Bit fields pack many small values into one integer-sized record. Think of them as a bit-packed struct.
Declare one with the bit_field keyword followed by the backing type: an integer or a fixed-length array of integers. Each field gives a type, then `|` and the number of bits it occupies. Field names and bit widths must add up to fit inside the backing type.
Fields are read and written with dot notation, just like a struct. Writing truncates the value to the field's bit width. Reading a signed field sign-extends it back to its declared type.
$ odin run bit-fields.odin -file flags size: 2 active: true mode: 2 count: -3 count after: 3 header size: 4 version: 9 kind: Green seq: 7 seq max: 15
Odin's bit_field layout is well-defined (least-significant bit first), unlike C's bit fields which are not portable across compilers and targets.
package main import "core:fmt" // 1 + 2 + 3 + 10 = 16 bits, the size of the u16 backing type. Flags :: bit_field u16 { active: bool | 1, mode: u16 | 2, count: i32 | 3, padding: u16 | 10, } // An enum can be stored in a bit field to save space. Color :: enum u8 {Red, Green, Blue} // 4 + 2 + 4 + 22 = 32 bits, filling the u32 exactly. Header :: bit_field u32 { version: u16 | 4, kind: Color | 2, seq: u8 | 4, rest: u32 | 22, } main :: proc() { // Fields default to their zero value. f := Flags{} fmt.println("flags size:", size_of(Flags)) // Write and read fields with dot notation. f.active = true f.mode = 2 f.count = -3 fmt.println("active:", f.active, "mode:", f.mode, "count:", f.count) // The count field is only 3 bits wide. Its signed range is // -4..3, so a value that does not fit is truncated. f.count = 3 fmt.println("count after:", f.count) // Enum and integer fields can share a record. h := Header{} h.version = 9 h.kind = .Green h.seq = 7 fmt.println("header size:", size_of(Header)) fmt.println("version:", h.version, "kind:", h.kind, "seq:", h.seq) // A 4-bit field holds values from 0 to 15. h.seq = 15 fmt.println("seq max:", h.seq) }