C++0x rvalue references and temporaries - c++

(I asked a variation of this question on comp.std.c++ but didn't get an answer.)
Why does the call to f(arg) in this code call the const ref overload of f?
void f(const std::string &); //less efficient
void f(std::string &&); //more efficient
void g(const char * arg)
{
f(arg);
}
My intuition says that the f(string &&) overload should be chosen, because arg needs to be converted to a temporary no matter what, and the temporary matches the rvalue reference better than the lvalue reference.
This is not what happens in GCC and MSVC (edit: Thanks Sumant: it doesn't happen in GCC 4.3-4.5). In at least G++ and MSVC, any lvalue does not bind to an rvalue reference argument, even if there is an intermediate temporary created. Indeed, if the const ref overload isn't present, the compilers diagnose an error. However, writing f(arg + 0) or f(std::string(arg)) does choose the rvalue reference overload as you would expect.
From my reading of the C++0x standard, it seems like the implicit conversion of a const char * to a string should be considered when considering if f(string &&) is viable, just as when passing a const lvalue ref arguments. Section 13.3 (overload resolution) doesn't differentiate between rvalue refs and const references in too many places. Also, it seems that the rule that prevents lvalues from binding to rvalue references (13.3.3.1.4/3) shouldn't apply if there's an intermediate temporary - after all, it's perfectly safe to move from the temporary.
Is this:
Me misreading/misunderstand the standard, where the implemented behavior is the intended behavior, and there's some good reason why my example should behave the way it does?
A mistake that the compiler vendors have somehow all made? Or a mistake based on common implementation strategies? Or a mistake in e.g. GCC (where this lvalue/rvalue reference binding rule was first implemented), that was copied by other vendors?
A defect in the standard, or an unintended consequence, or something that should be clarified?
EDIT: I have a follow-on question that is related: C++0x rvalue references - lvalues-rvalue binding

GCC is doing it wrong according the FCD. The FCD says at 8.5.3 about reference binding
If the reference is an lvalue reference and the initializer expression is an [lvalue / class type] ...
Otherwise, the reference shall be an lvalue reference to a non-volatile const type (i.e., cv1 shall be const), or the reference shall be an rvalue reference and the initializer expression shall be an rvalue or have a function type.
Your case for the call to the std::string && matches none of them, because the initializer is an lvalue. It doesn't get to the place to create a temporary rvalue, because that toplevel bullet already requires an rvalue.
Now, overload resolution doesn't directly use reference binding to see whether there exist an implicit conversion sequence. Instead, it says at 13.3.3.1.4/2
When a parameter of reference type is not bound directly to an argument expression, the conversion sequence is the one required to convert the argument expression to the underlying type of the reference according to 13.3.3.1.
Thus, overload resolution figures out a winner, even though that winner may actually not be able to bind to that argument. For example:
struct B { B(int) { /* ... */ } };
struct A { int bits: 1; };
void f(int&);
void f(B);
int main() { A a; f(a.bits); }
Reference binding at 8.5 forbids bitfields to bind to lvalue references. But overload resolution says that the conversion sequence is the one converting to int, thus succeeding even though when the call is made later, the call is ill-formed. Thus my bitfields example is ill-formed. If it was to choose the B version, it would have succeeded, but needed a user defined conversion.
However, there exist two exceptions for that rule. These are
Except for an implicit object parameter, for which see 13.3.1, a standard conversion sequence cannot be formed if it requires binding an lvalue reference to non-const to an rvalue or binding an rvalue reference to an lvalue.
Thus, the following call is valid:
struct B { B(int) { /* ... */ } };
struct A { int bits: 1; };
void f(int&); /* binding an lvalue ref to non-const to rvalue! */
void f(B);
int main() { A a; f(1); }
And thus, your example calls the const T& version
void f(const std::string &);
void f(std::string &&); // would bind to lvalue!
void g(const char * arg) { f(arg); }
However, if you say f(arg + 0), you create an rvalue, and thus the second function is viable.

