12 KiB
Dust Language Reference
!!! This is a work in progress and has incomplete information. !!!
This is an in-depth description of the syntax and abstractions used by the Dust language. It is not necessary to read or understand all of it before you start using Dust. Instead, refer to it when you need help with the syntax or understanding how the code is run.
Each section of this document corresponds to a node in the concrete syntax tree. Creating this tree is the first step in interpreting Dust code. Second, the syntax tree is traversed and an abstract tree is generated. Each node in the syntax tree corresponds to a node in the abstract tree. Third, the abstract tree is verified to ensure that it will not generate any values that violate the type restrictions. Finally, the abstract tree is run, beginning at the root.
You may reference the grammar file and the [Tree Sitter docs] (https://tree-sitter.github.io/) while reading this guide to understand how the language is parsed.
Root
The root node represents all of the source code. It is a sequence of statements that are executed synchronously, in order. The output of the program is always the result of the final statement or the first error encountered.
Values
There are ten kinds of value in Dust. Some are very simple and are parsed directly from the source code, some are collections and others are used in special ways, like functions and structures. All values can be assinged to an identifier.
Dust does not have a null type. Absent values are represented with the none value, which is a
kind of option. You may not create a variable without a value and no variable can ever
be in an 'undefined' state during execution.
Boolean
Booleans are true or false. They are represented by the literal tokens true and false.
Integer
Integers are whole numbers that may be positive, negative or zero. Internally, an integer is a signed 64-bit value.
42
Integers always overflow when their maximum or minimum value is reached. Overflowing means that
if the value is too high or low for the 64-bit integer, it will wrap around. You can use the built-
in values int:max and int:min to get the highest and lowest possible values.
assert_equal(int:max + 1, int:min)
assert_equal(int:min - 1, int:max)
Float
A float is a numeric value with a decimal. Floats are 64-bit and, like integers, will overflow at their bounds.
42.0
Range
A range represents a contiguous sequence of integers. Dust ranges are inclusive so both the high and low bounds will be represented.
0..100
String
A string is a utf-8 sequence used to represent text. Strings can be wrapped in single or double quotes as well as backticks.
'42'
"42"
`42`
'forty-two'
List
A list is collection of values stored as a sequence and accessible by indexing their position with an integer. Lists indexes begin at zero for the first item.
[ 42 'forty-two' ]
[ 123, 'one', 'two', 'three' ]
Note that the commas are optional, including trailing commas.
[1 2 3 4 5]:2
# Output: 3
Map
Maps are flexible collections with arbitrary key-value pairs, similar to JSON objects. A map is
created with a pair of curly braces and its entries are variables declared inside those braces. Map
contents can be accessed using a colon :. Commas may optionally be included after the key-value
pairs.
reminder = {
message = "Buy milk"
tags = ["groceries", "home"]
}
reminder:message
# Output: Buy milk
Internally a map is represented by a B-tree. The implicit advantage of using a B-tree instead of a
hash map is that a B-tree is sorted and therefore can be easily compared to another. Maps are also
used by the interpreter as the data structure for holding variables. You can even inspect the active
execution context by calling the built-in context() function.
The map stores each identifier's key with a value and the value's type. For internal use by the interpreter, a type can be set to a key without a value. This makes it possible to check the types of values before they are computed.
Function
A function encapsulates a section of the abstract tree so that it can be run seperately and with
different arguments. The function body is a block, so adding async will cause the body
to run like any other async block. Unlike some languages, there are no concepts like futures or
async functions in Dust.
Functions are first-class values in Dust, so they can be assigned to variables like any other value.
# This simple function has no arguments and no return value.
say_hi = () <none> {
output("hi") # The "output" function is a built-in that prints to stdout.
}
# This function has one argument and will return a value.
add_one = (number <num>) <num> {
number + 1
}
say_hi()
assert_equal(add_one(3), 4)
Functions can also be anonymous. This is useful for using callbacks (i.e. functions that are called by another function).
