4.7. Types

Type <- ConstType / PointerType / ArrayType / FunctionType / BasicType

ConstType <- 'const' BasicType

PointerType <- BasicType '*'

ArrayType <- BasicType '[' Expression ']'

FunctionType <- BasicType 'function' '(' FunctionTypeArgs? ')'

FunctionTypeArgs <- BasicType (',' BasicType)* ','?

Todo

define like any of these

BasicType <- 'void' / 'bool' / 'float32' / 'float64' / 'int8' / 'int16' / 'int32' / 'int64' / 'intaddr' / 'intmax' / 'intsize' / 'uint8' / 'uint16' / 'uint32' / 'uint64' / 'uintaddr' / 'uintmax' / 'uintsize' / 'struct' identifier / 'enum' identifier / 'union' identifier / '(' Type ')'

void denotes the empty type.

bool denotes the Boolean type, with two values: true and false, represented as 1 and 0, respectively.

float32 and float64 denote the binary32 and binary64 IEEE 754 floating-point types, respectively.

int8, int16, int32, and int64 denote signed, two’s-complement integers with sizes 8 bits, 16 bits, 32 bits, and 64 bits, respectively.

uint8, uint16, uint32, and uint64 denote unsigned integers with sizes 8 bits, 16 bits, 32 bits, and 64 bits, respectively.

intmax is a synonym for the largest signed integer type supported by the compiler.

uintmax is a synonym for the largest unsigned integer type supported by the compiler.

uintaddr is a synonym for an unsigned integer type large enough to hold any memory address valid on the compilation target; the specific type is implementation defined.

intaddr is a synonym for a signed integer type at least as large as uintaddr.

uintsize is a synonym for an unsigned integer type large enough to hold any number of bytes which may be contiguously allocated on the compilation target; the specific type is implementation defined.

intsize is a synonym for a signed integer type at least as large as uintsize.

struct, enum, or union followed by an identifier denotes the type with the given nature and name, which should be available in the compilation context when used.

Enclosing a Type in parentheses does not have semantic significance, but allows for syntactic disambiguation of constructs that would otherwise be visually ambiguous.

Multi-byte integer types should be represented as either big endian or little endian based on the preference of the compilation target platform, i.e. endianness is implementation defined. Compilers targeting less-than-64-bit CPUs may omit support for some explicitly sized basic types, but it would be nice if they provided software support for types not supported in hardware.