It was a defect in the standard draft you read. This defect got in as a side effect of some eager editing to disallow binding of rvalue references to lvalues for safety reasons.
Your intuition is right. Of course, there is no harm in allowing an rvalue reference to refer to some unnamed temporary even if the initializer was an lvalue expression. After all, this is what rvalue references are for. The issue you observed has been fixed last year. The upcoming standard will mandate that the second overload will be picked in your example where the rvalue reference will refer to some temporary string object.
The rule fix made it into the draft n3225.pdf (2010-11-27):
[...]
Otherwise, the reference shall be an lvalue reference to a non-volatile const type (i.e., cv1 shall be const), or the reference shall be an rvalue reference and the initializer expression shall be an rvalue or have a function type. [...]
[...]
Otherwise, a temporary of [...] is created [...]
double&& rrd3 = i; // rrd3 refers to temporary with value 2.0
But N3225 seems to have missed to say what i is in this example. The latest draft N3290 contains these examples:
double d2 = 1.0;
double&& rrd2 = d2; // error: copying lvalue of related type
int i3 = 2;
double&& rrd3 = i3; // rrd3 refers to temporary with value 2.0
Since your MSVC version was released before this issue got fixed, it still handles rvalue references according to the old rules. The next MSVC version is expected to implement the new rvalue reference rules (dubbed "rvalue references 2.1" by MSVC developers) see link.

I did not see the behavior mentioned by Doug on g++. g++ 4.5 and 4.4.3 both call f(string &&) as expected but VS2010 calls f(const string &). Which g++ version are you using?

A lot of things in the current draft of the standard need clarification, if you ask me. And the compilers are still developing, so it's hard to trust their help.
It looks pretty clear that your intuition is right… temporaries of any kind are supposed to bind to rvalue references. For example, §3.10, the new "taxonomy" section, categorically defines temporaries as rvalues.
The problem may be that the RR argument specification is insufficient to invoke the creation of a temporary. §5.2.2/5: "Where a parameter is of const reference type a temporary object is introduced if needed." That sounds suspiciously exclusive.
Seems to slip through the cracks again at §13.3.3.1/6: (emphasis mine)
When the parameter type is not a reference, the implicit conversion sequence models a copy-initialization of the parameter from the argument expression. The implicit conversion sequence is the one required to convert the argument expression to a prvalue of the type of the parameter.
Note that copy-initialization string &&rr = "hello"; works fine in GCC.
EDIT: Actually the problem doesn't exist on my version of GCC. I'm still trying to figure out how the second standard conversion of the user-defined conversion sequence relates to forming an rvalue reference. (Is RR formation a conversion at all? Or is it dictated by scattered tidbits like 5.2.2/5?)

Take a look at this:
http://blogs.msdn.com/vcblog/archive/2009/02/03/rvalue-references-c-0x-features-in-vc10-part-2.aspx
rvalue references: overload resolution
It looks like your case is: "Lvalues strongly prefer binding to lvalue references".

I don't know if that has changed in the latest versions of the standard, but it used to say something like "if in doubt, don't use the rvalue reference". Probably for compatibility reasons.
If you want the move semantics, use f(std::move(arg)), that works with both compilers.

