Resources
Resources are types that can only exist in one location at a time and must be used exactly once.
Resources must be created (instantiated) by using the create
keyword.
Before the closing bracket of a function that has resources created or moved into scope, those resources must explicitly be either moved to a valid storage location or destroyed.
They are moved when used as an initial value for a constant or variable, when assigned to a different variable, when passed as an argument to a function, and when returned from a function.
Resources can be explicitly destroyed using the destroy
keyword.
Accessing a field or calling a function of a resource does not move or destroy it.
When the resource is moved, the constant or variable that referred to the resource before the move becomes invalid. An invalid resource cannot be used again.
To make the usage and behavior of resource types explicit, the prefix @
must be used in type annotations of variable or constant declarations, parameters, and return types.
The move operator (<-
)
To make moves of resources explicit, the move operator <-
must be used when the resource is the initial value of a constant or variable, when it is moved to a different variable, when it is moved to a function as an argument, and when it is returned from a function.
Declare a resource named SomeResource
, with a variable-integer field:
_10access(all)_10resource SomeResource {_10 _10 access(all)_10 var value: Int_10_10 init(value: Int) {_10 self.value = value_10 }_10}
Declare a constant with a value of resource type SomeResource
:
_10let a: @SomeResource <- create SomeResource(value: 5)
Move the resource value to a new constant:
_11let b <- a_11_11_11// Invalid Line Below: Cannot use constant `a` anymore as the resource that it_11// referred to was moved to constant `b`._11_11a.value_11_11// Constant `b` owns the resource._11_11b.value // equals 5
Declare a function that accepts a resource. The parameter has a resource type, so the type annotation must be prefixed with @
:
_10access(all)_10fun use(resource: @SomeResource) {_10 // ..._10}
Call function use
, and move the resource into it:
_10use(resource: <-b)_10_10// Invalid Line Below: Cannot use constant `b` anymore as the resource it_10// referred to was moved into function `use`._10_10b.value
A resource object cannot go out of scope and be dynamically lost. The program must either explicitly destroy it or move it to another context.
Declare another, unrelated value of resource type SomeResource
:
_10{_10 let c <- create SomeResource(value: 10)_10_10 // Invalid: `c` is not moved or destroyed before the end of the scope, but must be._10 // It cannot be lost._10}
Declare another, unrelated value of resource type SomeResource
:
_10_10let d <- create SomeResource(value: 20)
Destroy the resource referred to by constant d
:
_10destroy d_10_10// Invalid: Cannot use constant `d` anymore as the resource_10// it referred to was destroyed._10//_10d.value
To make it explicit that the type is a resource type and must follow the rules associated with resources, it must be prefixed with @
in all type annotations (e.g., for variable declarations, parameters, or return types).
Declare a constant with an explicit type annotation. The constant has a resource type, so the type annotation must be prefixed with @
:
_10let someResource: @SomeResource <- create SomeResource(value: 5)
Declare a function that consumes a resource and destroys it. The parameter has a resource type, so the type annotation must be prefixed with @
:
_10access(all)_10fun use(resource: @SomeResource) {_10 destroy resource_10}
Declare a function that returns a resource:
- The return type is a resource type, so the type annotation must be prefixed with
@
. - The return statement must also use the
<-
operator to make it explicit the resource is moved.
_10access(all)_10fun get(): @SomeResource {_10 let newResource <- create SomeResource()_10 return <-newResource_10}
Resources must be used exactly once.
