• en


See also Stack Overflow, which is widely used by the OCaml community.

General Questions

What is OCaml?

OCaml is a programming language. It is a functional language, since the basic units of programs are functions. It is a strongly-typed language; it means that the objects that you use belong to a set that has a name, called its type. In OCaml, types are managed by the computer, the user has nothing to do about types (types are synthesized). The language is available on almost every Unix platform (including Linux and MacOS X) and on PCs under Windows. A brief tour on main features of OCaml.

Under what licensing terms is the OCaml software available?

The OCaml system is open source software. Since version 4.03 the compiler and the standard library are distributed under LGPL 2.1 with static linking exception, read the license for details. The software is also available under a BSD-style license for a fee through the OCaml Consortium.

What is the meaning of the name “OCaml”

CAML once was an acronym that stood for “Categorical Abstract Machine Language”, an abstract machine its early versions targeted. The evaluation model has changed since then, but the name stuck.

The “O” stands for “objective” and was added when the language got object-oriented programming capabilities.

Do you write “Caml”, “CAML”, “Ocaml”, “OCaML”, “OCAML” or “OCaml” ?

The official name of the language, capitalization included, is “OCaml”.

Is OCaml a compiled or interpreted language?

OCaml is compiled. However, the OCaml compiler offers a top-level interactive loop, that is similar to an interpreter. In fact, in the interactive system, the user may type in program chunks (we call these pieces OCaml “phrases”) that the system handles at once, compiling them, executing them, and writing their results.

What are the differences between Caml V3.1, Caml Light, and OCaml?

These are different Caml implementations that have been developed successively at Inria. These systems share many features since they all implement the core of the OCaml language; so the basic syntax is nearly the same. However, all these systems have their own extensions to the Caml core language.
Caml V3.1 is no longer maintained nor distributed. Caml Light is no longer developed, but still maintained. Because of its stable status, it is actively used in education. Most other users have switched to OCaml, the latest variant of the language. This is the version we suggest using in new software developments. See our brief history of the OCaml language.

How to report a bug in the compilers?

Use the bug tracking system to browse bug reports and features request, and submit new ones.

Core Language

Basic types

Is it possible to do computations with arbitrary precision arithmetic?

OCaml provides a library called Num that handles exact arithmetic computation for rational numbers.

Operations on big numbers gets the suffix /: addition is thus +/. You build big numbers using conversion from (small) integers or character strings. For printing in the interactive toplevel (aka REPL), a custom printer can be used. An example under OCaml is given below.

# #load "nums.cma";;
# open Num open Format;;
# let print_num ff n = fprintf ff "%s" (string_of_num n);;
val print_num : Format.formatter -> Num.num -> unit = <fun> # #install_printer print_num;;
# num_of_string "2/3";;
- : Num.num = 2/3 # let n = num_of_string "1/3" +/ num_of_string "2/3";;
val n : Num.num = 1 # let rec fact n = if n <= 0 then (num_of_int 1) else num_of_int n */ (fact (n - 1));;
val fact : int -> Num.num = <fun> # fact 100;;
- : Num.num = 93326215443944152681699238856266700490715968264381621468592963895217599993229915608941463976156518286253697920827223758251185210916864000000000000000000000000

Data structures

My array is modified, and I don't know why

This is due to the physical sharing of two arrays that you missed. In OCaml there is no implicit array copying. If you give two names to the same array, every modification on one array will be visible to the other:

# let v = Array.make 3 0;;
val v : int array = [|0; 0; 0|] # let w = v;;
val w : int array = [|0; 0; 0|] # w.(0) <- 4;;
- : unit = () # v;;
- : int array = [|4; 0; 0|]

The physical sharing effect also applies to elements stored in vectors: if these elements are also vectors, the sharing of these vectors implies that modifying one of these elements modifies the others (see also the entry below).

How to define multidimensional arrays?