Related

r-value parameters in a function

I was wondering about a c++ behaviour when an r-value is passed among functions.
Look at this simple code:
#include <string>
void foo(std::string&& str) {
// Accept a rvalue of str
}
void bar(std::string&& str) {
// foo(str); // Does not compile. Compiler says cannot bind lvalue into rvalue.
foo(std::move(str)); // It feels like a re-casting into a r-value?
}
int main(int argc, char *argv[]) {
bar(std::string("c++_rvalue"));
return 0;
}
I know when I'm inside bar function I need to use move function in order to invoke foo function. My question now is why?
When I'm inside the bar function the variable str should already be an r-value, but the compiler acts like it is a l-value.
Can somebody quote some reference to the standard about this behaviour?
Thanks!
str is a rvalue reference, i.e. it is a reference only to rvalues. But it is still a reference, which is a lvalue. You can use str as a variable, which also implies that it is an lvalue, not a temporary rvalue.
An lvalue is, according to §3.10.1.1:
An lvalue (so called, historically, because lvalues could appear on the left-hand side of an assignment expression) designates a function or an object. [ Example: If E is an expression of pointer type, then *E is an lvalue expression referring to the object or function to which E points. As another example, the result of calling a function whose return type is an lvalue reference is an lvalue. —end example ]
And an rvalue is, according to §3.10.1.4:
An rvalue (so called, historically, because rvalues could appear on the right-hand side of an assignment
expression) is an xvalue, a temporary object (12.2) or subobject thereof, or a value that is not associated with an object.
Based on this, str is not a temporary object, and it is associated with an object (with the object called str), and so it is not an rvalue.
The example for the lvalue uses a pointer, but it is the same thing for references, and naturally for rvalue references (which are only a special type of references).
So, in your example, str is an lvalue, so you have to std::move it to call foo (which only accepts rvalues, not lvalues).
The "rvalue" in "rvalue reference" refers to the kind of value that the reference can bind to:
lvalue references can bind to lvalues
rvalue references can bind to rvalues
(+ a bit more)
That's all there's to it. Importantly, it does not refer to the value that get when you use the reference. Once you have a reference variable (any kind of reference!), the id-expression naming that variable is always an lvalue. Rvalues occur in the wild only as either temporary values, or as the values of function call expressions, or as the value of a cast expression, or as the result of decay or of this.
There's a certain analogy here with dereferencing a pointer: dereferencing a pointer is always an lvalue, no matter how that pointer was obtained: *p, *(p + 1), *f() are all lvalues. It doesn't matter how you came by the thing; once you have it, it's an lvalue.
Stepping back a bit, maybe the most interesting aspect of all this is that rvalue references are a mechanism to convert an rvalue into an lvalue. No such mechanism had existed prior to C++11 that produced mutable lvalues. While lvalue-to-rvalue conversion has been part of the language since its very beginnings, it took much longer to discover the need for rvalue-to-lvalue conversion.
My question now is why?
I'm adding another answer because I want to emphasize an answer to the "why".
Even though named rvalue references can bind to an rvalue, they are treated as lvalues when used. For example:
struct A {};
void h(const A&);
void h(A&&);
void g(const A&);
void g(A&&);
void f(A&& a)
{
g(a); // calls g(const A&)
h(a); // calls h(const A&)
}
Although an rvalue can bind to the a parameter of f(), once bound, a is now treated as an lvalue. In particular, calls to the overloaded functions g() and h() resolve to the const A& (lvalue) overloads. Treating a as an rvalue within f would lead to error prone code: First the "move version" of g() would be called, which would likely pilfer a, and then the pilfered a would be sent to the move overload of h().
Reference.

Do *non*-const references prolong the lives of temporaries?

