Cohttp is an OCaml library for creating HTTP daemons. It has a portable HTTP parser, and implementations using various asynchronous programming libraries:
Cohttp_lwt_unix
uses the Lwt library, and specifically the UNIX bindings. It uses ocaml-tls as the TLS implementation to handle HTTPS connections.Cohttp_async
uses the Async library andasync_ssl
to handle HTTPS connections.Cohttp_lwt
exposes an OS-independent Lwt interface, which is used by the Mirage interface to generate standalone microkernels (use the cohttp-mirage subpackage).Cohttp_lwt_jsoo
compiles to a JavaScript module that maps the Cohttp calls to XMLHTTPRequests. This is used to compile OCaml libraries like the GitHub bindings to JavaScript and still run efficiently.
You can implement other targets using the parser very easily. Look at the IO
signature in lib/s.mli
and implement that in the desired backend.
You can find help from cohttp users and maintainers at the discuss.ocaml.org forum or on the OCaml discord server.
- Installation
- Client Tutorial
- Dealing with timeouts
- Managing sessions
- Multipart form data
- Creating custom resolver: a Docker Socket Client example
- Basic Server Tutorial
- Installed Binaries
- Debugging
- Important Links
Latest stable version should be obtained from opam
. Make sure to install the
specific backends you want as well. E.g.
$ opam install cohttp-lwt-unix cohttp-async
You can also obtain the development release:
$ opam pin add cohttp --dev-repo
Cohttp provides clients for Async, Lwt, and Js_of_ocaml (Lwt based). In this tutorial, we will use the lwt client but the example should be easily translatable to Async.
To create a simple request, use one of the methods in Cohttp_lwt_unix.Client
.
call
is the most general, there are also http method specialized such as
get
, post
, etc.
For example downloading the reddit frontpage:
cat - > client_example.ml <<EOF
open Lwt
open Cohttp
open Cohttp_lwt_unix
let body =
Client.get (Uri.of_string "https://www.reddit.com/") >>= fun (resp, body) ->
let code = resp |> Response.status |> Code.code_of_status in
Printf.printf "Response code: %d\n" code;
Printf.printf "Headers: %s\n" (resp |> Response.headers |> Header.to_string);
body |> Cohttp_lwt.Body.to_string >|= fun body ->
Printf.printf "Body of length: %d\n" (String.length body);
body
let () =
let body = Lwt_main.run body in
print_endline ("Received body\n" ^ body)
EOF
There are a few things to notice:
- We open 2 modules.
Cohttp
contains the backend independent modules andCohttp_lwt_unix
the lwt + unix specific ones. Client.get
accepts aUri.t
and makes an http request.Client.get
also accepts optional arguments for things like header information.- The http response is returned in a tuple. The first element of the tuple contains the response's status code, headers, http version, etc. The second element contains the body.
- The body is then converted to a string and is returned (after the length is
printed). Note that
Cohttp_lwt.Body.to_string
hence it's up to us to keep a reference to the result. - We must trigger lwt's event loop for the request to run.
Lwt_main.run
will run the event loop and return with final value ofbody
which we then print.
Note that Cohttp_lwt_unix
/Cohttp_async
are able to request an HTTPS page
by default. For Cohttp_lwt_unix
, we use ocaml-tls
(to use lwt_ssl
is enough to use Cohttp_lwt_unix_ssl
from the analogously
named package, the rest of the code does not change). For Cohttp_async
, we use
async_ssl
(but the user is able to use ocaml-tls
with some modifications).
Consult the following modules for reference:
The full documentation for the latest published version of the library is available on the repository github pages.
