path: root/docs/TYPES.md

# Typing in Puck

Puck has a comprehensive static type system.

## Basic types

Basic types can be one-of:
- `bool`: internally an enum.
- `int`: integer number. x bits of precision by default.
  - `uint`: unsigned integer for more precision.
  - `i8`, `i16`, `i32`, `i64`, `i28`: specified integer size
  - `u8`, `u16`, `u32`, `u64`, `u128`: specified integer size
- `float`: floating-point number.
  - `f32`, `f64`: specified float sizes
- `char`: a distinct 0-127 character. For working with ascii.
- `rune`: a Unicode character.
- `str`: a string type. mutable. internally a char-array? must also support unicode.
- `void`: an internal type designating the absence of a value.
  - possibly, the empty tuple. then would `empty` be better? or `unit`?
- `never`: a type that denotes functions that do not return.
  - distinct from returning nothing.
  - the bottom type.

`bool`, `int`/`uint` and siblings, `float` and siblings, `char`, and `rune` are all considered **primitive types** and are _always_ copied (unless passed as `var`). More on when parameters are passed by value (copied) vs. passed by reference can be found in the [memory management document](MEMORY_MANAGEMENT.md).

Basic types as a whole include the primitive types, as well as `str`, `void`, and `never`. Basic types can further be broken down into the following categories:
- boolean types: `bool`
- numeric types: `int`, `float`, and siblings
- textual types: `char`, `rune`, `str`
- funky types: `void`, `never`

Funky types will rarely be referenced by name: instead, the absence of a type typically implicitly denotes one or the other. Still, having a name is helpful in some situations.

## Container types

Container types, broadly speaking, are types that contain other types. These exclude the types in [advanced types](#advanced-types).

### Iterable types

Iterable types can be one-of:
- `array[S, T]`: Static arrays. Can only contain one type `T`. Of size `S` and cannot grow/shrink.
  - Initialize in-place with `array(a, b, c)`. Should we do this? otherwise `[a, b, c]`.
- `list[T]`: Dynamic arrays. Can only contain one type `T`. May grow/shrink dynamically.
  - Initialize in-place with `list(a, b, c)`. Should we do this? otherwise `@[a, b, c]`.
- `slice[T]`: Slices. Used to represent a "view" into some sequence of elements of type `T`.
  - Cannot be directly constructed. May be initialized from an array, list, or string, or may be used as a generic parameter on functions (more on that later).
  - Slices cannot grow/shrink. Their elements may be accessed and mutated. As they are underlyingly a reference to an array or list, they **must not** outlive the data they reference.
- `str`: Strings. Contain the `rune` type or alternatively `char`s or `bytes`?? {undecided}

All of these above types are some sort of sequence: and so have a length, and so can be _iterated_.
For convenience, a special `iterable` generic type is defined for use in parameters: that abstracts over all of the container types. This `iterable` type is also extended to any collection of a single type with a length.
- Is this redundant with `slice[T]`? todo

Elements of container types can be accessed by the `container[index]` syntax. Slices of container types can be accessed by the `container[lowerbound..upperbound]` syntax. Slices of non-consecutive elements can be accessed by the `container[a,b,c..d]` syntax, and indeed, the previous example expands to these. They can also be combined: `container[a,b,c..d]`.

### Abstract data types

There are an additional suite of related types: abstract data types. While falling under container types, these do not have a necessarily straightforward or best implementation, and so multiple implementations are provided.

Abstract data types can be one-of:
- `set[T]`: high-performance sets implemented as a bit array.
  - These have a maximum data size, at which point the compiler will suggest using a `HashSet[T]` instead.
- `table[T, U]`: simple symbol tables implemented as an association list.
  - These do not have a maximum size. However, at some point the compiler will suggest using a `HashTable[T, U]` instead.
- `HashSet[T]`: standard hash sets.
- `HashTable[T, U]`: standard hash tables.

Unlike iterable types, abstract data types are not iterable by default: as they are not ordered, and thus, it is not clear how they should be iterated. Despite this: for utility purposes, an `elems()` iterator based on a normalization of the elements is provided for `set` and `HashSet`, and `keys()`, `values()`, and `pairs()` iterators are provided for `table` and `HashTable` based on a normalization of the keys. This is deterministic to prevent user reliance on shoddy randomization, see Golang.