Once upon a time, I assumed that code like this would fail:
const MyClass& obj = MyClass();
obj.DoSomething();
because the MyClass object would be destroyed at the end of its full-expression, leaving obj as a dangling reference. However, I learned (here) that this isn't true; the standard actually has a special provision that allows const references to keep temporaries alive until said references are destroyed themselves. But, it was emphasized, only const references have this power. Today I ran the code below in VS2012 as an experiment.
struct Foo
{
Foo() { std::cout << "ctor" << std::endl; }
~Foo() { std::cout << "dtor" << std::endl; }
};
void f()
{
Foo& f = Foo();
std::cout << "Hello world" << std::endl;
}
The output when calling f() was:
ctor
Hello world
dtor
So I had a look at the C++11 draft standard, and only found this (§ 12.2/4):
There are two contexts in which temporaries are destroyed at a
different point than the end of the full-expression. The first context [doesn't
apply]. The second context is when a reference is bound to a
temporary. The temporary to which the reference is bound or the
temporary that is the complete object of a subobject to which the
reference is bound persists for the lifetime of the reference.
The word const is conspicuously absent from the above. So; has this behavior been changed for C++11, was I wrong about the const thing to begin with, or does VS2012 have a bug and I just haven't found the relevant part of the standard?
The behavior hasn't changed, you just need to turn your warning level up to /W4. VisualStudio implements the lifetime extension rule even for non-const lvalue references as a compiler extension. In this context, binding an rvalue to the non-const reference behaves the same as if you were binding it to a const reference.
With /W4 you'd see this:
warning C4239: nonstandard extension used : 'initializing' : conversion from 'Foo' to 'Foo &'
1> A non-const reference may only be bound to an lvalue
The text disallowing binding of an rvalue to a non-const lvalue reference can be found in §8.5.3/5
— Otherwise, the reference shall be an lvalue reference to a non-volatile const type (i.e., cv1 shall be
const), or the reference shall be an rvalue reference.[ Example:
double& rd2 = 2.0; // error: not an lvalue and reference not const
int i = 2;
double& rd3 = i; // error: type mismatch and reference not const
—end example ]
The second half of the quoted statement is what allows binding of a temporary to an rvalue reference, as shown in litb's answer.
string &&s = string("hello");
This, combined with the lifetime extension rule in §12.2/5, means the lifetime of the temporary will now match the lifetime of the (rvalue) reference it is bound to.
The word const was never present in this section. The rule
has always been (from as long as I can remember) that
a temporary used to initialize a reference has its lifetime
extended to match that of the reference, regardless of the type
of the reference.
Sometime in the late 1980's (very pre-standard), C++ introduced
the rule that a temporary could not be used to initialize
a non-const reference. Initializing a non-const reference with
a temporary would still extend the lifetime (presumably), but
since you couldn't do it... Most compilers implemented
a transition period, in which such an initialization would only
emit a warning (and the lifetime was extended).
For some reason, when Microsoft finally decided to implement
C++ (some time in the early 1990's), they decided not to
implement the new rule, and allowed initialization of
a non-const reference with a temporary (without even a warning,
at a time when most other vendors were gradually turning the
warning into an error). And of course, the implemented the
usual lifetime rule.
Finally, in C++11, new types of references were introduced,
which allowed (or even required) initialization with
a temporary. The rule about the lifetime of temporaries hasn't
changed, though; a temporary which is used to initialize
a reference (regardless of the type of reference) has its
lifetime extended.
(With a few exceptions: I would not recommend using a temporary
to initialize a class member reference in an initialization
list.)
No, because rvalue references don't need to be const, so the Standard quote is correct
string &&s = string("hello");
The lifetime is still enlargened. The constraints that make the code invalid for non-const lvalue reference is at clause 8 (notice that it is not the right place to just add "const" and "rvalue reference" etc in the paragraph you quoted. You need an active rejection of such bindings, not just saying that the lifetime of such bindings are not enlarged because you would leave the binding itself still wellformed).

If temporaries are implicitly non-modifiable, how does this work?

