path: root/erts/doc/src/match_spec.xml

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

<chapter>
  <header>
    <copyright>
      <year>1999</year><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>Match Specifications in Erlang</title>
    <prepared>Patrik Nyblom</prepared>
    <responsible></responsible>
    <docno></docno>
    <approved></approved>
    <checked></checked>
    <date>1999-06-01</date>
    <rev>PA1</rev>
    <file>match_spec.xml</file>
  </header>
  <p>A "match specification" (<c>match_spec</c>) is an Erlang term describing a
    small "program" that tries to match something. It can be used
    to either control tracing with
    <seealso marker="erlang#trace_pattern/3">erlang:trace_pattern/3</seealso>
    or to search for objects in an ETS table with for example
    <seealso marker="stdlib:ets#select/2">ets:select/2</seealso>.
    The match specification in many ways works like a small function in Erlang,
    but is interpreted/compiled by the Erlang runtime system to something much more
    efficient than calling an Erlang function. The match specification is also
    very limited compared to the expressiveness of real Erlang functions.</p>
  <p>The most notable difference between a match specification and an Erlang
    fun is the syntax. Match specifications are Erlang terms, not Erlang code.
    Also, a match specification has a strange concept of exceptions:</p>

  <list type="bulleted">
    <item>
      <p>An exception (such as <c><![CDATA[badarg]]></c>) in the
        <c><![CDATA[MatchCondition]]></c> part, which resembles an Erlang guard,
        generates immediate failure.</p>
    </item>
    <item>
      <p>An exception in the <c><![CDATA[MatchBody]]></c> part, which resembles
        the body of an Erlang function, is implicitly caught and results in the
        single atom <c><![CDATA['EXIT']]></c>.</p>
    </item>
  </list>

  <section>
    <title>Grammar</title>
    <p>A match specification used in tracing can be described in the following
      <em>informal</em> grammar:</p>

    <list type="bulleted">
      <item>MatchExpression ::= [ MatchFunction, ... ]
      </item>
      <item>MatchFunction ::= { MatchHead, MatchConditions, MatchBody }
      </item>
      <item>MatchHead ::=  MatchVariable | <c><![CDATA['_']]></c> |
        [ MatchHeadPart, ... ]
      </item>
      <item>MatchHeadPart ::= term() | MatchVariable | <c><![CDATA['_']]></c>
      </item>
      <item>MatchVariable ::= '$&lt;number&gt;'
      </item>
      <item>MatchConditions ::= [ MatchCondition, ...] | <c><![CDATA[[]]]></c>
      </item>
      <item>MatchCondition ::= { GuardFunction } | { GuardFunction,
        ConditionExpression, ... }
      </item>
      <item>BoolFunction ::= <c><![CDATA[is_atom]]></c> |
        <c><![CDATA[is_float]]></c> | <c><![CDATA[is_integer]]></c> |
        <c><![CDATA[is_list]]></c> | <c><![CDATA[is_number]]></c> |
        <c><![CDATA[is_pid]]></c> | <c><![CDATA[is_port]]></c> |
        <c><![CDATA[is_reference]]></c> | <c><![CDATA[is_tuple]]></c> |
        <c><![CDATA[is_map]]></c> | <c><![CDATA[is_binary]]></c> |
        <c><![CDATA[is_function]]></c> | <c><![CDATA[is_record]]></c> |
        <c><![CDATA[is_seq_trace]]></c> | <c><![CDATA['and']]></c> |
        <c><![CDATA['or']]></c> | <c><![CDATA['not']]></c> |
        <c><![CDATA['xor']]></c> | <c><![CDATA['andalso']]></c> |
        <c><![CDATA['orelse']]></c>
      </item>
      <item>ConditionExpression ::= ExprMatchVariable | { GuardFunction } |
        { GuardFunction, ConditionExpression, ... } | TermConstruct
      </item>
      <item>ExprMatchVariable ::= MatchVariable (bound in the MatchHead) |
        <c><![CDATA['$_']]></c> | <c><![CDATA['$$']]></c>
      </item>
      <item>TermConstruct = {{}} | {{ ConditionExpression, ... }} |
        <c><![CDATA[[]]]></c> | [ConditionExpression, ...] |
        <c><![CDATA[#{}]]></c> | #{term() => ConditionExpression, ...} |
        NonCompositeTerm | Constant
      </item>
      <item>NonCompositeTerm ::= term() (not list or tuple or map)
      </item>
      <item>Constant ::= {<c><![CDATA[const]]></c>, term()}
      </item>
      <item>GuardFunction ::= BoolFunction | <c><![CDATA[abs]]></c> |
        <c><![CDATA[element]]></c> | <c><![CDATA[hd]]></c> |
        <c><![CDATA[length]]></c> | <c><![CDATA[map_get]]></c> |
        <c><![CDATA[map_size]]></c> | <c><![CDATA[node]]></c> |
        <c><![CDATA[round]]></c> | <c><![CDATA[size]]></c> |
        <c><![CDATA[tl]]></c> | <c><![CDATA[trunc]]></c> |
        <c><![CDATA['+']]></c> | <c><![CDATA['-']]></c> |
        <c><![CDATA['*']]></c> | <c><![CDATA['div']]></c> |
        <c><![CDATA['rem']]></c> | <c><![CDATA['band']]></c> |
        <c><![CDATA['bor']]></c> | <c><![CDATA['bxor']]></c> |
        <c><![CDATA['bnot']]></c> | <c><![CDATA['bsl']]></c> |
        <c><![CDATA['bsr']]></c> | <c><![CDATA['>']]></c> |
        <c><![CDATA['>=']]></c> | <c><![CDATA['<']]></c> |
        <c><![CDATA['=<']]></c> | <c><![CDATA['=:=']]></c> |
        <c><![CDATA['==']]></c> | <c><![CDATA['=/=']]></c> |
        <c><![CDATA['/=']]></c> | <c><![CDATA[self]]></c> |
        <c><![CDATA[get_tcw]]></c>
      </item>
      <item>MatchBody ::= [ ActionTerm ]
      </item>
      <item>ActionTerm ::= ConditionExpression | ActionCall
      </item>
      <item>ActionCall ::= {ActionFunction} | {ActionFunction, ActionTerm, ...}
      </item>
      <item>ActionFunction ::= <c><![CDATA[set_seq_token]]></c> |
        <c><![CDATA[get_seq_token]]></c> | <c><![CDATA[message]]></c> |
        <c><![CDATA[return_trace]]></c> | <c><![CDATA[exception_trace]]></c> |
        <c><![CDATA[process_dump]]></c> | <c><![CDATA[enable_trace]]></c> |
        <c><![CDATA[disable_trace]]></c> | <c><![CDATA[trace]]></c> |
        <c><![CDATA[display]]></c> | <c><![CDATA[caller]]></c> |
        <c><![CDATA[set_tcw]]></c> | <c><![CDATA[silent]]></c>
      </item>
    </list>