# Use a callback to retain only the numeric characters in a string.
str:retain(
'a1b2c3'
(char <str>) <bool> {
is_some(int:parse(char))
}
)
Option
An option represents a value that may not be present. It has two variants: some and none.
say_something = (message <option(str)>) <str> {
either_or(message, "hiya")
}
say_something(some("goodbye"))
# goodbye
say_something(none)
# hiya
Dust includes built-in functions to work with option values: is_none, is_some and either_or.
Structure
A structure is a concrete type value. It is a value, like any other, and can be [assigned] (#assignment) to an identifier. It can then be instantiated as a map that will only allow the variables present in the structure. Default values may be provided for each variable in the structure, which will be propagated to the map it creates. Values without defaults must be given a value during instantiation.
struct User {
name <str>
email <str>
id <int> = generate_id()
}
bob = new User {
name = "Bob"
email = "bob@example.com"
}
# The variable "bob" is a structured map.
A map created by using new is called a structured map. In other languages it may be called a "homomorphic mapped type". Dust will generate errors if you try to set any values on the structured map that are not allowed by the structure.
Types
Dust enforces strict type checking. To make the language easier to write, type inference is used to allow variables to be declared without specifying the type. Instead, the interpreter will figure it out and set the strictest type possible.
To make the type-setting syntax easier to distinguish from the rest of your code, a type specification is wrapped in pointed brackets. So variable assignment using types looks like this:
my_float <float> = 666.0
Basic Types
The simple types, and their notation are:
- boolean
bool - integer
int - float
float - string
str
Number
The num type may represent a value of type int or float.
Any
The any type does not enforce type bounds.
None
The none type indicates that no value should be found after executing the statement or block, with
one expection: the none variant of the option type.
List Type
A list's contents can be specified to create type-safe lists. The list(str) type would only allow
string values. Writing list without the parentheses and content type is equivalent to writing
list(any).
Map Type
The map type is unstructured and can hold any key-value pair.
Iter
The iter type refers to types that can be used with a for loop. These include list,
range, string and map.
Function Type
A function's type specification is more complex than other types. A function value must always have its arguments and return type specified when the function value is created.
my_function = (number <int>, text <str>) <none> {
output(number)
output(text)
}
But what if we need to specify a function type without creating the function value? This is necessary when using callbacks or defining structures that have functions set at instantiation.
use_adder = (adder <(int) -> int>, number <int>) -> <int> {
adder(number)
}
use_adder(
(i <int>) <int> { i + 2 }
40
)
# Output: 42
struct Message {
send_n_times <(str, int) -> none>
}
stdout_message = new Message {
send_n_times = (content <str>, n <int>) <none> {
for _ in 0..n {
output(content)
}
}
}
Option Type
The option(type) type is expected to be either some(value) or none. The type of the value
inside the some is always specified.
result <option(str)> = none
for file in fs:read_dir("./") {
if file:size > 100 {
result = some(file:path)
break
}
}
output(result)
get_line_break_index(text <str>) <some(int)> {
str:find(text, '\n')
}
Custom Types
Custom types such as structures are referenced by their variable identifier.
File = struct {
path <str>
size <int>
type <str>
}
print_file_info(file <File>) <none> {
info = file:path
+ '\n'
+ file:size
+ '\n'
+ file:type
output(info)
}
Statements
TODO
Assignment
TODO
Blocks
TODO
Synchronous Blocks
TODO
Asynchronous Blocks
# An async block will run each statement in its own thread.
async {
output(random_integer())
output(random_float())
output(random_boolean())
}
data = async {
output("Reading a file...")
read("examples/assets/faithful.csv")
}
Break
TODO
For Loop
TODO
list = [ 1, 2, 3 ]
for number in list {
output(number + 1)
}
While Loop
TODO
A while loop continues until a predicate is false.
i = 0
while i < 10 {
output(i)
i += 1
}
If/Else
TODO
Match
TODO
Pipe
TODO
Expression
TODO
Expressions
TODO
Identifier
TODO
Index
TODO
Logic
TODO
Math
TODO
Value
TODO
New
TODO
Command
TODO
Built-In Values
TODO
Comments
TODO