I'm told that, in C++03, temporaries are implicitly non-modifiable.
However, the following compiles for me on GCC 4.3.4 (in C++03 mode):
cout << static_cast<stringstream&>(stringstream() << 3).str();
How is this compiling?
(I am not talking about the rules regarding temporaries binding to references.)
I'm told that, in C++03, temporaries are implicitly non-modifiable.
That is not correct. Temporaries are created, among other circumstances, by evaluating rvalues, and there are both non-const rvalues and const rvalues. The value category of an expression and the constness of the object it denotes are mostly orthogonal 1. Observe:
std::string foo();
const std::string bar();
Given the above function declarations, the expression foo() is a non-const rvalue whose evaluation creates a non-const temporary, and bar() is a const rvalue that creates a const temporary.
Note that you can call any member function on a non-const rvalue, allowing you to modify the object:
foo().append(" was created by foo") // okay, modifying a non-const temporary
bar().append(" was created by bar") // error, modifying a const temporary
Since operator= is a member function, you can even assign to non-const rvalues:
std::string("hello") = "world";
This should be enough evidence to convince you that temporaries are not implicitly const.
1: An exception are scalar rvalues such as 42. They are always non-const.
First, there's a difference between "modifying a temporary" and "modifying an object through an rvalue". I'll consider the latter, since the former is not really useful to discuss [1].
I found the following at 3.10/10 (3.10/5 in C++11):
An lvalue for an object is necessary
in order to modify the object except
that an rvalue of class type can also
be used to modify its referent under
certain circumstances. [Example: a
member function called for an object
(9.3) can modify the object. ]
So, rvalues are not const per-se but they are non-modifiable under all but some certain circumstances.
However, that a member function call can modify an rvalue would seem to indicate to me that the vast majority of cases for modifying an object through an rvalue are satisfied.
In particular, the assertion (in the original question I linked to) that (obj1+obj2).show() is not valid for non-const show() [ugh, why?!] was false.
So, the answer is (changing the question wording slightly for the conclusion) that rvalues, as accessed through member functions, are not inherently non-modifiable.
[1] - Notably, if you can obtain an lvalue to the temporary from the original rvalue, you can do whatever you like with it:
#include <cstring>
struct standard_layout {
standard_layout();
int i;
};
standard_layout* global;
standard_layout::standard_layout()
{
global = this;
}
void modifying_an_object_through_lvalue(standard_layout&&)
{
// Modifying through an *lvalue* here!
std::memset(global, 0, sizeof(standard_layout));
}
int main()
{
// we pass a temporary, but we only modify it through
// an lvalue, which is fine
modifying_an_object_through_lvalue(standard_layout{});
}
(Thanks to Luc Danton for the code!)

Why const for implicit conversion?

After extensive reading of ISO/IEC 14882, Programming language – C++ I'm still unsure why const is needed for implicit conversion to a user-defined type with a single argument constructor like the following
#include <iostream>
class X {
public:
X( int value ) {
printf("constructor initialized with %i",value);
}
}
void implicit_conversion_func( const X& value ) {
//produces "constructor initialized with 99"
}
int main (int argc, char * const argv[]) {
implicit_conversion_func(99);
}
Starting with section 4 line 3
An expression e can be implicitly converted to a type T if and only if the declaration T t=e; is well-formed, for some invented temporary variable t (8.5). Certain language constructs require that an expression be converted to a Boolean value. An expression e appearing in such a context is said to be contextually converted to bool and is well-formed if and only if the declaration bool t(e); is well-formed, for some invented temporary variable t (8.5). The effect of either implicit conversion is the same as performing the declaration and initialization and then using the temporary variable as the result of the conversion. The result is an lvalue if T is an lvalue reference type (8.3.2), and an rvalue otherwise. The expression e is used as an lvalue if and only if the initialization uses it as an lvalue.
Following that I found the section on initializers related to user-defined types in 8.5 line 6
If a program calls for the default initialization of an object of a const-qualified type T, T shall be a class type with a user-provided default constructor.
Finally I ended up at 12.3 line 2 about user-defined conversions which states
User-defined conversions are applied only where they are unambiguous (10.2, 12.3.2).
Needless to say, 10.2 and 12.3.2 didn't answer my question.
Can someone shed some light on what effect const has on implicit conversions?
Does the use of const make the conversion "unambiguous" per 12.3 line 2?
Does const somehow affect lvalue vs. rvalue talked about in section 4?
It doesn't really have much to do with the conversion being implicit. Moreover, it doesn't really have much to do with conversions. It is really about rvalues vs. lvalues.
When you convert 99 to type X, the result is an rvalue. In C++ results of conversions are always rvalues (unless you convert to reference type). It is illegal in C++ to attach non-const references to rvalues.
For example, this code will not compile
X& r = X(99); // ERROR
because it attempts to attach a non-const reference to an rvalue. On the other hand, this code is fine
const X& cr = X(99); // OK
because it is perfectly OK to attach a const reference to an rvalue.
The same thing happens in your code as well. The fact that it involves an implicit conversion is kinda beside the point. You can replace implicit conversion with an explicit
one
implicit_conversion_func(X(99));
and end up with the same situation: with const it compiles, without const it doesn't.
Again, the only role the conversion (explicit or implicit) plays here is that it helps us to produce an rvalue. In general, you can produce an rvalue in some other way and run into the same issue
int &ir = 3 + 2; // ERROR
const int &cir = 3 + 2; // OK
Per section 5.2.2 paragraph 5, when an argument to a function is of const reference type, a temporary variable is automatically introduced if needed. In your example, the rvalue result of X(99) has to be put into a temporary variable so that that variable can be passed by const reference to implicit_conversion_func.

