Should Pass-by-Value Be the Default for Large Objects in Modern C ?

Rethinking Pass-by-Value in C 11

In the world of C , pass-by-value has traditionally been discouraged for large objects due to performance concerns. However, the introduction of Rvalue references and move constructors in C 11 opens up new possibilities for passing objects efficiently.

Is Pass-by-Value Now the Best Default?

According to industry expert Dave Abrahams, it's reasonable to pass large objects like std::vector and std::string by value if you need to make a copy within the function body. This allows the compiler to perform optimization, eliminating the need for explicit copying by the programmer.

Advantages:

• Only minimal work is done for both lvalue and prvalue arguments.
• Simplifies the caller's syntax, allowing pass-by-reference, pass-by-prvalue, and pass-by-xvalue.

Custom Objects:

For custom objects, pass-by-reference to const is still a viable option, as it provides flexibility and minimizes the risk of unintentionally modifying the original object.

Best Practices:

• Pass objects by value if you need to make a copy within the function.
• Use Rvalue references only when necessary to optimize move semantics.
• Consider pass-by-reference to const for custom objects to ensure immutability and ownership clarity.

Example:

To implement the recommended pattern, valued constructors can be written as follows:

class T {
    U u;
    V v;
public:
    T(U u, V v)
        : u(std::move(u))
        , v(std::move(v))
    {}
};

The above is the detailed content of Should Pass-by-Value Be the Default for Large Objects in Modern C ?. For more information, please follow other related articles on the PHP Chinese website!