Build and execute with:
$ ocamlbuild -pkg cohttp-lwt-unix client_example.native
$ ./client_example.native
For manual builds, it is usually enough to remember that cohttp ships with 6
findlib (ocamlfind
) libraries:
cohttp
- BaseCohttp
module. No platform specific functionalitycohttp-async
- Async backendCohttp_async
cohttp-lwt
- Lwt backend without unix specificscohttp-lwt-unix
- Unix based lwt backend withtls
supportcohttp-lwt-unix-ssl
- Unix based lwt backend withlwt_ssl
supportcohttp-lwt-unix-nossl
- Unix based lwt backend (onlyhttp
)cohttp-lwt-jsoo
- Jsoo (XHR) clientcohttp-top
- Print cohttp types in the toplevel (#require "cohttp-top"
)
Create this dune
file
cat - > dune <<EOF
(executable
; (public_name client_example)
(name client_example)
(libraries cohttp-lwt-unix))
EOF
then build and execute the example with
$ dune exec ./client_example.exe
You can use Lwt.pick
to set a timeout
on the execution of a thread. For example, say that you want to set a timeout on
the Client.get
thread in the example above, then you could modify the get call
as follows
(* [...] *)
let compute ~time ~f =
Lwt.pick
[
(f () >|= fun v -> `Done v)
; (Lwt_unix.sleep time >|= fun () -> `Timeout)
]
let body =
let get () = Client.get (Uri.of_string "https://www.reddit.com/") in
compute ~time:0.1 ~f:get >>= function
| `Timeout -> Lwt.fail_with "Timeout expired"
| `Done (resp, body) ->
let code =
(* [...] *)
Executing the code, which you can actually try by calling
$ dune exec examples/lwt_unix_doc/client_lwt_timeout.exe
the call will most likely fail with the following output
Fatal error: exception (Failure "Timeout expired")
Similarly, in the case of cohttp-async
you can directly use Async's
with_timeout
function.
Managing sessions and saving cookies across requests is not directly supported by
cohttp
. It is not hard to roll out a custom solution, but an alternative is
to use the session
library,
which is compatible with cohttp
.
Multipart form data is not supported out of the box, but is provided by the external library
multipart-form-data
.
Cohttp provides a lot of utilities out of the box, but does not prevent the users to dig in and customise it for their needs. The following is an example of a unix socket client to communicate with Docker.
cat - > docker_example.ml <<EOF
open Lwt.Infix
let resolve_unix_socket : Conduit_lwt.Endpoint.t -> 'edn option Lwt.t = function
| IP _ -> Lwt.return_none
| Domain v -> (
match Domain_name.to_string v with
| "docker" -> Lwt.return_some (Unix.ADDR_UNIX "/var/run/docker.sock")
| _ -> Lwt.return_none )
let t =
let ctx =
Conduit_lwt.add ~priority:0 (* highest priority *) Conduit_lwt.TCP.protocol
resolve_unix_socket Conduit.empty
in
Cohttp_lwt_unix.Client.get ~ctx (Uri.of_string "http://docker/version")
>>= fun (resp, body) ->
let open Cohttp in
let code = resp |> Response.status |> Code.code_of_status in
Printf.printf "Response code: %d\n" code;
Printf.printf "Headers: %s\n" (resp |> Response.headers |> Header.to_string);
body |> Cohttp_lwt.Body.to_string >|= fun body ->
Printf.printf "Body of length: %d\n" (String.length body);
print_endline ("Received body\n" ^ body)
let _ = Lwt_main.run t
EOF
The main issue there is there no way to resolve a socket address, so you need to create a custom resolver to map a hostname to the Unix domain socket.
To build and execute with dune
, first create the following dune
file
cat - > dune <<EOF
(executable
;(public_name docker_example)
(name docker_example)
(libraries cohttp-lwt-unix conduit-lwt))
EOF
then run the example with
dune exec ./docker_example.exe
Even though conduit is transitively there, for this example we are explicitly mentioning it to emphasize that we are creating a new Conduit resolver. Refer to conduit's README for examples of use and links to up-to-date conduit documentation.
