path: root/lib/stdlib/doc/src/gen_statem.xml

<?xml version="1.0" encoding="utf-8" ?>
<!DOCTYPE erlref SYSTEM "erlref.dtd">

<erlref>
  <header>
    <copyright>
      <year>2016</year>
      <holder>Ericsson AB. All Rights Reserved.</holder>
    </copyright>
    <legalnotice>
      Licensed under the Apache License, Version 2.0 (the "License");
      you may not use this file except in compliance with the License.
      You may obtain a copy of the License at

          http://www.apache.org/licenses/LICENSE-2.0

      Unless required by applicable law or agreed to in writing, software
      distributed under the License is distributed on an "AS IS" BASIS,
      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
      See the License for the specific language governing permissions and
      limitations under the License.

    </legalnotice>

    <title>gen_statem</title>
    <prepared></prepared>
    <docno></docno>
    <date></date>
    <rev></rev>
  </header>
  <module>gen_statem</module>
  <modulesummary>Generic State Machine Behaviour</modulesummary>
  <description>
    <p>A behaviour module for implementing a state machine.
      Two callback modes are supported.  One for a finite state
      machine like <seealso marker="gen_fsm">gen_fsm</seealso>
      that require the state to be an atom and use that state as
      the name of the callback function for a particular state,
      and one without restriction on the state that use the same
      callback function for all states.
    </p>
    <p>
      A generic state machine process (<c>gen_statem</c>) implemented
      using this module will have a standard set of interface functions
      and include functionality for tracing and error reporting.
      It will also fit into an OTP supervision tree. Refer to
      <seealso marker="doc/design_principles:gen_server_concepts">
        OTP Design Principles</seealso> for more information.
    </p>
    <p>A <c>gen_statem</c> assumes all specific parts to be located in a
      callback module exporting a pre-defined set of functions.
      The relationship between the behaviour functions and the callback
      functions can be illustrated as follows:</p>
    <pre>
gen_statem module            Callback module
-----------------            ---------------
gen_statem:start
gen_statem:start_link -----> Module:init/1

gen_statem:stop       -----> Module:terminate/2

gen_statem:call
gen_statem:cast
erlang:send
erlang:'!'            -----> Module:StateName/5
                             Module:handle_event/5

-                     -----> Module:terminate/3

-                     -----> Module:code_change/3</pre>
    <p>Events are of different
      <seealso marker="#type-event_type">types</seealso>
      so the callback functions can know the origin of an event
      and how to respond.
    </p>
    <p>If a callback function fails or returns a bad value,
      the <c>gen_statem</c> will terminate.  An exception of class
      <seealso marker="erts:erlang#throw/1"><c>throw</c></seealso>,
      however, is not regarded as an error but as a valid return.
    </p>
    <marker id="state_function" />
    <p>The "<em>state function</em>" for a specific
      <seealso marker="#type-state">state</seealso>
      in a <c>gen_statem</c> is the callback function that is called
      for all events in this state, and is selected depending on
      <seealso marker="#type-callback_mode">callback_mode</seealso>
      that the implementation selects during <c>gen_statem</c> init.
    </p>
    <p>When
      <seealso marker="#type-callback_mode">callback_mode</seealso>
      is <c>state_functions</c> the state has to be an atom and
      is used as the state function name.
      See
      <seealso marker="#Module:StateName/5">
        <c>Module:StateName/5</c>
      </seealso>.
      This gathers all code for a specific state
      in one function and hence dispatches on state first.
    </p>
    <p>When
      <seealso marker="#type-callback_mode">callback_mode</seealso>
      is <c>handle_event_function</c> the state can be any term
      and the state function name is
      <seealso marker="#Module:handle_event/5">
        <c>Module:handle_event/5</c>
      </seealso>.
      This makes it easy to dispatch on state or on event as
      you like but you will have to implement it.
      Also be careful about which events you handle in which
      states so you do not accidentally postpone one event
      forever creating an infinite busy loop.
    </p>
    <p>Any state name or any state value (depending on
      <seealso marker="#type-callback_mode">callback_mode</seealso>)
      is permitted with a small gotcha regarding the state
      <c>undefined</c> that is used as the previous state when
      the first <c>gen_statem</c> state function is called.
      You might need to know about this faked state if you
      inspect the previous state argument in your state functions.
    </p>
    <p>The <c>gen_statem</c> enqueues incoming events in order of arrival
      and presents these to the
      <seealso marker="#state_function">state function</seealso>
      in that order.  The state function can postpone an event
      so it is not retried in the current state.
      After a state change all enqueued events (including postponed)
      are again presented to the state function.
    </p>
    <p>The <c>gen_statem</c> event queue model is sufficient
      to emulate the normal process message queue and selective receive
      with postponing an event corresponding to not matching
      it in a receive statement and changing states corresponding
      to entering a new receive statement.
    </p>
    <p>The
      <seealso marker="#state_function">state function</seealso>
      can insert events using the
      <seealso marker="#type-transition_action">
	<c>transition_action()</c> <c>next_event</c>
      </seealso>
      and such an event is inserted as the next to present
      to the state function.  That is: as if it is
      the oldest incoming event.  There is a dedicated
      <seealso marker="#type-event_type">
        <c>event_type()</c> <c>internal</c>
      </seealso>
      that can be used for such events making it impossible
      to mistake for an external event.
    </p>
    <p>Inserting an event replaces the trick of calling your own
      state handling functions that you often would have to
      resort to in e.g <seealso marker="gen_fsm">gen_fsm</seealso>
      to force processing a faked event before others.
      If you for example in <c>gen_statem</c> postpone an event
      in one state and then call some other state function of yours,
      you have not changed states and hence the postponed event
      will not be retried, which is logical but might be confusing.
    </p>
    <p>A <c>gen_statem</c> handles system messages as documented in
       <seealso marker="sys">sys</seealso>.
       The <seealso marker="sys">sys</seealso> module
       can be used for debugging a <c>gen_statem</c>.
    </p>
    <p>Note that a <c>gen_statem</c> does not trap exit signals
      automatically, this must be explicitly initiated by
      the callback module.
    </p>
    <p>Unless otherwise stated, all functions in this module fail if
      the specified <c>gen_statem</c> does not exist or
      if bad arguments are given.
    </p>
    <p>The <c>gen_statem</c> process can go into hibernation (see
      <seealso marker="erts:erlang#hibernate/3">
	<c>erlang:hibernate/3</c>
      </seealso>) if a
      <seealso marker="#state_function">state function</seealso> or
      <seealso marker="#Module:init/1"><c>Module:init/1</c></seealso>
      specifies <c>'hibernate'</c> in the returned
      <seealso marker="#type-transition_op"><c>Ops</c></seealso>
      list. This might be useful if the server is expected to be idle
      for a long time. However use this feature with care
      since hibernation implies at least two garbage collections
      (when hibernating and shortly after waking up) and that is not
      something you'd want to do between each event on a busy server.
    </p>
  </description>

