path: root/system/doc/reference_manual/expressions.xml

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

<chapter>
  <header>
    <copyright>
      <year>2003</year><year>2023</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>Expressions</title>
    <prepared></prepared>
    <docno></docno>
    <date></date>
    <rev></rev>
    <file>expressions.xml</file>
  </header>
  <p>In this section, all valid Erlang expressions are listed.
    When writing Erlang programs, it is also allowed to use macro-
    and record expressions. However, these expressions are expanded
    during compilation and are in that sense not true Erlang
    expressions. Macro- and record expressions are covered in
    separate sections:
   </p>
  <list type="bulleted">
    <item><p><seeguide marker="macros">Preprocessor</seeguide></p></item>
    <item><p><seeguide marker="records">Records</seeguide></p></item>
  </list>

  <section>
    <title>Expression Evaluation</title>
    <p>All subexpressions are evaluated before an expression itself is
      evaluated, unless explicitly stated otherwise. For example,
      consider the expression:</p>
    <code type="none">
Expr1 + Expr2</code>
    <p><c>Expr1</c> and <c>Expr2</c>, which are also expressions, are
      evaluated first - in any order - before the addition is
      performed.</p>
    <p>Many of the operators can only be applied to arguments of a
      certain type. For example, arithmetic operators can only be
      applied to numbers. An argument of the wrong type causes
      a <c>badarg</c> runtime error.</p>
  </section>

  <section>
    <marker id="term"></marker>
    <title>Terms</title>
    <p>The simplest form of expression is a term, that is an integer,
      float, atom, string, list, map, or tuple.
      The return value is the term itself.</p>
  </section>

  <section>
    <title>Variables</title>
    <p>A variable is an expression. If a variable is bound to a value,
      the return value is this value. Unbound variables are only
      allowed in patterns.</p>
    <p>Variables start with an uppercase letter or underscore (_).
      Variables can contain alphanumeric characters, underscore and <c>@</c>.
     </p>
     <p><em>Examples:</em></p>
    <pre>
X
Name1
PhoneNumber
Phone_number
_
_Height</pre>
    <p>Variables are bound to values using
      <seeguide marker="patterns">pattern matching</seeguide>. Erlang
      uses <em>single assignment</em>, that is, a variable can only be bound
      once.</p>
    <p>The <em>anonymous variable</em> is denoted by underscore (_) and
      can be used when a variable is required but its value can be
      ignored.</p>
      <p><em>Example:</em></p>
    <pre>
[H|_] = [1,2,3]</pre>
    <p>Variables starting with underscore (_), for example,
      <c>_Height</c>, are normal variables, not anonymous. However,
      they are ignored by the compiler in the sense that they do not
      generate warnings.</p>
      <p><em>Example:</em></p>
      <p>The following code:</p>
    <pre>
member(_, []) ->
    [].</pre>
    <p>can be rewritten to be more readable:</p>
    <pre>
member(Elem, []) ->
    [].</pre>
    <p>This causes a warning for an unused variable,
      <c>Elem</c>, if the code is compiled with the flag
      <c>warn_unused_vars</c> set. Instead, the code can be rewritten
      to:</p>
    <pre>
member(_Elem, []) ->
    [].</pre>
    <p>Notice that since variables starting with an underscore are
      not anonymous, this matches:</p>
    <pre>
{_,_} = {1,2}</pre>
    <p>But this fails:</p>
    <pre>
{_N,_N} = {1,2}</pre>
    <p>The scope for a variable is its function clause.
      Variables bound in a branch of an <c>if</c>, <c>case</c>, 
      or <c>receive</c> expression must be bound in all branches 
      to have a value outside the expression. Otherwise they
      are regarded as 'unsafe' outside the expression.</p>
    <p>For the <c>try</c> expression variable scoping is limited so that
      variables bound in the expression are always 'unsafe' outside 
      the expression.</p>
  </section>

  <section>
    <marker id="pattern"></marker>
    <title>Patterns</title>
    <p>A pattern has the same structure as a term but can contain
      unbound variables.</p>
      <p><em>Example:</em></p>
    <pre>
Name1
[H|T]
{error,Reason}</pre>
     <p>Patterns are allowed in clause heads,
     <seeguide marker="#case">case expressions</seeguide>,
     <seeguide marker="#receive">receive expressions</seeguide>,
     and
     <seeguide marker="#match_operator">match expressions</seeguide>.</p>

