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, primarily
      a finite state machine, but an infinite state space is possible.
      A generic state machine process (gen_statem) 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 gen_statem 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:State/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 gen_statem 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 gen_statem is the callback function that is called
      for all events in this state.
      An event can can be postponed causing it to be retried
      after the state has changed, or more precisely;
      after a state change all postponed events are retried
      in the new state.
    </p>
    <p>The state machine
      <seealso marker="#type-state"><c>State</c></seealso>
      is normally an atom in which case the
      <seealso marker="#state_function">state function</seealso>
      that will be called is
      <seealso marker="#Module:State/5"><c>Module:State/5</c></seealso>.
      For a
      <seealso marker="#type-state"><c>State</c></seealso>
      that is <em>not</em> an atom the
      <seealso marker="#state_function">state function</seealso>
      <seealso marker="#Module:handle_event/5">
	<c>Module:handle_event/5</c>
      </seealso> will be called.
      If you use <c>handle_event</c> as a
      <seealso marker="#type-state">state</seealso> and later
      decides to use non-atom states you will then have to
      rewrite your code to stop using that state.
    </p>
    <p>When the using an atom-only
      <seealso marker="#type-state">State</seealso>
      it becomes fairly obvious in the implementation code
      which events are handled in which state
      since there are different callback functions for different states.
    </p>
    <p>
      When using a non-atom <seealso marker="#type-state">State</seealso>
      all events are handled in the callback function
      <seealso marker="#Module:handle_event/5">
	<c>Module:handle_event/5</c>
      </seealso>
      so it may require more coding discipline to ensure what events
      are handled in which state.  Therefore it might be a wee bit
      easier to accidentally postpone an event in two or more states
      and not handling it in any of them causing a tight infinite
      loop when the event bounces to be retried between the states.
    </p>
    <p>A gen_statem handles system messages as documented in
       <seealso marker="sys">sys</seealso>.
       The <seealso marker="sys">sys</seealso> module
       can be used for debugging a gen_statem.
    </p>
    <p>Note that a gen_statem 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 gen_statem does not exist or if bad arguments are given.
    </p>
    <p>The gen_statem 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-state_op"><c>StateOps</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 gen_statem 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 gen_statem 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>
	<list type="bulleted">
          <item>the <c>pid()</c>,</item>
          <item><c><anno>Name</anno></c>,
	    if the gen_statem is locally registered,
	  </item>
          <item><c>{<anno>Name</anno>,<anno>Node</anno>}</c>,
	    if the gen_statem is locally registered at another node, or
	  </item>
          <item><c>{global,<anno>GlobalName</anno>}</c>,
	    if the gen_statem is globally registered.
	  </item>
          <item><c>{via,<anno>RegMod</anno>,<anno>ViaName</anno>}</c>,
	    if the gen_statem 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>
        </list>
      </desc>
    </datatype>
    <datatype>
      <name name="debug_opt" />
      <desc>
	<p>Debug option that can be used when starting
	  a gen_statem 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 gen_statem 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-state_op">state_op()</seealso>
	  <c>{reply,Client,Reply}</c> to a client
	  that has called the gen_statem server using
	  <seealso marker="#call/2">call/2</seealso>.
	</p>
      </desc>
    </datatype>
    <datatype>
      <name name="state" />
      <desc>
	<p>If the gen_statem <c>State</c> is an <c>atom()</c>, the
	  <seealso marker="#state_function">state function</seealso> is
	  <seealso marker="#Module:State/5">Module:State/5</seealso>.
	  If it is any other <c>term()</c> the
	  <seealso marker="#state_function">state function</seealso> is
	  <seealso marker="#Module:handle_event/5">
	    Module:handle_event/5
	  </seealso>.  After a state change (<c>NewState =/= State</c>)
	  all postponed events are retried.
	</p>
      </desc>
    </datatype>
    <datatype>
      <name name="state_data" />
      <desc>
	<p>A <c>term()</c> in which the state machine implementation
	  should store any state data it needs.  The difference between
	  this data and the
	  <seealso marker="#type-state">state()</seealso>
	  itself is that a change in this data does not cause
	  postponed events to be retried.
	</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 gen_statem 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-state_op">state_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="state_op" />
