A high-level companion language for Rust.
Important
This is a work in progress and under development. It currently is in a state of a hobby project and most features described here are not yet implemented. I am working on this project in my free time - if you like the ideas presented here, and want to help, feel free to contact me or start a discussion here on GitHub.
Galvan is not intended as a replacement for Rust. It is a companion language that transpiles to Rust. That means, Galvan's abstractions are not always zero-cost, instead Galvan tries to pick a sane default choice for most use cases.
Galvan is not intended for low-level programming - so you should not build a parser, compiler or audio compression library with it. Instead, Galvan is intended to be used for high-level applications, like CLI tools, web servers and so on. The ultimate goal is full Rust interoperability, so you can write your application in Galvan and rely on the full Rust ecosystem.
Rust is a great language, but it is not always the best choice for every task. Rust's syntax can be quite verbose and its type system and borrow checker - while extremely powerful - can be a burden for high-level applications and overwhelming for beginners. For low-level libraries and so-called "systems programming", Rust hits the sweet spot between helpful abstractions and being in control of implementation details. For application-level programming however, being provided with sensible choices for common use cases is important. This is where Galvan comes in: It provides a concise syntax and simplified way of writing Rust - without worrying about lifetimes, ownership and so on.
Galvan is a modern programming language that transpiles to Rust. It provides a concise syntax while leveraging the full power of Rust's ecosystem.
In Galvan, main
is not a function but an "entry point".
main {
let name = "Galvan"
print("Welcome to {name}, the modern language!")
}
Note that Galvan strings always support inline format arguments.
Like in Rust, functions are defined with the fn
keyword and return the value of the last expression:
fn add(a: Int, b: Int) -> Int {
a + b
}
Very short functions can also be defined with = and have their return type inferred:
Warning
Defining functions with '=' is not implemented yet
fn add(a: Int, b: Int) = a + b
Those functions are not allowed to have newlines in their body.
Types in Galvan are defined with the type
keyword.
/// A struct definition
pub type Color {
r: Int
g: Int
b: Int
}
// Structs can also use the named tuple syntax
pub type Person(name: String, age: Int)
/// A type alias
pub type Human = Person
/// A tuple type
pub type Couple(Person, Person)
Enums can have associated values, either for all variants or for specific variants. Enums are also declared using the type
keyword:
Warning
Enums are not implemented yet
/// An enum type
/// Enums can have general fields that are accessible to all enum variants
pub type Theme(name: String) {
Plain
/// Like in Rust, enum variants can have associated values, either named or unnamed
Monochrome(Color)
/// Unlike in Rust, '(' is also used for enum variants with named fields
Dark(background: Color, foreground: Color)
Light(background: Color, foreground: Color)
}
All functions are declared top-level. If their first parameter is named self
, they can be called as member functions:
pub type Dog(name: String)
fn bark(self: Dog) {
print("{self.name} barks")
}
main {
let dog = Dog(name: "Bello")
dog.bark()
}
Warning
Default values are not implemented yet
Struct types in Galvan can allow ommitting certain attributes when created with the default initializer. To do so,
type Book {
title: String = "Lorem Ipsum"
content: String = "Lorem ipsum dolor sit amet..."
}
main {
let book = Book()
}
In case the type can be constructed without arguments, Rust's Default
trait is automatically implemented.
Galvan features syntactic sugar for collection types:
pub type IntArray = [Int] // This is a Vec
pub type StringSet = {String} // This is a HashSet
pub type MyDict = {String: Int} // This is a HashMap
pub type OrderedDict = [String: Int] // This is an IndexMap
Ordered types use []
, unordered types use {}
.
Galvan provides concise syntax for optionals and result types:
type OptionalInt = Int?
type FileOrErr = File!
type FileOrIoErr = File!IoError
The error variant is specified after the !
symbol. If it is not given, a flexible error type is used.
Warning
!
and ?
are not implemented yet
fn open_file(path: String) -> File! {
let file = File::open(path)!
let contents = file.read_to_string()?.find("foo")?.uppercase() else { "" }
contents
}
!
operator unwraps the result and early returns if the result is an error. This is identical to the ?
operator in Rust.
?
is the safe call operator in Galvan. The subsequent expression is only evaluated if the result is not an error and not none.
else
can also be used as the null-coalescing operator (since you can use else after every expression that is an optional or a result type), you can use it to provide a default if the left-hand side expression is none.