The only way is to define an array whose elements are arrays themselves (OCaml arrays are unidimensional, they modelize mathematical vectors). The naive way to define multidimensional arrays is bogus: the result is not right because there is some unexpected physical sharing between the lines of the new array (see also previous entry):

# let m = Array.make 2 (Array.make 3 0);;
val m : int array array = [|[|0; 0; 0|]; [|0; 0; 0|]|] # m.(0).(0) <- 1;;
- : unit = () # m;;
- : int array array = [|[|1; 0; 0|]; [|1; 0; 0|]|]

The allocation of a new array has two phases. First, the initial value is computed; then this value is written in each element of the new array. That's why the line which is allocated by Array.make 3 0 is unique and physically shared by all the lines of the array m.
The solution is to use the make_matrix primitive that builds the matrix with all elements equal to the initial value provided. Alternatively, write the program that allocates a new line for each line of your matrix. For instance:

# let matrix n m init =
    let result = Array.make n (Array.make m init) in
    for i = 1 to n - 1 do
      result.(i) <- Array.make m init
val matrix : int -> int -> 'a -> 'a array array = <fun>

In the same vein, the copy_vect primitive gives strange results, when applied to matrices: you need to write a function that explicitly copies each line of the matrix at hand:

# let copy_matrix m =
    let l = Array.length m in
    if l = 0 then m else
      let result = Array.make l m.(0) in
      for i = 0 to l - 1 do
        result.(i) <- Array.copy m.(i)
val copy_matrix : 'a array array -> 'a array array = <fun>

Types definitions

How to define an enumerated type?

An enumerated type is a sum type with only constants. For instance, a type with 3 constants:

# type color = Blue | White | Red;;
type color = Blue | White | Red # Blue;;
- : color = Blue

The names Blue, White and Red are the constructors of the color type. One can define functions on this type by pattern matching:

# let string_of_color = function
    | Blue -> "blue"
    | White -> "white"
    | Red -> "red";;
val string_of_color : color -> string = <fun>

Is it possible to make a record value without declaring its type first?

No, Before making a record, you must give record's type a name, using the type keyword, or the type name for the record should at least be in scope. Ocaml needs to know the record type's name and associated field names before making a record value. Otherwise it results in error "Unbound record field".

# type person = { name: string; age: int }
  let p1 = { name="John"; age=30 };;
type person = { name : string; age : int; } val p1 : person = {name = "John"; age = 30}

How to share a field between two different record types?

When you define two types sharing a field name, the last defined type hides the field of the first type. For instance:

# type point_3d = {x : float; y : float; z : float};;
type point_3d = { x : float; y : float; z : float; } # type point_2d = {x : float; y : float};;
type point_2d = { x : float; y : float; } # {x = 10.; y = 20.; z = 30.};;
- : point_3d = {x = 10.; y = 20.; z = 30.}

Since OCaml 4.02, fields are automatically disambiguated when types are known. For example, in let u:point_3d = ... in u.x, u.x refers to the field of point_3d even though it is shadowed. However, field disambiguation does not work when type information is not available (e.g. in let get_x u = u.x where the type of get_x is not otherwise constrained), and may produce confusing results when types are omitted, so one may consider avoiding the problem entirely.

The simplest way to overcome this problem is to simply use different names! For instance

# type point3d = {x3d : float; y3d : float; z3d : float};;
type point3d = { x3d : float; y3d : float; z3d : float; } # type point2d = {x2d : float; y2d : float};;
type point2d = { x2d : float; y2d : float; }

One can propose two others solutions. First, it is possible to use modules to define the two types in different name spaces:

# module D3 = struct
    type point = {x : float; y : float; z : float}
  module D2 = struct
    type point = {x : float; y : float}
module D3 : sig type point = { x : float; y : float; z : float; } end module D2 : sig type point = { x : float; y : float; } end

This way fields can be fully qualified as D3.x D2.x:

# {D3.x = 10.; D3.y = 20.; D3.z = 30.};;
- : D3.point = {D3.x = 10.; y = 20.; z = 30.} # {D2.x = 10.; D2.y = 20.};;
- : D2.point = {D2.x = 10.; y = 20.}

You can also use objects that provide overloading on method names:

# class point_3d ~x ~y ~z = object
    method x : float = x
    method y : float = y
    method z : float = z
class point_3d : x:float -> y:float -> z:float -> object method x : float method y : float method z : float end # class point_2d ~x ~y = object method x : float = x method y : float = y end;;
class point_2d : x:float -> y:float -> object method x : float method y : float end

Note that objects provide you more than overloading: you can define truly polymorphic functions, working on both point_3d and point_2d, and you can even coerce a point_3d to a point_2d.

How to define two sum types that share constructor names?

Since OCaml 4.02, constructors are automatically disambiguated when types are known. For example, in type a = A;; type b = A of int;; let x:a = A, A is recognized as belonging to the type a even though its constructor is shadowed. However, constructor disambiguation does not work when type information is not available (e.g. in let get_n x = match x with A -> 1 where the type of get_n is not otherwise constrained), and may produce confusing results when types are omitted, so one may consider avoiding the problem entirely.