      <desc>
	<p>Either a
	  <seealso marker="#type-state_option">
	    <c>state_option()</c>
	  </seealso> of which the first occurence
	  in the containing list takes precedence, or a
	  <seealso marker="#type-state_operation">
	    <c>state_operation()</c>
	  </seealso> that are performed in order of
	  the containing list.
	</p>
	<p>These may be returned from the
	  <seealso marker="#state_function">state function</seealso>
	  or from <seealso marker="#Module:init/1">Module:init/1</seealso>.
	</p>
	<p>The gen_statem enqueues postponed events and
	  not yet processed events in order of arrival, except for
	  an event that a callback function inserts with
	  <c>{insert_event,EventType,EventContent}</c> that is inserted
	  as the next event to process.
	</p>
	<p>When the state machine changes states all enqueued events
	  becomes not yet processed to be processed before the old
	  not yet processed events.  In other words; the order of arrival
	  is retained.
	</p>
	<p>The processing order is:</p>
	<list type="ordered">
	  <item>If the option <c>retry</c> is <c>true</c>
	    the current event is enqueued as postponed to be retried.
	  </item>
	  <item>If the state changes all postponed events
	    are transferred to not yet processed to be processed
	    before other not yet processed events.
	  </item>
	  <item>All operations 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 inserted as if just received.
	  </item>
	  <item>The (possibly new)
	    <seealso marker="#state_function">state function</seealso>
	    is called with the next not yet processed event
	    if there is any, otherwise the gen_statem 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 gen_statem.
	  </item>
	</list>
      </desc>
    </datatype>
    <datatype>
      <name name="state_option" />
      <desc>
	<taglist>
	  <tag><c>retry</c></tag>
	  <tag><c>{retry,<anno>Retry</anno>}</c></tag>
	  <item>If <c><anno>Retry</anno> =:= true</c>
	    or plain <c>retry</c> postpone the current event
	    to be retried after a state change.
	  </item>
	  <tag><c>hibernate</c></tag>
	  <tag><c>{hibernate,<anno>Hibernate</anno>}</c></tag>
	  <item>If <c><anno>Hibernate</anno> =:= true</c>
	    or plain <c>hibernate</c> hibernate the gen_statem 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 not yet processed events the
	    <c>hibernate</c> operation is ignored as if an event
	    just arrived and awakened the gen_statem.
	  </item>
	  <tag>
	    <c>{timeout,<anno>Time</anno>,<anno>Msg</anno>}</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> or <c>Time =:= 0</c>
	    no timer is started but for zero time the timeout
	    event is enqued as just received after all
	    other already received events.
	    Also note that it is not possible
	    to cancel this timeout using the
	    <seealso marker="#type-state_operation">
	      <c>state_operation()</c>
	    </seealso> <c>cancel_timer</c>.
	    This timeout is cancelled automatically by any event.
	  </item>
	</taglist>
      </desc>
    </datatype>
    <datatype>
      <name name="state_operation" />
      <desc>
	<taglist>
	  <tag>
	   <c>{reply,<anno>Client</anno>,<anno>Reply</anno>}</c>
	  </tag>
	  <item>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>.
	  </item>
	  <tag><c>{stop,<anno>Reason</anno>}</c></tag>
	  <item>The gen_statem will call
	    <seealso marker="#Module:terminate/3">
	      <c>Module:terminate/3</c>
	    </seealso> with <c><anno>Reason</anno></c> and terminate.
	  </item>
	  <tag>
	    <c>
	      {insert_event,<anno>EventType</anno>,<anno>EventContent</anno>}
	    </c>
	  </tag>
	  <item>Insert the given event as the next to process
	    before any other not yet processed events.
	    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,<anno>EventType</anno>,<anno>EventContent</anno>}
	    </c>
	  </tag>
	  <item>Remove the oldest queued event
	    that matches equal to the given event.
	  </item>
	  <tag>
	    <c>
	      {remove_event,<anno>EventPredicate</anno>}
	    </c>
	  </tag>
	  <item>Remove the oldest queued event for which
	  the <c><anno>EventPredicate</anno></c> returns <c>true</c>.
	  </item>
	  <tag><c>{cancel_timer,<anno>TimerRef</anno>}</c></tag>
	  <item>Uses <c><anno>TimerRef</anno></c> when calling
	    <seealso marker="erts:erlang#cancel_timer/2">
	      <c>erlang:cancel_timer/2</c>
	    </seealso> to cancel a timer, cleans the gen_statem's
	    message queue from any late timeout message from
	    the timer, and removes any late timeout message
	    from the queued events 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,<anno>MonitorRef</anno>}</c></tag>
	  <item>Like <c>{cancel_timer,_}</c> above but for
	    <seealso marker="erts:erlang#demonitor/2">
	      <c>demonitor/2</c>
	    </seealso>.
	  </item>
	  <tag><c>{unlink,<anno>Id</anno>}</c></tag>
	  <item>Like <c>{cancel_timer,_}</c> above but for
	    <seealso marker="erts:erlang#unlink/1">
	      <c>unlink/1</c>
	    </seealso>.
	  </item>
	</taglist>
      </desc>
    </datatype>
  </datatypes>

