Michelson: the language of Smart Contracts in Tezos#

This specification gives a detailed formal semantics of the Michelson language and a short explanation of how smart contracts are executed and interact in the blockchain.

The language is stack-based, with high level data types and primitives, and strict static type checking. Its design cherry picks traits from several language families. Vigilant readers will notice direct references to Forth, Scheme, ML and Cat.

A Michelson program is a series of instructions that are run in sequence: each instruction receives as input the stack resulting from the previous instruction, and rewrites it for the next one. The stack contains both immediate values and heap allocated structures. All values are immutable and garbage collected.

The types of the input and output stack are fixed and monomorphic, and the program is typechecked before being introduced into the system. No smart contract execution can fail because an instruction has been executed on a stack of unexpected length or contents.

This specification gives the complete instruction set, type system and semantics of the language. It is meant as a precise reference manual, not an easy introduction. Even though, some examples are provided at the end of the document and can be read first or at the same time as the specification. The document also starts with a less formal explanation of the context: how Michelson code interacts with the blockchain.

Semantics of smart contracts and transactions#

The Tezos ledger currently has two types of accounts that can hold tokens (and be the destinations of transactions).

  • Implicit account: non-programmable account whose address is the public key hash, prefixed by tz and one digit.

  • Smart contract: programmable account associated to some Michelson code, whose address is a unique hash, prefixed by KT1. A transaction to such an address can provide data, and can fail for reasons detailed below.

See Accounts and addresses for more details.

Intra-transaction semantics#

Alongside their tokens, smart contracts keep a piece of storage. Both are ruled by a specific logic specified by a Michelson program. A transaction to a smart contract will provide an input value and in option some tokens, and in return, the smart contract can modify its storage and transfer its tokens.

The Michelson program receives as input a stack containing a single pair whose first element is an input value and second element the content of the storage space. It must return a stack containing a single pair whose first element is the list of internal operations that it wants to emit, and second element is the new contents of the storage space. Alternatively, a Michelson program can fail, explicitly using a specific opcode, or because something went wrong that could not be caught by the type system (e.g. gas exhaustion).

A bit of polymorphism can be used at contract level, with a lightweight system of named entrypoints: instead of an input value, the contract can be called with an entrypoint name and an argument, and these two components are transformed automatically in a simple and deterministic way to an input value. This feature is available both for users and from Michelson code. See the dedicated section.

Inter-transaction semantics#

An operation included in the blockchain is a sequence of “external operations” signed as a whole by a source address. These operations are of three kinds:

  • Transactions to transfer tokens to implicit accounts or tokens and parameters to a smart contract (or, optionally, to a specified entrypoint of a smart contract).

  • Originations to create new smart contracts from its Michelson source code, an initial amount of tokens transferred from the source, and an initial storage contents.

  • Delegations to assign the tokens of the source to the stake of another implicit account (without transferring any tokens).

Smart contracts can also emit “internal operations”. These are run in sequence after the external transaction completes, as in the following schema for a sequence of two external operations.

| op 1 | internal ops 1 |  op 2 | internal ops 2 |

Smart contracts called by internal transactions can in turn also emit internal operation. The interpretation of the internal operations of a given external operation uses a stack, as in the following example, also with two external operations.

| executing | emissions     | resulting stack          |
| op 1      | 1a, 1b, 1c    | 1a, 1b, 1c               |
| op 1a     | 1ai, 1aj      | 1ai, 1aj, 1b, 1c         |
| op 1ai    |               | 1aj, 1b, 1c              |
| op 1aj    |               | 1b, 1c                   |
| op 1b     | 1bi           | 1bi, 1c                  |
| op 1bi    |               | 1c                       |
| op 1c     |               |                          |
| op 2      | 2a, 2b        | 2a, 2b                   |
| op 2a     | 2ai           | 2ai, 2b                  |
| op 2ai    | 2ai1          | 2ai1, 2b                 |
| op 2ai1   |               | 2b                       |
| op 2b     | 2bi           | 2bi                      |
| op 2bi    | 2bi1          | 2bi1                     |
| op 2bi1   | 2bi2          | 2bi2                     |
| op 2bi2   |               |                          |


All transactions can fail for a few reasons, mostly:

  • Not enough tokens in the source to spend the specified amount.

  • The script took too many execution steps.

  • The script failed programmatically using the FAILWITH instruction.

External transactions can also fail for these additional reasons:

  • The signature of the external operations was wrong.

  • The code or initial storage in an origination did not typecheck.

  • The parameter in a transfer did not typecheck.

  • The destination did not exist.

  • The specified entrypoint did not exist.

All these errors cannot happen in internal transactions, as the type system catches them at operation creation time. In particular, Michelson has two types to talk about other accounts: address and contract t. The address type merely gives the guarantee that the value has the form of a Tezos address. The contract t type, on the other hand, guarantees that the value is indeed a valid, existing account whose parameter type is t. To make a transaction from Michelson, a value of type contract t must be provided, and the type system checks that the argument to the transaction is indeed of type t. Hence, all transactions made from Michelson are well formed by construction.

In any case, when a failure happens, either total success or total failure is guaranteed. If a transaction (internal or external) fails, then the whole sequence fails and all the effects up to the failure are reverted. These transactions can still be included in blocks, and the transaction fees are given to the implicit account who baked the block.

Language semantics#

This specification explains in a symbolic way the computation performed by the Michelson interpreter on a given program and initial stack to produce the corresponding resulting stack. The Michelson interpreter is a pure function: it only builds a result stack from the elements of an initial one, without affecting its environment. This semantics is then naturally given in what is called a big step form: a symbolic definition of a recursive reference interpreter. This definition takes the form of a list of rules that cover all the possible inputs of the interpreter (program and stack), and describe the computation of the corresponding resulting stacks.

Rules form and selection#

The rules have the main following form.

> (syntax pattern) / (initial stack pattern)  =>  (result stack pattern)
    iff (conditions)
    where (recursions)
    and (more recursions)

The left hand side of the => sign is used for selecting the rule. Given a program and an initial stack, one (and only one) rule can be selected using the following process. First, the toplevel structure of the program must match the syntax pattern. This is quite simple since there are only a few non-trivial patterns to deal with instruction sequences, and the rest is made of trivial patterns that match one specific instruction. Then, the initial stack must match the initial stack pattern. Finally, some rules add extra conditions over the values in the stack that follow the iff keyword. Sometimes, several rules may apply in a given context. In this case, the one that appears first in this specification is to be selected. If no rule applies, the result is equivalent to the one for the explicit FAILWITH instruction. This case does not happen on well-typed programs, as explained in the next section.

The right hand side describes the result of the interpreter if the rule applies. It consists in a stack pattern, whose parts are either constants, or elements of the context (program and initial stack) that have been named on the left hand side of the => sign.

Recursive rules (big step form)#

Sometimes, the result of interpreting a program is derived from the result of interpreting another one (as in conditionals or function calls). In these cases, the rule contains a clause of the following form.

where (intermediate program) / (intermediate stack)  =>  (partial result)

This means that this rule applies in case interpreting the intermediate state on the left gives the pattern on the right.

The left hand sign of the => sign is constructed from elements of the initial state or other partial results, and the right hand side identify parts that can be used to build the result stack of the rule.

If the partial result pattern does not actually match the result of the interpretation, then the result of the whole rule is equivalent to the one for the explicit FAILWITH instruction. Again, this case does not happen on well-typed programs, as explained in the next section.

Format of patterns#

Code patterns are of one of the following syntactical forms.

  • INSTR (an uppercase identifier) is a simple instruction (e.g. DROP).

  • INSTR (arg) ... is a compound instruction, whose arguments can be code, data or type patterns (e.g. PUSH nat 3).

  • { (instr) ; ... } is a possibly empty sequence of instructions, (e.g. IF { SWAP ; DROP } { DROP }), nested sequences can drop the braces.

  • name is a pattern that matches any program and names a part of the matched program that can be used to build the result.

  • _ is a pattern that matches any instruction.

Stack patterns are of one of the following syntactical forms.