Declare a function that consumes a resource but does not use it:
_10// This function is invalid, because it would cause a loss of the resource._10access(all)_10fun forgetToUse(resource: @SomeResource) {_10 // Invalid: The resource parameter `resource` is not used, but must be._10}
Declare a constant named res
that has the resource type SomeResource
:
_10let res <- create SomeResource()
Call the function use
and move the resource res
into it:
_11use(resource: <-res)_11_11// Invalid: The resource constant `res` cannot be used again,_11// as it was moved in the previous function call._11//_11use(resource: <-res)_11_11// Invalid: The resource constant `res` cannot be used again,_11// as it was moved in the previous function call._11//_11res.value
Declare a function that has a resource parameter:
_11// This function is invalid, because it does not always use the resource parameter,_11// which would cause a loss of the resource:_11access(all)_11fun sometimesDestroy(resource: @SomeResource, destroyResource: Bool) {_11 if destroyResource {_11 destroy resource_11 }_11 // Invalid: The resource parameter `resource` is not always used, but must be._11 // The destroy statement is not always executed, so at the end of this function_11 // it might have been destroyed or not._11}
Declare a function which has a resource parameter:
_11// This function is valid, as it always uses the resource parameter,_11// and does not cause a loss of the resource._11//_11access(all)_11fun alwaysUse(resource: @SomeResource, destroyResource: Bool) {_11 if destroyResource {_11 destroy resource_11 } else {_11 use(resource: <-resource)_11 }_11}
At the end of the function, the resource parameter was definitely used. It was either destroyed or moved in the call of function use
.
Declare a function that has a resource parameter:
_19// This function is invalid, because it does not always use the resource parameter,_19// which would cause a loss of the resource._19//_19access(all)_19fun returnBeforeDestroy(move: Bool) {_19 let res <- create SomeResource(value: 1)_19 if move {_19 use(resource: <-res)_19 return_19 } else {_19 // Invalid: When this function returns here, the resource variable_19 // `res` was not used, but must be._19 return_19 }_19 // Invalid: the resource variable `res` was potentially moved in the_19 // previous if-statement, and both branches definitely return,_19 // so this statement is unreachable._19 destroy res_19}
Resource variables
Resource variables cannot be assigned to, as that would lead to the loss of the variable's current resource value.
Instead, use a swap statement (<->
) or shift statement (<- target <-
) to replace the resource variable with another resource:
_10access(all)_10resource R {}_10_10var x <- create R()_10var y <- create R()_10_10// Invalid: Cannot assign to resource variable `x`,_10// as its current resource would be lost_10//_10x <- y
Instead, use a swap statement:
_10var replacement <- create R()_10x <-> replacement_10// `x` is the new resource._10// `replacement` is the old resource.
Or, use the shift statement (<- target <-
):
_10// This statement moves the resource out of `x` and into `oldX`,_10// and at the same time assigns `x` with the new value on the right-hand side._10let oldX <- x <- create R()_10// oldX still needs to be explicitly handled after this statement_10destroy oldX
Nested resources
Fields in composite types behave differently when they have a resource type.
Accessing a field or calling a function on a resource field is valid, however, moving a resource out of a variable resource field is not allowed. Instead, use a swap statement to replace the resource with another resource. For example:
_10let child <- create Child(name: "Child 1")_10child.name // is "Child 1"_10_10let parent <- create Parent(name: "Parent", child: <-child)_10parent.child.name // is "Child 1"_10_10// Invalid: Cannot move resource out of variable resource field._10let childAgain <- parent.child
Instead, use a swap statement:
_10var otherChild <- create Child(name: "Child 2")_10parent.child <-> otherChild_10// `parent.child` is the second child, Child 2._10// `otherChild` is the first child, Child 1.
When a resource containing nested resources in fields is destroyed with a destroy
statement, all the nested resources are also destroyed:
_11// Declare a resource with resource fields_11//_11access(all)_11resource Parent {_11 var child1: @Child_11 var child2: @Child_11 init(child1: @Child, child2: @Child) {_11 self.child1 <- child1_11 self.child2 <- child2_11 }_11}
The order in which the nested resources are destroyed is deterministic but unspecified, and cannot be influenced by the developer. In this example, when Parent
is destroyed, the child1
and child2
fields are also both destroyed in some unspecified order.
In previous versions of Cadence, it was possible to define a special destroy
function that would execute arbitrary code when a resource was destroyed, but this is no longer the case.