    <p>A match specification  used in
      <seealso marker="stdlib:ets"><c>ets(3)</c></seealso>
      can be described in the following <em>informal</em> grammar:</p>

    <list type="bulleted">
      <item>MatchExpression ::= [ MatchFunction, ... ]
      </item>
      <item>MatchFunction ::= { MatchHead, MatchConditions, MatchBody }
      </item>
      <item>MatchHead ::=  MatchVariable | <c><![CDATA['_']]></c> |
        { MatchHeadPart, ... }
      </item>
      <item>MatchHeadPart ::= term() | MatchVariable | <c><![CDATA['_']]></c>
      </item>
      <item>MatchVariable ::= '$&lt;number&gt;'
      </item>
      <item>MatchConditions ::= [ MatchCondition, ...] | <c><![CDATA[[]]]></c>
      </item>
      <item>MatchCondition ::= { GuardFunction } |
        { GuardFunction, ConditionExpression, ... }
      </item>
      <item>BoolFunction ::= <c><![CDATA[is_atom]]></c> |
        <c><![CDATA[is_float]]></c> | <c><![CDATA[is_integer]]></c> |
        <c><![CDATA[is_list]]></c> | <c><![CDATA[is_number]]></c> |
        <c><![CDATA[is_pid]]></c> | <c><![CDATA[is_port]]></c> |
        <c><![CDATA[is_reference]]></c> | <c><![CDATA[is_tuple]]></c> |
        <c><![CDATA[is_map]]></c> | <c><![CDATA[is_binary]]></c> |
        <c><![CDATA[is_function]]></c> | <c><![CDATA[is_record]]></c> |
        <c><![CDATA['and']]></c> | <c><![CDATA['or']]></c> |
        <c><![CDATA['not']]></c> | <c><![CDATA['xor']]></c> |
        <c><![CDATA['andalso']]></c> | <c><![CDATA['orelse']]></c>
      </item>
      <item>ConditionExpression ::= ExprMatchVariable | { GuardFunction } |
        { GuardFunction, ConditionExpression, ... } | TermConstruct
      </item>
      <item>ExprMatchVariable ::= MatchVariable (bound in the MatchHead) |
        <c><![CDATA['$_']]></c> | <c><![CDATA['$$']]></c>
      </item>
      <item>TermConstruct = {{}} | {{ ConditionExpression, ... }} |
        <c><![CDATA[[]]]></c> | [ConditionExpression, ...] | #{} |
        #{term() => ConditionExpression, ...} | NonCompositeTerm | Constant
      </item>
      <item>NonCompositeTerm ::= term() (not list or tuple or map)
      </item>
      <item>Constant ::= {<c><![CDATA[const]]></c>, term()}
      </item>
      <item>GuardFunction ::= BoolFunction | <c><![CDATA[abs]]></c> |
        <c><![CDATA[element]]></c> | <c><![CDATA[hd]]></c> |
        <c><![CDATA[length]]></c> | <c><![CDATA[map_get]]></c> |
        <c><![CDATA[map_size]]></c> | <c><![CDATA[node]]></c> |
        <c><![CDATA[round]]></c> | <c><![CDATA[size]]></c> |
        <c><![CDATA[tl]]></c> | <c><![CDATA[trunc]]></c> |
        <c><![CDATA['+']]></c> | <c><![CDATA['-']]></c> |
        <c><![CDATA['*']]></c> | <c><![CDATA['div']]></c> |
        <c><![CDATA['rem']]></c> | <c><![CDATA['band']]></c> |
        <c><![CDATA['bor']]></c> | <c><![CDATA['bxor']]></c> |
        <c><![CDATA['bnot']]></c> | <c><![CDATA['bsl']]></c> |
        <c><![CDATA['bsr']]></c> | <c><![CDATA['>']]></c> |
        <c><![CDATA['>=']]></c> | <c><![CDATA['<']]></c> |
        <c><![CDATA['=<']]></c> | <c><![CDATA['=:=']]></c> |
        <c><![CDATA['==']]></c> | <c><![CDATA['=/=']]></c> |
        <c><![CDATA['/=']]></c> | <c><![CDATA[self]]></c>
      </item>
      <item>MatchBody ::= [ ConditionExpression, ... ]
      </item>
    </list>
  </section>

