Server Basics

Creating a SoupServer

As with the client API, there is a single object that will encapsulate most of your interactions with libsoup. In this case, [class@Server].

You create the server with [ctor@Server.new], and as with the [class@Session] constructor, you can specify a few additional options:

[property@Server:tls-certificate]	A [class@Gio.TlsCertificate] (containing a private key) that will be used when handling HTTPS requests on the server.
[property@Server:raw-paths]	Set this to TRUE if you don't want libsoup to decode %-encoding in the Request-URI. (e.g. because you need to treat "/foo/bar" and "/foo%2Fbar" as different paths.
[property@Server:server-header]	Allows you to set a Server header string that will be sent on all responses.

Adding Listening Sockets

To tell the server where to listen, call [method@Server.listen] (to listen on a specific [class@Gio.SocketAddress]), [method@Server.listen_all] (to listen on a given port on all network interfaces), or [method@Server.listen_local] (to listen to a given port on the loopback interface only). You can call any of these functions multiple times, to set up multiple listening sockets.

To set up an HTTPS server, you must first either set the [property@Server:tls-certificate] property, or else call [method@Server.set_tls_certificate]. After that you can pass the SOUP_SERVER_LISTEN_HTTPS option to [method@Server.listen], etc.

By default, servers listen for both IPv4 and IPv6 connections; if you don't want this, use the SOUP_SERVER_LISTEN_IPV4_ONLY or SOUP_SERVER_LISTEN_IPV6_ONLY options.

The server runs asynchronously, in the thread-default [struct@GLib.MainContext] of the thread in which the listen calls were made.

Adding Handlers

By default, [class@Server] returns "404 Not Found" in response to all requests (except ones that it can't parse, which get "400 Bad Request"). To override this behavior, call [method@Server.add_handler] to set a callback to handle certain URI paths.

soup_server_add_handler (server, "/foo", server_callback,
                         data, destroy_notify);

The "/foo" indicates the base path for this handler. When a request comes in, if there is a handler registered for exactly the path in the request's Request-URI, then that handler will be called. Otherwise libsoup will strip path components one by one until it finds a matching handler. So for example, a request of the form GET /foo/bar/baz.html?a=1&b=2 HTTP/1.1 would look for handlers for /foo/bar/baz.html, /foo/bar, and /foo. If a handler has been registered with a NULL base path, then it is used as the default handler for any request that doesn't match any other handler.

Responding to Requests

A handler callback looks something like this:

static void
server_callback (SoupServer        *server,
                 SoupServerMessage *msg, 
                 const char        *path,
                 GHashTable        *query,
                 gpointer           user_data)
{
    // ...
}

msg is the request that has been received and user_data is the data that was passed to [method@Server.add_handler]. path is the path (from msg's URI), and query contains the result of parsing the URI query field. (It is NULL if there was no query.)

By default, libsoup assumes that you have completely finished processing the message when you return from the callback, and that it can therefore begin sending the response. If you are not ready to send a response immediately (e.g. you have to contact another server, or wait for data from a database), you must call [method@Server.pause_message] on the message before returning from the callback. This will delay sending a response until you call [method@Server.unpause_message]. (You must also connect to the [signal@ServerMessage::finished] signal on the message in this case, so that you can break off processing if the client unexpectedly disconnects before you start sending the data.)

To set the response status, call [method@ServerMessage.set_status]. If the response requires a body, you must decide whether to use Content-Length encoding (the default), or chunked encoding.