## Parameter types

Some types are only valid when being passed to a function, or in similar contexts.
No variables may be assigned these types, nor may any function return them.
These are monomorphized into more specific functions at compile-time if needed.

Parameter types can be one-of:
- mutable: `func foo(a: var str)`: Denotes the mutability of a parameter. Parameters are immutable by default.
  - Passed as a `ref` if not one already, and marked mutable.
- static: `func foo(a: static str)`: Denotes a parameter that's value must be known at compile-time. Useful with `when` for writing generic code.
- generic: `func foo[T](a: list[T], b: T)`: The standard implementation of generics, where a parameter's exact type is not listed, and instead statically dispatched based on usage.
- constrained: `func foo(a: str | int | float)`: A basic implementation of generics, where a parameter can be one-of several listed types. Makes for particularly straightforward monomorphization.
  - Separated with the bitwise or operator `|` rather than the symbolic or `||` or a raw `or` to give the impression that there isn't a corresponding "and" operation (the `&` operator is preoccupied with strings).
- functions: `func foo(a: func (x, y: int): int)`: First-class functions.
  - Functions may be prefixed with modifiers: one-of `pure`, `yeet`, `async`.
  - Syntactic sugar is available: `func foo(a: (int, int) -> int)`. This is not usable with modifiers.
- interfaces: `func foo(a: Stack[int])`: User-defined duck typing. Equivalent to and more general than Rust's traits or Haskell's typeclasses.
  - ex. for above: the `Stack[T]` interface specifies generic `push` and `pop` operations
  - More on interfaces in the [interfaces section](#interfaces).
- built-in interface: `func foo(a: enum)`: Included, special interfaces for being generic over [advanced types](#advanced-types).
  - Of note is how `slice[T]` functions: it is generic over `lists` and `arrays` of any length.

These parameter types (except `static`) share a common trait: they are not *sized*. The exact type is not generally known until compilation - and in the case of interfaces, sometimes not even during compilation! As the size is not always rigorously known, problems arise when attempting to construct parameter types or compose them with other types: and so this is disallowed. Not all is lost, however, as they may still be used with *indirection* - detailed in the [section on reference types](#reference-types).

### Generic types

Functions can take a _generic_ type, that is, be defined for a number of types at once:

```puck
func add[T](a: list[T], b: T) =
  return a.add(b)

func length[T](a: T) =
  return a.len # monomorphizes based on usage.
  # lots of things use .len, but only a few called by this do.
  # throws a warning if exported for lack of specitivity.

func length[T: str | list](a: T) =
  return a.len
```

The syntax for generics is `func`, `ident`, followed by the names of the generic parameters in brackets `[T, U, V]`, followed by the function's parameters (which may refer to the generic types).
Generics are replaced with concrete types at compile time (monomorphization) based on their usage in functions within the main function body.

Constrained generics have two syntaxes: the constraint can be directly on a parameter, leaving off the `[T]` box, or it may be defined within the box as `[T: int | float]` for easy reuse in the parameters.

## Reference types

Types are typically constructed by value on the stack. That is, without any level of indirection: and so type declarations that recursively refer to one another, or involve parameter types, would not be allowed. However, Puck provides two avenues for indirection.

Reference types can be one-of:
- `ref T`: An automatically-managed reference to type `T`. This is a pointer of size `uint` (native).
- `ptr T`: A manually-managed pointer to type `T`. (very) unsafe.

```puck
type Node = ref struct
  left: Node
  right: Node

type AnotherNode = struct
  left: ref AnotherNode
  right: ref AnotherNode

type BinaryTree = ref struct
  left: BinaryTree
  right: BinaryTree
```

The `ref` prefix may be placed at the top level of type declarations, or inside on a field of a structural type. `ref` types may often be more efficient when dealing with large data structures. They also provide for the usage of parameter types (except `static`, `var`, and standard generics) within type declarations.

The compiler abstracts over `ref` types to provide optimization for reference counts: and so neither a distinction between `Rc`/`Arc`/`Box`, nor a `*` dereference operator is needed.
Much care has been given to make references efficient and safe, and so `ptr` should be avoided if at all possible. The compiler will yell at you if you use it (or any other unsafe features).

The implementation of reference types is delved into in further detail in the [document on memory management](MEMORY_MANAGEMENT.md).

## Advanced Types

The `type` keyword is used to declare custom data types. These are *algebraic*: they function by composition.