  • [FAILED] is the special failed state.

  • [] is the empty stack.

  • (top) : (rest) is a stack whose top element is matched by the data pattern (top) on the left, and whose remaining elements are matched by the stack pattern (rest) on the right (e.g. x : y : rest).

  • name is a pattern that matches any stack and names it in order to use it to build the result.

  • _ is a pattern that matches any stack.

Data patterns are of one of the following syntactical forms.

  • integer/natural number literals, (e.g. 3).

  • string literals, (e.g. "contents").

  • raw byte sequence literals (e.g. 0xABCDEF42).

  • Tag (capitalized) is a symbolic constant, (e.g. Unit, True, False).

  • (Tag (arg) ...) tagged constructed data, (e.g. (Pair 3 4)).

  • a code pattern for first class code values.

  • name to name a value in order to use it to build the result.

  • _ to match any value.

The domain of instruction names, symbolic constants and data constructors is fixed by this specification. Michelson does not let the programmer introduce its own types.

Be aware that the syntax used in the specification may differ from the concrete syntax. In particular some instructions are annotated with types that are not present in the concrete language because they are synthesized by the typechecker.


Sometimes, it is easier to think (and shorter to write) in terms of program rewriting than in terms of big step semantics. When it is the case, and when both are equivalents, we write rules of the form:

p / S  =>  S''
where   p' / S'  =>  S''

using the following shortcut:

p / S  =>  p' / S'

The concrete language also has some syntax sugar to group some common sequences of operations as one. This is described in this specification using a simple regular expression style recursive instruction rewriting.

Introduction to the type system and notations#

This specification describes a type system for Michelson. To make things clear, in particular to readers that are not accustomed to reading formal programming language specifications, it does not give a typechecking or inference algorithm. It only gives an intentional definition of what we consider to be well-typed programs. For each syntactical form, it describes the stacks that are considered well-typed inputs, and the resulting outputs.

The type system is sound, meaning that if a program can be given a type, then if run on a well-typed input stack, the interpreter will never apply an interpretation rule on a stack of unexpected length or contents. Also, it will never reach a state where it cannot select an appropriate rule to continue the execution. Well-typed programs do not block, and do not go wrong.

Type notations#

The specification introduces notations for the types of values, terms and stacks. Apart from a subset of value types that appear in the form of type annotations in some places throughout the language, it is important to understand that this type language only exists in the specification.

A stack type can be written:

  • [] for the empty stack.

  • (top) : (rest) for the stack whose first value has type (top) and queue has stack type (rest).

Instructions, programs and primitives of the language are also typed, their types are written:

(type of stack before) -> (type of stack after)

The types of values in the stack are written:

  • identifier for a primitive data-type (e.g. bool).

  • identifier (arg) for a parametric data-type with one parameter type (arg) (e.g. list nat).

  • identifier (arg) ... for a parametric data-type with several parameters (e.g. map string int).

  • [ (type of stack before) -> (type of stack after) ] for a code quotation, (e.g. [ int : int : [] -> int : [] ]).

  • lambda (arg) (ret) is a shortcut for [ (arg) : [] -> (ret) : [] ].

Meta type variables#

The typing rules introduce meta type variables. To be clear, this has nothing to do with polymorphism, which Michelson does not have. These variables only live at the specification level, and are used to express the consistency between the parts of the program. For instance, the typing rule for the IF construct introduces meta variables to express that both branches must have the same type.

Here are the notations for meta type variables:

  • 'a for a type variable.

  • 'A for a stack type variable.

  • _ for an anonymous type or stack type variable.

Typing rules#

The system is syntax directed, meaning that it defines a single typing rule for each syntax construct. A typing rule restricts the type of input stacks that are authorized for this syntax construct, links the output type to the input type, and links both of them to the subexpressions when needed, using meta type variables.

Typing rules are of the form:

(syntax pattern)
:: (type of stack before) -> (type of stack after) [rule-name]
   iff (premises)

Where premises are typing requirements over subprograms or values in the stack, both of the form (x) :: (type), meaning that value (x) must have type (type).

A program is shown well-typed if one can find an instance of a rule that applies to the toplevel program expression, with all meta type variables replaced by non variable type expressions, and of which all type requirements in the premises can be proven well-typed in the same manner. For the reader unfamiliar with formal type systems, this is called building a typing derivation.

Here is an example typing derivation on a small program that computes (x+5)*10 for a given input x, obtained by instantiating the typing rules for instructions PUSH, ADD and for the sequence, as found in the next sections. When instantiating, we replace the iff with by.

{ PUSH nat 5 ; ADD ; PUSH nat 10 ; MUL }
:: [ nat : [] -> nat : [] ]
   by { PUSH nat 5 ; ADD }
      :: [ nat : [] -> nat : [] ]
         by PUSH nat 5
            :: [ nat : [] -> nat : nat : [] ]
               by 5 :: nat
        and ADD
            :: [ nat : nat : [] -> nat : [] ]
  and { PUSH nat 10 ; MUL }
      :: [ nat : [] -> nat : [] ]
         by PUSH nat 10
            :: [ nat : [] -> nat : nat : [] ]
               by 10 :: nat
        and MUL
            :: [ nat : nat : [] -> nat : [] ]

Producing such a typing derivation can be done in a number of manners, such as unification or abstract interpretation. In the implementation of Michelson, this is done by performing a recursive symbolic evaluation of the program on an abstract stack representing the input type provided by the programmer, and checking that the resulting symbolic stack is consistent with the expected result, also provided by the programmer.

Side note#

As with most type systems, it is incomplete. There are programs that cannot be given a type in this type system, yet that would not go wrong if executed. This is a necessary compromise to make the type system usable. Also, it is important to remember that the implementation of Michelson does not accept as many programs as the type system describes as well-typed. This is because the implementation uses a simple single pass typechecking algorithm, and does not handle any form of polymorphism.

Types and instructions#

The complete sets of Michelson types and instructions are detailed in the interactive Michelson reference page.

  • Specifically, it contains synthesis tables for types and for instructions.

  • Instructions are also organized by categories.

  • Each instruction is precisely defined using typing and semantic inference rules.

Removed instructions and types#

Protocol Babylon deprecated the following instructions. Because no smart contract used these on Mainnet before they got deprecated, they have been removed. The Michelson type-checker will reject any contract using them.

  • CREATE_CONTRACT { parameter 'p ; storage 'g ; code ... }: Forge a new contract from a literal.

    Γ ⊢ CREATE_CONTRACT { parameter 'p ; storage 'g ; code ... }
    :: key_hash : option key_hash : bool : bool : mutez : 'g : 'S
    ⇒ operation : address : 'S

    There is a new version of this instruction, see its documentation.

  • CREATE_ACCOUNT: Forge an account creation operation.

    Γ ⊢ CREATE_ACCOUNT :: key_hash : option key_hash : bool : mutez : 'S
    ⇒ operation : address : 'S

    Takes as argument the manager, optional delegate, the delegatable flag and finally the initial amount taken from the currently executed contract. This instruction originates a contract with two entrypoints; %default of type unit that does nothing and %do of type lambda unit (list operation) that executes and returns the parameter if the sender is the contract’s manager.

  • STEPS_TO_QUOTA: Push the remaining steps before the contract execution must terminate.

    Γ ⊢ STEPS_TO_QUOTA :: 'S ⇒ nat : 'S

Protocol Mumbai deprecated the following type. Because no smart contract used it on Mainnet before it got deprecated, it has been removed. The Michelson type-checker will reject any contract using it.

  • tx_rollup_l2_address: An address used to identify an account in a transaction rollup ledger. It is the hash of a BLS public key, used to authenticate layer-2 operations to transfer tickets from this account.


In addition to the instructions listed in the interactive Michelson reference manual, several extensions have been added to the language’s concrete syntax. If you are interacting with the node via RPC, bypassing the client, which expands away these macros, you will need to desugar them yourself.

These macros are designed to be unambiguous and reversible, meaning that errors are reported in terms of desugared syntax. Below you’ll see these macros defined in terms of other syntactic forms. That is how these macros are seen by the node.