  <datatypes>
    <datatype>
      <name name="server_name" />
      <desc>
	<p>Name specification to use when starting
	  a<c>gen_statem</c> server.
	  See <seealso marker="#start_link/3">
	    <c>start_link/3</c>
	  </seealso> and
	  <seealso marker="#type-server_ref">
	    <c>server_ref()</c>
	  </seealso> below.
	</p>
      </desc>
    </datatype>
    <datatype>
      <name name="server_ref" />
      <desc>
	<p>Server specification to use when addressing
	  a <c>gen_statem</c> server.
	  See <seealso marker="#call/2">call/2</seealso> and
	  <seealso marker="#type-server_name">
	    <c>server_name()</c>
	  </seealso> above.
	</p>
	<p>It can be:</p>
	<taglist>
	  <tag><c>pid()</c></tag>
	  <tag><c>LocalName</c></tag>
	  <item>The <c>gen_statem</c> is locally registered.</item>
	  <tag><c>Name, Node</c></tag>
	  <item>The <c>gen_statem</c> is locally registered
	    on another node.
	  </item>
	  <tag><c>GlobalName</c></tag>
	  <item>The <c>gen_statem</c> is globally registered
	    in <seealso marker="kernel:global"><c>global</c></seealso>.
	  </item>
	  <tag><c>RegMod, ViaName</c></tag>
	  <item>The <c>gen_statem</c> is registered through
	    an alternative process registry.
	    The registry callback module <c>RegMod</c>
	    should export the functions
	    <c>register_name/2</c>, <c>unregister_name/1</c>,
	    <c>whereis_name/1</c> and <c>send/2</c>,
	    which should behave like the corresponding functions
	    in <seealso marker="kernel:global"><c>global</c></seealso>.
	    Thus, <c>{via,global,GlobalName}</c> is the same as
	    <c>{global,GlobalName}</c>.
	  </item>
	</taglist>
      </desc>
    </datatype>
    <datatype>
      <name name="debug_opt" />
      <desc>
	<p>Debug option that can be used when starting
	  a <c>gen_statem</c> server through for example
	  <seealso marker="#enter_loop/4">enter_loop/4</seealso>.
	</p>
	<p>For every entry in <c><anno>Dbgs</anno></c>
	  the corresponding function in
	  <seealso marker="sys"><c>sys</c></seealso> will be called.
	</p>
      </desc>
    </datatype>
    <datatype>
      <name name="start_opt" />
      <desc>
	<p>Options that can be used when starting
	  a <c>gen_statem</c> server through for example
	  <seealso marker="#start_link/3">start_link/3</seealso>.
	</p>
      </desc>
    </datatype>
    <datatype>
      <name name="start_ret" />
      <desc>
	<p>Return value from the start functions for_example
	  <seealso marker="#start_link/3">start_link/3</seealso>.
	</p>
      </desc>
    </datatype>