    <section>
      <marker id="compound_pattern_operator"></marker>
      <title>The Compound Pattern Operator</title>
      <p>If <c>Pattern1</c> and <c>Pattern2</c> are valid patterns,
        the following is also a valid pattern:</p>
      <pre>
Pattern1 = Pattern2</pre>
      <p>When matched against a term, both <c>Pattern1</c> and
      <c>Pattern2</c> are matched against the term. The idea behind
      this feature is to avoid reconstruction of terms.</p>
       <p><em>Example:</em></p>
      <pre>
f({connect,From,To,Number,Options}, To) ->
    Signal = {connect,From,To,Number,Options},
    ...;
f(Signal, To) ->
    ignore.</pre>
      <p>can instead be written as</p>
      <pre>
f({connect,_,To,_,_} = Signal, To) ->
    ...;
f(Signal, To) ->
    ignore.</pre>

    <p>The compound pattern operator does not imply that its operands
    are matched in any particular order. That means that it is not
    legal to bind a variable in <c>Pattern1</c> and use it in <c>Pattern2</c>,
    or vice versa.</p>
    </section>

    <section>
      <title>String Prefix in Patterns</title>
      <p>When matching strings, the following is a valid pattern:</p>
      <pre>
f("prefix" ++ Str) -> ...</pre>
      <p>This is syntactic sugar for the equivalent, but harder to
        read:</p>
      <pre>
f([$p,$r,$e,$f,$i,$x | Str]) -> ...</pre>
    </section>

    <section>
      <title>Expressions in Patterns</title>
      <p>An arithmetic expression can be used within a pattern if
        it meets both of the following two conditions:</p>
      <list type="bulleted">
        <item>It uses only numeric or bitwise operators.</item>
        <item>Its value can be evaluated to a constant when complied.</item>
      </list>
      <p><em>Example:</em></p>
      <pre>
case {Value, Result} of
    {?THRESHOLD+1, ok} -> ...</pre>
    </section>
  </section>

  <section>
    <marker id="match_operator"></marker>
    <title>The Match Operator</title>
    <p>The following matches <c>Pattern</c> against
      <c>Expr</c>:</p>
    <pre>
Pattern = Expr</pre>
    <p>If the matching succeeds, any unbound variable in the pattern
    becomes bound and the value of <c>Expr</c> is returned.</p>
    <p>If multiple match operators are applied in sequence, they will be
    evaluated from right to left.</p>
    <p>If the matching fails, a <c>badmatch</c> run-time error occurs.</p>
    <p><em>Examples:</em></p>
    <pre>
1> <input>{A, B} = T = {answer, 42}.</input>
{answer,42}
2> <input>A.</input>
answer
3> <input>B.</input>
42
4> <input>T.</input>
{answer,42}
5> <input>{C, D} = [1, 2].</input>
** exception error: no match of right-hand side value [1,2]</pre>

    <p>Because multiple match operators are evaluated from right to left,
    it means that:</p>

    <pre>
Pattern1 = Pattern2 = . . . = PatternN = Expression</pre>

    <p>is equivalent to:</p>
    <pre>
Temporary = Expression,
PatternN = Temporary,
   .
   .
   .,
Pattern2 = Temporary,
Pattern = Temporary</pre>
  </section>

  <section>
    <title>The Match Operator and the Compound Pattern Operator</title>
    <note><p>This is an advanced section, which references to topics not
    yet introduced. It can safely be skipped on a first
    reading.</p></note>

    <p>The <c>=</c> character is used to denote two similar but
    distinct operators: the match operator and the compound pattern
    operator. Which one is meant is determined by context.</p>

    <p>The <em>compound pattern operator</em> is used to construct a
    compound pattern from two patterns. Compound patterns are accepted
    everywhere a pattern is accepted. A compound pattern matches if
    all of its constituent patterns match. It is not legal for a
    pattern that is part of a compound pattern to use variables (as
    keys in map patterns or sizes in binary patterns) bound in other
    sub patterns of the same compound pattern.</p>
    <p><em>Examples:</em></p>

    <pre>
1> <input>fun(#{Key := Value} = #{key := Key}) -> Value end.</input>
* 1:7: variable 'Key' is unbound
2> <input>F = fun({A, B} = E) -> {E, A + B} end, F({1,2}).</input>
{{1,2},3}
3> <input>G = fun(&lt;&lt;A:8,B:8>> = &lt;&lt;C:16>>) -> {A, B, C} end, G(&lt;&lt;42,43>>).</input>
{42,43,10795}</pre>