  <section>
    <title>Function Descriptions</title>
    <section>
      <title>Functions Allowed in All Types of Match Specifications</title>
      <p>The functions allowed in <c><![CDATA[match_spec]]></c> work as
        follows:</p>

      <taglist>
        <tag><c>is_atom</c>, <c>is_float</c>, <c>is_integer</c>, <c>is_list</c>,
          <c>is_number</c>, <c>is_pid</c>, <c>is_port</c>, <c>is_reference</c>,
          <c>is_tuple</c>, <c>is_map</c>, <c>is_binary</c>, <c>is_function</c>
        </tag>
        <item>
          <p>Same as the corresponding guard tests in Erlang, return
            <c><![CDATA[true]]></c> or <c><![CDATA[false]]></c>.</p>
        </item>
        <tag><c>is_record</c></tag>
        <item>
          <p>Takes an additional parameter, which <em>must</em> be the result
            of <c><![CDATA[record_info(size, <record_type>)]]></c>, like in
            <c><![CDATA[{is_record, '$1', rectype, record_info(size,
            rectype)}]]></c>.</p>
        </item>
        <tag><c>'not'</c></tag>
        <item>
          <p>Negates its single argument (anything other
            than <c><![CDATA[false]]></c> gives <c><![CDATA[false]]></c>).</p>
        </item>
        <tag><c>'and'</c></tag>
        <item>
          <p>Returns <c><![CDATA[true]]></c> if all its arguments (variable
            length argument list) evaluate to <c><![CDATA[true]]></c>, otherwise
            <c><![CDATA[false]]></c>. Evaluation order is undefined.</p>
        </item>
        <tag><c>'or'</c></tag>
        <item>
          <p>Returns <c><![CDATA[true]]></c> if any of its arguments
            evaluates to <c><![CDATA[true]]></c>. Variable length argument
            list. Evaluation order is undefined.</p>
        </item>
        <tag><c>'andalso'</c></tag>
        <item>
          <p>Works as <c><![CDATA['and']]></c>, but quits evaluating its
            arguments when one argument evaluates to something else
            than <c>true</c>. Arguments are evaluated left to right.</p>
        </item>
        <tag><c>'orelse'</c></tag>
        <item>
          <p>Works as <c><![CDATA['or']]></c>, but quits evaluating as soon
            as one of its arguments evaluates to <c><![CDATA[true]]></c>.
            Arguments are evaluated left to right.</p>
        </item>
        <tag><c>'xor'</c></tag>
        <item>
          <p>Only two arguments, of which one must be <c>true</c> and the
            other <c>false</c> to return <c><![CDATA[true]]></c>; otherwise
            <c><![CDATA['xor']]></c> returns false.</p>
        </item>
        <tag><c>abs</c>, <c>element</c>, <c>hd</c>, <c>length</c>, <c>node</c>,
          <c>round</c>, <c>size</c>, <c>tl</c>, <c>trunc</c>, <c>'+'</c>,
          <c>'-'</c>, <c>'*'</c>, <c>'div'</c>, <c>'rem'</c>, <c>'band'</c>,
          <c>'bor'</c>, <c>'bxor'</c>, <c>'bnot'</c>, <c>'bsl'</c>,
          <c>'bsr'</c>, <c>'>'</c>, <c>'>='</c>, <c>'&lt;'</c>, <c>'=&lt;'</c>,
          <c>'=:='</c>, <c>'=='</c>, <c>'=/='</c>, <c>'/='</c>,
          <c>self</c></tag>
        <item>
          <p>Same as the corresponding Erlang BIFs (or operators). In case of
            bad arguments, the result depends on the context. In the
            <c><![CDATA[MatchConditions]]></c> part of the expression, the test
            fails immediately (like in an Erlang guard). In the
            <c><![CDATA[MatchBody]]></c> part, exceptions are implicitly caught
            and the call results in the atom <c><![CDATA['EXIT']]></c>.</p>
        </item>
      </taglist>
    </section>