    <datatype>
      <name name="client" />
      <desc>
	<p>Client address to use when replying through for example the
	  <seealso marker="#type-transition_op">
	    transition_op() {reply,Client,Reply}
	  </seealso>
	  to a client that has called the <c>gen_statem</c> server using
	  <seealso marker="#call/2">call/2</seealso>.
	</p>
      </desc>
    </datatype>
    <datatype>
      <name name="state" />
      <desc>
	<p>After a state change (<c>NewState =/= State</c>)
	  all postponed events are retried.
	</p>
      </desc>
    </datatype>
    <datatype>
      <name name="state_name" />
      <desc>
	<p>If
	  <seealso marker="#type-callback_mode">
	    callback_mode
	  </seealso> is <c>state_functions</c>, which is the default,
	  the state has to be of this type.
	</p>
      </desc>
    </datatype>
    <datatype>
      <name name="data" />
      <desc>
	<p>A term in which the state machine implementation
	  should store any server data it needs.  The difference between
	  this and the <seealso marker="#type-state">state()</seealso>
	  itself is that a change in this data does not cause
	  postponed events to be retried.  Hence if a change
	  in this data would change the set of events that
	  are handled than that data item should be made
	  a part of the state.
	</p>
      </desc>
    </datatype>
    <datatype>
      <name name="event_type" />
      <desc>
	<p>External events are of 3 different type:
	  <c>{call,<anno>Client</anno>}</c>, <c>cast</c> or <c>info</c>.
	  Calls (synchronous) and casts (asynchronous)
	  originate from the corresponding API functions.
	  For calls the event contain whom to reply to.
	  Type <c>info</c> originates from normal messages sent
	  to the <c>gen_statem</c> process.
	  It is also possible for the state machine
	  implementation to insert events to itself,
	  in particular of types
	  <c>timeout</c> and <c>internal</c>.
	</p>
      </desc>
    </datatype>
    <datatype>
      <name name="event_predicate" />
      <desc>
	<p>A <c>fun()</c> of arity 2 that takes an event
	  and returns a boolean.
	  When used in <c>{remove_event,RemoveEventPredicate}</c> from
	  <seealso marker="#type-transition_op">transition_op()</seealso>.
	  The event for which the predicate returns <c>true</c> will
	  be removed.
	</p>
	<p>
	  The predicate may <em>not</em> use a throw exception
	  to return its result.
	</p>
      </desc>
    </datatype>
    <datatype>
      <name name="init_option" />
      <desc>
	<p>Option that only is valid when initializing
	  the <c>gen_statem</c> that is it can be returned from
	  <seealso marker="#Module:init/1">Module:init/1</seealso>
	  or given to
	  <seealso marker="#enter_loop/5">enter_loop/5,6</seealso>.
	</p>
      </desc>
    </datatype>
    <datatype>
      <name name="callback_mode" />
      <desc>
	<taglist>
	  <tag><c>state_functions</c></tag>
	  <item>The state has to be of type
	    <seealso marker="#type-state_name"><c>state_name()</c></seealso>
	    and one callback function per state that is
	    <seealso marker="#Module:StateName/5">
	      <c>Module:StateName/5</c>
	    </seealso> is used.  This is the default.
	  </item>
	  <tag><c>handle_event_function</c></tag>
	  <item>The state can be any term and the callback function
	    <seealso marker="#Module:handle_event/5">
	      <c>Module:handle_event/5</c>
	    </seealso> is used for all states.
	  </item>
	</taglist>
      </desc>
    </datatype>
    <datatype>
      <name name="transition_op" />
      <desc>
	<p>Either a
	  <seealso marker="#type-transition_option">
	    <c>transition_option()</c>
	  </seealso> of which the last occurence
	  in the containing list takes precedence, or a
	  <seealso marker="#type-transition_action">
	    <c>transition_action()</c>
	  </seealso> performed in the order of the containing list.
	</p>
	<p>These may be returned from the
	  <seealso marker="#state_function">state function</seealso>,
	  from <seealso marker="#Module:init/1">Module:init/1</seealso>
	  or given to
	  <seealso marker="#enter_loop/5">enter_loop/5,6</seealso>.
	</p>
	<p>The processing order for a state change is:</p>
	<list type="ordered">
	  <item>If the option <c>postpone</c> is <c>true</c>
	    the current event is postponed.
	  </item>
	  <item>If the state changes the queue of incoming events
	    is reset to start with the oldest postponed.
	  </item>
	  <item>All actions are processed in order of appearance.
	  </item>
	  <item>The <c>timeout</c> option is processed if present,
	    so a state timer may be started or a timeout zero event
	    may be enqueued as the newest incoming.
	  </item>
	  <item>The (possibly new)
	    <seealso marker="#state_function">state function</seealso>
	    is called with the oldest enqueued event if there is any,
	    otherwise the <c>gen_statem</c> goes into <c>receive</c>
	    or hibernation (if the option <c>hibernate</c> is <c>true</c>)
	    to wait for the next message.  In hibernation the next
	    non-system event awakens the <c>gen_statem</c>, or rather
	    the next incoming message awakens the <c>gen_statem</c>
	    but if it is a system event it goes back into hibernation.
	  </item>
	</list>
      </desc>
    </datatype>
    <datatype>
      <name name="transition_option" />
      <desc>
	<p>If multiple state options of the same kind are present
	  in the containing list these are set in the list order
	  and the last value is kept.
	</p>
	<taglist>
	  <tag><c>postpone</c></tag>
	  <item>If <c><anno>Postpone</anno> =:= true</c>
	    or plain <c>postpone</c> postpone the current event
	    to be retried after a state change.
	    This option is ignored when returned from
	    <seealso marker="#Module:init/1">Module:init/1</seealso>
	    or given to
	    <seealso marker="#enter_loop/5">enter_loop/5,6</seealso>
	    since there is no event to postpone in those cases.
	  </item>
	  <tag><c>hibernate</c></tag>
	  <item>If <c><anno>Hibernate</anno> =:= true</c>
	    or plain <c>hibernate</c> hibernate the <c>gen_statem</c>
	    by calling
	    <seealso marker="proc_lib#hibernate/3">
	      <c>proc_lib:hibernate/3</c>
	    </seealso> before <c>receive</c> to wait for a new event.
	    If there are enqueued events the <c>hibernate</c> action
	    is ignored as if an event just arrived and awakened
	    the <c>gen_statem</c>.
	  </item>
	  <tag><c>timeout</c>
	  </tag>
	  <item>Generate an event of
	    <seealso marker="#type-event_type">type <c>timeout</c></seealso>
	    after <c><anno>Time</anno></c> milliseconds unless some other
	    event is received before that time.  Note that a retried
	    event counts just like a new in this respect.
	    If <c>Time =:= infinity</c> no timer is started.
	    If <c>Time =:= 0</c> the timeout event
	    is immediately enqueued as the newest received.
	    Also note that it is not possible nor needed
	    to cancel this timeout using the
	    <seealso marker="#type-transition_action">
	      <c>transition_action()</c>
	    </seealso> <c>cancel_timer</c>.
	    This timeout is cancelled automatically by any event.
	  </item>
	</taglist>
      </desc>
    </datatype>
    <datatype>
      <name name="transition_action" />
      <desc>
	<p>The state transition actions are executed
	  in the containing list order.  This matters
	  for <c>next_event</c> where the last one in the list
	  will become the next event to present
	  to the state functions.  Regarding the other actions
	  it is only for <c>remove_event</c> with
	  <c><anno>EventPredicate</anno></c>
	  and for <c>reply_action()</c> that the order may matter.
	</p>
	<taglist>
	  <tag><c>reply_action()</c></tag>
	  <item>Reply to a calling client.</item>
	  <tag><c>next_event</c></tag>
	  <item>Insert the given <c><anno>EventType</anno></c>
	    and <c><anno>EventContent</anno></c> the next to process.
	    An event of type
	    <seealso marker="#type-event_type">
	      <c>internal</c>
	    </seealso> should be used when you want to reliably distinguish
	    an event inserted this way from any external event.
	  </item>
	  <tag><c>remove_event</c></tag>
	  <item>Remove the oldest queued event
	    that matches equal to <c><anno>EventType</anno></c>
	    and <c><anno>EventContent</anno></c> or for which
	    <c><anno>EventPredicate</anno></c> returns <c>true</c>.
	  </item>
	  <tag><c>cancel_timer</c></tag>
	  <item>Cancel the timer by calling
	    <seealso marker="erts:erlang#cancel_timer/2">
	      <c>erlang:cancel_timer/2</c>
	    </seealso> with <c><anno>TimerRef</anno></c>,
	    clean the process message queue from any late timeout message,
	    and removes any late timeout message
	    from the <c>gen_statem</c> event queue using
	    <c>{remove_event,<anno>EventPredicate</anno>}</c> above.
	    This is a convenience function that saves quite some
	    lines of code and testing time over doing it from
	    the primitives mentioned above.
	  </item>
	  <tag><c>demonitor</c></tag>
	  <item>Like <c>cancel_timer</c> above but for
	    <seealso marker="erts:erlang#demonitor/2">
	      <c>demonitor/2</c>
	    </seealso> with <c><anno>MonitorRef</anno></c>.
	  </item>
	  <tag><c>unlink</c></tag>
	  <item>Like <c>{cancel_timer,_}</c> above but for
	    <seealso marker="erts:erlang#unlink/1">
	      <c>unlink/1</c>
	    </seealso> with <c><anno>Id</anno></c>.
	  </item>
	</taglist>
      </desc>
    </datatype>
    <datatype>
      <name name="reply_action" />
      <desc>
	<p>Reply to a client that called
	  <seealso marker="#call/2"><c>call/2</c></seealso>.
          <c><anno>Client</anno></c> must be the term from the
	  <seealso marker="#type-event_type">
	    <c>{call,<anno>Client</anno>}</c>
	  </seealso> argument to the
	  <seealso marker="#state_function">state function</seealso>.
	</p>
      </desc>
    </datatype>
    <datatype>
      <name name="state_callback_result" />
      <desc>
	<taglist>
	  <tag><c>stop</c></tag>
	  <item>Send all <c><anno>Replies</anno></c> if given,
	    then terminate the <c>gen_statem</c> by calling
	    <seealso marker="#Module:terminate/3">
	      <c>Module:terminate/3</c>
	    </seealso> with <c>Reason</c> and
	    <c><anno>NewData</anno></c>, if given.
	  </item>
	  <tag><c>next_state</c></tag>
	  <item>The <c>gen_statem</c> will do a state transition to
	    <c><anno>NewState</anno></c>
	    (which may be the same as the current state),
	    set <c><anno>NewData</anno></c>
	    and execute all <c><anno>Ops</anno></c>
	  </item>
	  <tag><c>keep_state</c></tag>
	  <item>The <c>gen_statem</c> will keep the current state, or
	    do a state transition to the current state if you like,
	    set <c><anno>NewData</anno></c>
	    and execute all <c><anno>Ops</anno></c>
	  </item>
	  <tag><c>keep_state_and_data</c></tag>
	  <item>The <c>gen_statem</c> will keep the current state, or
	    do a state transition to the current state if you like,
	    keep the current server data,
	    and execute all <c><anno>Ops</anno></c>
	  </item>
	</taglist>
      </desc>
    </datatype>
  </datatypes>