Generally speaking, sharing names between two constructors is not possible. As for all other names, you must use distinct name constructors. However, you can define the two types in two different name spaces, i.e. into two different modules. As for fields discussed above, you obtain constructors that can be qualified by their module names. With OCaml you can alternatively use polymorphic variants, i.e. constructors that are, in some sense, predefined, since they are not defined by a type definition. For instance:

# type ids = [ `Name | `Val ];;
type ids = [ `Name | `Val ] # type person = [ `Name of string ];;
type person = [ `Name of string ] # let f : person -> string = function `Name s -> s;;
val f : person -> string = <fun> # let is_name : ids -> bool = function `Name -> true | _ -> false;;
val is_name : ids -> bool = <fun>

Functions and procedures

How to define a function?

In OCaml, the syntax to define functions is close to the mathematical usage: the definition is introduced by the keyword let, followed by the name of the function and its arguments; then the formula that computes the image of the argument is written after an = sign.

# let successor (n) = n + 1;;
val successor : int -> int = <fun>

In fact, parens surrounding the argument may be omitted, so we generally write:

# let successor n = n + 1;;
val successor : int -> int = <fun>

How to define a recursive function?

You need to explicitly tell that you want to define a recursive function: use let rec instead of let. For instance:

# let rec fact n =
    if n = 0 then 1 else n * fact (n - 1);;
val fact : int -> int = <fun> # let rec fib n = if n <= 1 then n else fib (n - 1) + fib (n - 2);;
val fib : int -> int = <fun>

Functions may be mutually recursive:

# let rec odd n =
    if n = 0 then false
    else if n = 1 then true else even (n - 1)
  and even n =
    if n = 0 then true
    else if n = 1 then false else odd (n - 1);;
val odd : int -> bool = <fun> val even : int -> bool = <fun>

How to apply a function?

Functions are applied as in mathematics: write the function's name, followed by its argument enclosed in parens: f (x). In practice, parens are omitted in case of constants or identifiers: we write fib 2 instead of fib (2), and fact x instead of fact (x).
When the argument of a function is more complex than just an identifier, you must enclose this argument between parentheses. In particular you need parens when the argument is a negative constant number: to apply f to -1 you must write f (-1) and not f -1 that is syntactically similar to f - 1 (hence it is a subtraction, not an application).

How to define a procedure?

Recall that procedures are commands that produce an effect (for instance printing something on the terminal or writing some memory location), but have no mathematically meaningful result.

In OCaml, there is no special treatment of procedures: they are just considered as special cases of functions that return the special “meaningless” value (). For instance, the print_string primitive that prints a character string on the terminal, just returns () as a way of indicating that its job has been properly completed.

Procedures that do not need any meaningful argument, get () as dummy argument. For instance, the print_newline procedure, that outputs a newline on the terminal, gets no meaningful argument: it has type unit -> unit.

Procedures with argument are defined exactly as ordinary functions. For instance:

# let message s = print_string s; print_newline();;
val message : string -> unit = <fun> # message "Hello world!";;
Hello world! - : unit = ()

How to define a procedure/function that takes no argument?

Note that it is impossible to define a procedure without any argument at all: its definition would imply to execute it, and there would be no way to call it afterwards. In the following fragment double_newline is bound to (), and its further evaluation never produces carriage returns as may be erroneously expected by the user.

# let double_newline = print_newline(); print_newline();;
val double_newline : unit = () # double_newline;;
- : unit = ()

The correct definition and usage of this procedure is:

# let double_newline () = print_newline(); print_newline();;
val double_newline : unit -> unit = <fun> # double_newline;;
- : unit -> unit = <fun> # double_newline ();;
- : unit = ()

How to define a function with more than one argument?

Just write the list of successive arguments when defining the function. For instance:

# let sum x y = x + y;;
val sum : int -> int -> int = <fun>

then gives the actual arguments in the same order when applying the function:

# sum 1 2;;
- : int = 3

These functions are named “curried” functions, as opposed to functions with tuples as argument:

# let sum' (x, y) = x + y;;
val sum' : int * int -> int = <fun> # sum' (1, 2);;
- : int = 3

By convention, OCaml code generally uses curried functions rather than functions accepting a tuple as an argument. Of course, this does not apply to cases where the tuple is denoting a data structure on its own (e.g. (float * float * float) being used to represent a point).

How to define a function that has several results?

