-
-
Notifications
You must be signed in to change notification settings - Fork 5
struct
A struct is a collection of primitives (int, string, ...) and other structs into a data tuple. While struct is a node type, when reading or writing a struct, the abi omits the node header and directly handles the primitives on stack or in memory.
Wasp uses wit as schema, all wit constructs are valid wasp, as "prefix notation" in wit files or with common wasp syntax:
struct point{
x:int
y:int
}
point(1,2).y == 2
structs and named tuples
structs are instances of classes without the header in memory. Or put the other way round: classes are thin meta wrappers around structs. While the type information for struct instances is only available at compile time, the header of class instances is part of the data in memory.
Todo pass by reference or multi-value?
The wasp abi can return structs either as multi-value or as references to linear memory
The struct keyword in Angle is very similar to c and swift.
The main differences are: Meta information about all structs is stored in a custom type table, and or carried via
return types. structs can be part of function dispatch, the only limitation being that they need to be determined at
compile time, because unlike class instances, structs do not know their own type.
struct pointers as type schemes
multi-value returns allows to carry meta information about results in the form of smart pointers.
Each class has one corresponding raw type corresponding to the struct definition of its fields.
? structs can contain non primitive values while classes can only contain reference types.
That is, in classes int is really resolved as type Integer (or int60?),
class example{
level:int
}
whereas in structs, int is really resolved as type int32
struct example{
level:int #int32
}
This internal distinction however is almost completely invisible to the developer.
Compare with NamedTuple(x: Int32, y: String) in ruby and record in C#
Todo: can angle distinguish between un-charged data and implicit struct template prototypes?
f:{time:now} # passive declaration syntax
f.time! ≈ now!
f{time:now} # active construction syntax (or passive if f is not know or passive)
f(time:tomorrow) # active construction depending on existing symbol f
f(tomorrow) == f{time:tomorrow}
interfaces are just classes without concrete fields and without implementations
structs are just classes without header internally, so they should ONLY be used in extern FFI abi calls, right?
structs are just named tuples?
Alternative one can call and access foreign libraries functions and data just by (IMPLICITLY) applying the right abi.
Currently record is just an alias for struct
record pair { x: u32, y: u32, }
record person { name: string, age: u32, has-lego-action-figure: bool, }
A flags statement defines a new record-like structure where all the fields
are booleans.
https://raw.githubusercontent.com/WebAssembly/component-model/main/design/mvp/WIT.md
A variant statement defines a new type where instances of the type match
exactly one of the variants listed for the type. This is similar to a "sum" type
in algebraic datatypes (or an enum in Rust if you're familiar with it).
Variants can be thought of as tagged unions as well.
Each case of a variant can have an optional type associated with it which is present when values have that particular case's tag.
All variant type must have at least one case specified.
variant filter {
all,
none,
some(list<string>),
}An enum statement defines a new type which is semantically equivalent to a
variant where none of the cases have a payload type.
A union statement defines a new type which is semantically equivalent to a
variant where all of the cases have a payload type and the case names are
numerical.
Resources represent a value that has a hidden representation not known to the outside world.
This means that the resource is operated on through a "handle" (a pointer of sorts). Resources also have ownership associated with them and languages will have to manage the lifetime of resources manually (they're similar to file descriptors).
Resources can also optionally have functions defined within them which adds an
implicit "self" argument as the first argument to each function of the same type
of the including resource, unless the function is flagged as static.
resource file-descriptor
resource request {
static new: func() -> request
body: func() -> future<list<u8>>
headers: func() -> list<string>
}Types
type number = u32
type fallible-function-result = result<u32, string>
type headers = list<string>Types in wit, however, cannot be recursive:
Example:
struct point{
int x
int y
}
void stretch_y(point& p){p->y*=2;}stretch_y {1,2} # unsave
stretch_y {1,2} # file if reflected upon c with canonical abi
stretch_y point{1,2} # extra file, don't pass class, but just POINTER to data according to abi
Both old and new gnu style initialization of instances is valid, the later giving more savety:
ParserOptions{use_tags = true, kebab_case = true}
ParserOptions{.use_tags = true, .kebab_case = true}
Expandable structures enforce a scheme, but also allow extra elements.
Example:
extendible struct Point{ int x, int y}
Point test{y=7, x=8, z=9} // ok extra elements are stored behind ordered fields
Upon using extendible structs in abis the behaviour for the extra elements needs to be specified.
⚠️ The extra data will be LOST in external receivers of the struct and are only valueable for read-compatibility or for internal meta handling.
⚠️ Angle classes are similar to extendible structs!
However since [[Angle]] [[class]]es are of type [[node]], the extra members will be handled transparently.
That is: while usage of structs in abi calls require strict memory layout, internal (or external) passing of node_pointer is always just a reference / pointer to a wasp node tree.
# aliases
extendible structs are just called `extendible` or `expandable`
extendible struct are also called `open struct`
Currently the keyword `record` is aliased to struct but may receive specific semantics in the future.
# wit interfaces types as implemented in v8 (node v20) :
(module
(type ${f64_f64} (struct (field f64) (field f64)))
(type $f64_f64_=>ref|{f64_f64}| (func (param f64 f64) (result (ref ${f64_f64}))))
(type $ref|{f64_f64}|=>_f64 (func (param (ref ${f64_f64})) (result f64)))
(export "create_point" (func $0))
(export "length" (func $1))
(func $0 (param $0 f64) (param