  <funcs>

    <func>
      <name name="start_link" arity="3" />
      <name name="start_link" arity="4" />
      <fsummary>Create a linked <c>gen_statem</c> process</fsummary>
      <desc>
        <p>Creates a <c>gen_statem</c> process according
	  to OTP design principles
	  (using
          <seealso marker="proc_lib"><c>proc_lib</c></seealso>
	  primitives)
	  that is linked to the calling process.
	  This is essential when the <c>gen_statem</c> shall be part of
	  a supervision tree so it gets linked to its supervisor.
	</p>
        <p>The <c>gen_statem</c> process calls
	  <seealso marker="#Module:init/1"><c>Module:init/1</c></seealso>
	  to initialize the server. To ensure a synchronized start-up
	  procedure, <c>start_link/3,4</c> does not return until
	  <seealso marker="#Module:init/1"><c>Module:init/1</c></seealso>
          has returned.
	</p>
	<p><c><anno>ServerName</anno></c> specifies the
	  <seealso marker="#type-server_name">
	    <c>server_name()</c>
	  </seealso> to register for the <c>gen_statem</c>.
	  If the <c>gen_statem</c> is started with  <c>start_link/3</c>
	  no <c><anno>ServerName</anno></c> is provided and
	  the <c>gen_statem</c> is not registered.
	</p>
        <p><c><anno>Module</anno></c> is the name of the callback module.</p>
        <p><c><anno>Args</anno></c> is an arbitrary term which is passed as
          the argument to
	  <seealso marker="#Module:init/1"><c>Module:init/1</c>
	  </seealso>.
	</p>
        <p>If the option <c>{timeout,Time}</c> is present in
	  <c><anno>Opts</anno></c>, the <c>gen_statem</c>
	  is allowed to spend <c>Time</c> milliseconds initializing
	  or it will be terminated and the start function will return
	  <seealso marker="#type-start_ret">
	    <c>{error,timeout}</c>
	  </seealso>.
	</p>
        <p>If the option
	  <seealso marker="#type-debug_opt"><c>{debug,Dbgs}</c></seealso>
	  is present in <c><anno>Opts</anno></c>, debugging through
	  <seealso marker="sys"><c>sys</c></seealso> is activated.
	</p>
        <p>If the option <c>{spawn_opt,SpawnOpts}</c> is present in
	  <c><anno>Opts</anno></c>, <c>SpawnOpts</c> will be passed
	  as option list to
          <seealso marker="erts:erlang#spawn_opt/2">spawn_opt/2</seealso>
	  which is used to spawn the <c>gen_statem</c> process.
	</p>
        <note>
          <p>Using the spawn option <c>monitor</c> is currently not
            allowed, but will cause this function to fail with reason
            <c>badarg</c>.</p>
        </note>
        <p>If the <c>gen_statem</c> is successfully created
	  and initialized this function returns
	  <seealso marker="#type-start_ret">
	    <c>{ok,Pid}</c>,
	  </seealso> where <c>Pid</c> is the <c>pid()</c>
	  of the <c>gen_statem</c>.
	  If there already exists a process with the specified
	  <c><anno>ServerName</anno></c> this function returns
	  <seealso marker="#type-start_ret">
            <c>{error,{already_started,Pid}}</c>
	  </seealso>, where <c>Pid</c> is the <c>pid()</c> of that process.
	</p>
        <p>If <c>Module:init/1</c> fails with <c>Reason</c>,
          this function returns
	  <seealso marker="#type-start_ret">
	    <c>{error,Reason}</c>
	  </seealso>. If <c>Module:init/1</c> returns
	  <seealso marker="#type-start_ret">
	    <c>{stop,Reason}</c>
	  </seealso>
	  or
	  <seealso marker="#type-start_ret">
            <c>ignore</c>
	  </seealso>, the process is terminated and this function
          returns
	  <seealso marker="#type-start_ret">
	    <c>{error,Reason}</c>
	  </seealso> or
	  <seealso marker="#type-start_ret">
	  <c>ignore</c>
	  </seealso>, respectively.
	</p>
      </desc>
    </func>