non-class rvalues always have cv-unqualified types

§3.10 section 9 says "non-class rvalues always have cv-unqualified types". That made me wonder...
int foo()
{
return 5;
}
const int bar()
{
return 5;
}
void pass_int(int&& i)
{
std::cout << "rvalue\n";
}
void pass_int(const int&& i)
{
std::cout << "const rvalue\n";
}
int main()
{
pass_int(foo()); // prints "rvalue"
pass_int(bar()); // prints "const rvalue"
}
According to the standard, there is no such thing as a const rvalue for non-class types, yet bar() prefers to bind to const int&&. Is this a compiler bug?
EDIT: Apparently, this is also a const rvalue :)
EDIT: This issue seems to be fixed in g++ 4.5.0, both lines print "rvalue" now.
The committee already seems to be aware that there's a problem in this part of the standard. CWG issue 690 talks about a somewhat similar problem with exactly the same part of the standard (in the "additional note" from September, 2009). I'd guess new language will be drafted for that part of the standard soon.
Edit: I've just submitted a post on comp.std.c++, noting the problem and suggesting new wording for the relevant piece of the standard. Unfortunately, being a moderated newsgroup, nearly everybody will probably have forgotten this question by the time it makes it through the approval queue there.
Good point. I guess there are two things to look at: 1) as you pointed out the non-class rvalue thingsy and 2) how overload resolution works:
The selection criteria for the best
function are the number of arguments,
how well the arguments match the
parameter-type-list of the candidate
function, [...]
I haven't seen anything in the standard that tells me non-class rvalues are treated specially during overload resolution.
Your question is covered in the draft of the standard I have though (N-4411) somewhat:
What does come into play is however a parallel reading of reference binding, implicit conversion sequences, references, and overload resolution in general:
13.3.3.1.4 Reference binding
2 When a parameter of reference type
is not bound directly to an argument
expression, the conversion sequence
is the one required to convert the argument expression to the underlying
type of the reference according
to 13.3.3.1.
and
13.3.3.2 Ranking implicit conversion sequences
3 Two implicit conversion sequences of
the same form are indistinguishable
conversion sequences unless one of the
following rules applies:
— Standard conversion sequence S1 is a better conversion sequence than
standard
conversion sequence S2 if
— S1 and S2 are reference bindings (8.5.3) and neither refers to an
implicit object parameter of a
nonstatic
member function declared without a ref-qualifier, and either S1 binds an
lvalue reference
to an lvalue and S2 binds an rvalue reference or S1 binds an rvalue
reference to an rvalue and S2
binds an lvalue reference.
[ Example:
int i;
int f();
int g(const int&);
int g(const int&&);
int j = g(i); // calls g(const int&)
int k = g(f()); // calls g(const int&&)