    <section>
      <title>Functions Allowed Only for Tracing</title>
      <p>The functions allowed only for tracing work as follows:</p>

      <taglist>
        <tag><c>is_seq_trace</c></tag>
        <item>
          <p>Returns <c><![CDATA[true]]></c> if a sequential trace token is set
            for the current process, otherwise <c><![CDATA[false]]></c>.</p>
        </item>
        <tag><c>set_seq_token</c></tag>
        <item>
          <p>Works as <c><![CDATA[seq_trace:set_token/2]]></c>, but returns
            <c><![CDATA[true]]></c> on success, and <c><![CDATA['EXIT']]></c>
            on error or bad argument. Only allowed in the
            <c><![CDATA[MatchBody]]></c> part and only allowed when tracing.</p>
        </item>
        <tag><c>get_seq_token</c></tag>
        <item>
          <p>Same as <c><![CDATA[seq_trace:get_token/0]]></c> and only
            allowed in the <c><![CDATA[MatchBody]]></c> part when tracing.</p>
        </item>
        <tag><c>message</c></tag>
        <item>
          <p>Sets an additional message appended to the
            trace message sent. One can only set one additional message in
            the body. Later calls replace the appended message.</p>
          <p>As a special case, <c><![CDATA[{message, false}]]></c> disables
            sending of trace messages ('call' and 'return_to') for this function
            call, just like if the match specification had not matched.
            This can be useful if only the side effects of 
            the <c><![CDATA[MatchBody]]></c> part are desired.</p>
          <p>Another special case is <c><![CDATA[{message, true}]]></c>, which
            sets the default behavior, as if the function had no match
            specification; trace message is sent with no extra information
            (if no other calls to <c><![CDATA[message]]></c> are placed before
            <c><![CDATA[{message, true}]]></c>, it is in fact a "noop").</p>
          <p>Takes one argument: the message. Returns <c><![CDATA[true]]></c>
            and can only be used in the <c><![CDATA[MatchBody]]></c> part and
            when tracing.</p>
        </item>
        <tag><c>return_trace</c></tag>
        <item>
          <p>Causes a <c><![CDATA[return_from]]></c> trace message to be sent
            upon return from the current function. Takes no arguments, returns
            <c><![CDATA[true]]></c> and can only be used in the
            <c><![CDATA[MatchBody]]></c> part when tracing.
            If the process trace flag <c><![CDATA[silent]]></c> is active, the
            <c><![CDATA[return_from]]></c> trace message is inhibited.</p>
          <p><em>Warning:</em> If the traced function is tail-recursive, this
            match specification function destroys that property. Hence, if a
            match specification executing this function is used on a
            perpetual server process, it can only be active for a limited
            period of time, or the emulator will eventually use all memory in
            the host machine and crash. If this match specification function is
            inhibited using process trace flag <c><![CDATA[silent]]></c>,
            tail-recursiveness still remains.</p>
        </item>
        <tag><c>exception_trace</c></tag>
        <item>
          <p>Works as <c>return_trace</c> plus; if the traced function exits
            because of an exception,
            an <c><![CDATA[exception_from]]></c> trace message is generated,
            regardless of the exception is caught or not.</p>
        </item>
        <tag><c>process_dump</c></tag>
        <item>
          <p>Returns some textual information about
            the current process as a binary. Takes no arguments and is only
            allowed in the <c><![CDATA[MatchBody]]></c> part when tracing.</p>
        </item>
        <tag><c>enable_trace</c></tag>
        <item>
          <p>With one parameter this function turns on tracing like the Erlang
            call <c><![CDATA[erlang:trace(self(), true, [P2])]]></c>, where
            <c><![CDATA[P2]]></c> is the parameter to
            <c><![CDATA[enable_trace]]></c>.</p>
          <p>With two parameters, the first parameter is to be either a process
            identifier or the registered name of a process. In this case tracing
            is turned on for the designated process in the same way as in the
            Erlang call <c><![CDATA[erlang:trace(P1, true, [P2])]]></c>, where
            <c>P1</c> is the first and <c>P2</c> is the second argument. The
            process <c><![CDATA[P1]]></c> gets its trace messages sent to the
            same tracer as the process executing the statement uses.
            <c><![CDATA[P1]]></c> <em>cannot</em> be one of the atoms
            <c><![CDATA[all]]></c>, <c><![CDATA[new]]></c> or 
            <c><![CDATA[existing]]></c> (unless they are registered names).
            <c><![CDATA[P2]]></c> <em>cannot</em> be
            <c><![CDATA[cpu_timestamp]]></c> or <c><![CDATA[tracer]]></c>.</p>
          <p>Returns <c><![CDATA[true]]></c> and can only be used in
            the <c><![CDATA[MatchBody]]></c> part when tracing.</p>
        </item>
        <tag><c>disable_trace</c></tag>
        <item>
          <p>With one parameter this function disables tracing like the Erlang
            call <c><![CDATA[erlang:trace(self(), false, [P2])]]></c>, where
            <c><![CDATA[P2]]></c> is the parameter to
