Interop

LIGO can work together with other smart contract languages on Tezos. However, data structures might have different representations in Michelson and not correctly match the standard LIGO types.

Michelson types and annotations

Michelson types consist of or's and pair's, combined with field annotations. Field annotations add constraints on a Michelson type, for example a pair of (pair (int %foo) (string %bar)) will only work with the exact equivalence or the same type without the field annotations.

To clarify:

(pair (int %foo) (string %bar))

works with

(pair (int %foo) (string %bar))

works with

(pair int string)

works not with

(pair (int %bar) (string %foo))

works not with

(pair (string %bar) (int %foo))

info

In the case of annotated entrypoints --- the annotated or tree directly under parameter in a contract --- you should use annotations, as otherwise it's unclear which entrypoint you are referring to.

Michelson layout of LIGO data structures

Right-comb tree by default

By default, the Michelson data representation of LIGO data structures is a location retaining right combed tree, like this:

  (or
    (unit %elephant)
    (or (unit %dog)
        (unit %cat)))

You can use the decorator @layout comb to make this choice explicitly:

type animal =
@layout("comb")
| ["Elephant"]
| ["Dog"]
| ["Cat"];

The decorator @layout("comb") can also be used on record types:

type artist =
@layout("comb")
{
  genre : string,
  since : timestamp,
  name  : string
};

The next section discusses an alternative layout, which used to be the default one until LIGO version 1.0.

Alternative alphabetically ordered left-balanced tree layout

Before version 1.0, LIGO used to translate its datatypes into an alphabetically ordered left balanced tree by default. So, for example:

type animal = | ["Elephant"] | ["Dog"] | ["Cat"];

will translate to:

(or
  (or
    (unit %cat)
    (unit %dog))
  (unit %elephant))

This behaviour can be obtained using @layout("tree").

Different Michelson annotations

If the Michelson annotation should be different from the LIGO attribute, give it as an argument to the LIGO attribute. For example:

type animal =
| @annot("memory") ["Elephant"]
| @annot("face") ["Dog"]
| @annot("fish") ["Cat"]

will result into:

(or
  (or
    (unit %fish)
    (unit %face))
  (unit %memory))

The decorator @annot("<name>") can also be used on record field annotations:

type artist = {
  @annot("style") genre: string,
  @annot("from") since: timestamp,
  @annot("performer") name: string
}

If the decorators @layout("comb") and @annot("<name>") are not adequate enough for your use-case, LIGO has more advanced advanced interoperability features, which we will we discuss next.

Advanced interoperability with Michelson

To interoperate with existing Michelson code or to be compatible with certain development tooling, LIGO has two special interoperation types: michelson_or and michelson_pair. These types give the flexibility to model the exact Michelson output, including field annotations.

Take for example the following Michelson type that we want to interoperate with:

(or
  (unit %z)
  (or %other
    (unit %y)
    (pair %other
      (string %x)
      (pair %other
        (int %w)
        (nat %v)))))

To reproduce this type we can use the following LIGO code:

type w_and_v = michelson_pair<[int, "w", nat, "v"]>;
type x_and = michelson_pair<[string, "x", w_and_v, "other"]>;
type y_or = michelson_or<[unit, "y", x_and, "other"]>;
type z_or = michelson_or<[unit, "z", y_or, "other"]>;

If you do not want to inject a Michelson annotation, the you simply provide an empty string.

info

Alternatively, if annotations are not important you can also use plain tuples for pair's instead. Plain tuples don't have any annotations.

To use variables of type michelson_or you have to use M_left and M_right. M_left picks the left or case while M_right picks the right or case. For michelson_pair you need to use tuples.

let z : z_or = M_left(unit);

let y_1 : y_or = M_left(unit);
let y : z_or = M_right(y_1);

let x_pair = ["foo", [2, 3n]];
let x_1 : y_or = M_right (x_pair);
let x : z_or = M_right (y_1);

Manual data structure conversion

If you want to get your hands dirty, it is also possible to do manual data structure conversion. The following code can be used as inspiration:

type z_to_v =
  ["Z"]
| ["Y"]
| ["X"]
| ["W"]
| ["V"];

type w_or_v = michelson_or<[unit, "w", unit, "v"]>;
type x_or = michelson_or<[unit, "x", w_or_v, "other"]>;
type y_or = michelson_or<[unit, "y", x_or, "other"]>;
type z_or = michelson_or<[unit, "z", y_or, "other"]>;

type test = {
  z: string,
  y: int,
  x: string,
  w: bool,
  v: int
};

let make_concrete_sum = (r: z_to_v): z_or =>
  match(r) {
    when(Z()): M_left(unit);
    when(Y()): M_right(M_left(unit));
    when(X()): M_right (M_right (M_left(unit)));
    when(W()): M_right (M_right (M_right(M_left(unit))));
    when(V()): M_right (M_right (M_right(M_right(unit))))
  };


let make_concrete_record = (r: test) =>
  [r.z, r.y, r.x, r.w, r.v];

let make_abstract_sum = (z_or: z_or): z_to_v =>
  match(z_or) {
    when(M_left(n)): Z();
    when(M_right(y_or)): match(y_or) {
        when(M_left(n)): Y();
        when(M_right(x_or)): match(x_or) {
            when(M_left(n)): X();
            when(M_right(w_or)): match(w_or) {
                when(M_left(n)): W();
                when(M_right(n)): V()
            }
        }
    }
  };

let make_abstract_record = (z: string, y: int, x: string, w: bool, v: int) =>  ({z,y,x,w,v});

Entrypoints and annotations

It's possible for a contract to have multiple entrypoints, which is implicitly translated in LIGO to a parameter with a variant type as shown below. The following contract:

type storage = int

@entry const left  = (i : int, x : storage) : [list<operation>, storage] => [list([]), x - i]
@entry const right = (i : int, x : storage) : [list<operation>, storage] => [list([]), x + i]

is translated internally to a contract similar to this one:

type storage = int;

type parameter =
   ["Left", int]
 | ["Right", int];

let main = (p: parameter, x: storage): [list<operation>, storage] =>
  [list ([]), match(p) {
    when(Left(i)): x - i;
    when(Right(i)): x + i
   }
  ];

This contract can be called by another contract, like this one:

type storage = int;

type parameter = int;

type x = | ["Left", int];

@entry
const main = (p: parameter, s: storage): [list<operation>, storage] => {
  let contract =
    match (Tezos.get_entrypoint_opt("%left", "tz1KqTpEZ7Yob7QbPE4Hy4Wo8fHG8LhKxZSx" as address)) {
      when(Some(c)): c;
      when(None()): failwith ("contract does not match")
    };
  return [
    list([Tezos.transaction(Left(2), 2mutez, contract)]),
    s];
};

Notice how we directly use the %left entrypoint without mentioning the %right entrypoint. This is done with the help of annotations. Without annotations it wouldn't be clear what our int would be referring to.

This currently only works for or's or variant types in LIGO.

Amendment

With the upcoming 007 amendment to Tezos this will change though, and also pairs can be ordered differently.