void operator=

[Per]

Hi!
I have a question about the operator=
I have seen two versions:

1.
void operator= (const SomeClass& C) {
....
}

2.
SomeClass& operator= (const SomeClass& C) {
....
return *this;
}

What is the difference and do one gain any performance using version
1. Is version 1 working with STL?

[Ivan Vecerina]

"Per" <pe*********@gmail.com> wrote in message
news:f6**************************@posting.google.c om...

Hi!
I have a question about the operator=
I have seen two versions:

1.
void operator= (const SomeClass& C) {
....
}

2.
SomeClass& operator= (const SomeClass& C) {
....
return *this;
} This is the correct form, equivalent to the built-in versions
of the operator.
See "Effective C++" item 15.
What is the difference and do one gain any performance
using version 1. A measurable gain is very unlikely, and it is even quite
possible that there would be no gain at all in code size
or execution speed.
Is version 1 working with STL?

It is not guaranteed to.
An STL implementation can legally assign objects as follows:
a = b = c;

hth,
Ivan
--
http://ivan.vecerina.com/contact/?subject=NG_POST <- email contact form
Brainbench MVP for C++ <> http://www.brainbench.com

[marieddu78]

In the first case you can't do:
SomeClass a,b,c;
...
a=b=c;
In the second case you can!

[Ron Natalie]

Per wrote:

What is the difference and do one gain any performance using version
1. Is version 1 working with STL?

Both are legal and probably have the same performance.
The problem with #1 is that it does't emulate the builtin operator=
which is generally regarded as bad style.

SomeClass a, b, c;
a = b = c; // won't compile with a void returning op=

[Swampmonster]

Yeh, like the others already stated, the performance gain will be
minimal to non-existend - depending on the compiler and whether it
inlines the function or not.
Therefor no reason I can see not to return the ref.