• BCS <ao pathlink.com> May 21 2009
• Frits van Bommel <fvbommel REMwOVExCAPSs.nl> May 21 2009
• BCS <ao pathlink.com> May 21 2009

BCS <ao pathlink.com> writes:

Are there any cases where the following cases both compile but are not
identical?

A a;
B b;

a = b;
a.Foo();

//// and

A a;
B b;

a = b;
b.Foo();

The reason I ask is I'm wondering if making the type (and value) of an
assignment 
expression the right hand side rather than the left hand side would silently 
break code. It would be handy for cases like this because it avoids the need 
for a temp variable:

class C { }
class D : C { int i();}

D newD();

C c;

void main()
{
   (c = newD()).i();
}

Frits van Bommel <fvbommel REMwOVExCAPSs.nl> writes:

BCS wrote:
 Are there any cases where the following cases both compile but are not 
 identical?
 
 A a;
 B b;
 
 a = b;
 a.Foo();
 
 //// and
 
 A a;
 B b;
 
 a = b;
 b.Foo();


struct A {
     int i;
     void Foo() { i = 42; }
}
alias A B;

The first case will set a.i to 42, the second will set b.i.

And with opAssign + different struct types they'd be calling two completely 
different Foo()s.

Then there's array types + global function Foo(ref A), union types (similar to 
struct types), class types + static Foo()s (different ones for A and B), ...

BCS <ao pathlink.com> writes:

Reply to Frits,

 BCS wrote:
 
 Are there any cases where the following cases both compile but are
 not identical?
 
 A a;
 B b;
 a = b;
 a.Foo();
 //// and
 
 A a;
 B b;
 a = b;
 b.Foo();


 int i;
 void Foo() { i = 42; }
 }
 alias A B;
 
 The first case will set a.i to 42, the second will set b.i.
 


I wasn't looking for side effect cases  but I guess that is correct.

 And with opAssign + different struct types they'd be calling two
 completely different Foo()s.
 


Oh, forgot opAssign (darn)

 Then there's array types + global function Foo(ref A), union types
 (similar to struct types), class types + static Foo()s (different ones
 for A and B), ...
 


ok looks like to much of a mess to make it work