Way back in chapter 1, we mentioned l-values and r-values, and then told you not to worry that much about them. That was fair advice prior to C++11. But understanding move semantics in C++11 requires a re-examination of the topic. So let’s do that now.
L-values and r-values
Despite having the word “value” in their names, l-values and r-values are actually not properties of values, but rather, properties of expressions.
Every expression in C++ has two properties: a type (which is used for type checking), and a value category (which is used for certain kinds of syntax checking, such as whether the result of the expression can be assigned to). In C++03 and earlier, l-values and r-values were the only two value categories available.
The actual definition of which expressions are l-values and which are r-values is surprisingly complicated, so we’ll take a simplified view of the subject that will largely suffice for our purposes.
It’s simplest to think of an l-value (also called a locator value) as a function or an object (or an expression that evaluates to a function or object). All l-values have assigned memory addresses.
When l-values were originally defined, they were defined as “values that are suitable to be on the left-hand side of an assignment expression”. However, later, the const keyword was added to the language, and l-values were split into two sub-categories: modifiable l-values, which can be changed, and non-modifiable l-values, which are const.
It’s simplest to think of an r-value as “everything that is not an l-value”. This notably includes literals (e.g. 5), temporary values (e.g. x+1), and anonymous objects (e.g. Fraction(5, 2)). r-values are typically evaluated for their values, have expression scope (they die at the end of the expression they are in), and cannot be assigned to. This non-assignment rule makes sense, because assigning a value applies a side-effect to the object. Since r-values have expression scope, if we were to assign a value to an r-value, then the r-value would either go out of scope before we had a chance to use the assigned value in the next expression (which makes the assignment useless) or we’d have to use a variable with a side effect applied more than once in an expression (which by now you should know causes undefined behavior!).
In order to support move semantics, C++11 introduces 3 new value categories: pr-values, x-values, and gl-values. We will largely ignore these since understanding them isn’t necessary to learn about or use move semantics effectively. If you’re interested, cppreference.com has an extensive list of expressions that qualify for each of the various value categories, as well as more detail about them.
L-value references
Prior to C++11, only one type of reference existed in C++, and so it was just called a “reference”. However, in C++11, it’s sometimes called an l-value reference. L-value references can only be initialized with modifiable l-values.
| L-value reference | Can be initialized with | Can modify |
|---|---|---|
| Modifiable l-values | Yes | Yes |
| Non-modifiable l-values | No | No |
| R-values | No | No |
L-value references to const objects can be initialized with l-values and r-values alike. However, those values can’t be modified.
| L-value reference to const | Can be initialized with | Can modify |
|---|---|---|
| Modifiable l-values | Yes | No |
| Non-modifiable l-values | Yes | No |
| R-values | Yes | No |
L-value references to const objects are particularly useful because they allow us to pass any type of argument (l-value or r-value) into a function without making a copy of the argument.
R-value references
C++11 adds a new type of reference called an r-value reference. An r-value reference is a reference that is designed to be initialized with an r-value (only). While an l-value reference is created using a single ampersand, an r-value reference is created using a double ampersand:
|
1 2 3 |
int x = 5; int &lref = x; // l-value reference initialized with l-value x int &&rref = 5; // r-value reference initialized with r-value 5 |
R-values references cannot be initialized with l-values.
| R-value reference | Can be initialized with | Can modify |
|---|---|---|
| Modifiable l-values | No | No |
| Non-modifiable l-values | No | No |
| R-values | Yes | Yes |
| R-value reference to const | Can be initialized with | Can modify |
|---|---|---|
| Modifiable l-values | No | No |
| Non-modifiable l-values | No | No |
| R-values | Yes | No |
R-value references have two properties that are useful. First, r-value references extend the lifespan of the object they are initialized with to the lifespan of the r-value reference (l-value references to const objects can do this too). Second, non-const r-value references allow you to modify the r-value!