    <p>The <em>match operator</em> is allowed everywhere an expression
    is allowed. It is used to match the value of an expression to a pattern.
    If multiple match operators are applied in sequence, they will be
    evaluated from right to left.</p>

    <p><em>Examples:</em></p>
    <pre>
1> <input>M = #{key => key2, key2 => value}.</input>
#{key => key2,key2 => value}
2> <input>f(Key), #{Key := Value} = #{key := Key} = M, Value.</input>
value
3> <input>f(Key), #{Key := Value} = (#{key := Key} = M), Value.</input>
value
4> <input>f(Key), (#{Key := Value} = #{key := Key}) = M, Value.</input>
* 1:12: variable 'Key' is unbound
5> <input>&lt;&lt;X:Y&gt;&gt; = begin Y = 8, &lt;&lt;42:8&gt;&gt; end, X.</input>
42</pre>

    <p>The expression at prompt <em>2&gt;</em> first matches the value of
    variable <c>M</c> against pattern <c>#{key := Key}</c>, binding
    variable <c>Key</c>. It then matches the value of <c>M</c> against
    pattern <c>#{Key := Value}</c> using variable <c>Key</c> as the
    key, binding variable <c>Value</c>.</p>

    <p>The expression at prompt <em>3&gt;</em> matches expression
    <c>(#{key := Key} = M)</c> against pattern <c>#{Key :=
    Value}</c>. The expression inside the parentheses is evaluated
    first. That is, <c>M</c> is matched against <c>#{key := Key}</c>,
    and then the value of <c>M</c> is matched against pattern <c>#{Key
    := Value}</c>. That is the same evaluation order as in <em>2</em>;
    therefore, the parentheses are redundant.</p>

    <p>In the expression at prompt <em>4&gt;</em> the expression <c>M</c>
    is matched against a pattern inside parentheses. Since the
    construct inside the parentheses is a pattern, the <c>=</c> that
    separates the two patterns is the compound pattern operator
    (<em>not</em> the match operator). The match fails because the two
    sub patterns are matched at the same time, and the variable
    <c>Key</c> is therefore not bound when matching against pattern
    <c>#{Key := Value}</c>.</p>

    <p>In the expression at prompt <em>5&gt;</em> the expressions
    inside the <seeguide marker="#block_expressions">block
    expression</seeguide> are evaluated first, binding variable
    <c>Y</c> and creating a binary.  The binary is then matched
    against pattern <c>&lt;&lt;X:Y&gt;&gt;</c> using the value of
    <c>Y</c> as the size of the segment.</p>
  </section>

  <section>
    <marker id="calls"></marker>
    <title>Function Calls</title>
    <pre>
ExprF(Expr1,...,ExprN)
ExprM:ExprF(Expr1,...,ExprN)</pre>
    <p>In the first form of function calls,
      <c>ExprM:ExprF(Expr1,...,ExprN)</c>, each of <c>ExprM</c> and
      <c>ExprF</c> must be an atom or an expression that evaluates to
      an atom. The function is said to be called by using the
      <em>fully qualified function name</em>. This is often referred
      to as a <em>remote</em> or <em>external function call</em>.</p>
      <p><em>Example:</em></p>
    <code type="none">
lists:keysearch(Name, 1, List)</code>
    <p>In the second form of function calls,
      <c>ExprF(Expr1,...,ExprN)</c>, <c>ExprF</c> must be an atom or
      evaluate to a fun.</p>

    <p>If <c>ExprF</c> is an atom, the function is said to be called by
      using the <em>implicitly qualified function name</em>.  If the
      function <c>ExprF</c> is locally defined, it is called.
      Alternatively, if <c>ExprF</c> is explicitly imported from the
      <c>M</c> module, <c>M:ExprF(Expr1,...,ExprN)</c> is called. If
      <c>ExprF</c> is neither declared locally nor explicitly
      imported, <c>ExprF</c> must be the name of an automatically
      imported BIF. </p>
      <p><em>Examples:</em></p>

    <code type="none">
handle(Msg, State)
spawn(m, init, [])</code>
    <p><em>Examples</em> where <c>ExprF</c> is a fun:</p>
    <pre>
1> <input>Fun1 = fun(X) -> X+1 end,</input>
<input>Fun1(3).</input>
4
2> <input>fun lists:append/2([1,2], [3,4]).</input>
[1,2,3,4]
3> </pre>