Syntactic sugar exists for merging COMPARE and comparison combinators, and also for branching.


> CMP(\op) / S  =>  COMPARE ; (\op) / S
  • IF{EQ|NEQ|LT|GT|LE|GE} bt bf

> IF(\op) bt bf / S  =>  (\op) ; IF bt bf / S

> IFCMP(\op) / S  =>  COMPARE ; (\op) ; IF bt bf / S


The FAIL macros is equivalent to UNIT; FAILWITH and is callable in any context since it does not use its input stack.

  • FAIL


Assertion macros#

All assertion operations are syntactic sugar for conditionals with a FAIL instruction in the appropriate branch. When possible, use them to increase clarity about illegal states.


> ASSERT  =>  IF {} {FAIL}

> ASSERT_(\op)  =>  IF(\op) {} {FAIL}

> ASSERT_CMP(\op)  =>  IFCMP(\op) {} {FAIL}





Syntactic Conveniences#

These macros are simply more convenient syntax for various common operations.

  • P(\left=A|P(\left)(\right))(\right=I|P(\left)(\right))R: A syntactic sugar for building nested pairs. In the case of right combs, PAIR n is more efficient.

> PA(\right)R / S => DIP ((\right)R) ; PAIR / S
> P(\left)IR / S => (\left)R ; PAIR / S
> P(\left)(\right)R =>  (\left)R ; DIP ((\right)R) ; PAIR / S

A good way to quickly figure which macro to use is to mentally parse the macro as P for pair constructor, A for left leaf and I for right leaf. The macro takes as many elements on the stack as there are leaves and constructs a nested pair with the shape given by its name.

Take the macro PAPPAIIR for instance:

P A  P P A  I    I R
( l, ( ( l, r ), r ))

A typing rule can be inferred:

:: 'a : 'b : 'c : 'd : 'S  ->  (pair 'a (pair (pair 'b 'c) 'd))
  • UNP(\left=A|P(\left)(\right))(\right=I|P(\left)(\right))R: A syntactic sugar for destructing nested pairs. These macros follow the same convention as the previous one.

> UNPA(\right)R / S => UNPAIR ; DIP (UN(\right)R) / S
> UNP(\left)IR / S => UNPAIR ; UN(\left)R / S
> UNP(\left)(\right)R => UNPAIR ; DIP (UN(\right)R) ; UN(\left)R / S
  • C[AD]+R: A syntactic sugar for accessing fields in nested pairs. In the case of right combs, CAR k and CDR k are more efficient.

> CA(\rest=[AD]+)R / S  =>  CAR ; C(\rest)R / S
> CD(\rest=[AD]+)R / S  =>  CDR ; C(\rest)R / S
  • CAR k: Access the k -th part of a right comb of size n > k + 1. CAR 0 is equivalent to CAR and in general CAR k is equivalent to k times the CDR instruction followed by once the CAR instruction. Note that this instruction cannot access the last element of a right comb; CDR k should be used for that.

> CAR n / S  =>  GET (2n+1) / S
  • CDR k: Access the rightmost element of a right comb of size k. CDR 0 is a no-op, CDR 1 is equivalent to CDR and in general CDR k is equivalent to k times the CDR instruction. Note that on a right comb of size n > k >= 2, CDR k will return the right comb composed of the same elements but the k leftmost ones.

> CDR n / S  =>  GET (2n) / S
  • IF_SOME bt bf: Inspect an optional value.

> IF_SOME bt bf / S  =>  IF_NONE bf bt / S
  • IF_RIGHT bt bf: Inspect a value of a union.

> IF_RIGHT bt bf / S  =>  IF_LEFT bf bt / S
  • SET_CAR: Set the left field of a pair. This is equivalent to SWAP; UPDATE 1.

  • SET_CDR: Set the right field of a pair. This is equivalent to SWAP; UPDATE 2.

  • SET_C[AD]+R: A syntactic sugar for setting fields in nested pairs. In the case of right combs, UPDATE n is more efficient.

> SET_CA(\rest=[AD]+)R / S   =>
    { DUP ; DIP { CAR ; SET_C(\rest)R } ; CDR ; SWAP ; PAIR } / S
> SET_CD(\rest=[AD]+)R / S   =>
    { DUP ; DIP { CDR ; SET_C(\rest)R } ; CAR ; PAIR } / S
  • MAP_CAR code: Transform the left field of a pair.

> MAP_CAR code  =>  DUP ; CDR ; DIP { CAR ; code } ; SWAP ; PAIR
  • MAP_CDR code: Transform the right field of a pair.

> MAP_CDR code  =>  DUP ; CDR ; code ; SWAP ; CAR ; PAIR
  • MAP_C[AD]+R code: A syntactic sugar for transforming fields in nested pairs.

> MAP_CA(\rest=[AD]+)R code / S   =>
    { DUP ; DIP { CAR ; MAP_C(\rest)R code } ; CDR ; SWAP ; PAIR } / S
> MAP_CD(\rest=[AD]+)R code / S   =>
    { DUP ; DIP { CDR ; MAP_C(\rest)R code } ; CAR ; PAIR } / S

Concrete syntax#

The concrete language is very close to the formal notation of the specification. Its structure is extremely simple: an expression in the language can only be one of the five following constructs.

  1. An integer in decimal notation.

  2. A character string.

  3. A byte sequence in hexadecimal notation prefixed by 0x.

  4. The application of a primitive to a sequence of expressions.

  5. A sequence of expressions.

This simple five cases notation is called Micheline.

In the Tezos protocol, the primitive constant with a single character string applied has special meaning. See Global Constants.


There are three kinds of constants:

  1. Integers or naturals in decimal notation.

  2. Strings, with some usual escape sequences: \n, \\, \". Unescaped line-breaks (both \n and \r) cannot appear in a Michelson string. Moreover, the current version of Michelson restricts strings to be the printable subset of 7-bit ASCII, namely characters with codes from within [32, 126] range, plus the escaped characters mentioned above.

  3. Byte sequences in hexadecimal notation, prefixed with 0x.

Differences with the formal notation#

The concrete syntax follows the same lexical conventions as the specification: instructions are represented by uppercase identifiers, type constructors by lowercase identifiers, and constant constructors are capitalized.

All domain specific constants are Micheline constants with specific formats. Some have two variants accepted by the data type checker: a readable one in a string, and an optimized one using a more compact encoding.

  • mutez amounts are written as naturals.

  • timestamps are written either using RFC3339 notation in a string (readable), or as the number of seconds since Epoch (when positive) or before Epoch (when negative) (optimized).

  • contracts, addresses, keys and signatures are written as strings, in their usual Base58 encoded versions (readable), or as their raw bytes (optimized).

  • bls12_381_g1s and bls12_381_g2s are written as their raw bytes, using a big-endian point encoding, as specified here.

  • bls12_381_frs are written as their raw bytes, using a little-endian encoding.

The optimized versions should not reach the RPCs, the protocol code will convert to optimized by itself when forging operations, storing to the database, and before hashing to get a canonical representation of a datum for a given type.

To prevent errors, control flow primitives that take instructions as parameters require sequences in the concrete syntax.

IF { instr1_true ; instr2_true ; ... }
   { instr1_false ; instr2_false ; ... }

Main program structure#

The toplevel of a smart contract file must be an un-delimited sequence of three primitive applications (in no particular order) that provide its code, parameter and storage fields.

See the next section for a concrete example.


The annotation mechanism of Michelson provides ways to better track data on the stack and to give additional type constraints. Except for a single exception specified just after, annotations are only here to add constraints, i.e. they cannot turn an otherwise rejected program into an accepted one. The notable exception to this rule is for entrypoints: the semantics of the CONTRACT and SELF instructions vary depending on their constructor annotations, and some contract origination may fail due to invalid entrypoint constructor annotations.

Stack visualization tools like the Michelson’s Emacs mode print annotations associated with each type in the program, as propagated by the typechecker as well as variable annotations on the types of elements in the stack. This is useful as a debugging aid.

