The record type intends to be immutable, with init-only
public properties that can be modified using with-expression
. To optimize for that common case, records change the defaults of what a simple member declaration of the type string Name
means.
Instead of an implicitly private field, as you declare it in other classes and structs, in records you can use the shorthand for a public, init-only auto-property as shown below.
public record Customer { string Name; int Age; string Address; string Country; }
The above declaration is similar to the following declaration.
public record Customer
{
public string Name { get; init; }
public int Age { get; init; }
public string Address { get; init; }
public string Country { get; init; }
}
If you want a private field, you can just add the private modifier explicitly as shown below.
private string Name;
Sometimes it's useful to have a more positional approach to a record, where its content is given via constructor arguments and can be extracted with positional deconstruction.
It is also possible to specify your constructor and deconstructor in a record.
public record Customer
{
string Name;
int Age;
string Address;
string Country;
public Customer(string name, int age, string address, string country)
=> (Name, Age, Address, Country) = (name, age, address, country);
public void Deconstruct(out string name, out int age, out string address, out string country)
=> (name, age, address, country) = (Name, Age, Address, Country);
}
You can express the same thing in much shorter syntax as shown below.
public record Customer ( string Name, int Age, string Address, string Country );
It declares the public init-only auto-properties and the constructor and the deconstructor.
// positional construction
Customer customer = new ("Mark", 34, "23 ashdown", "UK");
// positional deconstruction
var (name, age, address, country) = customer;