    <func>
      <name name="start" arity="3" />
      <name name="start" arity="4" />
      <fsummary>Create a stand-alone <c>gen_statem</c> process</fsummary>
      <desc>
        <p>Creates a stand-alone <c>gen_statem</c> process according to
	  OTP design principles (using
          <seealso marker="proc_lib"><c>proc_lib</c></seealso>
	  primitives).
	  Since it does not get linked to the calling process
	  this start function can not be used by a supervisor
	  to start a child.
	</p>
        <p>See <seealso marker="#start_link/3">start_link/3,4</seealso>
          for a description of arguments and return values.
	</p>
      </desc>
    </func>

    <func>
      <name name="stop" arity="1" />
      <fsummary>Synchronously stop a generic server</fsummary>
      <desc>
	<p>The same as
	  <seealso marker="#stop/3">
	    <c>stop(<anno>ServerRef</anno>, normal, infinity)</c>
	  </seealso>.
	</p>
      </desc>
    </func>
    <func>
      <name name="stop" arity="3" />
      <fsummary>Synchronously stop a generic server</fsummary>
      <desc>
	<p>Orders the <c>gen_statem</c>
	  <seealso marker="#type-server_ref">
	    <c><anno>ServerRef</anno></c>
	  </seealso> to exit with the given <c><anno>Reason</anno></c>
	  and waits for it to terminate.
	  The <c>gen_statem</c> will call
          <seealso marker="#Module:terminate/3">
	    Module:terminate/3
	  </seealso> before exiting.
	</p>
	<p>This function returns <c>ok</c> if the server terminates
	  with the expected reason. Any other reason than <c>normal</c>,
          <c>shutdown</c>, or <c>{shutdown,Term}</c> will cause an
          error report to be issued through
          <seealso marker="kernel:error_logger#format/2">
	    error_logger:format/2
	  </seealso>.
	  The default <c><anno>Reason</anno></c> is <c>normal</c>.
	</p>
	<p><c><anno>Timeout</anno></c> is an integer greater than zero
	  which specifies how many milliseconds to wait for the server to
          terminate, or the atom <c>infinity</c> to wait indefinitely.
	  The default value is <c>infinity</c>.
	  If the server has not terminated within the specified time,
	  a <c>timeout</c> exception is raised.
	</p>
	<p>If the process does not exist, a <c>noproc</c> exception
	  is raised.
	</p>
      </desc>
    </func>

    <func>
      <name name="call" arity="2" />
      <name name="call" arity="3" />
      <fsummary>Make a synchronous call to a <c>gen_statem</c></fsummary>
      <desc>
        <p>Makes a synchronous call to the <c>gen_statem</c>
	  <seealso marker="#type-server_ref">
	    <c><anno>ServerRef</anno></c>
	  </seealso> by sending a request
	  and waiting until its reply arrives.
	  The <c>gen_statem</c> will call the
	  <seealso marker="#state_function">state function</seealso> with
	  <seealso marker="#type-event_type"><c>event_type()</c></seealso>
	  <c>{call,Client}</c> and event content
	  <c><anno>Request</anno></c>.
	</p>
	<p>A <c><anno>Reply</anno></c> is generated when a
	  <seealso marker="#state_function">state function</seealso>
	  returns with
	  <c>{reply,Client,<anno>Reply</anno>}</c> as one
	  <seealso marker="#type-transition_op">
	    <c>transition_op()</c>
	  </seealso>,
	  and that <c><anno>Reply</anno></c> becomes the return value
	  of this function.
	</p>
        <p><c><anno>Timeout</anno></c> is an integer greater than zero
	  which specifies how many milliseconds to wait for a reply,
	  or the atom <c>infinity</c> to wait indefinitely,
	  which is the default.  If no reply is received within
	  the specified time, the function call fails.
	  <note>
	    <p>To avoid getting a late reply in the caller's
	      inbox this function spawns a proxy process that
	      does the call.  A late reply gets delivered to the
	      dead proxy process hence gets discarded.  This is
	      less efficient than using
	      <c><anno>Timeout</anno> =:= infinity</c>.
	    </p>
	  </note>
	</p>
        <p>The call may fail for example if the <c>gen_statem</c> dies
	  before or during this function call.
	</p>
      </desc>
    </func>

    <func>
      <name name="cast" arity="2" />
      <fsummary>Send an asynchronous event to a <c>gen_statem</c></fsummary>
      <desc>
	<p>Sends an asynchronous event to the <c>gen_statem</c>
	  <seealso marker="#type-server_ref">
	    <c><anno>ServerRef</anno></c>
	  </seealso> and returns <c>ok</c> immediately,
	  ignoring if the destination node or <c>gen_statem</c>
	  does not exist.
	  The <c>gen_statem</c> will call the
	  <seealso marker="#state_function">state function</seealso> with
	  <seealso marker="#type-event_type"><c>event_type()</c></seealso>
	  <c>cast</c> and event content
	  <c><anno>Msg</anno></c>.
	</p>
      </desc>
    </func>