Let’s take a look at some examples:
|
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 |
#include <iostream> class Fraction { private: int m_numerator; int m_denominator; public: Fraction(int numerator = 0, int denominator = 1) : m_numerator(numerator), m_denominator(denominator) { } friend std::ostream& operator<<(std::ostream& out, const Fraction &f1) { out << f1.m_numerator << "/" << f1.m_denominator; return out; } }; int main() { Fraction &&rref = Fraction(3, 5); // r-value reference to temporary Fraction std::cout << rref << '\n'; return 0; } // rref (and the temporary Fraction) goes out of scope here |
This program prints:
3/5
As an anonymous object, Fraction(3, 5) would normally go out of scope at the end of the expression in which it is defined. However, since we’re initializing an r-value reference with it, its duration is extended until the end of the block. We can then use that r-value reference to print the Fraction’s value.
Now let’s take a look at a less intuitive example:
|
1 2 3 4 5 6 7 8 9 10 |
#include <iostream> int main() { int &&rref = 5; // because we're initializing an r-value reference with a literal, a temporary with value 5 is created here rref = 10; std::cout << rref; return 0; } |
This program prints:
10
While it may seem weird to initialize an r-value reference with a literal value and then be able to change that value, when initializing an r-value with a literal, a temporary is constructed from the literal so that the reference is referencing a temporary object, not a literal value.
R-value references are not very often used in either of the manners illustrated above.
R-value references as function parameters
R-value references are more often used as function parameters. This is most useful for function overloads when you want to have different behavior for l-value and r-value arguments.
|
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 |
void fun(const int &lref) // l-value arguments will select this function { std::cout << "l-value reference to const\n"; } void fun(int &&rref) // r-value argument will select this function { std::cout << "r-value reference\n"; } int main() { int x = 5; fun(x); // l-value argument calls l-value version of function fun(5); // r-value argument calls r-value version of function return 0; } |
This prints:
l-value reference to const r-value reference
As you can see, when passed an l-value, the overloaded function resolved to the version with the l-value reference. When passed an r-value, the overloaded function resolved to the version with the r-value reference (this is considered a better match than a l-value reference to const).
Why would you ever want to do this? We’ll discuss this in more detail in the next lesson. Needless to say, it’s an important part of move semantics.
Returning an r-value reference
You should almost never return an r-value reference, for the same reason you should almost never return an l-value reference. In most cases, you’ll end up returning a hanging reference when the referenced object goes out of scope at the end of the function.
Quiz time
1) State which of the following lettered statements will not compile:
|
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 |
int main() { int x; // l-value references int &ref1 = x; // A int &ref2 = 5; // B const int &ref3 = x; // C const int &ref4 = 5; // D // r-value references int &&ref5 = x; // E int &&ref6 = 5; // F const int &&ref7 = x; // G const int &&ref8 = 5; // H return 0; } |
 15.3 -- Move constructors and move assignment
|
 Index
|
 15.1 -- Intro to smart pointers and move semantics
|
What do you mean when "However, since we’re initializing an r-value reference with it, its duration is extended until the end of the block." (referring to this code -> Fraction &&rref = Fraction(3, 5);)
Even if you initialize Fraction rref = Fraction(3, 5), or Fraction& rref = Fraction(3, 5) doesn't it call the same constructor and extends its lifetime as well?
Creates a temporary fraction and then uses the copy constructor to initialize rref. No extension necessary since the temporary fraction can die after being copied.
is illegal because you can't have a non-const l-value reference pointing to an r-value.
Both of these extend the duration of the temporary Fraction.
Hi Alex,
As you can see, the r-value and the local class variable both produce the same outputs.
So what is the purpose of using r-value over the local class variable?
Thanks,
TC
There is zero benefit to using references in this case. This is equally true of using a const l-value reference in said case.
As the lesson notes, r-value references are more typically used as function parameters when you want to differentiate function behaviors for l-value and r-value arguments.
Thank you Alex
Hey Alex, quick question about these two sentences:
"First, r-value references extend the lifespan of the object they are initialized with to the lifespan of the r-value reference. l-value references to const objects can do this too, but it’s far more useful for r-value references since r-values have expression scope otherwise."
What exactly do you mean by "since r-values have expression scope otherwise"? If an object with expression scope is initialized with a l-value ref to a const object, then its lifespan is also extended past expression scope, just like with a r-value ref.
