Module Astring

Alternative Char and String modules.

Open the module to use it. This defines one value in your scope, redefines the (^) operator, the Char module and the String module.

Consult the differences with the OCaml String module, the porting guide and a few examples.

v0.8.3 - homepage


val strf : ('aFormat.formatter, unit, string) Pervasives.format4 -> 'a

strf is Format.asprintf.

val (^) : string -> string -> string

s ^ s' is String:append.

module Char : sig ... end

Characters (bytes in fact).

module String : sig ... end

Strings, substrings, string sets and maps.

Differences with the OCaml String module

First note that it is not a goal of Astring to maintain compatibility with the OCaml String module.

In Astring:

Porting guide

Opening Astring at the top of a module that uses the OCaml standard library in a project that compiles with -safe-string will either result in typing errors or compatible behaviour except for uses of the String.trim function, see below.

If for some reason you can't compile your project with -safe-string this may not be a problem. However you have to make sure that your code does not depend on fresh strings being returned by functions of the String module. The functions of Astring.String assume strings to be immutable and thus do not always allocate fresh strings for their results. This is the case for example for the ( ^ ) operator redefinition: no string is allocated whenever one of its arguments is an empty string. That being said it is still better to first make your project compile with -safe-string and then port to Astring.

The String.sub function is renamed to String.with_range. If you are working with String.find you may find it easier to use String.with_index_range which takes indices as arguments and is thus directly usable with the result of String.find. But in general index based string processing should be frown upon and replaced by substring extraction combinators.

Porting String.trim usages

The standard OCaml String.trim function only trims the characters ' ', '\t', '\n', '\012', '\r'. In Astring the default set adds vertical tab (0x0B) to the set to match the behaviour of the C isspace(3) function.

If you want to preserve the behaviour of the original function you can replace any use of String.trim with the following std_ocaml_trim function:

let std_ocaml_trim s =
  let drop = function
  | ' ' | '\n' | '\012' | '\r' | '\t' -> true
  | _ -> false
  String.trim ~drop s


We show how to use substrings to quickly devise LL(1) parsers. To keep it simple we do not implement precise error report, but note that it would be easy to add it by replacing the raise Exit calls by an exception with more information: we have everything at hand at these call points to report good error messages.

The first example parses version numbers structured as follows:


an unreadable Str regular expression for this would be:


Using substrings is certainly less terse but note that the parser is made of reusable sub-functions.

let parse_version : string -> (int * int * int * string option) option =
fun s -> try
  let parse_opt_v s = match String.Sub.head s with
  | Some ('v'|'V') -> String.Sub.tail s
  | Some _ -> s
  | None -> raise Exit
  let parse_dot s = match String.Sub.head s with
  | Some '.' -> String.Sub.tail s
  | Some _ | None -> raise Exit
  let parse_int s =
    match String.Sub.span ~min:1 ~sat:Char.Ascii.is_digit s with
    | (i, _) when String.Sub.is_empty i -> raise Exit
    | (i, s) ->
        match String.Sub.to_int i with
        | None -> raise Exit | Some i -> i, s
  let maj, s = parse_int (parse_opt_v (String.sub s)) in
  let min, s = parse_int (parse_dot s) in
  let patch, s = match String.Sub.head s with
  | Some '.' -> parse_int (parse_dot s)
  | _ -> 0, s
  let info = match String.Sub.head s with
  | Some ('+' | '-') -> Some (String.Sub.(to_string (tail s)))
  | Some _ -> raise Exit
  | None -> None
  Some (maj, min, patch, info)
with Exit -> None

The second example parses space separated key-value bindings environments of the form:

key0 = value0 key2 = value2 ...

To support values with spaces, values can be quoted between two '"' characters. If they are quoted then any "\\\"" subsequence (0x2F,0x22) is interpreted as the character '"' (0x22) and "\\\\" (0x2F,0x2F) is interpreted as the character '\\' (0x22).

let parse_env : string -> string option =
fun s -> try
  let skip_white s = String.Sub.drop ~sat:Char.Ascii.is_white s in
  let parse_key s =
    let id_char c = Char.Ascii.is_letter c || c = '_' in
    match String.Sub.span ~min:1 ~sat:id_char s with
    | (key, _) when String.Sub.is_empty key -> raise Exit
    | (key, rem) -> (String.Sub.to_string key), rem
  let parse_eq s = match String.Sub.head s with
  | Some '=' -> String.Sub.tail s
  | Some _ | None -> raise Exit
  let parse_value s = match String.Sub.head s with
  | Some '"' -> (* quoted *)
      let is_data = function '\\' | '"' -> false | _ -> true in
      let rec loop acc s =
        let data, rem = String.Sub.span ~sat:is_data s in
        match String.Sub.head rem with
        | Some '"' ->
            let acc = List.rev (data :: acc) in
            String.Sub.(to_string @@ concat acc), (String.Sub.tail rem)
        | Some '\\' ->
            let rem = String.Sub.tail rem in
            begin match String.Sub.head rem with
            | Some ('"' | '\\' as c) ->
                let acc = String.(sub (of_char c)) :: data :: acc in
                loop acc (String.Sub.tail rem)
            | Some _ | None -> raise Exit
        | None | Some _ -> raise Exit
      loop [] (String.Sub.tail s)
  | Some _ ->
      let is_data c = not (Char.Ascii.is_white c) in
      let data, rem = String.Sub.span ~sat:is_data s in
      String.Sub.to_string data, rem
  | None -> "", s
  let rec parse_bindings acc s =
    if String.Sub.is_empty s then acc else
    let key, s = parse_key s in
    let value, s = s |> skip_white |> parse_eq |> skip_white |> parse_value in
    parse_bindings (String.Map.add key value acc) (skip_white s)
  Some (String.sub s |> skip_white |> parse_bindings String.Map.empty)
with Exit -> None