You can define a function that return a pair or a tuple:

# let div_mod x y = (x / y, x mod y);;
val div_mod : int -> int -> int * int = <fun> # div_mod 15 7;;
- : int * int = (2, 1)

What is an “anonymous function”?

You may use functions that have no names: we call them functional values or anonymous functions. A functional value is introduced by the keyword fun, followed by its argument, then an arrow -> and the function body. For instance:

# fun x -> x + 1;;
- : int -> int = <fun> # (fun x -> x + 1) 2;;
- : int = 3

What is the difference between fun and function?

Functions are usually introduced by the keyword fun. Each parameter is introduced by its own fun construct. For instance, the construct:

fun x -> fun y -> ...

defines a function with two parameters x and y. An equivalent but shorter form is:

fun x y -> ...

Functions that use pattern-matching are introduced by the keyword function. For example:

# function None -> false | Some _ -> true;;
- : 'a option -> bool = <fun>

My function is never applied

This is probably due to a missing argument: since OCaml is a functional programming language, there is no error when you evaluate a function with missing arguments: in this case, a functional value is returned, but the function is evidently not applied. Example: if you evaluate print_newline without argument, there is no error, but nothing happens. The compiler issues a warning in case of a blatant misuse.

# print_newline;;
- : unit -> unit = <fun> # print_newline ();;
- : unit = ()

Pattern matching

How to do nested pattern matching?

You imperatively need to enclose between parens a pattern matching which is written inside another pattern matching. In effect, the internal pattern matching “catches” all the pattern matching clauses that are written after it. For instance:

let f = function
  | 0 -> match ... with | a -> ... | b -> ...
  | 1 -> ...
  | 2 -> ...

is parsed as

let f = function
  | 0 ->
     match ... with
     | a -> ...
     | b -> ...
     | 1 -> ...
     | 2 -> ...

This error may occur for every syntactic construct that involves pattern matching: function, match .. with and try ... with. The usual trick is to enclose inner pattern matchings with begin and end. One write:

let f = function
  | 0 ->
     begin match ... with
     | a -> ...
     | b -> ...
  | 1 -> ...
  | 2 -> ...


Error message: a type is not compatible with itself

You may obtain the message: This expression has type “some type” but is used with type “the same some type”. This may occur very often when using the interactive system.
The reason is that two types with the same name have been defined the compiler does not confuse the two types, but the types are evidently written the same. Consider for instance:

# type t = T of int;;
type t = T of int # let x = T 1;;
val x : t = T 1 # type t = T of int;;
type t = T of int # let incr = function T x -> T (x+1);;
val incr : t -> t = <fun> # incr x;;
Error: This expression has type t/2 but an expression was expected of type t/1 Line 1, characters 0-17: Definition of type t/1 Line 1, characters 0-17: Definition of type t/2

This phenomenon appears when you load many times the same file into the interactive system, since each reloading redefines the types. The solution is to quit your interactive system and reload your files in a new session.

A function obtained through partial application is not polymorphic enough

The more common case to get a “not polymorphic enough” definition is when defining a function via partial application of a general polymorphic function. In OCaml polymorphism is introduced only through the “let” construct, and results from application are weakly polymorph; hence the function resulting from the application is not polymorph. In this case, you recover a fully polymorphic definition by clearly exhibiting the functionality to the type-checker : define the function with an explicit functional abstraction, that is, add a function construct or an extra parameter (this rewriting is known as eta-expansion):

# let map_id = List.map (function x -> x) (* Result is weakly polymorphic *);;
val map_id : '_weak1 list -> '_weak1 list = <fun> # map_id [1;2];;
- : int list = [1; 2] # map_id (* No longer polymorphic *);;
- : int list -> int list = <fun> # let map_id' l = List.map (function x -> x) l;;
val map_id' : 'a list -> 'a list = <fun> # map_id' [1;2];;
- : int list = [1; 2] # map_id' (* Still fully polymorphic *);;
- : 'a list -> 'a list = <fun>

The two definitions are semantically equivalent, and the new one can be assigned a polymorphic type scheme, since it is no more a function application.

The type of this expression contains type variables that cannot be generalized

This message appears when the OCaml compiler tries to compile a function or a value which is monomorphic, but for which some types have not been completely inferred. Some types variables are left in the type, which are are called “weak” (and are displayed by an underscore: '_a); they will disappear thanks to type inference as soon as enough information will be given.