We distinguish three kinds of annotations:

  • type annotations, written :type_annot,

  • variable annotations, written @var_annot,

  • and field or constructors annotations, written %field_annot.

Type annotations#

Each type can be annotated with at most one type annotation. They are used to give names to types. For types to be equal, their unnamed version must be equal and their names must be the same or at least one type must be unnamed.

For instance, the following Michelson program which put its integer parameter in the storage is not well typed:

parameter (int :p) ;
storage (int :s) ;
code { UNPAIR ; SWAP ; DROP ; NIL operation ; PAIR }

Whereas this one is:

parameter (int :p) ;
storage int ;
code { UNPAIR ; SWAP ; DROP ; NIL operation ; PAIR }

Inner components of composed typed can also be named.

(pair :point (int :x_pos) (int :y_pos))

Push-like instructions, that act as constructors, can also be given a type annotation. The stack type will then have on top a type with a corresponding name.

:: 'A -> (unit :t) : 'A

:: 'a : 'b : 'S -> (pair :t 'a 'b) : 'S

:: 'a : 'S -> (option :t 'a) : 'S

NONE :t 'a
:: 'S -> (option :t 'a) : 'S

LEFT :t 'b
:: 'a : 'S -> (or :t 'a 'b) : 'S

RIGHT :t 'a
:: 'b : 'S -> (or :t 'a 'b) : 'S

NIL :t 'a
:: 'S -> (list :t 'a) : 'S

EMPTY_SET :t 'elt
:: 'S -> (set :t 'elt) : 'S

EMPTY_MAP :t 'key 'val
:: 'S -> (map :t 'key 'val) : 'S

EMPTY_BIG_MAP :t 'key 'val
:: 'S -> (big_map :t 'key 'val) : 'S

A no-op instruction CAST ensures the top of the stack has the specified type, and change its type if it is compatible. In particular, this allows to change or remove type names explicitly.

:: 'a : 'S   ->   'b : 'S
   iff  'a = 'b

> CAST t / a : S  =>  a : S

Variable annotations#

Variable annotations can only be used on instructions that produce elements on the stack. An instruction that produces n elements on the stack can be given at most n variable annotations.

The stack type contains both the types of each element in the stack, as well as an optional variable annotation for each element. In this sub-section we note:

  • [] for the empty stack,

  • @annot (top) : (rest) for the stack whose first value has type (top) and is annotated with variable annotation @annot and whose queue has stack type (rest).

The instructions which do not accept any variable annotations are:


The instructions which accept at most one variable annotation are:


The instructions which accept at most two variable annotations are:


Annotations on instructions that produce multiple elements on the stack will be used in order, where the first variable annotation is given to the top-most element on the resulting stack. Instructions that produce n elements on the stack but are given less than n variable annotations will see only their top-most stack type elements annotated.

UNPAIR @first @second
:: pair 'a 'b : 'S
   ->  @first 'a : @second 'b : 'S

UNPAIR @first
:: pair 'a 'b : 'S
   ->  @first 'a : 'b : 'S

A no-op instruction RENAME allows to rename variables in the stack or to erase variable annotations in the stack.

:: @old 'a ; 'S -> @new 'a : 'S

:: @old 'a ; 'S -> 'a : 'S

Field and constructor annotations#

Components of pair types, option types and or types can be annotated with a field or constructor annotation. This feature is useful to encode records fields and constructors of sum types.

(pair :point
      (int %x)
      (int %y))

The previous Michelson type can be used as visual aid to represent the record type (given in OCaml-like syntax):

type point = { x : int ; y : int }


(or :t
    (int %A)
       (bool %B)
       (pair %C
             (nat %n1)
             (nat %n2))))

can be used to represent the algebraic data type (in OCaml-like syntax):

type t =
  | A of int
  | B of bool
  | C of { n1 : nat ; n2 : nat }

Field annotations are part of the type (at the same level as type name annotations), and so types with differing field names (if present) are not considered equal.

Instructions that construct elements of composed types can also be annotated with one or multiple field annotations (in addition to type and variable annotations).

PAIR %fst %snd
:: 'a : 'b : 'S -> (pair ('a %fst) ('b %snd)) : 'S

LEFT %left %right 'b
:: 'a : 'S -> (or ('a %left) ('b %right)) : 'S

RIGHT %left %right 'a
:: 'b : 'S -> (or ('a %left) ('b %right)) : 'S

To improve readability and robustness, instructions CAR and CDR accept one field annotation. For the contract to type check, the name of the accessed field in the destructed pair must match the one given here.

CAR %fst
:: (pair ('a %fst) 'b) : S -> 'a : 'S

CDR %snd
:: (pair 'a ('b %snd)) : S -> 'b : 'S


Primitive applications can receive one or many annotations.

An annotation is a sequence of characters that matches the regular expression @%|@%%|%@|[@:%][_0-9a-zA-Z][_0-9a-zA-Z\.%@]*. Note however that @%, @%% and %@ are special annotations and are not allowed everywhere.

Annotations come after the primitive name and before its potential arguments.

(prim @v :t %x arg1 arg2 ...)

Ordering between different kinds of annotations is not significant, but ordering among annotations of the same kind is. Annotations of the same kind must be grouped together.

For instance these two annotated instructions are equivalent:

PAIR :t @my_pair %x %y

PAIR %x %y :t @my_pair

An annotation can be empty, in this case it will mean no annotation and can be used as a wildcard. For instance, it is useful to annotate only the right field of a pair instruction PAIR % %right or to ignore field access constraints, e.g. in the macro UNPPAIPAIR %x1 % %x3 %x4.

Annotations and macros#

Macros also support annotations, which are propagated on their expanded forms. As with instructions, macros that produce n values on the stack accept n variable annotations.

DUU+P @annot
> DUU(\rest=U*)P @annot / S  =>  DIP (DU(\rest)P @annot) ; SWAP / S

C[AD]+R @annot %field_name
> CA(\rest=[AD]+)R @annot %field_name / S  =>  CAR ; C(\rest)R @annot %field_name / S
> CD(\rest=[AD]+)R @annot %field_name / S  =>  CDR ; C(\rest)R @annot %field_name / S

> CMP(\op) @annot / S  =>  COMPARE ; (\op) @annot / S

The variable annotation on SET_C[AD]+R and MAP_C[AD]+R annotates the resulting toplevel pair while its field annotation is used to check that the modified field is the expected one.

SET_C[AD]+R @var %field
> SET_CAR @var %field =>  CDR %field ; SWAP ; PAIR @var
> SET_CDR @var %field =>  CAR %field ; PAIR @var
> SET_CA(\rest=[AD]+)R @var %field / S   =>
  { DUP ; DIP { CAR ; SET_C(\rest)R %field } ; CDR ; SWAP ; PAIR @var } / S
> SET_CD(\rest=[AD]+)R  @var %field/ S   =>
  { DUP ; DIP { CDR ; SET_C(\rest)R %field } ; CAR ; PAIR @var } / S

MAP_C[AD]+R @var %field code
> MAP_CAR code  =>  DUP ; CDR ; DIP { CAR %field ; code } ; SWAP ; PAIR @var
> MAP_CDR code  =>  DUP ; CDR %field ; code ; SWAP ; CAR ; PAIR @var
> MAP_CA(\rest=[AD]+)R @var %field code / S   =>
  { DUP ; DIP { CAR ; MAP_C(\rest)R %field code } ; CDR ; SWAP ; PAIR @var} / S
> MAP_CD(\rest=[AD]+)R @var %field code / S   =>
 { DUP ; DIP { CDR ; MAP_C(\rest)R %field code } ; CAR ; PAIR @var} / S

Macros for nested PAIR accept multiple annotations. Field annotations for PAIR give names to leaves of the constructed nested pair, in order. This next snippet gives examples instead of generic rewrite rules for readability purposes.

