STOMP on Erlang
stomp.erl
is a simple Erlang client for the STOMP protocol in version 1.2.
Currently only subscribing to queues is supported.
It provides an application called stomp
which takes configuration of the form:
[{stomp, #{host => "stomp-server.somedomain.sexy", % required port => 61613, % optional login => <<"someuser">>, % optional passcode => <<"hunter2>>, % optional }}].
Types
The following types are used in stomp.erl
, you can include them from
stomp.hrl
:
%% Client ack modes, refer to the STOMP protocol documentation -type ack_mode() :: client | client_individual | auto. %% Subscriptions are enumerated from 0 -type sub_id() :: integer(). %% Message IDs (for acknowledgements) are simple strings. They are %% extracted from the 'ack' field of the header in client or client-individual %% mode, and from the 'message-id' field in auto mode. -type message_id() :: binary(). %% A STOMP message as received from a queue subscription -record(stomp_msg, { headers :: #{ binary() => binary() }, body :: binary() }. -type stomp_msg() :: #stomp_msg{}.
Once the application starts it will register a process under the name
stomp_worker
and expose the following API:
Subscribing to a queue
%% Subscribe to a destination, receive the subscription ID -spec subscribe(binary(), % Destination (e.g. <<"/queue/lizards">>) ack_mode(), % Client-acknowledgement mode -> {ok, sub_id()}.
This synchronous call subscribes to a message queue. The stomp_worker
will
link itself to the caller and forward received messages as
{msg, sub_id(), stomp_msg()}
.
Depending on the acknowledgement mode specified on connecting, the subscriber may have to acknowledge receival of messages.
Acknowledging messages
%% Acknowledge a message ID. %% This is not required in auto mode. In client mode it will acknowledge the %% received messages up to the ID specified. In client-individual mode every %% single message has to be acknowledged. -spec ack(sub_id(), message_id()) -> ok. %% Explicitly "unacknowledge" a message -spec nack(sub_id(), message_id()) -> ok.
Both of these calls are asynchronous and will return immediately. Note that in
the case of the stomp_worker
crashing before a message acknowledgement is
handled, the message may be delivered again. Your consumer needs to be able to
handle this.