  <funcs>

    <func>
      <name name="start_link" arity="3" />
      <name name="start_link" arity="4" />
      <fsummary>Create a linked gen_statem process</fsummary>
      <desc>
        <p>Creates a gen_statem 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 gen_statem shall be part of
	  a supervision tree so it gets linked to its supervisor.
	</p>
        <p>The gen_statem 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 gen_statem.
	  If the gen_statem is started with  <c>start_link/3</c>
	  no <c><anno>ServerName</anno></c> is provided and
	  the gen_statem 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>Options</anno></c>, the gen_statem 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>Options</anno></c>, debugging through
	  <seealso marker="sys"><c>sys</c></seealso> is activated.
	</p>
        <p>If the option <c>{spawn_opt,SOpts}</c> is present in
	  <c><anno>Options</anno></c>, <c>SOpts</c> will be passed
	  as option list to the <c>spawn_opt</c> BIF
	  which is used to
          <seealso marker="erts:erlang#spawn_opt/2">spawn</seealso>
	  the gen_statem.
	</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 gen_statem 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 gen_statem.
	  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 gen_statem process</fsummary>
      <desc>
        <p>Creates a stand-alone gen_statem 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 gen_statem
	  <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 gen_statem 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 gen_statem</fsummary>
      <desc>
        <p>Makes a synchronous call to the gen_statem
	  <seealso marker="#type-server_ref">
	    <c><anno>ServerRef</anno></c>
	  </seealso> by sending a request
	  and waiting until its reply arrives.
	  The gen_statem 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-state_op"><c>state_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 gen_statem dies
	  before or during this function call.
	</p>
      </desc>
    </func>

    <func>
      <name name="cast" arity="2" />
      <fsummary>Send an asynchronous event to a gen_statem</fsummary>
      <desc>
	<p>Sends an asynchronous event to the gen_statem
	  <seealso marker="#type-server_ref">
	    <c><anno>ServerRef</anno></c>
	  </seealso> and returns <c>ok</c> immediately,
	  ignoring if the destination node or gen_statem does not exist.
	  The gen_statem 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="2" />
      <fsummary>Send a reply to a client</fsummary>
      <desc>
        <p>This function can be used by a gen_statem 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>.
	</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 gen_statem 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 gen_statem receive loop</fsummary>
      <desc>
	<p>If <c><anno>Server_or_StateOps</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>StateOps = <anno>Server_or_StateOps</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_StateOps</anno></c> and
	  <c>StateOps = []</c>.
	</p>
      </desc>
    </func>
    <func>
      <name name="enter_loop" arity="6" />
      <fsummary>Enter the gen_statem receive loop</fsummary>
      <desc>
        <p>Makes an the calling process become a gen_statem. Does not return,
          instead the calling process will enter the gen_statem receive
          loop and become a gen_statem 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 gen_statem behaviour provides.
	</p>
        <p><c><anno>Module</anno></c>, <c><anno>Options</anno></c> and
	  <c><anno>Server</anno></c> have the same meanings
	  as when calling
          <seealso marker="#start_link/3">gen_statem: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> and <c><anno>StateData</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>gen_statem</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,StateData}</v>
	<v>&nbsp;| {ok,State,StateData,StateOps}</v>
        <v>&nbsp;| {stop,Reason} | ignore</v>
	<v>State = <seealso marker="#type-state">state()</seealso></v>
	<v>StateData = <seealso marker="#type-state_data">state_data()</seealso></v>
	<v>StateOps = [<seealso marker="#type-state_op">state_op()</seealso>]</v>
	<v>Reason = term()</v>
      </type>
      <desc>
	<marker id="Module:init-1" />
        <p>Whenever a gen_statem is started using
          <seealso marker="#start_link/3">gen_statem:start_link/3,4</seealso>
	  or
          <seealso marker="#start/3">gen_statem:start/3,4</seealso>,
          this function is called by the new process to initialize
	  the implementation loop 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,StateData}</c> or
	  <c>{ok,State,StateData,StateOps}</c>.
	  <c>State</c> is the <seealso marker="#type-state">state</seealso>
	  of the gen_statem.
	</p>
	<p>The <seealso marker="#type-state_op"><c>StateOps</c></seealso>
	  are executed before entering the first
	  <seealso marker="#type-state">state</seealso> just as for a
	  <seealso marker="#state_function">state function</seealso>.
	</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">gen_statem:start_link/3,4</seealso>.
	</p>
      </desc>
    </func>

