer
Sheriff
Joined: Jan 30, 2000
Posts: 18670
posted
0
As Hassan says, the - is a unary operator. However the JLS doesn't say anything about associativity - instead it laboriously defines the grammar of Java using grammar notation, spread throughout the spec. Concepts like operator precedence and associativity are not explicitly defined the the spec; rather, they are implicit in the grammatical definitions found in the spec. To parse the statement <code><pre> z = x + - y;</pre></code> we look first to JSL 14.8: <pre> ExpressionStatement: StatementExpression ; StatementExpression: Assignment PreIncrementExpression PreDecrementExpression PostIncrementExpression PostDecrementExpression MethodInvocation ClassInstanceCreationExpression </pre> This means that an ExpressionStatement is defined as a StatementExpression followed by a semicolon, and a StatementExpression may consist of a single Assignment (among other possibilities). Now from section JLS 15.26: <pre> Assignment: LeftHandSide AssignmentOperator AssignmentExpression LeftHandSide: ExpressionName FieldAccess ArrayAccess AssignmentOperator: one of = *= /= %= += -= <<= >>= >>>= &= ^= |= </pre> The LeftHandSide is the ExpressionName "z". The AssimentOperator is "=". And the AssignmentExpression is "x + - y". Specifically, it must be a ConditionalExpression, since we've already parsed the "=" which made the overall expression an Assignment. Think of the name as meaning "an expression which might have a conditional" rather than "an expression which does have a conditional" and it will make a little more sense - the names can be misleading. From JLS 15.25: <pre> ConditionalExpression: ConditionalOrExpression ConditionalOrExpression ? Expression : ConditionalExpression </pre> We don't have a '?' or ':', so "x + - y" must be a ConditionalOrExpression. We will descend through multiple expression types from here, in order (JLS sections 15.24 - 15.19): <pre> ConditionalOrExpression ConditionalAndExpression InclusiveOrExpression ExclusiveOrExpression AndExpression EqualityExpression RelationalExpression ShiftExpression </pre> This eventually leads to AdditiveExpression in JLS 15.18: <pre> AdditiveExpression: MultiplicativeExpression AdditiveExpression + MultiplicativeExpression AdditiveExpression - MultiplicativeExpression </pre> Suddenly, we have two possible matches: AdditiveExpression (x) + MultiplicativeExpression (- y) or AdditiveExpression (x -) + MultiplicativeExpression (y) However, if you continue to descend down the definitions, there will never be a way of parsing "x -" as an AdditiveExpression. There is however a way to parse "- y" as a MultiplicativeExpression in JLS 15.17: <pre> MultiplicativeExpression: UnaryExpression MultiplicativeExpression * UnaryExpression MultiplicativeExpression / UnaryExpression MultiplicativeExpression % UnaryExpression </pre> And JLS 15.15: <pre> UnaryExpression: PreIncrementExpression PreDecrementExpression + UnaryExpression - UnaryExpression UnaryExpressionNotPlusMinus </pre> You can further trace the definitions of "x" and "y" to: <pre>UnaryExpressionNotPlusMinus PostfixExpression ExpressionName</pre> I know this is tedious, and I omitted a number of steps, but hopefully it gives you an understanding of how the official grammar works. Enjoy.
[This message has been edited by Jim Yingst (edited May 07, 2001).]
"I'm not back." - Bill Harding, Twister
I agree. Here's the link: http://ej-technologies/jprofiler - if it wasn't for jprofiler, we would need to
run our stuff on 16 servers instead of 3.