Algebraic data types can be one-of:
- `tuple`: An ordered collection of types. Optionally named.
- `struct`: An unordered, named collection of types. May have default values.
- `enum`: Ordinal labels, that may hold values. Their default values are their ordinality.
- `union`: Powerful matchable tagged unions a la Rust. Sum types.
- `interface`: Usage-based typeclasses. User-defined duck typing.
- `distinct`: a type that must be explicitly converted
- type aliases, declared as `type Identifier = Alias`

### tuples

Tuples are an *ordered* collection of either named or unnamed types.

They are declared with `tuple[Type, identifier: Type, ...]` and initialized with parentheses: `(413, "hello", value: 40000)`.

```puck
```

Tuples are particularly useful for "on-the-fly" types.
Structs are generally a better choice for custom type declarations.

### structs

Structs are an *unordered* collection of named types.

They are declared with `struct[identifier: Type, ...]` and initialized with their name, followed by their fields in brackets: `Struct(field: "value", another: 500)`.

```puck
```

Structs are *structural* and so structs composed entirely of fields with the same signature (identical in name and type) are considered *equivalent*.
This is part of a broader structural trend in the type system, and is discussed in detail in [subtyping](#subtyping).

### enums

Enums are *ordinal labels* that may have associated values.

They are declared with `enum[Label, AnotherLabel = 4, ...]` and are never initialized (their values are known statically).
Enums may be accessed directly by their label, and are ordinal and iterable regardless of their associated value.

In the case of an identifier conflict (with other enum labels, or types, or...) they must be prefixed with the name of their associated type (separated by a dot). (this is standard for identifier conflicts: and is discussed in more detail in the [modules document](MODULES.md).)

### unions

Unions are *tagged* type unions. They provide a high-level wrapper over an inner type that must be accessed via pattern matching.

They are declared with `union[Variant: Type, ...]` and initialized with the name of a variant followed by its inner type constructor in brackets: `Square(side: 5)`. Tuple and struct types are special-cased to eliminate extraneous parentheses.

```puck
```

They take up as much space in memory as the largest variant, plus the size of the tag (typically one byte).

#### pattern matching

todo...

```puck
```

The `of` operator is similar to the `is` operator in that it queries type equality. However, unbound identifiers within `of` expressions are bound to appropriate values (if matched) and injected into the scope.

### interfaces

todo...

### distinct types

todo...

### type aliases

Finally, any type can be declared as an *alias* to a type simply by assigning it to such.


## Errata

### composition of types

Algebraic data types are, well, *algebraic*, and we've mentioned that they function by composition but have yet to show an extended example of what this means for the end user.

```puck
```

Note the existence of *anonymous* algebraic types: structs, tuples, unions, even interfaces may be nested within other algebraic data types and declared without an identifier. todo...

### default values

Puck does not have any concept of `null`: all values *must* be initialized.
But always explicitly initializing types is loud, and so most types have an associated "default value".

**Default values**:
- `bool`: `false`
- `int`, `uint`, etc: `0`
- `float`, etc: `0.0`
- `char`: `char(0)`
- `rune`: `rune(0)`
- `str`: `""`
- `void`, `never`: n/a
- `array[T]`, `list[T]`: `[]`/`@[]`
- `set[T]`, `table[T, U]`: `{}`
- `tuple[T, U, ...]`: `(default values of its fields)`
- `struct[T, U, ...]`: `Struct(default values of its fields)`
- `enum[One, Two, ...]`: `One` (the first label)
  - todo: is this correct?
- `union[T, U, ...]`: todo
- `slice[T]`, `func`: **disallowed**
- `ref`, `ptr`: **disallowed**

For `ref` and `ptr` types, however, this is trickier. There is no reasonable "default" for these types *aside from* null.
Instead of giving in, the compiler will instead disallow any non-initializations or other cases in which a default value would be inserted.
(`slice[T]` and `func` types are references under the hood and also behave in this way)

### subtyping

todo...

### inheritance

Puck is not an object-oriented language. Idiomatic design patterns in object-oriented languages will be harder to accomplish and not quite as idiomatic.

But, Puck has a number of features that somewhat support the object-oriented paradigm, including:
- uniform function call syntax
- structural typing / subtyping
- interfaces

```puck
```

These features may *compose* into code that closely resembles its object-oriented counterpart. But make no mistake! Puck is static first and functional somewhere in there: dynamic dispatch and the like are not accessible (currently).