# let r = ref [];;
val r : '_weak2 list ref = {contents = []} # let f = List.map (fun x -> x);;
val f : '_weak3 list -> '_weak3 list = <fun>

Since the expression mentioned in the error message cannot be compiled as is, two cases must be envisioned:

  • The expression can really not be turned into a polymorphic expression, as in r above. You must use an explicit type annotation, in order to turn it into something completely monomorphic.
  • The expression can be transformed into something polymorphic through rewriting some part of the code (for example using eta-expansion) as in the case of f.

How to write a function with polymorphic arguments?

In ML, an argument of a function cannot be polymorphic inside the body of the function; hence the following typing:

# let f (g : 'a -> 'a) x y = g x, g y;;
val f : ('a -> 'a) -> 'a -> 'a -> 'a * 'a = <fun>

The function is not as polymorphic as we could have hoped.
Nevertheless, in OCaml it is possible to use first-order polymorphism. For this, you can use either records or objects; in the case of records, you need to declare the type before using it in the function.

# let f (o : < g : 'a. 'a -> 'a >) x y = o#g x, o#g y;;
val f : < g : 'a. 'a -> 'a > -> 'b -> 'c -> 'b * 'c = <fun> # type id = { g : 'a. 'a -> 'a };;
type id = { g : 'a. 'a -> 'a; } # let f r x y = r.g x, r.g y;;
val f : id -> 'a -> 'b -> 'a * 'b = <fun>


Why some printing material is mixed up and does not appear in the right order?

If you use printing functions of the format module, you might not mix printing commands from format with printing commands from the basic I/O system. In effect, the material printed by functions from the format module is delayed (stored into the pretty-printing queue) in order to find out the proper line breaking to perform with the material at hand. By contrast low level output is performed with no more buffering than usual I/O buffering.

# print_endline "before";
  Format.print_string "MIDDLE";
  print_endline "after";;
before after MIDDLE- : unit = ()

To avoid this kind of problems you should not mix printing orders from format and basic printing commands; that's the reason why when using functions from the format module, it is considered good programming habit to open format globally in order to completely mask low level printing functions by the high level printing functions provided by format.

Module Language

Can I have two mutually recursive structures, signatures, functors inside a single compilation unit?

Yes, but structures always have to have an explcit signature. Recursive structures may be defined as follows:

# module rec A : sig
    type a = { x: int }
  end = struct
    type a = { x: int }
    let b : B.b = { y = 1.0 }
  end and B : sig
    type b = { y: float }
  end = struct
    type b = { y: float }
    let a : A.a = { x = 1 }
Line 1: Error: Reference to undefined global `CamlinternalMod'

In a similar way, mutually recursive signatures and functors can also be defined.

Can I have two mutually recursive compilation units?

With any two compilation units (.ml or .mli files), there must always exist an order in which it is possible to compile them sequentially. This precludes most kinds of recursion between compilation units.

However, two implementations can be recursive on types by exporting abstract versions of the types in the interfaces, with the manifest versions in the implementations referring to the actual types.

For example, consider these x.ml/x.mli files:

# type a (* only in x.ml: *) = Y.a
  type b =
  | BNone
  | BOne of a;;
Error: Unbound module Y

and y.ml/y.mli files:

# type b (* only in y.ml: *) = X.b
  type a =
  | ANone
  | AOne of b;;
Error: Unbound module X

In this way, cooperation between the X and Y modules allows the recursive value X.BOne (Y.AOne (X.BOne ...)) to be produced.

How do I express sharing constraints between modules?

Use manifest type specifications in the arguments of the functor. For instance, assume defined the following signatures:

module type S1 = sig ... type t ... end
module type S2 = sig ... type u ... end

To define a functor F that takes two arguments X:S1 and Y:S2 such that X.t and Y.u are the same, write:

module F (X: S1) (Y: S2 with type u = X.t) =
  struct ... end

Indeed, internally this expands to

module F (X: S1) (Y: sig ... type u = X.t ... end) =
  struct ... end

Compilation units are forced to be modules. What if I want to make a unit with a functor or a signature instead?

In OCaml, functors and signatures (module types) can be components of modules. So, just make the functor or signature be a component of a compilation unit. A good example is the module Set from the standard library.

Development Tools

Interactive toplevel (aka REPL)

How to stop the evaluation (execution) of an expression?

It is often possible to interrupt a program or the OCaml system by typing some combination of keys that is operating system dependent: under Unix send an interrupt signal (generally Control-C), under Macintosh OS type Command-., under Windows use the “OCaml” menu.

How to quit the interactive loop?

Type #quit;;. You can also quit it by inputting an end-of-file character with Ctrl-D on Unix, and Ctrl-Z on Windows.