Derive Macro typed_builder::TypedBuilder

source ·
#[derive(TypedBuilder)]
{
    // Attributes available to this derive:
    #[builder]
}
Expand description

TypedBuilder is not a real type - deriving it will generate a ::builder() method on your struct that will return a compile-time checked builder. Set the fields using setters with the same name as the struct’s fields and call .build() when you are done to create your object.

Trying to set the same fields twice will generate a compile-time error. Trying to build without setting one of the fields will also generate a compile-time error - unless that field is marked as #[builder(default)], in which case the ::default() value of it’s type will be picked. If you want to set a different default, use #[builder(default=...)].

§Examples

use typed_builder::TypedBuilder;

#[derive(PartialEq, TypedBuilder)]
struct Foo {
    // Mandatory Field:
    x: i32,

    // #[builder(default)] without parameter - use the type's default
    // #[builder(setter(strip_option))] - wrap the setter argument with `Some(...)`
    #[builder(default, setter(strip_option))]
    y: Option<i32>,

    // Or you can set the default
    #[builder(default=20)]
    z: i32,
}

assert!(
    Foo::builder().x(1).y(2).z(3).build()
    == Foo { x: 1, y: Some(2), z: 3, });

// Change the order of construction:
assert!(
    Foo::builder().z(1).x(2).y(3).build()
    == Foo { x: 2, y: Some(3), z: 1, });

// Optional fields are optional:
assert!(
    Foo::builder().x(1).build()
    == Foo { x: 1, y: None, z: 20, });

// This will not compile - because we did not set x:
// Foo::builder().build();

// This will not compile - because we set y twice:
// Foo::builder().x(1).y(2).y(3);

§Customization with attributes

In addition to putting #[derive(TypedBuilder)] on a type, you can specify a #[builder(...)] attribute on the type, and on any fields in it.

On the type, the following values are permitted:

  • doc: enable documentation of the builder type. By default, the builder type is given #[doc(hidden)], so that the builder() method will show FooBuilder as its return type, but it won’t be a link. If you turn this on, the builder type and its build method will get sane defaults. The field methods on the builder will be undocumented by default.

  • crate_module_path: This is only needed when typed_builder is reexported from another crate - which usually happens when another macro uses it. In that case, it is the reponsibility of that macro to set the crate_module_path to the unquoted module path from which the typed_builder crate can be accessed, so that the TypedBuilder macro will be able to access the typed declared in it.

    Defaults to #[builder(crate_module_path=::typed_builder)].

  • The following subsections:

    • builder_method(...): customize the builder method that creates the builder type
    • builder_type(...): customize the builder type
    • build_method(...): customize the final build method

    All have the same fields:

    • vis = "...": sets the visibility of the build method, default is pub
    • name = ...: sets the fn name of the build method, default is build
    • doc = "..." replaces the default documentation that will be generated for the build() method of the builder type. Setting this implies doc.
  • The build_method(...) subsection also has:

    • into or into = ...: change the output type of the builder. When a specific value/type is set via the assignment, this will be the output type of the builder. If no specific type is set, but into is specified, the return type will be generic and the user can decide which type shall be constructed. In both cases an Into conversion is required to be defined from the original type to the target type.
  • field_defaults(...) is structured like the #[builder(...)] attribute you can put on the fields and sets default options for fields of the type. If specific field need to revert some options to the default defaults they can prepend ! to the option they need to revert, and it would ignore the field defaults for that option in that field.

    use typed_builder::TypedBuilder;
    
    #[derive(TypedBuilder)]
    #[builder(field_defaults(default, setter(strip_option)))]
    struct Foo {
        // Defaults to None, options-stripping is performed:
        x: Option<i32>,
    
        // Defaults to 0, option-stripping is not performed:
        #[builder(setter(!strip_option))]
        y: i32,
    
        // Defaults to Some(13), option-stripping is performed:
        #[builder(default = Some(13))]
        z: Option<i32>,
    
        // Accepts params `(x: f32, y: f32)`
        #[builder(setter(!strip_option, transform = |x: f32, y: f32| Point { x, y }))]
        w: Point,
    }
    
    #[derive(Default)]
    struct Point { x: f32, y: f32 }
  • mutators(...) takes functions, that can mutate fields inside of the builder. See mutators for details.