PAPPAIIR @p %x1 %x2 %x3 %x4
:: 'a : 'b : 'c : 'd : 'S
   -> @p (pair ('a %x1) (pair (pair ('b %x) ('c %x3)) ('d %x4))) : 'S

PAPAIR @p %x1 %x2 %x3
:: 'a : 'b : 'c : 'S  ->  @p (pair ('a %x1) (pair ('b %x) ('c %x3))) : 'S

Annotations for nested UNPAIR are deprecated.

Automatic variable and field annotations inferring#

When no annotation is provided by the Michelson programmer, the typechecker infers some annotations in specific cases. This greatly helps users track information in the stack for bare contracts.

For unannotated accesses with CAR and CDR to fields that are named will be appended (with an additional . character) to the pair variable annotation.

:: @p (pair ('a %foo) (pair %bar ('b %x) ('c %y))) : 'S ->  @p.bar.x 'b : 'S

If fields are not named but the pair is still named in the stack then .car or .cdr will be appended.

:: @p (pair 'a (pair 'b 'c)) : 'S ->  @p.cdr.car 'b : 'S

If the original pair is not named in the stack, but a field annotation is present in the pair type the accessed value will be annotated with a variable annotation corresponding to the field annotation alone.

:: (pair ('a %foo) (pair %bar ('b %x) ('c %y))) : 'S ->  @bar.x 'b : 'S

A similar mechanism is used for context dependent instructions:

ADDRESS  :: @c contract _ : 'S   ->   @c.address address : 'S

CONTRACT 'p  :: @a address : 'S   ->   @a.contract contract 'p : 'S

BALANCE :: 'S   ->   @balance mutez : 'S

SOURCE  :: 'S   ->   @source address : 'S

SENDER  :: 'S   ->   @sender address : 'S

SELF  :: 'S   ->   @self contract 'p : 'S

SELF_ADDRESS  :: 'S   ->   @self address : 'S

AMOUNT  :: 'S   ->   @amount mutez : 'S

NOW  :: 'S   ->   @now timestamp : 'S

LEVEL :: 'S  ->   @level nat : 'S

Inside nested code blocks, bound items on the stack will be given a default variable name annotation depending on the instruction and stack type (which can be changed). For instance the annotated typing rule for ITER on lists is:

ITER body
:: @l (list 'e) : 'A  ->  'A
   iff body :: [ @l.elt e' : 'A -> 'A ]

Special annotations#

The special variable annotations @% and @%% can be used on instructions CAR, CDR, and UNPAIR. It means to use the accessed field name (if any) as a name for the value on the stack. The following typing rule demonstrates their use for instruction CAR.

CAR @%
:: @p (pair ('a %fst) ('b %snd)) : 'S   ->   @fst 'a : 'S

CAR @%%
:: @p (pair ('a %fst) ('b %snd)) : 'S   ->   @p.fst 'a : 'S

The special field annotation %@ can be used on instructions PAIR, LEFT and RIGHT. It means to use the variable name annotation in the stack as a field name for the constructed element. Two examples with PAIR follows, notice the special treatment of annotations with ..

PAIR %@ %@
:: @x 'a : @y 'b : 'S   ->   (pair ('a %x) ('b %y)) : 'S

PAIR %@ %@
:: @p.x 'a : @p.y 'b : 'S   ->  @p (pair ('a %x) ('b %y)) : 'S
:: @p.x 'a : @q.y 'b : 'S   ->  (pair ('a %x) ('b %y)) : 'S


The specification up to this point has been mostly ignoring existence of entrypoints: a mechanism of contract level polymorphism. This mechanism is optional, non intrusive, and transparent to smart contracts that don’t use them. This section is to be read as a patch over the rest of the specification, introducing rules that apply only in presence of contracts that make use of entrypoints.

Defining and calling entrypoints#

Entrypoints piggyback on the constructor annotations. A contract with entrypoints is basically a contract that takes a disjunctive type (a nesting of or types) as the root of its input parameter, decorated with constructor annotations. An extra check is performed on these constructor annotations: a contract cannot define two entrypoints with the same name.

An external transaction can include an entrypoint name alongside the parameter value. In that case, if there is a constructor annotation with this name at any position in the nesting of or types, the value is automatically wrapped into the according constructors. If the transaction specifies an entrypoint, but there is no such constructor annotation, the transaction fails.

For instance, suppose the following input type.

parameter (or (or (nat %A) (bool %B)) (or %maybe_C (unit %Z) (string %C)))

The input values will be wrapped as in the following examples.

| entrypoint | input     | wrapped input                   |
| %A         | 3         | Left (Left 3)                   |
| %B         | False     | Left (Right False)              |
| %C         | "bob"     | Right (Right "bob")             |
| %Z         | Unit      | Right (Left Unit)               |
| %maybe_C   | Right "x" | Right (Right "x")               |
| %maybe_C   | Left Unit | Right (Left Unit)               |
| not given  | value     | value (untouched)               |
| %BAD       | _         | failure, contract not called    |

The default entrypoint#

A special semantics is assigned to the default entrypoint. If the contract does not explicitly declare a default entrypoint, then it is automatically assigned to the root of the parameter type. Conversely, if the contract is called without specifying an entrypoint, then it is assumed to be called with the default entrypoint. This behaviour makes the entrypoint system completely transparent to contracts that do not use it.

This is the case for the previous example, for instance. If a value is passed to such a contract specifying entrypoint default, then the value is fed to the contract untouched, exactly as if no entrypoint was given.

A non enforced convention is to make the entrypoint default of type unit, and to implement the crediting operation (just receive the transferred tokens).

A consequence of this semantics is that if the contract uses the entrypoint system and defines a default entrypoint somewhere else than at the root of the parameter type, then it must provide an entrypoint for all the paths in the toplevel disjunction. Otherwise, some parts of the contracts would be dead code.

Another consequence of setting the entrypoint somewhere else than at the root is that it makes it impossible to send the raw values of the full parameter type to a contract. A trivial solution for that is to name the root of the type. The conventional name for that is root.

Let us recapitulate this by tweaking the names of the previous example.

parameter (or %root (or (nat %A) (bool %B)) (or (unit %default) string))

The input values will be wrapped as in the following examples.

| entrypoint | input               | wrapped input         |
| %A         | 3                   | Left (Left 3)         |
| %B         | False               | Left (Right False)    |
| %default   | Unit                | Right (Left Unit)     |
| %root      | Right (Right "bob") | Right (Right "bob")   |
| not given  | Unit                | Right (Left Unit)     |
| %BAD       | _                   | failure, contract not |

Calling entrypoints from Michelson#

Michelson code can also produce transactions to a specific entrypoint.

For this, both types address and contract have the ability to denote not just an address, but a pair of an address and an entrypoint. The concrete notation is "address%entrypoint". Note that "address" is strictly equivalent to "address%default", and for clarity, the second variant is forbidden in the concrete syntax.

When the TRANSFER_TOKENS instruction is called, it places the entrypoint provided in the contract handle in the transaction.

The CONTRACT t instruction has a variant CONTRACT %entrypoint t, that works as follows. Note that CONTRACT t is strictly equivalent to CONTRACT %default t, and for clarity, the second variant is forbidden in the concrete syntax.

| input address | instruction         | output contract                          |
| "addr"        | CONTRACT t          | (Some "addr") if contract exists, has a  |
|               |                     | default entrypoint of type t, or has no  |
|               |                     | default entrypoint and parameter type t  |
| "addr%name"   | CONTRACT t          | (Some "addr%name") if addr exists and    |
+---------------+---------------------+ has an entrypoint %name of type t        |
| "addr"        | CONTRACT %name t    |                                          |
| "addr%_"      | CONTRACT %_ t       | None                                     |

Similarly, the SELF instruction has a variant SELF %entrypoint, that is only well-typed if the current contract has an entrypoint named %entrypoint.

  • SELF %entrypoint

:: 'S   ->   contract 'p : 'S
   where   contract 'p is the type of the entrypoint %entrypoint of the current contract

Implicit accounts are considered to have a single default entrypoint of type Unit.

JSON syntax#

Micheline expressions are encoded in JSON like this:

  • An integer N is an object with a single field "int" whose value is the decimal representation as a string.

