The associativity of stream insertion operator is rtl, forgetting this fact sometimes cause to runtime or logical errors. for example:
1st-
int F()
{
static int internal_counter c=0;
return ++c;
}
in the main function:
//....here is main()
cout<<”1st=”<<F()<<”,2nd=”<<F()<<”,3rd=”<<F();
and the output is:
1st=3,2nd=2,3rd=1
that is different from what we expect at first look.
2nd- suppose that we have an implementation of stack data structure like this:
//
//... a Stack<DataType> class ……
//
Stack<int> st(10);
for(int i=1;i<11;i++)
开发者_JAVA百科st.push(i);
cout<<st.pop()<<endl<<st.pop()<<endl<<st.pop()<<endl<<st.pop()<<endl;
expected output is something like:
10
9
8
7
but we have:
7
8
9
10
There is no internal bug of << implementation but it can be so confusing... and finally[:-)] my question: is there any way to change associativity of an operator by overloading it?
do you think this could be not reverse? i mean is it possible to change order by modifying or changing an open source STL?
No there isn't. But I think you may be mixing up associativity with evaluation order. The only operators that specify an evalualtion order are &&, || and , (comma). When you say:
cout<<st.pop()<<endl<<st.pop()<<endl<<st.pop()<<endl<<st.pop()<<endl;
the compiler can evaluate sub-expressions such as st.pop() in any order it likes, which is what causes the unexpected output.
The only things that are right-associative are the assignment operators. See §5.4 to 5.18 of the standard. The <<
operators are evaluated left-to-right or the messages would be backward in grammar, not in content. The content is due to side effects, which are unordered in C++ except (as Neil mentions) for "short-circuit" && and ||, and comma.
To see how this is an order of evaluation issue and not an associativity issue, modify your code to this:
int a = st.pop();
int b = st.pop();
int c = st.pop();
cout << a << endl << b << endl << c << endl;
精彩评论