    <p>Notice that when calling a local function, there is a difference
    between using the implicitly or fully qualified function name.
    The latter always refers to the latest version of the module.
    See <seeguide marker="code_loading">Compilation and Code Loading
    </seeguide> and <seeguide marker="functions#eval">
    Function Evaluation</seeguide>.</p>

    <section>
      <title>Local Function Names Clashing With  Auto-Imported BIFs</title>
    <p>If a local function has the same name as an auto-imported BIF,
    the semantics is that implicitly qualified function calls are
    directed to the locally defined function, not to the BIF. To avoid
    confusion, there is a compiler directive available,
    <c>-compile({no_auto_import,[F/A]})</c>, that makes a BIF not
    being auto-imported. In certain situations, such a compile-directive
    is mandatory.</p>

    <change><p>Before OTP R14A (ERTS version 5.8), an implicitly
    qualified function call to a function having the same name as an
    auto-imported BIF always resulted in the BIF being called. In
    newer versions of the compiler, the local function is called instead.
    This is to avoid that future additions to the
    set of auto-imported BIFs do not silently change the behavior
    of old code.</p></change>

    <p>However, to avoid that old (pre R14) code changed its
    behavior when compiled with OTP version R14A or later, the
    following restriction applies: If you override the name of a BIF
    that was auto-imported in OTP versions prior to R14A (ERTS version
    5.8) and have an implicitly qualified call to that function in
    your code, you either need to explicitly remove the auto-import
    using a compiler directive, or replace the call with a fully
    qualified function call. Otherwise you get a compilation
    error. See the following example:</p>

    <code type="none">
-export([length/1,f/1]).

-compile({no_auto_import,[length/1]}). % erlang:length/1 no longer autoimported

length([]) ->
    0;
length([H|T]) ->
    1 + length(T). %% Calls the local function length/1

f(X) when erlang:length(X) > 3 -> %% Calls erlang:length/1,
                                  %% which is allowed in guards
    long.</code>

    <p>The same logic applies to explicitly imported functions from
    other modules, as to locally defined functions.
    It is not allowed to both import a
    function from another module and have the function declared in the
    module at the same time:</p>

    <code type="none">
-export([f/1]).

-compile({no_auto_import,[length/1]}). % erlang:length/1 no longer autoimported

-import(mod,[length/1]).

f(X) when erlang:length(X) > 33 -> %% Calls erlang:length/1,
                                   %% which is allowed in guards

    erlang:length(X);              %% Explicit call to erlang:length in body

f(X) ->
    length(X).                     %% mod:length/1 is called</code>


    <p>For auto-imported BIFs added in Erlang/OTP R14A and thereafter,
    overriding the name with a local function or explicit import is always
    allowed. However, if the <c>-compile({no_auto_import,[F/A])</c>
    directive is not used, the compiler issues a warning whenever
    the function is called in the module using the implicitly qualified
    function name.</p>
    </section>
  </section>

  <section>
    <title>If</title>
    <pre>
if
    GuardSeq1 ->
        Body1;
    ...;
    GuardSeqN ->
        BodyN
end</pre>
    <p>The branches of an <c>if</c>-expression are scanned sequentially
      until a guard sequence <c>GuardSeq</c> that evaluates to true is
      found. Then the corresponding <c>Body</c> (sequence of expressions
      separated by ',') is evaluated.</p>
    <p>The return value of <c>Body</c> is the return value of
      the <c>if</c> expression.</p>
    <p>If no guard sequence is evaluated as true,
      an <c>if_clause</c> run-time error
      occurs. If necessary, the guard expression <c>true</c> can be
      used in the last branch, as that guard sequence is always true.</p>
    <p><em>Example:</em></p>
    <pre>
is_greater_than(X, Y) ->
    if
        X>Y ->
            true;
        true -> % works as an 'else' branch
            false
    end</pre>
  </section>