Responding with `Content-Length` Encoding

This is the simpler way to set a response body, if you have all of the data available at once.

static void
server_callback (SoupServer        *server,
                 SoupServerMessage *msg, 
                 const char        *path,
                 GHashTable        *query,
                 gpointer           user_data)
{
    MyServerData *server_data = user_data;
    const char *mime_type;
    GByteArray *body;

    if (soup_server_message_get_method (msg) != SOUP_METHOD_GET) {
        soup_server_message_set_status (msg, SOUP_STATUS_NOT_IMPLEMENTED, NULL);
        return;
    }

    /* This is somewhat silly. Presumably your server will do
     * something more interesting.
     */
    body = g_hash_table_lookup (server_data->bodies, path);
    mime_type = g_hash_table_lookup (server_data->mime_types, path);
    if (!body || !mime_type) {
        soup_server_message_set_status (msg, SOUP_STATUS_NOT_FOUND, NULL);
        return;
    }

    soup_server_message_set_status (msg, SOUP_STATUS_OK, NULL);
    soup_server_message_set_response (msg, mime_type, SOUP_MEMORY_COPY,
                                      body->data, body->len);
}

Responding with `chunked` Encoding

If you want to supply the response body in chunks as it becomes available, use chunked encoding instead. In this case, first call soup_message_headers_set_encoding (msg->response_headers, SOUP_ENCODING_CHUNKED) to tell libsoup that you'll be using chunked encoding. Then call [method@MessageBody.append] (or [method@MessageBody.append_bytes]) on msg->response_body with each chunk of the response body as it becomes available, and call [method@MessageBody.complete] when the response is complete. After each of these calls, you must also call [method@Server.unpause_message] to cause the chunk to be sent. (You do not normally need to call [method@Server.pause_message], because I/O is automatically paused when doing a chunked transfer if no chunks are available.)

When using chunked encoding, you must also connect to the [signal@ServerMessage::finished] signal on the message, so that you will be notified if the client disconnects between two chunks; [class@Server] will unref the message if that happens, so you must stop adding new chunks to the response at that point. (An alternate possibility is to write each new chunk only when the [signal@ServerMessage::wrote_chunk] signal is emitted indicating that the previous one was written successfully.)

The simple-proxy example in the examples/ directory gives an example of using chunked encoding.

Handling Authentication

To have [class@Server] handle HTTP authentication for you, create a [class@AuthDomainBasic] or [class@AuthDomainDigest], and pass it to [method@Server.add_auth_domain]:

SoupAuthDomain *domain;

domain = soup_auth_domain_basic_new (
    "realm", "My Realm",
    "auth-callback", auth_callback,
    "auth-data", auth_data,
    "add-path", "/foo",
    "add-path", "/bar/private",
    NULL);
soup_server_add_auth_domain (server, domain);
g_object_unref (domain);

Then, every request under one of the auth domain's paths will be passed to the auth_callback first before being passed to the server_callback:

static gboolean
auth_callback (SoupAuthDomain *domain, SoupServerMessage *msg,
               const char *username, const char *password,
               gpointer user_data)
{
    MyServerData *server_data = user_data;
    MyUserData *user;

    user = my_server_data_lookup_user (server_data, username);
    if (!user)
        return FALSE;

    /* FIXME: Don't do this. Keeping a cleartext password database
     * is bad.
     */
    return strcmp (password, user->password) == 0;
}

The [callback@AuthDomainBasicAuthCallback] is given the username and password from the Authorization header and must determine, in some server-specific manner, whether or not to accept them. (In this example we compare the password against a cleartext password database, but it would be better to store the password somehow encoded, as in the UNIX password database. Alternatively, you may need to delegate the password check to PAM or some other service.)

If you are using Digest authentication, note that [callback@AuthDomainDigestAuthCallback] works completely differently (since the server doesn't receive the cleartext password from the client in that case, so there's no way to compare it directly). See the documentation for [class@AuthDomainDigest] for more details.

You can have multiple [class@AuthDomain]s attached to a [class@Server], either in separate parts of the path hierarchy, or overlapping. (e.g. you might want to accept either Basic or Digest authentication for a given path.) When more than one auth domain covers a given path, the request will be accepted if the user authenticates successfully against any of the domains.

If you want to require authentication for some requests under a certain path, but not all of them (e.g. you want to authenticate PUT requests, but not GET requests), use a [callback@AuthDomainFilter].