Galvan supports union types everywhere where a type identifier is expected:
Warning
Union types are not implemented yet
fn print_value(value: Int | String) {
print("Value: {value}")
}
By default, arguments are passed by value. If the argument needs to be mutated, the mut
keyword can be used to pass it by reference:
For consistency, the let
keyword is allowed as well but redundant as parameters are passed by value by default.
fn add_one(mut value: Int) {
value += 1
}
// Using `let` is not necessary here but allowed
fn incremented(let value : Int) -> Int {
value + 1
}
Galvan's mut value: T
would be equivalent to Rust's value: &mut T
. Galvan does not have immutable references, as all values are copy-on-write.
// No copy is happening here as the value is not mutated
// Arguments are passed by value by default
fn bark_at(self: Dog, other: Dog) {
print("{self.name} barks at {other.name}")
}
// A copy is happening here as the value is mutated
fn shout_at(self: Dog, other: Dog) {
// Redeclaring is neccessary as value parameters cannot be mutated
let other = other
// Copy is happening here
other.name = other.name.uppercase()
print("{self.name} shouts at {other.name}")
}
fn grow(mut self: Dog) {
// This mutates the original value as it is passed by reference
self.age += 1
}
References that are allowed to be stored in structs have to be declared as heap references. This is done by prefixing the declaration with ref
:
pub type Person {
name: String,
age: Int,
// This is a heap reference
ref dog: Dog
}
main {
// Note that constructors use '(' with named arguments
ref dog = Dog(name: "Bello", age: 5)
// The `dog` field now points to the same entity as the `dog` variable
let person = Person(name: "Jochen", age: 67, dog: ref dog)
dog.age += 1
print(person.dog.age) // 6
print(dog.age) // 6
}
Heap references use atomic reference counting to be auto-freed when no longer needed and are always mutable.
In contrast to let
and mut
values, ref
values. They follow reference semantics, meaning that they point to the same object. For this reason, they are always mutable.
When calling a function with mut
or ref
parameters, you have to annotate the argument respectively. This is not the case for the receiver of a member function.
fn make_uppercase(mut arg: String) { ... }
fn store(ref arg: String) { ... }
main {
ref my_string = "This is a heap ref"
// Argument must be annotated as mutable
make_uppercase(mut my_string)
// Argument must be annotated as ref
store(ref my_string)
}
By annotating the argument as mut
, the caller acknowledges that the given argument might be mutated in-place when calling this function.
Immutable variables or members of immutable struct instances (declared with let
) cannot be passed as mut
.
By annotating the argument as ref
, the caller acknowledges that the function might store a mutable (heap) reference.
Only variables and members declared as ref
can be passed as ref
Like in Rust, loops can yield a value:
Warning
Loops are not implemented yet
mut i = 0
let j = loop {
if i == 15 {
return i
}
i += 1
}
print(j) // 15
print(i) // 15
For loops are also supported:
Warning
For loops are not implemented yet
for 0..<n {
print(it)
}
The loop variable is available via the it
keyword, but can also be named explicitly using closure parameter syntax:
for 0..<n |i| {
print(i)
}
Note that ranges are declared using ..<
(exclusive upper bound) or ..=
(inclusive upper bound).
Warning
Nested if-else is not implemented yet
if condition {
print("Condition is true")
} else if other_condition {
print("Other condition is true")
} else {
print("No condition is true")
}
You can use try to unwrap a result or optional:
Warning
Implicit arguments via it
are not implemented yet
try potential_error {
print("Optional was {it}")
} else {
print("Error occured: {it}")
}
The unwrapped variant is available via the it keyword, like in closures. You can also name it using closure parameter syntax to declare them explicitly:
try potential_error |value| {
print("Optional was {value}")
} else |error| {
print("Error occured: {error}")
}
Like if
, you can also use try
without an else branch:
try potential_error |value| {
print("Optional was {value}")
}
let optional = try potential_error |successful| { successful }
This can be useful to pass the unwrapped value to a function, or to convert a Result to an Optional.
