A zero-cost wrapper adding type-safety to resource identifiers.
This solves two problems:
- If two resources are identified by the same type (
Foo
andBar
both being identified by ai32
for instance), they might be swapped by mistake in application code. The compiler cannot help you with this, but it is of course incorrect. - When writting code that needs to manipulate resource identifiers, you often need to lookup what the concrete type of the ID is in order to write function signatures and type annotations. In most cases, the nature of the ID does not matter, so it is more convenient to just treat it as an opaque type.
use tagged_id::{Id, Identify};
#[derive(Id)] // note the derive macro
#[tagged_id(i32)] // this attribute specifies the underlying type
struct Foo {
id: Id<Foo>, // id is a i32
some_field: String,
}
struct Bar {
id: Id<Bar>, // id is also a i32, see impl below
some_value: i32,
}
// This is what the derive macro generates
impl Identify for Bar {
type InnerId = i32;
}
fn main() {
let foo_id: Id<Foo> = Id::new(42);
let bar_id: Id<Bar> = Id::new(42);
// This does not compile since the tags are different.
assert_eq!(foo_id, bar_id);
}
Id<T>
inherits relevant core trait implementations of the inner type. For instance, if InnerId
is Copy
, then Id<T>
is also Copy
.
Id<T>
is just a newtype wrapper of the inner type with a trait bound, which makes it a zero cost abstraction.
derive
: Enable the derive macro, this feature is enabled by default.smartstring
: EnableFrom<&str>
instance converting to aCompactString
. When disabled, an instance forString
is enabled instead.uuid
: EnableFrom<Uuid>
instance for convenience.serde
: Enable serde support fortagged-id
and dependencies that support it likesmartstring
.sqlx-{postgres,mysql,sqlite}
: EnableEncode
andDecode
instances for transparent use with the correspondingsqlx
backend.