Fyr supports the following data types:
uint8, uint16, uint32, uint64 // Unsigned integers of fixed size
int8, int16, int32, int64 // Signed integers of fixed size
float, double // 32bit and 64bit floating point numbers
byte // Platform dependend
// Smallest addressable unsigned unit
// Usually 8 bits
char // Platform dependend
// Smallest addressable signed unit
// Usually 8 bits
bool // A boolean of 8 bits size that is false or true
string // Immutable UTF-8 strings
rune // A 32-bit unicode character
int // Platform dependend
// Signed value, usually 32 bits even on 64-bit platforms
uint // Platform dependend
// Signed value, usually 32 bits even on 64-bit platforms
WebAssembly uses a 32-bit address space by default, hence Fyr treats int as an alias for int32 in WebAssembly. On Arduinos, int is an alias for int16. When compiling for other targets than WebAssembly MVP, int can be an alias for int16 or int32 or int64 depending on the target platform, but the default is int32.
The char type is required to interface with C functions.
Idiomatic Fyr code should not use char otherwise, since it is a platform-dependent type by definition.
The types float and double might be renamed to float32 and float64 in a future version of the language.
All data types have a default value of 0. If a value is not explicitly initialized, Fyr ensures that it is initialized with zeros.
Fyr does not perform any implicit type conversion. Explicit casts are required as in the following example:
var x int32 = 42
var y int16 = <int16>x
All numeric types including bool can be explicitly casted into each other.
Boolean literals are true and false.
Integer literals are either written as decimals as in 1234 or in hexadecimal as in 0xffef.
Floating point literal are of the form 12.34 or .34.
Fyr strings are immutable and store data in UTF-8 encoding. Strings are stored as pointers to string data. Hence, string assignment is very fast. String comparison, however, compares the value of the strings these internal pointers are pointing at.
The index operator [index] returns bytes of the UTF-8 encoding.
To obtain the byte count of a string, call its len() member function.
To access the single unicode characters (of type rune) encoded in the UTF-8 string, a for loop can be used.
func print(i int, r rune) {
...
}
export func main() int {
var str = "Übung"
for(let i, r in str) {
print(i, r) // Loops 5 times
}
return str.len() // Returns 6
}
The above example would call print 5 times although the string is 6 bytes long.
The first two bytes together represent the rune 'Ü'.
There are several ways of denoting runes:
'a' // value 0x61
'\a' // value 7
'\b' // value 8
'\t' // value 9
'\n' // value 10
'\v' // value 11
'\f' // value 12
'\r' // value 13
'\\' // value 0x5c
'\'' // value 0x27
'\"' // value 0x22
'\x3f' // value 0x3f
'\u123f' // value of 0x123f
'\U0012345f' // value of 0x12345f
Runes can be used inside a string as well as in "\xdcbung" which is equivalent to "Übung".
Strings can be coverted to []byte and the other way round as in the following example:
var arr []byte = <[]byte>"Hallo"
var str string = <string>arr
Note that this conversion implies a copy because strings are immutable and slices are not.
Arrays are value types of fixed length. Assigning one array to another results in a copy of this array. When accessing an array, Fyr checks that the index is within the bounds of the array. An out-of-bound index causes the application to abort.
func search(arr [4]int, val int) int {
for(let i, v in arr) {
if (v == val) {
return i
}
}
return 0
}
export func main() int {
var arr [4]int = [...]
arr[0] = 123
arr[1] = 234
return search(arr, 234)
}
Note that variables need explicit initialization before the variable can be used.
The initialization expression [...] initializes an array or slice with its default value, which is zero in this case.
In the above example, the array is copied when search is called.
This is inefficient, and the search function works only for arrays of a fixed size.
Therefore, Fyr supports slices, which are essentially pointers to an underlying array.
func search(arr &[]int, val int) int {
for(let i, v in arr) {
if (v == val) {
return i
}
}
return 0
}
export func main() int {
var arr [4]int = [123, 234, 345, 456]
return search(arr[:], 234)
}
The above example uses an array initialization, which is just a shortcut to populate an array.
It is possible to combine this with the default initialization with [12, 23, ...].case “cl
In this case the first two array elements are initialized with 12 and 23 and all remaining elements are initialized with the default value.
The slice operator [:] creates a slice which points to the underlying array and contains all elements of arr.
For example, [2:] would include all elements of the array starting by the array element at position 2.
An access to a slice checks that the index is within the bounds of the slice. Out-of-bound access aborts the application.
A slice is denoted for example as []int.
Note that slices are pointers to an array.
Assigning one slice to another just assigns this internal pointer, but the array is not copied.
However, in this example the search function accepts a local reference slice of type &[]int.
When a function accepts a local reference type as a parameter, the compiler verifies that the function will not store any pointers to this data that live longer than the function invocation.
This allows main to hand out a slice to an array which is on the stack, because Fyr can verify that no pointers to arr live longer than the stack frame in which arr is stored.
If search accepts []int instead, main must allocate the array on the heap, because otherwise Fyr cannot ensure that there are no dangling pointers to the array after main returns.