Destroy events
While it is not possible to specify arbitrary code to execute upon the destruction of a resource, it is possible to specify a special event to be automatically emitted when a resource is destroyed. The event has a reserved name — ResourceDestroyed
— and it uses a special syntax:
_10resource R {_10 event ResourceDestroyed(id: UInt64 = self.id) _10_10 let id: UInt64_10_10 init(_ id: UInt64) {_10 self.id = id_10 }_10}
Whenever a value of type R
defined this way is destroyed, a special R.ResourceDestroyed
event is emitted. The special syntax used in the definition of the ResourceDestroyed
specifies what the values associated with each event parameter will be; in this case, the id
field of the R.ResourceDestroyed
event will be the value that the id
field held immediately before the resource was destroyed. In general, for a ResourceDestroyed
event defined as:
_10event ResourceDestroyed(field1: T1 = e1, field2: T2 = e2, ...)
- The value of
field1
on the event will be the result of evaluatinge1
before destroying the resource. - The value of
field2
on the event will be the result of evaluatinge2
before destroying the resource, and so on.
As one might expect, e1
and e2
must also be expressions of type T1
and T2
, respectively.
In order to guarantee that these events can be emitted with no chance of failure at runtime, there are restrictions placed on which kinds of types and expressions can be used in their definitions. In general, an expression defining the value of a field (the e
in the general definition above) can only be a member or indexed access on self
(or, base
in the case of an attachment), or a literal. The types of event fields are restricted to number types, String
s, Boolean
s, Address
es, and Path
s.
Resources in closures
Resources cannot be captured in closures, as that could potentially result in duplications:
_13resource R {}_13_13// Invalid: Declare a function which returns a closure which refers to_13// the resource parameter `resource`. Each call to the returned function_13// would return the resource, which should not be possible._13//_13fun makeCloner(resource: @R): fun(): @R {_13 return fun (): @R {_13 return <-resource_13 }_13}_13_13let test = makeCloner(resource: <-create R())
Resources in arrays and dictionaries
Arrays and dictionaries behave differently when they contain resources: it is not allowed to index into an array to read an element at a certain index or assign to it, or index into a dictionary to read a value for a certain key or set a value for the key.
Instead, use a swap statement (<->
) or shift statement (<- target <-
) to replace the accessed resource with another resource.
Declare a constant for an array of resources. Then, create two resources and move them into the array (resources
has type @[R]
):
_15resource R {}_15_15let resources <- [_15 <-create R(),_15 <-create R()_15]_15_15// Invalid: Reading an element from a resource array is not allowed._15//_15let firstResource <- resources[0]_15_15// Invalid: Setting an element in a resource array is not allowed,_15// as it would result in the loss of the current value._15//_15resources[0] <- create R()
Instead, when attempting to either read an element or update an element in a resource array, use a swap statement with a variable to replace the accessed element:
_10var res <- create R()_10resources[0] <-> res_10// `resources[0]` now contains the new resource._10// `res` now contains the old resource.
Use the shift statement to move the new resource into the array at the same time that the old resource is being moved out:
_10let oldRes <- resources[0] <- create R()_10// The old object still needs to be handled_10destroy oldRes
The same applies to dictionaries.
Declare a constant for a dictionary of resources. Then, create two resources and move them into the dictionary (resources
has type @{String: R}
):
_10let resources <- {_10 "r1": <-create R(),_10 "r2": <-create R()_10}_10_10// Invalid: Reading an element from a resource dictionary is not allowed._10// It's not obvious that an access like this would have to remove_10// the key from the dictionary._10//_10let firstResource <- resources["r1"]
Instead, make the removal explicit by using the remove
function:
_10let firstResource <- resources.remove(key: "r1")_10_10// Invalid: Setting an element in a resource dictionary is not allowed,_10// as it would result in the loss of the current value._10//_10resources["r1"] <- create R()
When attempting to either read an element or update an element in a resource dictionary, use a swap statement with a variable to replace the accessed element.