  <section>
    <marker id="case"></marker>
    <title>Case</title>
    <pre>
case Expr of
    Pattern1 [when GuardSeq1] ->
        Body1;
    ...;
    PatternN [when GuardSeqN] ->
        BodyN
end</pre>
    <p>The expression <c>Expr</c> is evaluated and the patterns
      <c>Pattern</c> are sequentially matched against the result. If a
      match succeeds and the optional guard sequence <c>GuardSeq</c> is
      true, the corresponding <c>Body</c> is evaluated.</p>
    <p>The return value of <c>Body</c> is the return value of
      the <c>case</c> expression.</p>
    <p>If there is no matching pattern with a true guard sequence,
      a <c>case_clause</c> run-time error occurs.</p>
    <p><em>Example:</em></p>
    <pre>
is_valid_signal(Signal) ->
    case Signal of
        {signal, _What, _From, _To} ->
            true;
        {signal, _What, _To} ->
            true;
        _Else ->
            false
    end.</pre>
  </section>

  <section>
    <marker id="maybe"></marker>
    <title>Maybe</title>
    <change><p><c>maybe</c> is an experimental <seeguide
    marker="system/reference_manual:features#features">feature</seeguide>
    introduced in Erlang/OTP 25. By default, it is disabled. To enable
    <c>maybe</c>, either use the <c>-feature(maybe_expr,enable)</c>
    directive (from within source code), or the compiler option
    <c>{feature,maybe_expr,enable}</c>.</p>
    </change>

    <code type="erl"><![CDATA[
maybe
    Expr1,
    ...,
    ExprN
end]]></code>

    <p>The expressions in a <c>maybe</c> block are evaluated sequentially. If all
    expressions are evaluated successfully, the return value of the <c>maybe</c>
    block is <c>ExprN</c>. However, execution can be short-circuited by a
    conditional match expression:</p>

    <code type="erl"><![CDATA[
Expr1 ?= Expr2]]></code>

    <p><c>?=</c> is called the conditional match operator. It is only
    allowed to be used at the top-level of a <c>maybe</c> block. It
    matches the pattern <c>Expr1</c> against <c>Expr2</c>. If the
    matching succeeds, any unbound variable in the pattern becomes
    bound. If the expression is the last expression in the
    <c>maybe</c> block, it also returns the value of <c>Expr2</c>. If the
    matching is unsuccessful, the rest of the expressions in the <c>maybe</c>
    block are skipped and the return value of the <c>maybe</c>
    block is <c>Expr2</c>.</p>

    <p>None of the variables bound in a <c>maybe</c> block must be
    used in the code that follows the block.</p>

    <p>Here is an example:</p>

    <code type="erl"><![CDATA[
maybe
    {ok, A} ?= a(),
    true = A >= 0,
    {ok, B} ?= b(),
    A + B
end]]></code>

    <p>Let us first assume that <c>a()</c> returns <c>{ok,42}</c> and
    <c>b()</c> returns <c>{ok,58}</c>. With those return values, all
    of the match operators will succeed, and the return value of the
    <c>maybe</c> block is <c>A + B</c>, which is equal to <c>42 +
    58 = 100</c>.</p>

    <p>Now let us assume that <c>a()</c> returns <c>error</c>. The
    conditional match operator in <c>{ok, A} ?= a()</c> fails to
    match, and the return value of the <c>maybe</c> block is the
    value of the expression that failed to match, namely <c>error</c>.
    Similarly, if <c>b()</c> returns <c>wrong</c>, the return value of
    the <c>maybe</c> block is <c>wrong</c>.</p>

    <p>Finally, let us assume that <c>a()</c> returns
    <c>-1</c>. Because <c>true = A >= 0</c> uses the match operator
    `=`, a <c>{badmatch,false}</c> run-time error occurs when the
    expression fails to match the pattern.</p>

    <p>The example can be written in a less succient way using nested
    case expressions:</p>

    <code type="erl"><![CDATA[
case a() of
    {ok, A} ->
        true = A >= 0,
        case b() of
            {ok, B} ->
                A + B;
            Other1 ->
                Other1
        end;
    Other2 ->
        Other2
end]]></code>

    <p>The <c>maybe</c> block can be augmented with <c>else</c> clauses:</p>

    <code type="erl"><![CDATA[
maybe
    Expr1,
    ...,
    ExprN
else
    Pattern1 [when GuardSeq1] ->
        Body1;
    ...;
    PatternN [when GuardSeqN] ->
        BodyN
end]]></code>

    <p>If a conditional match operator fails, the failed expression is
    matched against the patterns in all clauses between the
    <c>else</c> and <c>end</c> keywords. If a match succeeds and the
    optional guard sequence <c>GuardSeq</c> is true, the corresponding
    <c>Body</c> is evaluated. The value returned from the body is the
    return value of the <c>maybe</c> block.</p>