            <c><![CDATA[disable_trace]]></c>.</p>
          <p>With two parameters this function works as the Erlang call
            <c><![CDATA[erlang:trace(P1, false, [P2])]]></c>, where <c>P1</c>
            can be either a process identifier or a registered name and is
            specified as the first argument to the match specification function.
            <c><![CDATA[P2]]></c> <em>cannot</em> be
            <c><![CDATA[cpu_timestamp]]></c> or <c><![CDATA[tracer]]></c>.</p>
          <p>Returns <c><![CDATA[true]]></c> and can only be used in the
            <c><![CDATA[MatchBody]]></c> part when tracing.</p>
        </item>
        <tag><c>trace</c></tag>
        <item>
          <p>With two parameters this function takes a list
            of trace flags to disable as first parameter and a list
            of trace flags to enable as second parameter. Logically, the
            disable list is applied first, but effectively all changes
            are applied atomically. The trace flags
            are the same as for <c><![CDATA[erlang:trace/3]]></c>,
            not including <c><![CDATA[cpu_timestamp]]></c>, but including
            <c><![CDATA[tracer]]></c>.</p>
          <p>If a tracer is specified in both lists, the tracer in the
            enable list takes precedence. If no tracer is specified, the same
            tracer as the process executing the match specification is used (not the meta tracer).
            If that process doesn't have tracer either, then trace flags are ignored.</p>
          <p>When using a <seealso marker="erl_tracer">tracer module</seealso>,
            the module must be loaded before the match specification is
            executed. If it is not loaded, the match fails.</p>
          <p>With three parameters to this function, the first is
            either a process identifier or the registered name of a
            process to set trace flags on, the second is the disable
            list, and the third is the enable list.</p>
          <p>Returns <c><![CDATA[true]]></c> if any trace property was changed
            for the trace target process, otherwise <c><![CDATA[false]]></c>.
            Can only be used in the <c><![CDATA[MatchBody]]></c> part when
            tracing.</p>
        </item>
        <tag><c>caller</c></tag>
        <item>
          <p>Returns the calling function as a tuple <c>{Module, Function,
             Arity}</c> or the atom <c><![CDATA[undefined]]></c> if the calling
             function cannot be determined. Can only be used in the
             <c><![CDATA[MatchBody]]></c> part when tracing.</p>
           <p>Notice that if a "technically built in function" (that is, a
             function not written in Erlang) is traced, the
             <c><![CDATA[caller]]></c> function sometimes returns the atom
             <c><![CDATA[undefined]]></c>. The calling
             Erlang function is not available during such calls.</p>
        </item>
        <tag><c>display</c></tag>
          <item>
            <p>For debugging purposes only. Displays the single argument as an
              Erlang term on <c>stdout</c>, which is seldom what is wanted.
              Returns <c><![CDATA[true]]></c> and can only be used in the
              <c><![CDATA[MatchBody]]></c> part when tracing.</p>
          </item>
        <tag><marker id="get_tcw"/><c>get_tcw</c></tag>
        <item>
          <p>Takes no argument and returns the value of the node's trace
            control word. The same is done by
            <c><![CDATA[erlang:system_info(trace_control_word)]]></c>.</p>
          <p>The trace control word is a 32-bit unsigned integer intended for
            generic trace control. The trace control word can be tested and
            set both from within trace match specifications and with BIFs.
            This call is only allowed when tracing.</p>
        </item>
        <tag><marker id="set_tcw"/><c>set_tcw</c></tag>
        <item>
          <p>Takes one unsigned integer argument, sets the value of
            the node's trace control word to the value of the argument,
            and returns the previous value. The same is done by
            <c><![CDATA[erlang:system_flag(trace_control_word, Value)]]></c>.
            It is only allowed to use <c><![CDATA[set_tcw]]></c> in the
            <c><![CDATA[MatchBody]]></c> part when tracing.</p>
        </item>
        <tag><c>silent</c></tag>
        <item>
          <p>Takes one argument. If the argument is <c><![CDATA[true]]></c>,
            the call trace message mode for the current process is set to
            silent for this call and all later calls, that is, call trace
            messages are inhibited even if
            <c><![CDATA[{message, true}]]></c> is called in the
            <c><![CDATA[MatchBody]]></c> part for a traced function.</p>
          <p>This mode can also be activated with flag
            <c><![CDATA[silent]]></c> to
            <c><![CDATA[erlang:trace/3]]></c>.</p>
          <p>If the argument is <c><![CDATA[false]]></c>, the call trace
            message mode for the current process is set to normal
            (non-silent) for this call and all later calls.</p>
          <p>If the argument is not <c><![CDATA[true]]></c> or
            <c><![CDATA[false]]></c>, the call trace message mode is
            unaffected.</p>
        </item>
      </taglist>
    </section>