The result of a dictionary read is optional, as the given key might not exist in the dictionary. The types on both sides of the swap operator must be the same, so also declare the variable as an optional:
_10var res: @R? <- create R()_10resources["r1"] <-> res_10// `resources["r1"]` now contains the new resource._10// `res` now contains the old resource.
Use the shift statement to move the new resource into the dictionary at the same time that the old resource is being moved out:
_10let oldRes <- resources["r2"] <- create R()_10// The old object still needs to be handled_10destroy oldRes
Resources cannot be moved into arrays and dictionaries multiple times, as that would cause a duplication:
_10let resource <- create R()_10_10// Invalid: The resource variable `resource` can only be moved into the array once._10//_10let resources <- [_10 <-resource,_10 <-resource_10]
_10let resource <- create R()_10_10// Invalid: The resource variable `resource` can only be moved into the dictionary once._10let resources <- {_10 "res1": <-resource,_10 "res2": <-resource_10}
Resource arrays and dictionaries can be destroyed:
_10let resources <- [_10 <-create R(),_10 <-create R()_10]_10destroy resources
_10let resources <- {_10 "r1": <-create R(),_10 "r2": <-create R()_10}_10destroy resources
The variable array functions, like append
, insert
, and remove
, behave like non-resource arrays. Please note, however, that the result of the remove
functions must be used:
_18let resources <- [<-create R()]_18// `resources.length` is `1`_18_18resources.append(<-create R())_18// `resources.length` is `2`_18_18let first <- resource.remove(at: 0)_18// `resources.length` is `1`_18destroy first_18_18resources.insert(at: 0, <-create R())_18// `resources.length` is `2`_18_18// Invalid: The statement ignores the result of the call to `remove`,_18// which would result in a loss._18resource.remove(at: 0)_18_18destroy resources
- The variable array function
contains
is not available, as it is impossible: if the resource can be passed to thecontains
function, it is by definition not in the array. - The variable array function
concat
is not available, as it would result in the duplication of resources. - The dictionary functions like
insert
andremove
behave like non-resource dictionaries. Please note, however, that the result of these functions must be used:
_18let resources <- {"r1": <-create R()}_18// `resources.length` is `1`_18_18let first <- resource.remove(key: "r1")_18// `resources.length` is `0`_18destroy first_18_18let old <- resources.insert(key: "r1", <-create R())_18// `old` is nil, as there was no value for the key "r1"_18// `resources.length` is `1`_18_18let old2 <- resources.insert(key: "r1", <-create R())_18// `old2` is the old value for the key "r1"_18// `resources.length` is `1`_18_18destroy old_18destroy old2_18destroy resources
Resource identifier
Resources have an implicit unique identifier associated with them, implemented by a predeclared public field let uuid: UInt64
on each resource.
This identifier is automatically set when the resource is created, before the resource's initializer is called (i.e., the identifier can be used in the initializer), and will be unique even after the resource is destroyed (i.e., no two resources will ever have the same identifier).
-
Declare a resource without any fields:
_10resource R {} -
Create two resources:
_10let r1 <- create R()_10let r2 <- create R() -
Get each resource's unique identifier:
_10let id1 = r1.uuid_10let id2 = r2.uuid -
Destroy the first resource:
_10destroy r1 -
Create a third resource:
_10let r3 <- create R()_10_10let id3 = r3.uuid_10_10id1 != id2 // true_10id2 != id3 // true_10id3 != id1 // true
The details of how the identifiers are generated is an implementation detail.
Do not rely on or assume any particular behavior in Cadence programs.
Resource owner
Resources have the implicit field let owner: &Account?
. If the resource is currently stored in an account, then the field contains the publicly accessible portion of the account. Otherwise the field is nil
.
The field's value changes when the resource is moved from outside account storage into account storage, when it is moved from the storage of one account to the storage of another account, and when it is moved out of account storage.