Category Fixed Width Types

Home › All Categories › Fixed-Width Types

Fixed-Width Types

The names

Ten type names that say their width, each the same type as another built-in name:

tiko	Same type as	Bytes	Range
int8	Byte	1	-128 to 127
uint8	UByte	1	0 to 255
int16	Short	2	-32768 to 32767
uint16	UShort	2	0 to 65535
int32	Long	4	-2147483648 to 2147483647
uint32	ULong	4	0 to 4294967295
int64	LongInt	8	-9223372036854775808 to 9223372036854775807
uint64	ULongInt	8	0 to 18446744073709551615
float32	Single	4	IEEE 754 single precision
float64	Double	8	IEEE 754 double precision

The sizes are the same on every target.

Why they exist

The names in the second column don't say how wide they are, and the same names mean different widths in C. long is 32 bits in tiko on every target, but in C it is 32 bits on Windows and 64 on Linux and macOS. The names above follow Go, Nim and C's <stdint.h> (int32_t), so the width is in the name.

dim as uint8 red = 255
dim as int64 total = 9223372036854775807
dim as float32 ratio = 1 / 3
print red; total; ratio

They are the same types

Each name is a typedef, not a new type: int32 is long. Values mix freely with no conversion, a procedure that takes long accepts an int32, and an overload on int64 is the overload on longint. The names leave no trace in the ABI, so a procedure's linker symbol is the one it gets with the names they stand for:

namespace N
    sub F( byval a as int8, byval b as uint16, byval c as int32, _
           byval d as uint64, byval e as float32, byval g as float64 )
    end sub
end namespace
'' _ZN1N1FEatiyfd under tiko

A program's own declaration wins

The names are not keywords. tiko declares them for every program, but anything the program declares with the same name at module level replaces tiko's, with no diagnostic. Some headers declare these names themselves: win/basetsd.bi declares type INT32 as long. Such code compiles:

type INT32 as long                      '' the same type: nothing changes
type uint8 as integer                   '' a different type: uint8 is now 8 bytes
dim shared float32 as long = 5          '' not a type: float32 is now this variable
function uint16( byval x as long ) as long
    return x + 1
end function
print sizeof( INT32 ); sizeof( uint8 ); float32; uint16( 6 )

This prints 4 8 5 7. Because tiko ignores case, INT32 and int32 are the same name.

A name used above the program's declaration is fine when the declaration gives it the type tiko does, as win/basetsd.bi does: dim as int32 n and then #include "windows.bi" compiles, and so does a module whose interface uses int32, imported by a program that includes windows.bi, and so does an alias such as type base_id as int32 used above it. Above a declaration of a different type, such as a type uint8 as integer written below a dim as uint8, the use is error TKO0784 (TKO0811 through an alias): the code above it would otherwise change meaning.

The program's declaration replaces the name throughout the program, not just in its own file. jpeglib.bi declares type INT32 as clong, which is 8 bytes on Linux x86-64. A Linux program that includes it therefore gets an 8-byte int32.

What is not included

• integer and uinteger have no fixed-width name: they are the width of a pointer on the target.
• boolean stays as it is, and there is no bool. tiko ignores case, so bool would be BOOL, which win/windef.bi declares as a 4-byte long.
• Pointers are spelled with ptr: int32 ptr, any ptr.
• Strings keep their own names: string, zstring, wstring and ustring.

---

Home · Keyword Index