    { "int": "N" }

  • A string "contents" is an object with a single field "string" whose value is the decimal representation as a string.

    { "string": "contents" }

  • A sequence is a JSON array.

    [ expr, ... ]

  • A primitive application is an object with two fields "prim" for the primitive name and "args" for the arguments (that must contain an array). A third optional field "annots" contains a list of annotations, including their leading @, % or : sign.

    { "prim": "pair", "args": [ { "prim": "nat", "args": [] }, { "prim": "nat", "args": [] } ], "annots": [":t"] }

As in the concrete syntax, all domain specific constants are encoded as strings.

Development tools#

To ease the development of Michelson scripts, some tools are provided to Michelson developers.

Emacs mode#

Emacs can be used as a practical environment for writing, editing and debugging Michelson programs. Install it and follow the configuration instructions in the Michelson Emacs README here.

Interactive toplevel#

An interactive Michelson toplevel (also known as a REPL) built on the Mockup mode mode of Octez client is available in scripts/michelson_repl.sh, the typical usage is:

$ octez-client --mode mockup --base-dir /tmp/mockup create mockup
$ rlwrap scripts/michelson_repl.sh
  { Stack_elt unit Unit }
  { Stack_elt unit Unit ; Stack_elt unit Unit }
  { Stack_elt int 0 }


Contracts in the system are stored as a piece of code and a global data storage. The type of the global data of the storage is fixed for each contract at origination time. This is ensured statically by checking on origination that the code preserves the type of the global data. For this, the code of the contract is checked to be of type lambda (pair 'arg 'global) -> (pair (list operation) 'global) where 'global is the type of the original global store given on origination. The contract also takes a parameter and returns a list of internal operations, hence the complete calling convention above. The internal operations are queued for execution when the contract returns.

Empty contract#

The simplest contract is the contract for which the parameter and storage are all of type unit. This contract is as follows:

code { CDR ;           # keep the storage
       NIL operation ; # return no internal operation
       PAIR };         # respect the calling convention
storage unit;
parameter unit;

Example contract with entrypoints#

The following contract maintains a number in its storage. It has two entrypoints add and sub to modify it, and the default entrypoint, of type unit will reset it to 0.

{ parameter (or (or (nat %add) (nat %sub)) (unit %default)) ;
  storage int ;
  code { AMOUNT ; PUSH mutez 0 ; ASSERT_CMPEQ ; UNPAIR ;
           { IF_LEFT { ADD } { SWAP ; SUB } }
           { DROP ; DROP ; PUSH int 0 } ;
         NIL operation ; PAIR } }

Example contract with recursive lambda#

The following contract computes the factorial of the given parameter using a recursive function and then saves the result in the storage.

In Michelson regular functions start with a stack containing a single value, the function argument. If the function is of type lambda int int, when calling the function the stack will have just an int. Recursive functions start with two values, the argument and the function itself. Therefore, if the recursive function is of type lambda int int then, when it is being called, the stack will have an int at the top and a lambda int int at the bottom.

In this recursive factorial we can see the first branch of the IF, this is the base case. The second one performs the recursive call. To do that, we need to access the function. This is what the DUP 3 instruction does. Then we decrement the argument and finally make the recursive call with EXEC.

{ parameter int;
  storage int;
  code { CAR ;
         LAMBDA_REC  int int
                     { DUP;
                       IF { PUSH int 1 }
                          { DUP;
                            DUP 3;
                            PUSH int 1;
                            DUP 4;
                       DIP { DROP 2 }};
         NIL operation;

Multisig contract#

The multisig is a typical access control contract. The ownership of the multisig contract is shared between N participants represented by their public keys in the contract’s storage. Any action on the multisig contract needs to be signed by K participants where the threshold K is also stored in the storage.

To avoid replay of the signatures sent to the contract, the signed data include not only a description of the action to perform but also the address of the multisig contract and a counter that gets incremented at each successful call to the contract.

The multisig commands of Octez command line client use this smart contract. Moreover, functional correctness of this contract has been verified using the Coq proof assistant.

parameter (pair
             (pair :payload
                (nat %counter) # counter, used to prevent replay attacks
                (or :action    # payload to sign, represents the requested action
                   (pair :transfer    # transfer tokens
                      (mutez %amount) # amount to transfer
                      (contract %dest unit)) # destination to transfer to
                      (option %delegate key_hash) # change the delegate to this address
                      (pair %change_keys          # change the keys controlling the multisig
                         (nat %threshold)         # new threshold
                         (list %keys key)))))     # new list of keys
             (list %sigs (option signature)));    # signatures

storage (pair (nat %stored_counter) (pair (nat %threshold) (list %keys key))) ;

    UNPAIR ; SWAP ; DUP ; DIP { SWAP } ;
        UNPAIR ;
        # pair the payload with the current contract address, to ensure signatures
        # can't be replayed across different contracts if a key is reused.
        PACK ; # form the binary payload that we expect to be signed
        DIP { UNPAIR @counter ; DIP { SWAP } } ; SWAP
      } ;

    # Check that the counters match
    UNPAIR @stored_counter; DIP { SWAP };

    # Compute the number of valid signatures
    DIP { SWAP } ; UNPAIR @threshold @keys;
        # Running count of valid signatures
        PUSH @valid nat 0; SWAP ;
            DIP { SWAP } ; SWAP ;
                  { SWAP ;
                        SWAP ; DIIP { DIP { DUP } ; SWAP } ;
                        # Checks signatures, fails if invalid
                        CHECK_SIGNATURE ; ASSERT ;
                        PUSH nat 1 ; ADD @valid } }
                  { SWAP ; DROP }
                # There were fewer signatures in the list
                # than keys. Not all signatures must be present, but
                # they should be marked as absent using the option type.
              } ;
      } ;
    # Assert that the threshold is less than or equal to the
    # number of valid signatures.
    DROP ; DROP ;

    # Increment counter and place in storage
    DIP { UNPAIR ; PUSH nat 1 ; ADD @new_counter ; PAIR} ;

    # We have now handled the signature verification part,
    # produce the operation requested by the signers.
    NIL operation ; SWAP ;
      { # Transfer tokens
      { IF_LEFT {
                  # Change delegate
                  SET_DELEGATE ; CONS }
                  # Change set of signatures
                  DIP { SWAP ; CAR } ; SWAP ; PAIR ; SWAP }} ;
    PAIR }


Here is an example using views, consisting of two contracts. The first contract defines two views at toplevel that are named add_v and mul_v.

{ parameter nat;
  storage nat;
  code { CAR; NIL operation ; PAIR };
  view "add_v" nat nat { UNPAIR; ADD };
  view "mul_v" nat nat { UNPAIR; MUL };

The second contract calls the add_v view of the above contract and obtains a result immediately.

{ parameter (pair nat address) ;
  storage nat ;
  code { CAR ; UNPAIR; VIEW "add_v" nat ;
         IF_SOME { } { FAIL }; NIL operation; PAIR }; }

