Module Tiny_httpd

Tiny Http Server

This library implements a very simple, basic HTTP/1.1 server using blocking IOs and threads. Basic routing based is provided for convenience, so that several handlers can be registered.

It is possible to use a thread pool, see create's argument new_thread.

The echo example (see src/examples/echo.ml) demonstrates some of the features by declaring a few endpoints, including one for uploading files:

module S = Tiny_httpd

let () =
  let server = S.create () in

  (* say hello *)
  S.add_route_handler ~meth:`GET server
    S.Route.(exact "hello" @/ string @/ return)
    (fun name _req -> S.Response.make_string (Ok ("hello " ^name ^"!\n")));

  (* echo request *)
  S.add_route_handler server
    S.Route.(exact "echo" @/ return)
    (fun req -> S.Response.make_string
        (Ok (Format.asprintf "echo:@ %a@." S.Request.pp req)));

  (* file upload *)
  S.add_route_handler ~meth:`PUT server
    S.Route.(exact "upload" @/ string_urlencoded @/ return)
    (fun path req ->
        try
          let oc = open_out @@ "/tmp/" ^ path in
          output_string oc req.S.Request.body;
          flush oc;
          S.Response.make_string (Ok "uploaded file")
        with e ->
          S.Response.fail ~code:500 "couldn't upload file: %s"
            (Printexc.to_string e)
      );

  (* run the server *)
  Printf.printf "listening on http://%s:%d\n%!" (S.addr server) (S.port server);
  match S.run server with
  | Ok () -> ()
  | Error e -> raise e

It is then possible to query it using curl:

$ dune exec src/examples/echo.exe &
listening on http://127.0.0.1:8080

# the path "hello/name" greets you.
$ curl -X GET http://localhost:8080/hello/quadrarotaphile
hello quadrarotaphile!

# the path "echo" just prints the request.
$ curl -X GET http://localhost:8080/echo --data "howdy y'all"
echo:
{meth=GET;
 headers=Host: localhost:8080
         User-Agent: curl/7.66.0
         Accept: */*
         Content-Length: 10
         Content-Type: application/x-www-form-urlencoded;
 path="/echo"; body="howdy y'all"}

Tiny buffer implementation

These buffers are used to avoid allocating too many byte arrays when processing streams and parsing requests.

module Buf = Tiny_httpd_core.Buf

IO Abstraction

module IO = Tiny_httpd_core.IO

Logging

module Log = Tiny_httpd_core.Log

Utils

module Util = Tiny_httpd_core.Util

Resource pool

module Pool = Tiny_httpd_core.Pool

Static directory serving

module Dir = Tiny_httpd_unix.Dir
module type VFS = Tiny_httpd_unix.Dir.VFS

HTML combinators

module Html = Tiny_httpd_html

Alias to Tiny_httpd_html

Main server types

module Request = Tiny_httpd_core.Request
module Response = Tiny_httpd_core.Response
module Response_code = Tiny_httpd_core.Response_code
module Route = Tiny_httpd_core.Route
module Headers = Tiny_httpd_core.Headers
module Meth = Tiny_httpd_core.Meth
module Server = Tiny_httpd_core.Server
exception Bad_req of int * string

Exception raised to exit request handlers with a code+error message

Middlewares

A middleware can be inserted in a handler to modify or observe its behavior.

module Middleware = Server.Middleware
module Head_middleware = Server.Head_middleware

A middleware that only considers the request's head+headers.

Main Server type

A HTTP server. See create for more details.

module type IO_BACKEND = Server.IO_BACKEND

A backend that provides IO operations, network operations, etc.

val create_from : ?enable_logging:bool -> ?buf_size:int -> ?head_middlewares:Head_middleware.t list -> ?middlewares:([ `Encoding | `Stage of int ] * Middleware.t) list -> backend:(module IO_BACKEND) -> unit -> t

Create a new webserver using provided backend.

The server will not do anything until run is called on it. Before starting the server, one can use add_path_handler and set_top_handler to specify how to handle incoming requests.