On each field, the following values are permitted:

  • default: make the field optional, defaulting to Default::default(). This requires that the field type implement Default. Mutually exclusive with any other form of default.

  • default = ...: make the field optional, defaulting to the expression ....

  • default_code = "...": make the field optional, defaulting to the expression .... Note that you need to enclose it in quotes, which allows you to use it together with other custom derive proc-macro crates that complain about “expected literal”. Note that if ... contains a string, you can use raw string literals to avoid escaping the double quotes - e.g. #[builder(default_code = r#""default text".to_owned()"#)].

  • via_mutators: initialize the field when constructing the builder, useful in combination with mutators.

  • via_mutators = ... or via_mutators(init = ...): initialies the field with the expression ... when constructing the builder, useful in combination with mutators.

  • mutators(...) takes functions, that can mutate fields inside of the builder. Mutators specified on a field, mark this field as required, see mutators for details.

  • setter(...): settings for the field setters. The following values are permitted inside:

    • doc = "...": sets the documentation for the field’s setter on the builder type. This will be of no value unless you enable docs for the builder type with #[builder(doc)] or similar on the type.

    • skip: do not define a method on the builder for this field. This requires that a default be set.

    • into: automatically convert the argument of the setter method to the type of the field. Note that this conversion interferes with Rust’s type inference and integer literal detection, so this may reduce ergonomics if the field type is generic or an unsigned integer.

    • strip_option: for Option<...> fields only, this makes the setter wrap its argument with Some(...), relieving the caller from having to do this. Note that with this setting on one cannot set the field to None with the setter - so the only way to get it to be None is by using #[builder(default)] and not calling the field’s setter.

    • strip_bool: for bool fields only, this makes the setter receive no arguments and simply set the field’s value to true. When used, the default is automatically set to false.

    • transform = |param1: Type1, param2: Type2 ...| expr: this makes the setter accept param1: Type1, param2: Type2 ... instead of the field type itself. The parameters are transformed into the field type using the expression expr. The transformation is performed when the setter is called.

    • prefix = "..." prepends the setter method with the specified prefix. For example, setting prefix = "with_" results in setters like with_x or with_y. This option is combinable with suffix = "...".

    • suffix = "..." appends the setter method with the specified suffix. For example, setting suffix = "_value" results in setters like x_value or y_value. This option is combinable with prefix = "...".

    • mutable_during_default_resolution: when expressions in default = ... field attributes are evaluated, this field will be mutable, allowing earlier-defined fields to be mutated by later-defined fields. Warning - Use this feature with care! If the field that mutates the previous field in its default expression is set via a setter, that mutation will not happen.

§Mutators

Set fields can be mutated using mutators, these can be defind via mutators(...).

Fields annotated with #[builder(via_mutators)] are always available to mutators. Additional fields, that the mutator accesses need to be delcared using #[mutator(requires = [field1, field2, ...])]. The mutator will only be availible to call when they are set.

Mutators on a field, result in them automatically making the field required, i.e., it needs to be marked as via_mutators, or its setter be called. Appart from that, they behave identically.

use typed_builder::TypedBuilder;

#[derive(PartialEq, Debug, TypedBuilder)]
#[builder(mutators(
    // Mutator has only acces to fields marked as `via_mutators`.
    fn inc_a(&mut self, a: i32){
        self.a += a;
    }
    // Mutator has access to `x` additionally.
    #[mutator(requires = [x])]
    fn x_into_b(&mut self) {
        self.b.push(self.x)
    }
))]
struct Struct {
    // Does not require explicit `requires = [x]`, as the field
    // the mutator is specifed on, is required implicitly.
    #[builder(mutators(
        fn x_into_b_field(self) {
            self.b.push(self.x)
        }
    ))]
    x: i32,
    #[builder(via_mutators(init = 1))]
    a: i32,
    #[builder(via_mutators)]
    b: Vec<i32>
}

// Mutators do not enforce only being called once
assert_eq!(
    Struct::builder().x(2).x_into_b().x_into_b().x_into_b_field().inc_a(2).build(),
    Struct {x: 2, a: 3, b: vec![2, 2, 2]});