Implementing a server in cohttp using the Lwt backend (for Async is very similar) is mostly equivalent to implementing a function of type :
conn -> Cohttp.Request.t -> Cohttp_lwt.Body.t -> (Cohttp.Response.t * Cohttp_lwt.Body.t) Lwt.t
The parameters are self explanatory but we'll summarize them quickly here:
conn
- contains connection informationCohttp.Request.t
- Request information such as method, uri, headers, etc.Cohttp_lwt.Body.t
- Contains the request body. You must manually decode the request body into json, form encoded pairs, etc. For cohttp, the body is simply binary data.
Here's an example of a simple cohttp server that outputs back request information.
cat - > server_example.ml <<EOF
open Lwt
open Cohttp
open Cohttp_lwt_unix
let server =
let callback _conn req body =
let uri = req |> Request.uri |> Uri.to_string in
let meth = req |> Request.meth |> Code.string_of_method in
let headers = req |> Request.headers |> Header.to_string in
( body |> Cohttp_lwt.Body.to_string >|= fun body ->
Printf.sprintf "Uri: %s\nMethod: %s\nHeaders\nHeaders: %s\nBody: %s" uri
meth headers body )
>>= fun body -> Server.respond_string ~status:`OK ~body ()
in
let tcp_config =
{
Conduit_lwt.TCP.sockaddr = Unix.ADDR_INET (Unix.inet_addr_loopback, 8000);
capacity = 40;
}
in
Server.create tcp_config Conduit_lwt.TCP.protocol Conduit_lwt.TCP.service
(Server.make ~callback ())
let _ = Lwt_main.run (server ())
EOF
Build and execute with:
$ ocamlbuild -pkg cohttp-lwt-unix server_example.native
$ ./server_example.native
Create this dune
file
cat - > dune <<EOF
(executable
; (public_name server_example)
(name server_example)
(libraries cohttp-lwt-unix conduit-lwt))
EOF
then build and execute the example with
$ dune exec ./client_example.exe
As in the previous example, here we are explicitly mentioning conduit-lwt to emphasize that we are relying on Conduit to specify the protocols and the services. Refer to conduit's README for examples of use and links to up-to-date conduit documentation.
Cohttp comes with a few simple binaries that are handy, useful also to test cohttp itself, and can serve as examples of how to use the library. All binaries come in two flavours - Async and Lwt.
$ cohttp-curl-{lwt,async}
This is a simple curl utility implemented using cohttp. An example of an invocation is:
$ cohttp-curl-lwt -v -X GET "https://www.reddit.com/"
$ cohttp-server-{lwt,async}
This binary acts in a similar fashion to the Python SimpleHTTPServer
. Just
run cohttp-server-async
in a directory and it will open up a local port and
serve the files over HTTP.
$ cohttp-server-async
Assuming that the server is running in cohttp's source directory:
$ cohttp-curl-lwt 'http://0.0.0.0:8080/README.md'
Other examples using the async api are avaliable in the examples/async folder in the sources
You can activate some runtime debugging for the servers by setting COHTTP_DEBUG
to any value different from 0
or false
, and it will set a default debug-level logger on stdout. Note: If you turn on the debugging on the cohttp-lwt-server
example, you need to make sure you also pass the -vvv
option, which forces the debug level of the logger.
Since both Cohttp and Conduit use Logs
for debugging output, you can enable custom debugging in your code (if needed) by adding something like the following to your code (courtesy of @dinosaure)
let reporter ppf =
let report src level ~over k msgf =
let k _ =
over () ;
k () in
let with_metadata header _tags k ppf fmt =
Format.kfprintf k ppf
("%a[%a]: " ^^ fmt ^^ "\n%!")
Logs_fmt.pp_header (level, header)
Fmt.(styled `Magenta string)
(Logs.Src.name src) in
msgf @@ fun ?header ?tags fmt -> with_metadata header tags k ppf fmt in
{ Logs.report }
let () = Fmt_tty.setup_std_outputs ~style_renderer:`Ansi_tty ~utf_8:true ()
let () = Logs.set_reporter (reporter Fmt.stderr)
let () = Logs.set_level ~all:true (Some Logs.Debug)