    <func>
      <name name="reply" arity="1" />
      <name name="reply" arity="2" />
      <fsummary>Send a reply to a client</fsummary>
      <desc>
        <p>This function can be used by a <c>gen_statem</c>
	  to explicitly send a reply to a client that called
	  <seealso marker="#call/2"><c>call/2</c></seealso>
          when the reply cannot be defined in
          the return value of the
	  <seealso marker="#state_function">state function</seealso>.
	</p>
        <p><c><anno>Client</anno></c> must be the term from the
	  <seealso marker="#type-event_type">
	    <c>{call,<anno>Client</anno>}</c>
	  </seealso> argument to the
	  <seealso marker="#state_function">state function</seealso>.
	  <c><anno>Client</anno></c> and <c><anno>Reply</anno></c>
	  can also be specified using a
	  <seealso marker="#type-reply_action">
	    <c>reply_action()</c>
	  </seealso> and multiple replies with a list of them.
	</p>
	<note>
	  <p>A reply sent with this function will not be visible
	    in <seealso marker="sys">sys</seealso> debug output.
	  </p>
	</note>
      </desc>
    </func>

    <func>
      <name name="enter_loop" arity="4" />
      <fsummary>Enter the <c>gen_statem</c> receive loop</fsummary>
      <desc>
	<p>The same as
	  <seealso marker="#enter_loop/6"><c>enter_loop/6</c></seealso>
	  except that no
	  <seealso marker="#type-server_name">
	    <c>server_name()</c>
	  </seealso> must have been registered.
	</p>
      </desc>
    </func>
    <func>
      <name name="enter_loop" arity="5" />
      <fsummary>Enter the <c>gen_statem</c> receive loop</fsummary>
      <desc>
	<p>If <c><anno>Server_or_Ops</anno></c> is a <c>list()</c>
	  the same as
	  <seealso marker="#enter_loop/6"><c>enter_loop/6</c></seealso>
	  except that no
	  <seealso marker="#type-server_name">
	    <c>server_name()</c>
	  </seealso> must have been registered and
	  <c>Ops = <anno>Server_or_Ops</anno></c>.
	</p>
	<p>Otherwise the same as
	  <seealso marker="#enter_loop/6"><c>enter_loop/6</c></seealso>
	  with
	  <c>Server = <anno>Server_or_Ops</anno></c> and
	  <c>Ops = []</c>.
	</p>
      </desc>
    </func>
    <func>
      <name name="enter_loop" arity="6" />
      <fsummary>Enter the <c>gen_statem</c> receive loop</fsummary>
      <desc>
        <p>Makes an the calling process become a <c>gen_statem</c>.
	  Does not return, instead the calling process will enter
	  the <c>gen_statem</c> receive loop and become
	  a <c>gen_statem</c> server.
	  The process <em>must</em> have been started
	  using one of the start functions in
          <seealso marker="proc_lib"><c>proc_lib</c></seealso>.
	  The user is responsible for any initialization of the process,
	  including registering a name for it.
	</p>
        <p>This function is useful when a more complex initialization
          procedure is needed than
	  the <c>gen_statem</c> behaviour provides.
	</p>
        <p><c><anno>Module</anno></c>, <c><anno>Opts</anno></c> and
	  <c><anno>Server</anno></c> have the same meanings
	  as when calling
          <seealso marker="#start_link/3">
	    <c>gen_statem</c>:start[_link]/3,4
	  </seealso>.
          However, the
	  <seealso marker="#type-server_name">
	    <c>server_name()</c>
	  </seealso> name must have been registered accordingly
	  <em>before</em> this function is called.</p>
        <p><c><anno>State</anno></c>, <c><anno>Data</anno></c>
	  and <c><anno>Ops</anno></c>
	  have the same meanings as in the return value of
          <seealso marker="#Module:init/1">Module:init/1</seealso>.
          Also, the callback module <c><anno>Module</anno></c>
	  does not need to export an <c>init/1</c> function.
	</p>
        <p>Failure: If the calling process was not started by a
          <seealso marker="proc_lib"><c>proc_lib</c></seealso>
	  start function, or if it is not registered
          according to
	  <seealso marker="#type-server_name">
	    <c>server_name()</c>
	  </seealso>.
	</p>
      </desc>
    </func>

  </funcs>



  <section>
    <title>CALLBACK FUNCTIONS</title>
    <p>The following functions should be exported from a
      <c><c>gen_statem</c></c> callback module.
    </p>
  </section>
  <funcs>

    <func>
      <name>Module:init(Args) -> Result</name>
      <fsummary>Initialize process and internal state</fsummary>
      <type>
        <v>Args = term()</v>
        <v>Result = {ok,State,Data}</v>
	<v>&nbsp;| {ok,State,Data,Ops}</v>
        <v>&nbsp;| {stop,Reason} | ignore</v>
	<v>State = <seealso marker="#type-state">state()</seealso></v>
	<v>Data = <seealso marker="#type-data">data()</seealso>
	</v>
	<v>Ops =
	  [<seealso marker="#type-transition_op">transition_op()</seealso>
	   | <seealso marker="#type-init_option">init_option()</seealso>]
	</v>
	<v>Reason = term()</v>
      </type>
      <desc>
	<marker id="Module:init-1" />
        <p>Whenever a <c>gen_statem</c> is started using
          <seealso marker="#start_link/3">start_link/3,4</seealso>
	  or
          <seealso marker="#start/3">start/3,4</seealso>,
          this function is called by the new process to initialize
	  the implementation state and server data.
	</p>
        <p><c>Args</c> is the <c>Args</c> argument provided to the start
          function.</p>
        <p>If the initialization is successful, the function should
          return <c>{ok,State,Data}</c> or
	  <c>{ok,State,Data,Ops}</c>.
	  <c>State</c> is the <seealso marker="#type-state">state</seealso>
	  of the <c>gen_statem</c>.
	</p>
	<p>The <seealso marker="#type-transition_op"><c>Ops</c></seealso>
	  are executed when entering the first
	  <seealso marker="#type-state">state</seealso> just as for a
	  <seealso marker="#state_function">state function</seealso>.
	</p>
	<p>This function allows an option to select the callback mode
	  of the <c>gen_statem</c>.  See
	  <seealso marker="#type-init_option">init_option</seealso>.
	  This option is not allowed from the state function(s).
	</p>
        <p>If something goes wrong during the initialization
          the function should return <c>{stop,Reason}</c>
	  or <c>ignore</c>. See
          <seealso marker="#start_link/3">start_link/3,4</seealso>.
	</p>
	<p>This function may use
	  <seealso marker="erts:erlang#throw/1"><c>throw</c></seealso>,
	  to return its value.
	</p>
      </desc>
    </func>