    <p>If there is no matching pattern with a true guard sequence,
    an <c>else_clause</c> run-time error occurs.</p>

    <p>None of the variables bound in a <c>maybe</c> block must be used in
    the <c>else</c> clauses. None of the variables bound in the <c>else</c> clauses
    must be used in the code that follows the <c>maybe</c> block.</p>

    <p>Here is the previous example augmented with a <c>else</c> clauses:</p>

    <code type="erl"><![CDATA[
maybe
    {ok, A} ?= a(),
    true = A >= 0,
    {ok, B} ?= b(),
    A + B
else
    error -> error;
    wrong -> error
end]]></code>

    <p>The <c>else</c> clauses translate the failing value from
    the conditional match operators to the value <c>error</c>. If the
    failing value is not one of the recognized values, a
    <c>else_clause</c> run-time error occurs.</p>
  </section>

  <section>
    <marker id="send"></marker>
    <title>Send</title>
    <pre>
Expr1 ! Expr2</pre>
    <p>Sends the value of <c>Expr2</c> as a message to the process
      specified by <c>Expr1</c>. The value of <c>Expr2</c> is also
      the return value of the expression.</p>
    <p><c>Expr1</c> must evaluate to a pid, an alias (reference),
      a port, a registered name (atom), or
      a tuple <c>{Name,Node}</c>. <c>Name</c> is an atom and
      <c>Node</c> is a node name, also an atom.</p>
    <list type="bulleted">
      <item>If <c>Expr1</c> evaluates to a name, but this name is not
       registered, a <c>badarg</c> run-time error occurs.</item>
      <item>Sending a message to a reference never fails, even if the
       reference is no longer (or never was) an alias.</item>
      <item>Sending a message to a pid never fails, even if the pid
       identifies a non-existing process.</item>
      <item>Distributed message sending, that is, if <c>Expr1</c>
       evaluates to a tuple <c>{Name,Node}</c> (or a pid located at
       another node), also never fails.</item>
    </list>
  </section>

  <section>
    <marker id="receive"></marker>
    <title>Receive</title>
    <pre>
receive
    Pattern1 [when GuardSeq1] ->
        Body1;
    ...;
    PatternN [when GuardSeqN] ->
        BodyN
end</pre>
    <p>
      Fetches a received message present in the message queue of
      the process. The first message in the message queue is matched
      sequentially against the patterns from top to bottom. If no
      match was found, the matching sequence is repeated for the second
      message in the queue, and so on. Messages are queued in the
      <seeguide marker="processes#message-queue-order">order they were
      received</seeguide>. If a match succeeds, that
      is, if the <c>Pattern</c> matches and the optional guard sequence
      <c>GuardSeq</c> is true, then the message is removed from the message
      queue and the corresponding <c>Body</c> is evaluated. All other
      messages in the message queue remain unchanged.
    </p>
    <p>The return value of <c>Body</c> is the return value of
      the <c>receive</c> expression.</p>
    <p><c>receive</c> never fails. The execution is suspended, possibly
      indefinitely, until a message arrives that matches one of
      the patterns and with a true guard sequence. </p>
    <p><em>Example:</em></p>
    <pre>
wait_for_onhook() ->
    receive
        onhook ->
            disconnect(),
            idle();
        {connect, B} ->
            B ! {busy, self()},
            wait_for_onhook()
    end.</pre>
    <p>The <c>receive</c> expression can be augmented with a
      timeout:</p>
    <pre>
receive
    Pattern1 [when GuardSeq1] ->
        Body1;
    ...;
    PatternN [when GuardSeqN] ->
        BodyN
after
    ExprT ->
        BodyT
end</pre>
    <p><c>receive..after</c> works exactly as <c>receive</c>, except
      that if no matching message has arrived within <c>ExprT</c>
      milliseconds, then <c>BodyT</c> is evaluated instead. The
      return value of <c>BodyT</c> then becomes the return value
      of the <c>receive..after</c> expression. <c>ExprT</c> is to
      evaluate to an integer, or the atom <c>infinity</c>. The allowed
      integer range is from 0 to 4294967295, that is, the longest possible
      timeout is almost 50 days. With a zero value the timeout occurs
      immediately if there is no matching message in the message queue.
    </p>
    <p>
      The atom <c>infinity</c> will make the process wait indefinitely for a
      matching message. This is the same as not using a timeout. It can be
      useful for timeout values that are calculated at runtime.
    </p>
    <p><em>Example:</em></p>
    <pre>
wait_for_onhook() ->
    receive
        onhook ->
            disconnect(),
            idle();
        {connect, B} ->
            B ! {busy, self()},
            wait_for_onhook()
    after
        60000 ->
            disconnect(),
            error()
    end.</pre>
    <p>It is legal to use a <c>receive..after</c> expression with no
      branches:</p>
    <pre>
receive
after
    ExprT ->
        BodyT
end</pre>
    <p>This construction does not consume any messages, only suspends
      execution in the process for <c>ExprT</c> milliseconds. This can be
      used to implement simple timers.</p>
    <p><em>Example:</em></p>
    <pre>
timer() ->
    spawn(m, timer, [self()]).