  • parameter buf_size

    size for buffers (since 0.11)

  • parameter enable_logging

    if true and Logs is installed, emit logs via Logs (since 0.18). Default true.

  • since 0.14
val addr : t -> string

Address on which the server listens.

val is_ipv6 : t -> bool

is_ipv6 server returns true iff the address of the server is an IPv6 address.

  • since 0.3
val port : t -> int

Port on which the server listens. Note that this might be different than the port initially given if the port was 0 (meaning that the OS picks a port for us).

val active_connections : t -> int

Number of currently active connections.

val add_decode_request_cb : t -> (unit Tiny_httpd_core.Request.t -> (unit Tiny_httpd_core.Request.t * (Tiny_httpd_core.IO.Input.t -> Tiny_httpd_core.IO.Input.t)) option) -> unit

Add a callback for every request. The callback can provide a stream transformer and a new request (with modified headers, typically). A possible use is to handle decompression by looking for a Transfer-Encoding header and returning a stream transformer that decompresses on the fly.

val add_encode_response_cb : t -> (unit Tiny_httpd_core.Request.t -> Tiny_httpd_core.Response.t -> Tiny_httpd_core.Response.t option) -> unit

Add a callback for every request/response pair. Similarly to add_encode_response_cb the callback can return a new response, for example to compress it. The callback is given the query with only its headers, as well as the current response.

val add_middleware : stage:[ `Encoding | `Stage of int ] -> t -> Middleware.t -> unit

Add a middleware to every request/response pair.

  • parameter stage

    specify when middleware applies. Encoding comes first (outermost layer), then stages in increasing order.

  • raises Invalid_argument

    if stage is `Stage n where n < 1

  • since 0.11
val add_head_middleware : t -> Head_middleware.t -> unit

Add a request-header only Head_middleware.t. This is called on requests, to modify them, and returns a new request immediately.

  • since 0.18
Request handlers

Setup a handler called by default.

This handler is called with any request not accepted by any handler installed via add_path_handler. If no top handler is installed, unhandled paths will return a 404 not found

This used to take a string Request.t but it now takes a byte_stream Request.t since 0.14 . Use Request.read_body_full to read the body into a string if needed.

val add_route_handler : ?accept: (unit Tiny_httpd_core.Request.t -> (unit, Tiny_httpd_core.Response_code.t * string) result) -> ?middlewares:Middleware.t list -> ?meth:Tiny_httpd_core.Meth.t -> t -> ('a, string Tiny_httpd_core.Request.t -> Tiny_httpd_core.Response.t) Tiny_httpd_core.Route.t -> 'a -> unit

add_route_handler server Route.(exact "path" @/ string @/ int @/ return) f calls f "foo" 42 request when a request with path "path/foo/42/" is received.

Note that the handlers are called in the reverse order of their addition, so the last registered handler can override previously registered ones.

  • parameter meth

    if provided, only accept requests with the given method. Typically one could react to `GET or `PUT.

  • parameter accept

    should return Ok() if the given request (before its body is read) should be accepted, Error (code,message) if it's to be rejected (e.g. because its content is too big, or for some permission error). See the http_of_dir program for an example of how to use accept to filter uploads that are too large before the upload even starts. The default always returns Ok(), i.e. it accepts all requests.

  • since 0.6
val add_route_handler_stream : ?accept: (unit Tiny_httpd_core.Request.t -> (unit, Tiny_httpd_core.Response_code.t * string) result) -> ?middlewares:Middleware.t list -> ?meth:Tiny_httpd_core.Meth.t -> t -> ('a, Tiny_httpd_core.IO.Input.t Tiny_httpd_core.Request.t -> Tiny_httpd_core.Response.t) Tiny_httpd_core.Route.t -> 'a -> unit

Similar to add_route_handler, but where the body of the request is a stream of bytes that has not been read yet. This is useful when one wants to stream the body directly into a parser, json decoder (such as Jsonm) or into a file.