    <func>
      <name>Module:handle_event(EventType, EventContent,
        PrevState, State, StateData) -> Result
      </name>
      <name>Module:State(EventType, EventContent,
        PrevState, State, StateData) -> Result
      </name>
      <fsummary>Handle an event</fsummary>
      <type>
	<v>EventType =
	  <seealso marker="#type-event_type">event_type()</seealso>
	</v>
	<v>EventContent = term()</v>
	<v>Result = {NewState,NewStateData,StateOps}</v>
	<v>&nbsp;&nbsp;| {NewState,NewStateData}</v>
	<d>&nbsp;&nbsp;The same as <c>{NewState,NewStateData,[]}</c></d>
	<v>&nbsp;&nbsp;| {NewStateData}</v>
	<d>&nbsp;&nbsp;The same as <c>{State,NewStateData,[retry]}</c></d>
	<v>&nbsp;&nbsp;| {}</v>
	<d>&nbsp;&nbsp;The same as <c>{State,StateData,[]}</c></d>
	<v>&nbsp;&nbsp;| StateOps</v>
	<d>&nbsp;&nbsp;The same as <c>{State,StateData,StateOps}</c></d>

	<v>PrevState = State = NewState =
	  <seealso marker="#type-state">state()</seealso>
	</v>
	<v>StateData = NewStateData =
	  <seealso marker="#type-state_data">state_data()</seealso>
	</v>
	<v>StateOps =
	  [<seealso marker="#type-state_op">state_op()</seealso>]
	</v>
      </type>
      <desc>
        <p>Whenever a gen_statem receives an event from
          <seealso marker="#call/2">gen_statem:call/2</seealso>,
          <seealso marker="#cast/2">gen_statem:cast/2</seealso> or
	  as a normal process message this function is called.
	  If the <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-state_op">StateOps</seealso>
	  or by calling
	  <seealso marker="#reply/2">
	    <c>gen_statem:reply(Client, Reply)</c>
	  </seealso>.
	</p>
        <p><seealso marker="#type-state"><c>State</c></seealso>
	  is the internal state of the gen_statem which
	  when <c>State</c> is an <c>atom()</c>
	  is the same as this function's name, so it is seldom useful,
	  except for example when comparing with <c>PrevState</c>
	  that is the gen_statem's previous state, or in
	  <seealso marker="#Module:handle_event/5">
	    Module:handle_event/5
	  </seealso> since that function is common for all states
	  that are not an <c>atom()</c>.
	</p>
	<p>If this function returns with
	  <seealso marker="#type-state"><c>NewState =/= State</c></seealso>
	  all postponed events will be retried in the new state.
	</p>
	<p>See <seealso marker="#type-state_op">state_op()</seealso>
	  for the operations that can be done by gen_statem
	  after returning from this function.
	</p>
      </desc>
    </func>

    <func>
      <name>Module:terminate(Reason, State, StateData)</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>StateData =
	  <seealso marker="#type-state_data">
	    state_data()
	  </seealso>
	</v>
      </type>
      <desc>
        <p>This function is called by a gen_statem 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 gen_statem 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 gen_statem.
	</p>
        <p><c>Reason</c> depends on why the gen_statem is terminating.
          If it is because another callback function has returned a
          stop tuple <c>{stop,Reason}</c> in
	  <seealso marker="#type-state_op">StateOps</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 gen_statem 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 gen_statem 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 gen_statem 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 gen_statem 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 gen_statem 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>
      </desc>
    </func>

    <func>
      <name>Module:code_change(OldVsn, OldState, OldStateData, 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,NewStateData}} | Reason</v>
	<v>OldState = NewState =
	  <seealso marker="#type-state">state()</seealso>
	</v>
	<v>OldStateData = NewStateData =
	  <seealso marker="#type-state_data">state_data()</seealso>
	</v>
	<v>Reason = term()</v>
      </type>
      <desc>
        <p>This function is called by a gen_statem 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>OldStateData</c> is the internal state
	  of the gen_statem.
	</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,NewStateData}}</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>
    </func>

    <func>
      <name>Module:format_status(Opt, [PDict,State,StateData]) ->
        Status
      </name>
      <fsummary>Optional function for providing a term describing the
        current gen_statem status</fsummary>
      <type>
        <v>Opt = normal | terminate</v>
        <v>PDict = [{Key, Value}]</v>
	<v>State =
	  <seealso marker="#type-state">state()</seealso>
	</v>
	<v>StateData =
	  <seealso marker="#type-state_data">state_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 gen_statem module provides a default
            implementation of this function that returns the callback
            module state.
	  </p>
        </note>
        <p>This function is called by a gen_statem 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 gen_statem status. <c>Opt</c> is set
            to the atom <c>normal</c> for this case.
	  </item>
          <item>The gen_statem 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 gen_statem 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 gen_statem.
	</p>
        <p><c>PDict</c> is the current value of the gen_statem's
          process dictionary.
	</p>
        <p><seealso marker="#type-state"><c>State</c></seealso>
	  is the internal state of the gen_statem.
	</p>
        <p>The function should return <c>Status</c>, a term that
          customises the details of the current state and status of
          the gen_statem. 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 gen_statem 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>
      </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>