timer(Pid) ->
    receive
    after
        5000 ->
            Pid ! timeout
    end.</pre>
  </section>

  <section>
    <title>Term Comparisons</title>
    <pre>
Expr1 <input>op</input> Expr2</pre>
    <table>
      <row>
        <cell align="left" valign="middle"><em>op</em></cell>
        <cell align="left" valign="middle"><em>Description</em></cell>
      </row>
      <row>
        <cell align="left" valign="middle">==</cell>
        <cell align="left" valign="middle">Equal to</cell>
      </row>
      <row>
        <cell align="left" valign="middle">/=</cell>
        <cell align="left" valign="middle">Not equal to</cell>
      </row>
      <row>
        <cell align="left" valign="middle">=&lt;</cell>
        <cell align="left" valign="middle">Less than or equal to</cell>
      </row>
      <row>
        <cell align="left" valign="middle">&lt;</cell>
        <cell align="left" valign="middle">Less than</cell>
      </row>
      <row>
        <cell align="left" valign="middle">&gt;=</cell>
        <cell align="left" valign="middle">Greater than or equal to</cell>
      </row>
      <row>
        <cell align="left" valign="middle">&gt;</cell>
        <cell align="left" valign="middle">Greater than</cell>
      </row>
      <row>
        <cell align="left" valign="middle">=:=</cell>
        <cell align="left" valign="middle">Exactly equal to</cell>
      </row>
      <row>
        <cell align="left" valign="middle">=/=</cell>
        <cell align="left" valign="middle">Exactly not equal to</cell>
      </row>
      <tcaption>Term Comparison Operators.</tcaption>
    </table>
    <p>The arguments can be of different data types. The following
      order is defined:</p>
    <pre>
number &lt; atom &lt; reference &lt; fun &lt; port &lt; pid &lt; tuple &lt; map &lt; nil &lt; list &lt; bit string</pre>
    <p><c>nil</c> in the previous expression represents the empty list
      (<c>[]</c>), which is regarded as a separate type from
      <c>list/0</c>. That is why <c>nil &lt; list</c>.
    </p>
    <p>Lists are compared element by element. Tuples are ordered by
      size, two tuples with the same size are compared element by
      element.</p>
    <p>Bit strings are compared bit by bit. If one bit string is a
      prefix of the other, the shorter bit string is considered smaller.</p>
    <p>Maps are ordered by size, two maps with the same size are compared by keys in
        ascending term order and then by values in key order.
        In maps key order integers types are considered less than floats types.
    </p>
    <p>Atoms are compared using their string value, codepoint by codepoint.</p>
      <p>When comparing an integer to a float, the term with the lesser
      precision is converted into the type of the other term, unless the
      operator is one of <c>=:=</c> or <c>=/=</c>. A float is more precise than
      an integer until all significant figures of the float are to the left of
      the decimal point. This happens when the float is larger/smaller than
      +/-9007199254740992.0. The conversion strategy is changed
      depending on the size of the float because otherwise comparison of large
      floats and integers would lose their transitivity.</p>

    <p>Term comparison operators return the Boolean value of the
      expression, <c>true</c> or <c>false</c>.</p>

    <p><em>Examples:</em></p>
    <pre>
1> <input>1==1.0.</input>
true
2> <input>1=:=1.0.</input>
false
3> <input>1 > a.</input>
false
4> <input>#{c => 3} > #{a => 1, b => 2}.</input>
false
5> <input>#{a => 1, b => 2} == #{a => 1.0, b => 2.0}.</input>
true
6> <input>&lt;&lt;2:2>> &lt; &lt;&lt;128>>.</input>
true
7> <input>&lt;&lt;3:2>> &lt; &lt;&lt;128>>.</input>
false
</pre>
  </section>