Return values are implicit, however you can use the return
keyword to return early:
Warning
Return keyword is not implemented yet
fn fib(n: Int) -> Int {
if n <= 1 {
return n
}
fib(n - 1) + fib(n - 2)
}
Returning an error early is done using the throw
keyword:
Warning
Throw keyword is not implemented yet
fn checked_divide(a: Float, b: Float) -> Float! {
if b == 0 {
throw "Division by zero"
}
a / b
}
In Galvan, type identifiers always start with an upper case letter. Using a lower case letter instead introduces a type parameter:
Warning
Generics are not implemented yet
type Container {
value: t
}
fn get_value(self: Container<t>) -> t {
self.value
}
Bounds can be specified using the where
keyword:
fn concat_hash(self: t, other: t) -> t where t: Hash {
self.hash() ++ other.hash()
}
Galvan offers a wide range of builtin operators. While all of them have an ASCII variant, Galvan also accepts a unicode symbol where it makes sense.
Arithmetic operators:
+
: Addition-
: Subtraction*
: Multiplication/
: Division%
: Remainder^
: Exponentiation
Note
Galvan does not offer unicode alternatives for logical operators
as ∧
and ∨
could be confused with v
and ^
respectively.
If you want to use unicode operators, you can define them yourself.
Logical operators:
and
,&&
: Logical andor
,||
: Logical orxor
,^^
: Logical xornot
,!
: Logical not
Warning
Bitwise operators are implemented yet Bitwise operators are prefixed with b:
b|
: Bitwise orb&
: Bitwise andb^
: Bitwise xorb<<
: Bitwise left shiftb>>
: Bitwise right shiftb~
: Bitwise not
Comparison operators:
==
: Equality!=
,≠
: Inequality<
: Less than<=
,≤
: Less than or equal>
Greater than>=
,≥
:: Greater than or equal===
,≡
: Pointer equality, only works for heap references!==
,≢
: Pointer inequality, only works for heap references
Collection operators:
++
: Concatenation--
: Removal**
: Repetition[]
: Indexing[:]
: Slicingin
,∈
,∊
: Membership
Range operators:
..<
: Exclusive Range..=
: Inclusive Range+-
,±
: Inclusive Range around a value (tolerance)
Galvan supports Unicode and custom operators:
Warning
Custom operators are not implemented yet
@infix("⨁")
fn xor(lhs: n, rhs: n) = lhs ^^ rhs
@prefix("√")
fn sqrt(n: Float) = n.sqrt()
main {
let a_bool = true
let other_bool = false
let value = if a_bool ⨁ other_bool { √16.0 } else { 3.0 }
}
This section defines custom infix ⨁
and prefix √
operators.
Note that no whitespace is allowed between a prefix operator and the operands.
Infix operators have to be surrounded by whitespace.
Prefix operators generally take precedence over infix operators.
Closures are defined using the parameter list syntax:
let add = |a, b| a + b
Closure types use the arrow syntax:
fn map(self: [t], f: t -> u) -> [u] {
mut result = []
for self {
result.push(f(it))
}
result
}
Functions with trailing closures are allowed to omit the parameter list and the () around the parameter list:
iter
.map { it * 2 }
// Trailing closures with only one parameter can use the it keyword instead of naming it explicitly
.filter { it % 2 == 0 }
// The parameter list before the trailing closure can be omitted
.reduce start |acc, e| { acc + e }
Trailing closures can also use numbered parameters instead of giving a parameter list
iter
.map { #0 * 2 }
.filter { #0 % 2 == 0 }
.reduce start { #0 + #1 }
In a statement or as the right-hand side of an assignment, parentheses for function calls can be omitted:
fn add(a: Int, b: Int) -> Int {
a + b
}
main {
let result = add 2, 3
print result // 5
}
As this syntax is only allowed in assignments and statements, you cannot use it in i.e. function parameters. This avoids ambiguity. It is also not allowed to use this syntax for functions that take no arguments to avoid confusing it with a variable.
Every obstacle to writing unit tests is a unit test that is not written. For this reason, Galvan provides a concise syntax to quickly write unit tests in any .galvan file:
test {
assert 2 == 2
}
test "Ensure that addition works correctly" {
assert 2 + 2 == 4
}
Like 'main', 'test' is not a function but an entry point. Tests can take a string as a description. Although this is optional, adding a brief description to your unit tests is highly encouraged.
While Galvan uses semicolons to separate statements, Galvan infers semicolons on newlines when:
- the next line starts with an alpha-character (or an underscore) as the first non-whitespace character
- the next line starts with
{
,(
,[
, or'
,"
as the first non-whitespace character
Regardless of the rules above, Galvan does not infer a semicolon when the current line itself is not a valid statement. Galvan also infers commas for struct type declarations if a newline is used after each field instead.