Simple_httpd library to write web server and application


This library implements a HTTP/1.1 server for Linux using domains and algebraic effects. It uses epoll and eventfd to schedule the treatment of clients efficiently, with one domain used to accept request and several domains to handle these requests. Simple_httpd can listen to several addresses and ports and use this information together with the headers Host field to decide how to answer the request.

It is relatively fast and can handle thousands of simultaneous connections. It may use memory cache, sendfile system call and other TCP options to reach an efficiency comparable or better than nginx on static files.

Through a bunch of modules, described here Simple_httpd, we provide:


As we said, Simple_httpd is fast. Here is a plot of the latency obtained using vegeta with 1000 requests per seconds for 15s, on a small static file. Simple_httpd uses Simple_httpd.Dir.add_vfs using vfs_pack from the example/ file shown below:

Latencies for static files

Here is the same graphic for a small chaml/php dynamic file.

Latencies for dynamic files

Here are some other graphs showing the maximum number of requests possible, using wrk

comparison for small files

comparison for large files

apache and nginx are the usual servers with php-fpm and their default configuration on debian 12 except for accepting more connections than the default. There is a (reproducible) problem with nginx and php without ssl with a few extreme value that impact the mean.

Here is a similar graph showing the difference between .php and .chaml on a similar file. It shows that simple_httpd can serve at least than 5 times more requests.

comparison chaml versus php

Quick start

A good start is the template folder of the distribution, documented here. It contains the squeletton for a server serving two sites with status report and statistics.

See also examples/ below, that demonstrates some of the features by declaring a few endpoints, including one for uploading files and a virtual file system.

To go further, you should start reading Simple_httpd the main module of the library.

(* a fairly complete example *)
open Simple_httpd
open Response_code
module H = Headers

(** Parse command line options *)

(** Default address, port and maximum number of connections *)
let addr = ref ""
let port = ref 8080
let top_dir = ref None
let ssl_cert = ref ""
let ssl_priv = ref ""

(** Server.args provides a bunch and standard option to control the
    maximum number of connections, logs, etc... *)
let args, parameters = Server.args ()

let _ =
  Arg.parse (Arg.align ([
      "--addr", Arg.Set_string addr, " set address";
      "-a", Arg.Set_string addr, " set address";
      "--port", Arg.Set_int port, " set port";
      "-p", Arg.Set_int port, " set port";
      "--dir", Arg.String (fun s -> top_dir := Some s), " set the top dir for file path";
      "--ssl", Tuple[Set_string ssl_cert; Set_string ssl_priv], " give ssl certificate and private key";

    ] @ args)) (fun _ -> raise (Arg.Bad "")) "echo [option]*"

let ssl =
  if !ssl_cert <> "" then
    Some Address.{ cert = !ssl_cert; priv = !ssl_priv; protocol = Ssl.TLSv1_3 }
  else None

(** Server initialisation *)
let listens = [Address.make ~addr:!addr ~port:!port ?ssl ()]
let server = Server.create parameters ~listens

(** Compose the above filter with the compression filter
    provided by [Simple_httpd.Camlzip], than will compress output
    when [deflate] is accepted *)
let filter, get_stats =
  let filter_stat, get_stats = Stats.filter () in
  let filter_zip =
    Camlzip.filter ~compress_above:1024 ~buf_size:(16*1024) () in
  (Filter.compose_cross filter_zip filter_stat, get_stats)

(** Add a route answering 'Hello world' to [http://localhost/hello/world] *)
let _ =
  Server.add_route_handler ~meth:GET server ~filter
    Route.(exact "hello" @/ string @/ return)
    (fun name _req -> Response.make_string (Printf.sprintf "hello %s" name))

(** Add an echo request *)
let _ =
  Server.add_route_handler server ~filter
    Route.(exact "echo" @/ return)
    (fun req ->
      let q =
        Request.query req |> (fun (k,v) -> Printf.sprintf "%S = %S" k v)
        |> String.concat ";"
        (Format.asprintf "echo:@ %a@ (query: %s)@." Request.pp req q))

(** Add file upload *)
let _ =
  Server.add_route_handler_stream ~meth:PUT server ~filter
    Route.(exact "upload" @/ string @/ return)
    (fun path req ->
        Log.f (Req 0) (fun k->k "start upload %S, headers:\n%s\n\n%!" path
                     (Format.asprintf "%a" Headers.pp (Request.headers req)));
          let oc = open_out @@ "/tmp/" ^ path in
          Input.to_chan oc (Request.body req);
          flush oc;
          Response.make_string "uploaded file"
        with e ->
            "couldn't upload file: %s" (Printexc.to_string e)

(** Access to the statistics *)
let _ =
  Server.add_route_handler_chaml server ~filter Route.(exact "stats" @/ return)

(** Add a virtual file system VFS, produced by [simple-httpd-vfs-pack] from
    an actual folger *)
let _ =
  let vfs = Vfs.make ?top_dir:!top_dir () in
  Dir.add_vfs server
    ~config:(Dir.config ~download:true
               ~dir_behavior:Dir.Index_or_lists ())
    ~vfs:vfs ~prefix:"vfs"

(** Add a route sending a compressed stream for the given file in the current
    directory *)
let _ =
  Server.add_route_handler ~meth:GET server ~filter
    Route.(exact "zcat" @/ string @/ return)
    (fun path _req ->
        let ic = open_in path in
        let str = Input.of_chan ic in
        let mime_type =
            let p = Unix.open_process_in (Printf.sprintf "file -i -b %S" path) in
              let s = [H.Content_Type, String.trim (input_line p)] in
              ignore @@ Unix.close_process_in p;
            with _ -> ignore @@ Unix.close_process_in p; []
          with _ -> []
        Response.make_stream ~headers:mime_type str

(** Main pagen using the Html module (deprecated by vfs_pack and many other
    solutions *)
let _ =
  Server.add_route_handler_chaml server ~filter Route.return
     <!DOCTYPE html>
         <title>index of echo</title>
	 <p><b>endpoints are</b></p>
	   <li><pre>/ (GET)</pre> this file!</li>
           <li><pre>/hello/:name (GET)</pre> echo message</li>
           <li><pre><a href="/echo">echo</a></pre> echo back query</li>
           <li><pre>/upload/:path (PUT)</pre> to upload a file</li>
           <li><pre>/zcat/:path (GET)</pre> to download a file (deflate transfer-encoding)</li>
           <li><pre><a href="/stats/">/stats (GET)</a></pre> to access statistics</li>
           <li><pre><a href="/vfs/">/vfs (GET)</a></pre> to access a VFS
             embedded in the binary</li>

(** Start the server *)
let _ = server