Type system

Every dynamic field in a RAD file is described by a type. Types are what let the format store each field at the narrowest width that fits and let a reader parse a value it has never seen before. A type is encoded as a single u8 identifier in a tag descriptor.

Type identifiers

ID	Type	Width	Notes
0	bool	1 byte
1	u8	1 byte
2	u16	2 bytes
3	u32	4 bytes
4	u64	8 bytes
5	f32	4 bytes	IEEE-754
6	f64	8 bytes	IEEE-754
7	array	variable	see below
8	string	variable	u16 length + bytes
9	u128	16 bytes
10	i8	1 byte
11	i16	2 bytes
12	i32	4 bytes
13	i64	8 bytes
14	i128	16 bytes

All multi-byte integers and floats are little-endian.

Note

The identifier values are not contiguous by width — the fixed-width unsigned integers and floats (0–6) and the two variable-length types (7, 8) were assigned first; the wider and signed integers (9–14) were added later. Use the table, not the ordering, as the source of truth.

Fixed-width types

The scalar types (bool, the signed/unsigned integers, and the floats) are stored as their natural little-endian byte representation, with no length prefix. A bool is a single byte, conventionally 0 or 1.

Strings (ID 8)

A string is:

[ length : u16 ][ length bytes ]

the bytes being the (UTF-8/ASCII) characters, with no trailing NUL. Reference names and tag names are encoded this way.

Arrays (ID 7)

An array descriptor records two further type identifiers — the type of its length field and the type of its elements:

descriptor:  [ 7 ][ length_type : u8 ][ element_type : u8 ]
value:       [ length : length_type ][ element … ]   # `length` elements

For example, an array of u32 with a u32 length is descriptor 7, 3, 3, and a value is a u32 count followed by that many u32s. Arrays do not nest — the element type must be a scalar, not another array or string.

Signed and wide integers

The signed integers (i8–i128) and the wide unsigned integer (u128) round out the set for fields that need them — for instance a barcode that does not fit in 64 bits. They are stored little-endian like the other integers. Not every consumer supports every width in every position; a reader that encounters a width it cannot handle for a given field should report a clear error rather than misparse.