  • since 0.6
Server-sent events

EXPERIMENTAL: this API is not stable yet.

module type SERVER_SENT_GENERATOR = Server.SERVER_SENT_GENERATOR

A server-side function to generate of Server-sent events.

type server_sent_generator = (module SERVER_SENT_GENERATOR)

Server-sent event generator. This generates events that are forwarded to the client (e.g. the browser).

  • since 0.9
val add_route_server_sent_handler : ?accept: (unit Tiny_httpd_core.Request.t -> (unit, Tiny_httpd_core.Response_code.t * string) result) -> ?middlewares:Head_middleware.t list -> t -> ('a, string Tiny_httpd_core.Request.t -> server_sent_generator -> unit) Tiny_httpd_core.Route.t -> 'a -> unit

Add a handler on an endpoint, that serves server-sent events.

The callback is given a generator that can be used to send events as it pleases. The connection is always closed by the client, and the accepted method is always GET. This will set the header "content-type" to "text/event-stream" automatically and reply with a 200 immediately. See server_sent_generator for more details.

This handler stays on the original thread (it is synchronous).

  • since 0.9
Upgrade handlers

These handlers upgrade the connection to another protocol.

module type UPGRADE_HANDLER = Server.UPGRADE_HANDLER

Handler that upgrades to another protocol.

type upgrade_handler = (module UPGRADE_HANDLER)
  • since 0.17
val add_upgrade_handler : ?accept: (unit Tiny_httpd_core.Request.t -> (unit, Tiny_httpd_core.Response_code.t * string) result) -> ?middlewares:Head_middleware.t list -> t -> ('a, upgrade_handler) Tiny_httpd_core.Route.t -> 'a -> unit
Run the server
val running : t -> bool

Is the server running?

  • since 0.14
val stop : t -> unit

Ask the server to stop. This might not have an immediate effect as run might currently be waiting on IO.

val run : ?after_init:(unit -> unit) -> t -> (unit, exn) result

Run the main loop of the server, listening on a socket described at the server's creation time, using new_thread to start a thread for each new client.

This returns Ok () if the server exits gracefully, or Error e if it exits with an error.

  • parameter after_init

    is called after the server starts listening. since 0.13 .

val run_exn : ?after_init:(unit -> unit) -> t -> unit

run_exn s is like run s but re-raises an exception if the server exits with an error.

  • since 0.14
val create : ?enable_logging:bool -> ?masksigpipe:bool -> ?max_connections:int -> ?timeout:float -> ?buf_size:int -> ?get_time_s:(unit -> float) -> ?new_thread:((unit -> unit) -> unit) -> ?addr:string -> ?port:int -> ?sock:Unix.file_descr -> ?head_middlewares:Head_middleware.t list -> ?middlewares:([ `Encoding | `Stage of int ] * Middleware.t) list -> unit -> t

Create a new webserver using UNIX abstractions.

The server will not do anything until run is called on it. Before starting the server, one can use add_path_handler and set_top_handler to specify how to handle incoming requests.

  • parameter masksigpipe

    if true, block the signal Sys.sigpipe which otherwise tends to kill client threads when they try to write on broken sockets. Default: true except when on Windows, which defaults to false.

  • parameter buf_size

    size for buffers (since 0.11)

  • parameter new_thread

    a function used to spawn a new thread to handle a new client connection. By default it is Thread.create but one could use a thread pool instead. See for example this use of moonpool.

  • parameter max_connections

    maximum number of simultaneous connections.

  • parameter timeout

    connection is closed if the socket does not do read or write for the amount of second. Default: 0.0 which means no timeout. timeout is not recommended when using proxy.

  • parameter addr

    address (IPv4 or IPv6) to listen on. Default "127.0.0.1".

  • parameter port

    to listen on. Default 8080.

  • parameter sock

    an existing socket given to the server to listen on, e.g. by systemd on Linux (or launchd on macOS). If passed in, this socket will be used instead of the addr and port. If not passed in, those will be used. This parameter exists since 0.10.

  • parameter enable_logging

    if true and Logs is installed, log requests. Default true. This parameter exists since 0.18. Does not affect debug-level logs.

  • parameter get_time_s

    obtain the current timestamp in seconds. This parameter exists since 0.11.