    <note>
      <p>All "function calls" must be tuples, even if they take no arguments.
        The value of <c><![CDATA[self]]></c> is the atom()
        <c><![CDATA[self]]></c>, but the value of <c><![CDATA[{self}]]></c> is
        the pid() of the current process.</p>
    </note>
  </section>

  <section>
    <marker id="match_target"/>
    <title>Match target</title>
    <p>Each execution of a match specification is done against
      a match target term. The format and content of the target term
      depends on the context in which the match is done. The match
      target for ETS is always a full table tuple. The match target
      for call trace is always a list of all function arguments. The
      match target for event trace depends on the event type, see
      table below.</p>
      <table>
      <row>
        <cell align="left" valign="middle">Context</cell>
        <cell align="left" valign="middle">Type</cell>
        <cell align="left" valign="middle">Match target</cell>
        <cell align="left" valign="middle">Description</cell>
      </row>
      <row>
        <cell align="left" valign="middle">ETS</cell>
        <cell align="left" valign="middle"></cell>
        <cell align="left" valign="middle">{Key, Value1, Value2, ...}</cell>
        <cell align="left" valign="middle">A table object</cell>
      </row>
      <row>
        <cell align="left" valign="middle">Trace</cell>
        <cell align="left" valign="middle">call</cell>
        <cell align="left" valign="middle">[Arg1, Arg2, ...]</cell>
        <cell align="left" valign="middle">Function arguments</cell>
      </row>
      <row>
        <cell align="left" valign="middle">Trace</cell>
        <cell align="left" valign="middle">send</cell>
        <cell align="left" valign="middle">[Receiver, Message]</cell>
        <cell align="left" valign="middle">Receiving process/port and message term</cell>
      </row>
      <row>
        <cell align="left" valign="middle">Trace</cell>
        <cell align="left" valign="middle">'receive'</cell>
        <cell align="left" valign="middle">[Node, Sender, Message]</cell>
        <cell align="left" valign="middle">Sending node, process/port and message term</cell>
      </row>
      <tcaption>Match target depending on context</tcaption>
      </table>
  </section>

  <section>
    <title>Variables and Literals</title>
    <p>Variables take the form <c><![CDATA['$<number>']]></c>, where
      <c><![CDATA[<number>]]></c> is an integer between 0 and
      100,000,000 (1e+8). The behavior if the number is outside these limits
      is <em>undefined</em>. In the <c><![CDATA[MatchHead]]></c> part, the
      special variable <c><![CDATA['_']]></c> matches anything, and never gets
      bound (like <c><![CDATA[_]]></c> in Erlang).</p>

    <list type="bulleted">
      <item>
        <p>In the <c><![CDATA[MatchCondition/MatchBody]]></c> parts,
          no unbound variables are allowed, so <c><![CDATA['_']]></c> is
          interpreted as itself (an atom). Variables can only be bound in the
          <c><![CDATA[MatchHead]]></c> part.</p>
      </item>
      <item>
        <p>In the <c><![CDATA[MatchBody]]></c> and
          <c><![CDATA[MatchCondition]]></c> parts, only variables bound
          previously can be used.</p>
      </item>
      <item>
        <p>As a special case, the following apply in the
          <c><![CDATA[MatchCondition/MatchBody]]></c> parts:</p>
        <list type="bulleted">
          <item>
            <p>The variable <c><![CDATA['$_']]></c> expands to the whole
            <seealso marker="#match_target">match target</seealso> term.
            </p>
          </item>
          <item>
            <p>The variable <c><![CDATA['$$']]></c> expands to a list of the
              values of all bound variables in order (that is,
              <c><![CDATA[['$1','$2', ...]]]></c>).</p>
          </item>
        </list>
      </item>
    </list>

    <p>In the <c><![CDATA[MatchHead]]></c> part, all literals (except the
      variables above) are interpreted "as is".</p>

    <p>In the <c><![CDATA[MatchCondition/MatchBody]]></c> parts, the
      interpretation is in some ways different. Literals in these parts
      can either be written "as is", which works for all literals except
      tuples, or by using the special form <c><![CDATA[{const, T}]]></c>,
      where <c><![CDATA[T]]></c> is any Erlang term.</p>

    <p>For tuple literals in the match specification, double tuple parentheses
      can also be used, that is, construct them as a tuple of
      arity one containing a single tuple, which is the one to be
      constructed. The "double tuple parenthesis" syntax is useful to
      construct tuples from already bound variables, like in
      <c><![CDATA[{{'$1', [a,b,'$2']}}]]></c>. Examples:</p>

