Types

Cloth distinguishes values from managed references. A type determines assignment compatibility, available operations, and how copying behaves.

A class, interface, enum, or struct has nominal identity: matching members alone do not make two named types interchangeable.

Numeric types

Signed integers are int8, int16, int32, and int64. Unsigned integers are uint8, uint16, uint32, uint64, and byte. Floating-point types are float32 and float64.

The aliases int = int32, uint = uint32, and float = float32 are independent of the target. bool and char are separate primitive types, not numeric operands.

References and absence

References are non-null by default. Append ? to any value or reference type to admit absence: User?, string?, int32?, Status?, and Point? are all valid. Nullable values use tagged inline storage until converted to object?.

For arrays, element and array nullability are separate: User?[] contains nullable users; User[]? is a nullable array of non-null users. See nullability.

Inference and conversion

var infers a local’s exact type from its initializer. It needs a value: var value = null; and an initializer returning void are invalid.

Unsuffixed numeric literals can adopt an expected type when representable. Suffixes such as i8, u64, and f32 select an exact initial type instead. Existing numeric values widen implicitly only through specified lossless conversions. Reference widening follows class inheritance, declared interfaces, and object. Primitive, enum, and struct values box only when converted to object or object?. Arrays remain invariant. See conversions.