Full grammar#

<data> ::=
  | <int constant>
  | <string constant>
  | <byte sequence constant>
  | Unit
  | True
  | False
  | Pair <data> <data> ...
  | Left <data>
  | Right <data>
  | Some <data>
  | None
  | Lambda_rec <instruction>
  | { <data> ; ... }
  | { Elt <data> <data> ; ... }
  | instruction
<natural number constant> ::=
  | [0-9]+
<int constant> ::=
  | <natural number constant>
  | -<natural number constant>
<string constant> ::=
  | "<string content>*"
<string content> ::=
  | \"
  | \r
  | \n
  | \t
  | \b
  | \\
  | [^"\]
<byte sequence constant> ::=
  | 0x[0-9a-fA-F]+
<instruction> ::=
  | { <instruction> ... }
  | DROP
  | DROP <natural number constant>
  | DUP
  | DUP <natural number constant>
  | SWAP
  | DIG <natural number constant>
  | DUG <natural number constant>
  | PUSH <type> <data>
  | SOME
  | NONE <type>
  | UNIT
  | IF_NONE { <instruction> ... } { <instruction> ... }
  | PAIR
  | PAIR <natural number constant>
  | CAR
  | CDR
  | UNPAIR <natural number constant>
  | LEFT <type>
  | RIGHT <type>
  | IF_LEFT { <instruction> ... } { <instruction> ... }
  | NIL <type>
  | CONS
  | IF_CONS { <instruction> ... } { <instruction> ... }
  | SIZE
  | EMPTY_SET <comparable type>
  | EMPTY_MAP <comparable type> <type>
  | EMPTY_BIG_MAP <comparable type> <type>
  | MAP { <instruction> ... }
  | ITER { <instruction> ... }
  | MEM
  | GET
  | GET <natural number constant>
  | UPDATE <natural number constant>
  | IF { <instruction> ... } { <instruction> ... }
  | LOOP { <instruction> ... }
  | LOOP_LEFT { <instruction> ... }
  | LAMBDA <type> <type> { <instruction> ... }
  | LAMBDA_REC <type> <type> { <instruction> ... }
  | EXEC
  | DIP { <instruction> ... }
  | DIP <natural number constant> { <instruction> ... }
  | CAST
  | PACK
  | UNPACK <type>
  | ADD
  | SUB
  | MUL
  | EDIV
  | ABS
  | INT
  | NEG
  | LSL
  | LSR
  | OR
  | AND
  | XOR
  | NOT
  | EQ
  | NEQ
  | LT
  | GT
  | LE
  | GE
  | SELF
  | CONTRACT <type>
  | CREATE_CONTRACT { <instruction> ... }
  | NOW
  | SHA3
  | SHA256
  | SHA512
  | SAPLING_EMPTY_STATE <natural number constant>
  | NAT
<type> ::=
  | <comparable type>
  | option <type>
  | list <type>
  | set <comparable type>
  | operation
  | contract <type>
  | ticket <comparable type>
  | pair <type> <type> ...
  | or <type> <type>
  | lambda <type> <type>
  | map <comparable type> <type>
  | big_map <comparable type> <type>
  | bls12_381_g1
  | bls12_381_g2
  | bls12_381_fr
  | sapling_transaction <natural number constant>
  | sapling_state <natural number constant>
  | chest
  | chest_key
<comparable type> ::=
  | unit
  | never
  | bool
  | int
  | nat
  | string
  | chain_id
  | bytes
  | mutez
  | key_hash
  | key
  | signature
  | timestamp
  | address
  | option <comparable type>
  | or <comparable type> <comparable type>
  | pair <comparable type> <comparable type> ...

Reference implementation#

The language is implemented in OCaml as follows:

  • The lower internal representation is written as a GADT whose type parameters encode exactly the typing rules given in this specification. In other words, if a program written in this representation is accepted by OCaml’s typechecker, it is guaranteed type-safe. This is of course also valid for programs not handwritten but generated by OCaml code, so we are sure that any manipulated code is type-safe.

    In the end, what remains to be checked is the encoding of the typing rules as OCaml types, which boils down to half a line of code for each instruction. Everything else is left to the venerable and well trusted OCaml.

  • The interpreter is basically the direct transcription of the rewriting rules presented above. It takes an instruction, a stack and transforms it. OCaml’s typechecker ensures that the transformation respects the pre and post stack types declared by the GADT case for each instruction.

    The only things that remain to be reviewed are value dependent choices, such as we did not swap true and false when interpreting the IF instruction.

  • The input, untyped internal representation is an OCaml ADT with only 5 grammar constructions: String, Int, Bytes, Seq and Prim. It is the target language for the parser, since not all parsable programs are well typed, and thus could simply not be constructed using the GADT.

  • The typechecker is a simple function that recognizes the abstract grammar described in section X by pattern matching, producing the well-typed, corresponding GADT expressions. It is mostly a checker, not a full inferrer, and thus takes some annotations (basically the input and output of the program, of lambdas and of uninitialized maps and sets). It works by performing a symbolic evaluation of the program, transforming a symbolic stack. It only needs one pass over the whole program.

    Here again, OCaml does most of the checking, the structure of the function is very simple, what we have to check is that we transform a Prim ("If", ...) into an If, a Prim ("Dup", ...) into a Dup, etc.

TZT, a Syntax extension for writing unit tests#

This section describes the TZT format, an extension of the Michelson language allowing to run Michelson unit tests at a finer level than a full smart contract script. This extension adds syntax to specify an instruction (or sequence of instructions) to test, a concrete input stack and the expected output stack.

These unit tests can be useful for both smart contract developers who need to independently test various parts of the smart contracts they develop and to the developers of new implementations of the Michelson interpreter who need to check that their new implementations behave as the reference implementation by passing a conformance test suite.

Similarly to Michelson scripts, the concrete syntax of TZT unit tests is Micheline.

TZT unit test files usually have the extension .tzt. A unit test file describes a single unit test. It consists of a Micheline sequence of primitive applications (see Micheline), in no particular order. This is similar to Michelson scripts but the set of primitives allowed at the toplevel differ; in Michelson scripts, the allowed toplevel primitives are parameter (mandatory), storage (mandatory), code (mandatory), and view (optional and repeated). For TZT unit tests, the toplevel primitives which can be used are:

  • input,

  • code,

  • output,

  • now,

  • sender,

  • source,

  • chain_id,

  • self,

  • parameter,

  • amount,

  • balance,

  • other_contracts, and

  • big_maps.

Mandatory primitives#

Each of the mandatory primitives input, code, and output must occur exactly once in a unit test file in no particular order.

The input primitive is used to declare the input stack (see the syntax of concrete stacks).

The code primitive is used to declare the instruction or sequence of instructions to execute.

The output primitive is used to declare if the execution is expected to succeed or fail and what result is expected from the execution. For executions expected to succeed, the argument of the output primitive is simply the expected output stack (see the syntax of errors). For executions expected to fail, the argument is the expected error. In both cases, the wildcard pattern can be used to omit part of the expected output.

The simplest test which can be written asserts that executing no instruction on the empty stack successfully returns the empty stack:

input {};
code {};
output {}

Here is a slightly more involved test which demonstrates the effect of the SWAP instruction:

    Stack_elt nat 8 ;
    Stack_elt bool False
code SWAP;
    Stack_elt bool False ;
    Stack_elt nat 8

It is possible to test the effect of several instructions by wrapping them in a sequence:

    Stack_elt nat 8 ;
    Stack_elt bool False
code { SWAP ; SWAP };
    Stack_elt nat 8 ;
    Stack_elt bool False

Here is an example showing how to test the FAILWITH instruction:

input {Stack_elt nat 2};
output (Failed 2)

Optional primitives#

Optional primitives are used to set the execution context for the test. Each of the optional primitives can be used at most once, in no particular order.

  • amount (optional, defaults to 0): the amount, expressed in mutez, that should be pushed by the AMOUNT instruction

  • balance (optional, defaults to 0): the balance, expressed in mutez, that should be pushed by the BALANCE instruction

  • now (optional, defaults to "1970-01-01T00:00:00Z"): the timestamp that should be pushed by the NOW instruction

  • sender (optional, defaults to "tz1KqTpEZ7Yob7QbPE4Hy4Wo8fHG8LhKxZSx"): the sender address that should be pushed by the SENDER instruction

  • source (optional, defaults to "tz1KqTpEZ7Yob7QbPE4Hy4Wo8fHG8LhKxZSx"): the source address that should be pushed by the SOURCE instruction

  • chain_id (optional, defaults to "NetXdQprcVkpaWU"): the chain identifier that should be pushed by the CHAIN_ID instruction

  • self (optional, defaults to "KT1BEqzn5Wx8uJrZNvuS9DVHmLvG9td3fDLi"): the address that should be pushed by the SELF and SELF_ADDRESS instructions

  • parameter (optional, defaults to unit): the type of the parameter of the contract pushed by the SELF instruction

  • other_contracts (optional, defaults to {}): mapping between the contract addresses that are assumed to exist and their parameter types (see the syntax of other contracts specifications)

  • big_maps (optional, defaults to {}): mapping between integers representing big_map indices and descriptions of big maps (see the syntax of extra big maps specifications)

The following test example asserts that the default value for the NOW instruction is the unix epoch:

input {};
code NOW;
output { Stack_elt timestamp "1970-01-01T00:00:00Z" }

The following example shows how to use the now toplevel primitive to make the NOW instruction return a chosen timestamp:

input {};
now "2020-01-08T07:13:51Z";
code NOW;
output { Stack_elt timestamp "2020-01-08T07:13:51Z" }

Syntax of concrete stacks#

A concrete stack is written as a Micheline sequence whose elements are of the form Stack_elt <ty> <x> where <x> is a Michelson value and <ty> is its type. For example, { Stack_elt bool True ; Stack_elt nat 42 } is a concrete stack of length 2 whose top element is the boolean True and the bottom element is the natural number 42.

