e2 vs OTP

e2_service vs gen_server

gen_server is the workhorse OTP behavior – it’s the primary OTP interface to Erlang processes.

e2_service is the equivalent in e2.

e2_service is identical to gen_server under the covers – it is a gen_server behavior. The additional overhead in using an e2 service is the cost of an additional functional call for each gen_server operation.

e2_service differs from gen_server as follows:

  • init/1 and terminate/2 are optional in e2

  • handle_call/3, handle_cast/2, and handle_info/2 are consolidated into a single handle_msg/3 callback in e2

  • e2_service does not support code_change/3 1

The minimum foot print of gen_server looks like this:




-export([init/1, handle_call/3, handle_cast/2, handle_info/2,
         terminate/2, code_change/3]).

-define(SERVER, ?MODULE).

-record(state, {}).

start_link() ->
    gen_server:start_link({local, ?SERVER}, ?MODULE, [], []).

init([]) ->
    {ok, #state{}}.

handle_call(_Request, _From, State) ->
    {reply, ok, State}.

handle_cast(_Msg, State) ->
    {noreply, State}.

handle_info(_Info, State) ->
    {noreply, State}.

terminate(_Reason, _State) ->

code_change(_OldVsn, State, _Extra) ->
    {ok, State}.

The minimum foot print of e2_service looks like this:


-export([start_link/0, handle_msg/3]).

-record(state, {}).

start_link() ->
    e2_service:start_link(?MODULE, #state{}).

handle_msg(_Msg, _From, State) ->
    {noreply, State}.

e2_application vs application

The application OTP behavior is represented by the e2_application behavior.

e2_application implicitly uses a top-level supervisor rather than require a separate module. e2 application modules provide the top level list of supervised child specs.

In OTP, applications typically start the top level supervisor.

e2_task vs gen_server

An e2 task is a type of sevice that runs actively after it’s started. It’s a cleaner alternative to this pattern used in gen_server behaviors:

start_link() ->
    gen_server:start_link({local, ?MODULE}, []).

init(Args) ->
    {ok, init_state(Args), 0}.

handle_info(timeout, State) ->
    %% TODO: start my task here
    {stop, normal, State}.

Returning 0 for the timeout value in init/1 will cause gen_server to call handle_info/2 with a timeout message before processing any messages sent to the process from external sources. This can be used to start process work immediately after init/1 returns without the risk of receiving unwanted messages.

In e2, this is handled with the task behavior:

start_link() ->
    e2_task:start_link(?MODULE, []).

init(Args) ->
    {ok, init_state(Args)}.

handle_task(State) ->
    %% TODO: start my task here
    {stop, normal, State}.

e2_task_supervisor vs supervisor

Task supervisors in e2 are simple_one_for_one supervisors. This supervisor restart strategy automatically removes children when they’re completed, which is typically the desired behavior when managing concurrently running tasks of a particular type.



This will likely be exposed in the future.