    <table>
      <row>
        <cell align="left" valign="middle"><em>Expression</em></cell>
        <cell align="left" valign="middle"><em>Variable Bindings</em></cell>
        <cell align="left" valign="middle"><em>Result</em></cell>
      </row>
      <row>
        <cell align="left" valign="middle">{{'$1','$2'}}</cell>
        <cell align="left" valign="middle">'$1' = a, '$2' = b</cell>
        <cell align="left" valign="middle">{a,b}</cell>
      </row>
      <row>
        <cell align="left" valign="middle">{const, {'$1', '$2'}}</cell>
        <cell align="left" valign="middle">Irrelevant</cell>
        <cell align="left" valign="middle">{'$1', '$2'}</cell>
      </row>
      <row>
        <cell align="left" valign="middle">a</cell>
        <cell align="left" valign="middle">Irrelevant</cell>
        <cell align="left" valign="middle">a</cell>
      </row>
      <row>
        <cell align="left" valign="middle">'$1'</cell>
        <cell align="left" valign="middle">'$1' = []</cell>
        <cell align="left" valign="middle">[]</cell>
      </row>
      <row>
        <cell align="left" valign="middle">['$1']</cell>
        <cell align="left" valign="middle">'$1' = []</cell>
        <cell align="left" valign="middle">[[]]</cell>
      </row>
      <row>
        <cell align="left" valign="middle">[{{a}}]</cell>
        <cell align="left" valign="middle">Irrelevant</cell>
        <cell align="left" valign="middle">[{a}]</cell>
      </row>
      <row>
        <cell align="left" valign="middle">42</cell>
        <cell align="left" valign="middle">Irrelevant</cell>
        <cell align="left" valign="middle">42</cell>
      </row>
      <row>
        <cell align="left" valign="middle">"hello"</cell>
        <cell align="left" valign="middle">Irrelevant</cell>
        <cell align="left" valign="middle">"hello"</cell>
      </row>
      <row>
        <cell align="left" valign="middle">$1</cell>
        <cell align="left" valign="middle">Irrelevant</cell>
        <cell align="left" valign="middle">49 (the ASCII value for
          character '1')</cell>
      </row>
      <tcaption>Literals in MatchCondition/MatchBody Parts of a Match
        Specification</tcaption>
    </table>
  </section>

  <section>
    <title>Execution of the Match</title>
    <p>The execution of the match expression, when the runtime system
      decides whether a trace message is to be sent, is as follows:</p>

    <p>For each tuple in the <c><![CDATA[MatchExpression]]></c> list and while
      no match has succeeded:</p>

    <list type="ordered">
      <item>
        <p>Match the <c><![CDATA[MatchHead]]></c> part against the match target
          term, binding the <c><![CDATA['$<number>']]></c> variables
          (much like in <c><![CDATA[ets:match/2]]></c>). If the
          <c><![CDATA[MatchHead]]></c> part cannot match the arguments, the
          match fails.</p>
      </item>
      <item>
        <p>Evaluate each <c><![CDATA[MatchCondition]]></c> (where only
          <c><![CDATA['$<number>']]></c> variables previously bound in the
          <c><![CDATA[MatchHead]]></c> part can occur) and expect it to return
          the atom <c><![CDATA[true]]></c>. When a condition does not evaluate
          to <c><![CDATA[true]]></c>, the match fails. If any BIF call
          generates an exception, the match also fails.</p>
      </item>
      <item>
        <p>Two cases can occur:</p>
        <list type="bulleted">
          <item>
            <p>If the match specification is executing when tracing:</p>
            <p>Evaluate each <c><![CDATA[ActionTerm]]></c> in the same way as
              the <c><![CDATA[MatchConditions]]></c>, but ignore the return
              values. Regardless of what happens in this part, the match has
              succeeded.</p>
          </item>
          <item>
            <p>If the match specification is executed when selecting objects
              from an ETS table:</p>
            <p>Evaluate the expressions in order and return the value of
              the last expression (typically there is only one expression
              in this context).</p>
          </item>
        </list>
      </item>
    </list>
  </section>

  <section>
    <marker id="differences_ets_tracing"/>
    <title>Differences between Match Specifications in ETS and Tracing</title>
    <p>ETS match specifications produce a return value.
        Usually the <c><![CDATA[MatchBody]]></c> contains one single
        <c><![CDATA[ConditionExpression]]></c> that defines the return value
        without any side effects. Calls with side effects are not allowed in
        the ETS context.</p>

    <p>When tracing there is no return value to produce, the
      match specification either matches or does not. The effect when the
      expression matches is a trace message rather than a returned
      term. The <c><![CDATA[ActionTerm]]></c>s are executed as in an imperative
      language, that is, for their side effects. Functions with side effects
      are also allowed when tracing.</p>
  </section>

  <section>
    <title>Tracing Examples</title>
    <p>Match an argument list of three, where the first and third arguments
      are equal:</p>

    <code type="none"><![CDATA[
[{['$1', '_', '$1'],
  [],
  []}]
    ]]></code>

    <p>Match an argument list of three, where the second argument is
      a number &gt; 3:</p>

    <code type="none"><![CDATA[
[{['_', '$1', '_'],
  [{ '>', '$1', 3}],
  []}]
    ]]></code>