Omitting parts of the output#

Any part of the output specification can be replaced with the wildcard pattern _.

For example, let’s consider the following test of the PAIR instruction:

input {Stack_elt bool True; Stack_elt string "foo"};
code PAIR;
output {Stack_elt (pair bool string) (Pair True "foo")}

Omitting the True argument to the Pair primitive can be done as follows:

input {Stack_elt bool True; Stack_elt string "foo"};
code PAIR;
output {Stack_elt (pair bool string) (Pair _ "foo")}

Omitting the Pair primitive:

input {Stack_elt bool True; Stack_elt string "foo"};
code PAIR;
output {Stack_elt (pair bool string) (_ True "foo")}

Omitting the pair bool string type:

input {Stack_elt bool True; Stack_elt string "foo"};
code PAIR;
output {Stack_elt _ (Pair True "foo")}

Omitting the resulting stack element:

input {Stack_elt bool True; Stack_elt string "foo"};
code PAIR;
output {_}

Omitting all of the output:

input {Stack_elt bool True; Stack_elt string "foo"};
code PAIR;
output _

The difference between the last two examples is that output {_} means that the instruction is expected to successfully return a stack of length 1 while output _ means that nothing in particular is expected from the execution of the instruction, not even being successful.

The wildcard pattern is typically used to omit unspecified aspects of the Michelson language when writing portable tests; in particular the cryptographic nonces in values of type operation (see the syntax of concrete operations) or implementation-specific parts of error outputs (see the syntax of errors).

Output normalization#

The input and output stacks can use the readable and optimized formats for Michelson values and even mix the formats; for a test to pass, the expected output does not need to syntactically match the result of the execution but only to match up to conversion between optimized and readable formats; the TZT test runner is responsible for normalizing the actual output and the expected one to common format. This means in particular that conversion between readable and optimized formats can be tested by using {} as the code instruction sequence to test; for example these two tests pass:

input {Stack_elt address 0x0000e7670f32038107a59a2b9cfefae36ea21f5aa63c};
code {};
output {Stack_elt address "tz1gjaF81ZRRvdzjobyfVNsAeSC6PScjfQwN"}
input {Stack_elt address "tz1gjaF81ZRRvdzjobyfVNsAeSC6PScjfQwN"};
code {};
output {Stack_elt address 0x0000e7670f32038107a59a2b9cfefae36ea21f5aa63c}

This normalization feature is however incompatible with using the wildcard pattern in the output; when using wildcards the output must be formatted using the readable format so the following test does not pass:

input {Stack_elt address "tz1gjaF81ZRRvdzjobyfVNsAeSC6PScjfQwN"};
code {};
output {Stack_elt _ 0x0000e7670f32038107a59a2b9cfefae36ea21f5aa63c}

but the following test does pass:

input {Stack_elt address 0x0000e7670f32038107a59a2b9cfefae36ea21f5aa63c};
code {};
output {Stack_elt _ "tz1gjaF81ZRRvdzjobyfVNsAeSC6PScjfQwN"}

Syntax of errors#

To test that the execution of an instruction fails, the following syntaxes can be used instead of the output stack as the argument of the output toplevel primitive to specify which error the instruction is expected to raise:

  • (StaticError <error description>): an error occurred before the instruction was executed; the error description format is unspecified so consider using a wildcard such as (StaticError _) to write portable tests;

  • (Failed <value>): the execution reached a FAILWITH instruction and the topmost element of the stack at this point was <value>;

  • MutezUnderflow: a mutez subtraction resulted in a negative value. This should only happen in the case of the deprecated mutez case of the SUB instruction;

  • Overflow: an overflow was detected. This can happen when an addition or multiplication on type mutez produces a result which is too large to be represented as a value of type mutez, or when the number of bits to shift using the LSL or LSR instruction is too large.

The following example shows how to test a runtime failure; it asserts that the FAILWITH instruction produces a runtime error containing the top of the stack.

input { Stack_elt nat 4 ; Stack_elt bytes 0x };
output (Failed 4)

The following example shows how to test type checking failure; it asserts that the DUP instruction cannot be used on an empty stack.

input {};
code DUP;
output (StaticError _)

The following example shows another kind of static failure: a string cannot be passed as argument to the DUP instruction.

input { Stack_elt nat 8 };
code { DUP "foo" };
output (StaticError _)

Syntax of concrete operations#

The operation type has no concrete syntax in Michelson. In order to specify the result of the operation forging instructions TRANSFER_TOKENS, CREATE_CONTRACT, and SET_DELEGATE , the following data constructors are added:

  • Transfer_tokens,

  • Create_contract, and

  • Set_delegate.

They take as arguments the inputs of the corresponding operation forging instructions plus a cryptographic nonce represented as a byte sequence. The result of TRANSFER_TOKENS, CREATE_CONTRACT, and SET_DELEGATE have respectively the following shapes:

  • Transfer_tokens <argument> <amount in mutez> <address of destination> <nonce>,

  • Create_contract { <script> } <optional delegate> <initial balance in mutez> <initial storage> <nonce>, and

  • Set_delegate <optional delegate> <nonce>.

The computation of the cryptographic nonce is not specified. To write portable tests, the nonces appearing in output stack expectations should be replaced by a wildcard pattern.

Here is an example unit test for the SET_DELEGATE instruction used to set the delegate of the current contract to the account at address tz1NwQ6hkenkn6aYYio8VnJvjtb4K1pfeU1Z:

input { Stack_elt (option key_hash) (Some "tz1NwQ6hkenkn6aYYio8VnJvjtb4K1pfeU1Z") } ;
output { Stack_elt operation (Set_delegate (Some "tz1NwQ6hkenkn6aYYio8VnJvjtb4K1pfeU1Z") _) }

Syntax of other contracts specifications#

The behaviour of the CONTRACT instruction depends on whether or not its input is the address of an originated contract accepting the expected type as parameter. To test it, the other_contract toplevel primitive can be used to specify which contracts are assumed to be originated and which type they accept as parameter.

The mapping given to the other_contract toplevel primitive is a Micheline sequence whose elements have the form Contract "KT1..." <ty> where "KT1..." is a valid smart contract address and <ty> is the type of its parameter. Each address should appear at most once and the order is irrelevant.

Syntax of extra big maps specifications#

The behaviour of the instructions operating on type big_map depend on the contents of big maps stored in the context. To test them, the big_maps toplevel primitive can be used to specify the types and contents of the big maps which are assumed to be present.

The mapping given to the big_maps toplevel primitive is a Micheline sequence whose elements have the form Big_map <i> <kty> <vty> { Elt <k1> <v1>; Elt <k2> <v2>; ...} where <i> is an integer (the identifier of the big map), <kty> is the comparable type of keys, <vty> is the type of values, each <ki> is of type <kty> and each <vi> is of type <vty>. Each identifier should appear at most once and the order in which big maps are specified is irrelevant but each { Elt <k1> <v1>; Elt <k2> <v2>; ...} description of big map contents should be given in increasing order of keys.

The following example tests the GET instruction in the big_map case:

big_maps { Big_map 4 string nat { Elt "bar" 42 } };
input { Stack_elt (big_map string nat) 4 };
code { PUSH string "foo"; GET };
output { Stack_elt (option nat) None }