    <func>
      <name>Module:StateName(EventType, EventContent,
        PrevStateName, StateName, Data) -> Result
      </name>
      <name>Module:handle_event(EventType, EventContent,
        PrevState, State, Data) -> Result
      </name>
      <fsummary>Handle an event</fsummary>
      <type>
	<v>EventType =
	  <seealso marker="#type-event_type">event_type()</seealso>
	</v>
	<v>EventContent = term()</v>
	<v>PrevStateName =
	  <seealso marker="#type-state_name">state_name()</seealso>
	</v>
	<v>StateName =
	  <seealso marker="#type-state_name">state_name()</seealso>
	</v>
	<v>PrevState = State =
	  <seealso marker="#type-state">state()</seealso>
	</v>
	<v>Data = NewData =
	  <seealso marker="#type-data">data()</seealso>
	</v>
	<v>Result =
	  <seealso marker="#type-state_callback_result">
	    state_callback_result()
	  </seealso>
	</v>
      </type>
      <desc>
        <p>Whenever a <c>gen_statem</c> receives an event from
          <seealso marker="#call/2">call/2</seealso>,
          <seealso marker="#cast/2">cast/2</seealso> or
	  as a normal process message this function is called.
	  If
	  <seealso marker="#type-callback_mode">callback_mode</seealso>
	  is <c>state_functions</c> then <c>Module:StateName/5</c> is called,
	  and if it is <c>handle_event_function</c>
	  then <c>Module:handle_event/5</c> is called.
	</p>
	<p>If <c>EventType</c> is
	  <seealso marker="#type-event_type"><c>{call,Client}</c></seealso>
	  the client is waiting for a reply.  The reply can be sent
	  from this or from any other
	  <seealso marker="#state_function">state function</seealso>
	  by returning with <c>{reply,Client,Reply}</c> in
	  <seealso marker="#type-transition_op">Ops</seealso>, in
	  <seealso marker="#type-reply_action">Replies</seealso>
	  or by calling
	  <seealso marker="#reply/2">
	    <c>reply(Client, Reply)</c>
	  </seealso>.
	</p>
        <p><c>StateName</c> is useful in some odd cases for example
	  if you call a common event handling function from your state
	  function then you might want to pass <c>StateName</c>.
	</p>
	<p><c>PrevStateName</c> and <c>PrevState</c> are useful
	  in some odd cases for example when you want to do something
	  only at the first event in a state.
	  Note that when <c>gen_statem</c> enters its first state
	  this is set to <c>undefined</c>.
	</p>
	<p>If this function returns with a new state that
	  does not match equal (<c>=/=</c>) to the current state
	  all postponed events will be retried in the new state.
	</p>
	<p>See
	  <seealso marker="#type-transition_op">transition_op()</seealso>
	  for options that can be set and actions that can be done
	  by <c>gen_statem</c> after returning from this function.
	</p>
	<p>These functions may use
	  <seealso marker="erts:erlang#throw/1"><c>throw</c></seealso>,
	  to return its value.
	</p>
      </desc>
    </func>

    <func>
      <name>Module:terminate(Reason, State, Data) -> Ignored</name>
      <fsummary>Clean up before termination</fsummary>
      <type>
        <v>Reason = normal | shutdown | {shutdown,term()} | term()</v>
	<v>State = <seealso marker="#type-state">state()</seealso></v>
	<v>Data = <seealso marker="#type-data">data()</seealso>
	</v>
	<v>Ignored = term()</v>
      </type>
      <desc>
        <p>This function is called by a <c>gen_statem</c>
	  when it is about to terminate. It should be the opposite of
	  <seealso marker="#Module:init/1"><c>Module:init/1</c></seealso>
          and do any necessary cleaning up. When it returns,
          the <c>gen_statem</c> terminates with <c>Reason</c>. The return
          value is ignored.</p>
        <p><c>Reason</c> is a term denoting the stop reason and
          <seealso marker="#type-state">State</seealso>
	  is the internal state of the <c>gen_statem</c>.
	</p>
        <p><c>Reason</c> depends on why the <c>gen_statem</c>
	  is terminating.
          If it is because another callback function has returned a
          stop tuple <c>{stop,Reason}</c> in
	  <seealso marker="#type-transition_op">Ops</seealso>,
	  <c>Reason</c> will have the value specified in that tuple.
	  If it is due to a failure, <c>Reason</c> is the error reason.
	</p>
        <p>If the <c>gen_statem</c> is part of a supervision tree and is
          ordered by its supervisor to terminate, this function will be
          called with <c>Reason = shutdown</c> if the following
          conditions apply:</p>
        <list type="bulleted">
          <item>the <c>gen_statem</c> has been set
	    to trap exit signals, and
	  </item>
          <item>the shutdown strategy as defined in the supervisor's
            child specification is an integer timeout value, not
            <c>brutal_kill</c>.
	  </item>
        </list>
        <p>Even if the <c>gen_statem</c> is <em>not</em>
	  part of a supervision tree, this function will be called
	  if it receives an <c>'EXIT'</c> message from its parent.
	  <c>Reason</c> will be the same as
	  in the <c>'EXIT'</c> message.
	</p>
        <p>Otherwise, the <c>gen_statem</c> will be immediately terminated.
	</p>
        <p>Note that for any other reason than <c>normal</c>,
          <c>shutdown</c>, or <c>{shutdown,Term}</c>
	  the <c>gen_statem</c> is assumed to terminate due to an error
	  and an error report is issued using
          <seealso marker="kernel:error_logger#format/2">
	    error_logger:format/2
	  </seealso>.
	</p>
	<p>This function may use
	  <seealso marker="erts:erlang#throw/1"><c>throw</c></seealso>,
	  to return its value.
	</p>
      </desc>
    </func>

