Skip to main content



FunC documentation was initially written by @akifoq.

FunC has the following built-in types.

Atomic types

  • int is the type of 257-bit signed integers. By default, overflow checks are enabled and lead to integer overflow exceptions.
  • cell is the type of TVM cells. All persistent data in TON Blockchain is stored in trees of cells. Every cell has up to 1023 bits of arbitrary data and up to four references to other cells. Cells serve as memory in stack-based TVMs.
  • slice is the type of cell slices. A cell can be transformed into a slice, and then the data bits and references to other cells from the cell can be obtained by loading them from the slice.
  • builder is the type of cell builders. Data bits and references to other cells can be stored in a builder, and then the builder can be finalized to a new cell.
  • tuple is the type of TVM tuples. Tuple is an ordered collection of up to 255 components with arbitrary value types that are possibly distinct.
  • cont is the type of TVM continuations. Continuations are used for controlling the flow of the TVM program execution. It is rather low-level object from the perspective of FunC, although paradoxically quite general.

Note that any of the types above occupy only a single entry in the TVM stack.

Absence of boolean type

In FunC, booleans are represented as integers; false is represented as 0 and true is represented as -1 (257 ones in binary notation). Logical operations are done as bitwise operations. When a condition is checked, every non-zero integer is considered a true value.

Null values

By the value null of TVM type Null, FunC represents the absence of a value of some atomic type. Some primitives from the standard library may be typed as ones returning an atomic type and actually return nulls in some cases. Others may be typed as ones excepting a value of an atomic type but work fine with null values too. Such behavior is explicitly stated in the primitive specification. By default, null values are prohibited and lead to a run-time exception.

In such a way, the atomic type A may be implicitly transformed into type A^? a.k.a. Maybe A (the type checker is agnostic to such a transformation).

Hole type

FunC has support for type inference. Types _ and var represent type "holes" which can later be filled with some actual type during type checking. For example, var x = 2; is a definition of variable x equal to 2. The type checker can infer that x has type int, because 2 has type int, and the left and right sides of the assignment must have equal types.

Composite types

Types can be composed into more complex ones.

Functional type

Types of the form A -> B represent functions with specified domain and codomain. For example, int -> cell is the type of function that takes one integer argument and returns a TVM cell.

Internally, values of such types are represented as continuations.

Tensor types

Types of the form (A, B, ...) essentially represent ordered collections of values of types A, B, ..., which all together occupy more than one TVM stack entry.

For example, if a function foo has type int -> (int, int), it means that the function takes one integer and returns a pair of them.

A call of this function may look like (int a, int b) = foo(42);. Internally, the function consumes one stack entry and leaves two of them.

Note that from a low-level perspective, value (2, (3, 9)) of type (int, (int, int)) and value (2, 3, 9) of type (int, int, int), are represented in the same way as three stack entries 2, 3 and 9. For the FunC type checker they are values of different types. For example, code (int a, int b, int c) = (2, (3, 9)); wouldn't be compiled.

A special case of the tensor type is the unit type (). It is usually used to represent the fact that a function doesn't return any value or has no arguments. For example, a function print_int would have type int -> () and the function random has type () -> int. It has a unique inhabitant () which occupies 0 stack entries.

Type of form (A) is considered by type checker as the same type as A.

Tuples types

Types of the form [A, B, ...] represent TVM tuples with specific lengths and types of components known in compile time. For example, [int, cell] is the type of TVM tuple which length is exactly 2, and where the first component is an integer and the second is a cell. [] is the type of empty tuples (having the unique inhabitant—the empty tuple). Note that in contrast to the unit type (), the value of [] occupies one stack entry.

Polymorphism with type variables

FunC has Miller-Rabin type system with support for polymorphic functions. For example, the following function:

forall X -> (X, X) duplicate(X value) {
return (value, value);

is a polymorphic function which takes a (single stack entry) value and returns two copies of this value. duplicate(6) will produce values 6 6, and duplicate([]) will produce two copies [] [] of an empty tuple.

In this example, X is a type variable.

See more info on this topic in the functions section.

User-defined types

Currently, FunC has no support for defining types except for the type constructions described above.

Type width

As you may have noticed, every value of a type occupies some number of stack entries. If it is the same number for all values of the type, this number is called type width. Polymorphic functions can currently be defined only for types with fixed and known in advance type width.