    <p>Match an argument list of three, where the third argument is
      either a tuple containing argument one and two, <em>or</em> a list
      beginning with argument one and two (that is,
      <c><![CDATA[[a,b,[a,b,c]]]]></c> or <c><![CDATA[[a,b,{a,b}]]]></c>):</p>

    <code type="none"><![CDATA[
[{['$1', '$2', '$3'],
  [{'orelse', 
      {'=:=', '$3', {{'$1','$2'}}},
      {'and', 
        {'=:=', '$1', {hd, '$3'}},
        {'=:=', '$2', {hd, {tl, '$3'}}}}}],
  []}]
    ]]></code>

    <p>The above problem can also be solved as follows:</p>

    <code type="none"><![CDATA[
[{['$1', '$2', {'$1', '$2}], [], []},
 {['$1', '$2', ['$1', '$2' | '_']], [], []}]
    ]]></code>

    <p>Match two arguments, where the first is a tuple beginning with
      a list that in turn begins with the second argument times
      two (that is, <c>[{[4,x],y},2]</c> or <c>[{[8], y, z},4])</c>:</p>

    <code type="none"><![CDATA[
[{['$1', '$2'],[{'=:=', {'*', 2, '$2'}, {hd, {element, 1, '$1'}}}],
  []}]
    ]]></code>

    <p>Match three arguments. When all three are equal and are
      numbers, append the process dump to the trace message, otherwise
      let the trace message be "as is", but set the sequential trace
      token label to 4711:</p>

    <code type="none"><![CDATA[
[{['$1', '$1', '$1'],
  [{is_number, '$1'}],
  [{message, {process_dump}}]},
 {'_', [], [{set_seq_token, label, 4711}]}]
    ]]></code>

    <p>As can be noted above, the parameter list can be matched against a
      single <c><![CDATA[MatchVariable]]></c> or an <c><![CDATA['_']]></c>.
      To replace the whole parameter list with a single variable is a special
      case. In all other cases the <c><![CDATA[MatchHead]]></c> must be a
      <em>proper</em> list.</p>

    <p>Generate a trace message only if the trace control word is set to 1:</p>

    <code type="none"><![CDATA[
[{'_',
  [{'==',{get_tcw},{const, 1}}],
  []}]
    ]]></code>

    <p>Generate a trace message only if there is a <c>seq_trace</c> token:</p>

    <code type="none"><![CDATA[
[{'_',
  [{'==',{is_seq_trace},{const, 1}}],
  []}]
    ]]></code>

    <p>Remove the <c>'silent'</c> trace flag when the first argument is
      <c>'verbose'</c>, and add it when it is <c>'silent':</c></p>

    <code type="none"><![CDATA[
[{'$1',
  [{'==',{hd, '$1'},verbose}],
  [{trace, [silent],[]}]},
 {'$1',
  [{'==',{hd, '$1'},silent}],
  [{trace, [],[silent]}]}]
    ]]></code>

    <p>Add a <c>return_trace</c> message if the function is of arity 3:</p>

    <code type="none"><![CDATA[
[{'$1',
  [{'==',{length, '$1'},3}],
  [{return_trace}]},
 {'_',[],[]}]
    ]]></code>

    <p>Generate a trace message only if the function is of arity 3 and the
      first argument is <c>'trace'</c>:</p>

    <code type="none"><![CDATA[
[{['trace','$2','$3'],
  [],
  []},
 {'_',[],[]}]
    ]]></code>
  </section>

  <section>
    <title>ETS Examples</title>
    <p>Match all objects in an ETS table, where the first element is
      the atom <c>'strider'</c> and the tuple arity is 3, and return the whole
      object:</p>

    <code type="none"><![CDATA[
[{{strider,'_','_'},
  [],
  ['$_']}]
    ]]></code>

    <p>Match all objects in an ETS table with arity &gt; 1 and the first
      element is 'gandalf', and return element 2:</p>

    <code type="none"><![CDATA[
[{'$1',
  [{'==', gandalf, {element, 1, '$1'}},{'>=',{size, '$1'},2}],
  [{element,2,'$1'}]}]
    ]]></code>

    <p>In this example, if the first element had been the key, it is
      much more efficient to match that key in the <c><![CDATA[MatchHead]]></c>
      part than in the <c><![CDATA[MatchConditions]]></c> part.
      The search space of the tables is restricted with regards to the
      <c><![CDATA[MatchHead]]></c> so
      that only objects with the matching key are searched.</p>

    <p>Match tuples of three elements, where the second element is either
      <c>'merry'</c> or <c>'pippin'</c>, and return the whole objects:</p>

    <code type="none"><![CDATA[
[{{'_',merry,'_'},
  [],
  ['$_']},
 {{'_',pippin,'_'},
  [],
  ['$_']}]
    ]]></code>

    <p>Function <seealso marker="stdlib:ets#test_ms/2"><c>ets:test_ms/2></c></seealso>
      can be useful for testing complicated ETS matches.</p>
  </section>
</chapter>