    <func>
      <name>Module:code_change(OldVsn, OldState, OldData, Extra) ->
        Result
      </name>
      <fsummary>Update the internal state during upgrade/downgrade</fsummary>
      <type>
        <v>OldVsn = Vsn | {down,Vsn}</v>
        <v>&nbsp;&nbsp;Vsn = term()</v>
        <v>OldState = NewState = term()</v>
        <v>Extra = term()</v>
	<v>Result = {ok,{NewState,NewData}} | Reason</v>
	<v>OldState = NewState =
	  <seealso marker="#type-state">state()</seealso>
	</v>
	<v>OldData = NewData =
	  <seealso marker="#type-data">data()</seealso>
	</v>
	<v>Reason = term()</v>
      </type>
      <desc>
        <p>This function is called by a <c>gen_statem</c> when it should
          update its internal state during a release upgrade/downgrade,
          i.e. when the instruction <c>{update,Module,Change,...}</c>
          where <c>Change={advanced,Extra}</c> is given in
          the <c>appup</c> file. See
          <seealso marker="doc/design_principles:release_handling#instr">
	    OTP Design Principles
	  </seealso>
          for more information.
	</p>
        <p>In the case of an upgrade, <c>OldVsn</c> is <c>Vsn</c>, and
          in the case of a downgrade, <c>OldVsn</c> is
          <c>{down,Vsn}</c>. <c>Vsn</c> is defined by the <c>vsn</c>
          attribute(s) of the old version of the callback module
          <c>Module</c>. If no such attribute is defined, the version
          is the checksum of the BEAM file.
	</p>
        <p><c>OldState</c> and <c>OldData</c> is the internal state
	  of the <c>gen_statem</c>.
	</p>
        <p><c>Extra</c> is passed as-is from the <c>{advanced,Extra}</c>
          part of the update instruction.
	</p>
        <p>If successful, the function shall return the updated
          internal state in an
	  <c>{ok,{NewState,NewData}}</c> tuple.
	</p>
	<p>If the function returns <c>Reason</c>, the ongoing
	  upgrade will fail and roll back to the old release.</p>
      </desc>
	<p>This function may use
	  <seealso marker="erts:erlang#throw/1"><c>throw</c></seealso>,
	  to return its value.
	</p>
    </func>

    <func>
      <name>Module:format_status(Opt, [PDict,State,Data]) ->
        Status
      </name>
      <fsummary>Optional function for providing a term describing the
        current <c>gen_statem</c> status</fsummary>
      <type>
        <v>Opt = normal | terminate</v>
        <v>PDict = [{Key, Value}]</v>
	<v>State =
	  <seealso marker="#type-state">state()</seealso>
	</v>
	<v>Data =
	  <seealso marker="#type-data">data()</seealso>
	</v>
	<v>Key = term()</v>
	<v>Value = term()</v>
        <v>Status = term()</v>
      </type>
      <desc>
        <note>
          <p>This callback is optional, so callback modules need not
            export it. The <c>gen_statem</c> module provides a default
            implementation of this function that returns the callback
            module state and data.  The default function will also
	    be used if this callback fails.
	  </p>
        </note>
        <p>This function is called by a <c>gen_statem</c> process when:</p>
        <list type="bulleted">
          <item>One of
            <seealso marker="sys#get_status/1">
	      <c>sys:get_status/1,2</c>
	    </seealso>
            is invoked to get the <c>gen_statem</c> status. <c>Opt</c> is set
            to the atom <c>normal</c> for this case.
	  </item>
          <item>The <c>gen_statem</c> terminates abnormally and logs an error.
	    <c>Opt</c> is set to the atom <c>terminate</c> for this case.
	  </item>
        </list>
        <p>This function is useful for customising the form and
          appearance of the <c>gen_statem</c> status for these cases. A
          callback module wishing to customise the
          <seealso marker="sys#get_status/1">
	    <c>sys:get_status/1,2</c>
	  </seealso> return value as well as how
          its status appears in termination error logs exports an
          instance of <c>format_status/2</c> that returns a term
          describing the current status of the <c>gen_statem</c>.
	</p>
        <p><c>PDict</c> is the current value of the <c>gen_statem</c>'s
          process dictionary.
	</p>
        <p><seealso marker="#type-state"><c>State</c></seealso>
	  is the internal state of the <c>gen_statem</c>.
	</p>
        <p><seealso marker="#type-data"><c>Data</c></seealso>
	  is the internal server data of the <c>gen_statem</c>.
	</p>
        <p>The function should return <c>Status</c>, a term that
          customises the details of the current state and status of
          the <c>gen_statem</c>. There are no restrictions on the
          form <c>Status</c> can take, but for the
          <seealso marker="sys#get_status/1">
	    <c>sys:get_status/1,2</c>
	    </seealso> case (when <c>Opt</c>
          is <c>normal</c>), the recommended form for
          the <c>Status</c> value is <c>[{data, [{"State",
          Term}]}]</c> where <c>Term</c> provides relevant details of
          the <c>gen_statem</c> state. Following this recommendation isn't
          required, but doing so will make the callback module status
          consistent with the rest of the
          <seealso marker="sys#get_status/1">
	    <c>sys:get_status/1,2</c>
	  </seealso> return value.
	</p>
        <p>One use for this function is to return compact alternative
          state representations to avoid having large state terms
          printed in logfiles.
	</p>
	<p>This function may use
	  <seealso marker="erts:erlang#throw/1"><c>throw</c></seealso>,
	  to return its value.
	</p>
      </desc>
    </func>

  </funcs>

  <section>
    <title>SEE ALSO</title>
    <p><seealso marker="gen_event">gen_event</seealso>,
      <seealso marker="gen_fsm">gen_fsm</seealso>,
      <seealso marker="gen_server">gen_server</seealso>,
      <seealso marker="supervisor">supervisor</seealso>,
      <seealso marker="proc_lib">proc_lib</seealso>,
      <seealso marker="sys">sys</seealso></p>
  </section>
</erlref>