Basically, what's confusing me is that with respect to extending lifespans, I don't see the advantage that r-value refs have over l-value refs to const objects. On the other hand, the fact that r-value refs can modify their corresponding r-values is a clear advantage. Thanks in advance 🙂
By "expression scope", I mean that r-values are temporary values that typically only live until the end of the expression they're in.
I take your second point -- there's no real advantage of r-values references over const l-value references in terms of lifespan extension. I've removed the latter part of the original sentence accordingly.
Thanks for pointing that out!
"You should almost never return an r-value reference, for the same reason you should almost never return an l-value reference. In most cases, you’ll end up returning a hanging reference when the referenced object goes out of scope at the end of the function."
But I thought that "r-value references extend the lifespan of the object they are initialized with to the lifespan of the r-value reference." So, shouldn't returning an r-value reference extend the lifespan of the r-value to the control of the function caller code that is receiving the return value?
Great question, and this is a confusing point.
Basically, if you do this:
What happens is: inside getFoo() a temporary Foo is constructed. Since this function returns by value, a copy of the Foo is returned, and the original Foo is destroyed. That returned Foo then has its lifetime extended by being bound to the rvalue reference in main.
Now, look at this:
In this example, inside getFoo() a temporary Foo is constructed. At the end of the function, this Foo gets destroyed, and an r-value reference to the destroyed Foo is returned. Thus ref is left as a dangling pointer.
In the first case, the lifetime can be extended before the Foo is destroyed. In the second case, the Foo gets destroyed before its lifetime can be extended.
Good article!!!
Alex, could you plz explain why R-value references for 'user-defined data types' are modifiable but not for 'built-in data types'?
Hi skv!
Mind giving an example? Alex showed that r-value references to built-in data types can be modified.
Prints 10
> should almost never return an l-value reference
shouldn't this be:
> should never return a local l-value refernece
Both are true. You should never return a reference to a local l-value, as it will go out of scope and leave you with a hanging reference. You should almost never return an l-value reference period, as there are only a few cases where it really makes sense to do so. The main time I see this occur is when overloading operators, when we want to return either the implicit object (or std::ostream) by reference so we can chain operators.
Thanks for explaining!
Hey Alex, I have a question, why don't you mention the knowledge about 'universal reference' and 'perfect forwarding' in this chapter? I think this part is quite related. But I don't know if they are not important at all.
Thank you very much for the wonderful tutorial. Whenever I seek for something I come to your tutorial firstly to find the answer, it is so clearly explained and solved me many problems. That's also the reason why I'm curious about my question above.
I didn't include material on universal (forwarding) references or perfect forwarding because these are slightly more advanced techniques and I simply haven't had time to write those articles. I'd like to get to them some day.
I've found a very good article about forward declaration Alex, maybe it can help you someday to write something about it , and some other readers interested in it.
[link]http://thbecker.net/articles/rvalue_references/section_01.html.
Hi alex.thank you for this tutorial.
Can you introduce me a tutorial to learn Socket programming in c++?
I'm not aware of whether any good socket programming tutorials exist or not. Sorry!
Hello, Alex,
I added a r-value reference y to your example and then called fun with it:
When I run the program, this is what I get:
l-value reference to const
r-value reference
l-value reference to const
I don't understand - why fun(y) calls l-value version?
Because a named r-value reference is actually an l-value. It has a name, and persists beyond an expression.
Instead of “everything this is not an l-value” you probably intended to write
“everything that is not an l-value”.
What is the difference between const reference and lvalue reference to const objects? Can we use the second one for the copy constructor? The reason that I am asking is that following code gives the error "member function already defined"
PS: Thanks Alex for this excellent tutorial. I almost done with all!!!
A "const reference" is just a shorthand name for a "l-value reference to const object".
"const Foo&" and "Foo const&" are the same thing to the compiler.
hello alex,
in the quiz, why the //D will compile
Const l-value references can be initialized with either l-value or r-values.
Hi Alex,
1) In the Fraction class: r-reference should have '&&' not '&'
2) Quiz: "State whether ... will compile or not!" or "State which .. won't compile!"
Fixed and fixed! Thanks for the suggestions!