Fyr features pointers and is memory safe, i.e. access to free’d memory areas always causes an abort.
func search(arr []int, val int) int {
for(let i, v in arr) {
if (v == val) {
return i
}
}
return 0
}
export func main() int {
let arr []int = [123, 234, 345, 456]
return search(arr, 234)
}
In the above example, [123, 234, 345, 456] allocates the array on the heap and returns a slice to it.
Hence, arr is now a slice type.
Due to type inference, the following two statements are equivalent:
let arr []int = [123, 234, 345, 456]
let arr = [123, 234, 345, 456]
To allocate a slice of variable size, use the make operator:
let a = 100
let arr []int = make<int>(100) // len 100
let arr2 []int = make<int>(100, 200) // len 100, cap 200
The length of an array, string or slice can be obtained with the len operator.
The length of the array a slice is pointing to can be obtained with the cap operator.
The elements of a slice can be copied using the clone operator.
It returns a slice which points to the cloned array.
The append operator can append to a slice, while copy allows copying elements between slices.
A tuple type is an anonymous struct.
var result (string, bool) = (null, false)
In the above example (string, bool) is a tuple type.
The expression (null, false) is a literal that is assignable to a tuple.
The main use case of tuples is complex return values.
func lookup(name string) (string, bool) {
...
}
The above function returns a tuple and could thus be assigned to the result variable.
result = lookup("foobar")
Furthermore, tuples returned by a function can be decomposed upon assignment.
id, ok = lookup("foobar")
Tuples can be used in other cases as well, but structs are often easier to use. The advantage of tuples is that they are anonymous. If every function returning a tuple uses a struct instead, the code would be cluttered with structs.
type lookupResult struct {
id string
ok bool
}
func lookup(name string) lookupResult {
}
In the above case tuples are easier to use.
An Or-Type is a union of types with a discriminator. A value of an Or-Type must match one of these types and the discriminator stores which type that is.
var v string | bool = false
v = "Hello"
Using the is keyword, we can test the type of value being stored in the Or-Type.
var v string | bool = false
if (v is bool) {
...
}
v = "Hello"
if (v is string) {
...
}
A value can be extracted using a cast operation. If the Or-Type stores a different value, the program will panic.
var v string | bool = "Joe"
var name = <string>v
Or types can be used together with symbols to construct enumerations. For example, we can define a type to store a person’s gender.
type Gender "male" | "female" | string
var g Gender = "male"
if (g is "male") {
...
}
g = "female"
if (g == "female") {
...
}
g = "complicated"
if (g is string) {
...
}
In the above example "male" is a symbol type.
Note that is denotes a type and not a string value.
A symbol is a type that has only one possible value: a string literal of its own name, e.g. "male" in the above example.
Internally, a symbol is not necessarily stored as a string.
The compiler may opt to enumerate the symbols and store a number instead.
Assigning a string value (not string literal!) to the Gender type, sets g's value to a string type, no matter what the content of the string.
type Gender "male" | "female" | string
var s = "male" // The type of s is now string
var g Gender = s // The type stored in g is now a string
if (g is string) { // True
...
}
if (g is "male") { // False
...
}
The default value of an Or-Type is the default value of its first type option. In the case of
type Gender "male" | "female" | string
type Person struct {
name string
gender Gender
}
var p Person = {name: "Joe"}
the field gender is not explicitly initialized and therefore defaults to "male", because the symbol type "male" is the first type option on Gender and "male" is the default value of this symbol type.
A slice of bytes or chars can be converted to a string via <string>slice.
In this case the memory owned by the slice is used for the string.
The expression of the typecast (slice in the above example) will therefore become inaccessible, because this typecast freezes the slice since strings are immutable.
A take might be required to take ownership of the byte array.
The last byte of the slice must be zero, otherwise the program aborts, because Fyr strings are always zero terminated.
If the slice start is not equivalent to the start of the underlying array, the conversion moves the data to the beginning of the array. This has O(n) complexity as in the following example.
let slice []byte = [65, 66, 67, 68, 0]
let str = <string>slice[1:]
To get O(1) complexity, the slice start must be equivalent with the start of the underlying array, as in the following example.
let slice []byte = [65, 66, 67, 68, 0]
let str = <string>slice
Converting a null slice results in a null string.
Converting a slice of length zero aborts the program, because the last byte of the slice must be zero.
A slice of bytes or chars can be converted to a string by allocating new memory via <string>clone(slice).
This is a special case, because the compiler will allocate one additional byte when cloning the slice and set it to zero.
This allows the following string conversion to succeed.
In this construction, the slice does not have to end with a zero byte.
// No trailing zero in the slice
let slice []byte = [65, 66, 67, 68]
let str = <string>clone(slice)
Converting a null slice results in a null string.
Converting a slice of length zero results in a string of length zero.
A string can be casted to a slice or unique slice. This copies the slice and returns ownership of the new slice. The last byte of the slice is a zero, because Fyr strings are zero terminated.
let slice = <[]byte>"Hallo"
let slice2 = <^[]byte>"Hallo"
A null string results in a null slice.
A pure value can be copied byte by byte. This is true for for all data-types except pointer-like types. Unsafe pointers are pure values, however. Pointer-like types can either not be copied, because copying could result in an object having multiple owning pointers, or an object is not destructed when its pointer is overwritten, or reference-counting does not happen correctly in case of reference pointers.
When using the copy or clone operators, only pure values can be copied or cloned.