  <section>
    <title>Arithmetic Expressions</title>
    <pre>
<input>op</input> Expr
Expr1 <input>op</input> Expr2</pre>
    <table>
      <row>
        <cell align="left" valign="middle"><em>Operator</em></cell>
        <cell align="left" valign="middle"><em>Description</em></cell>
        <cell align="left" valign="middle"><em>Argument Type</em></cell>
      </row>
      <row>
        <cell align="left" valign="middle">+</cell>
        <cell align="left" valign="middle">Unary +</cell>
        <cell align="left" valign="middle">Number</cell>
      </row>
      <row>
        <cell align="left" valign="middle">-</cell>
        <cell align="left" valign="middle">Unary -</cell>
        <cell align="left" valign="middle">Number</cell>
      </row>
      <row>
        <cell align="left" valign="middle">+</cell>
        <cell align="left" valign="middle">&nbsp;</cell>
        <cell align="left" valign="middle">number</cell>
      </row>
      <row>
        <cell align="left" valign="middle">-</cell>
        <cell align="left" valign="middle">&nbsp;</cell>
        <cell align="left" valign="middle">Number</cell>
      </row>
      <row>
        <cell align="left" valign="middle">*</cell>
        <cell align="left" valign="middle">&nbsp;</cell>
        <cell align="left" valign="middle">Number</cell>
      </row>
      <row>
        <cell align="left" valign="middle">/</cell>
        <cell align="left" valign="middle">Floating point division</cell>
        <cell align="left" valign="middle">Number</cell>
      </row>
      <row>
        <cell align="left" valign="middle">bnot</cell>
        <cell align="left" valign="middle">Unary bitwise NOT</cell>
        <cell align="left" valign="middle">Integer</cell>
      </row>
      <row>
        <cell align="left" valign="middle">div</cell>
        <cell align="left" valign="middle">Integer division</cell>
        <cell align="left" valign="middle">Integer</cell>
      </row>
      <row>
        <cell align="left" valign="middle">rem</cell>
        <cell align="left" valign="middle">Integer remainder of X/Y</cell>
        <cell align="left" valign="middle">Integer</cell>
      </row>
      <row>
        <cell align="left" valign="middle">band</cell>
        <cell align="left" valign="middle">Bitwise AND</cell>
        <cell align="left" valign="middle">Integer</cell>
      </row>
      <row>
        <cell align="left" valign="middle">bor</cell>
        <cell align="left" valign="middle">Bitwise OR</cell>
        <cell align="left" valign="middle">Integer</cell>
      </row>
      <row>
        <cell align="left" valign="middle">bxor</cell>
        <cell align="left" valign="middle">Arithmetic bitwise XOR</cell>
        <cell align="left" valign="middle">Integer</cell>
      </row>
      <row>
        <cell align="left" valign="middle">bsl</cell>
        <cell align="left" valign="middle">Arithmetic bitshift left</cell>
        <cell align="left" valign="middle">Integer</cell>
      </row>
      <row>
        <cell align="left" valign="middle">bsr</cell>
        <cell align="left" valign="middle">Bitshift right</cell>
        <cell align="left" valign="middle">Integer</cell>
      </row>
      <tcaption>Arithmetic Operators.</tcaption>
    </table>

    <p><em>Examples:</em></p>
    <pre>
1> <input>+1.</input>
1
2> <input>-1.</input>
-1
3> <input>1+1.</input>
2
4> <input>4/2.</input>
2.0
5> <input>5 div 2.</input>
2
6> <input>5 rem 2.</input>
1
7> <input>2#10 band 2#01.</input>
0
8> <input>2#10 bor 2#01.</input>
3
9> <input>a + 10.</input>
** exception error: an error occurred when evaluating an arithmetic expression
     in operator  +/2
        called as a + 10
10> <input>1 bsl (1 bsl 64).</input>
** exception error: a system limit has been reached
     in operator  bsl/2
        called as 1 bsl 18446744073709551616</pre>
  </section>

  <section>
    <title>Boolean Expressions</title>
    <pre>
<input>op</input> Expr
Expr1 <input>op</input> Expr2</pre>
    <table>
      <row>
        <cell align="left" valign="middle"><em>Operator</em></cell>
        <cell align="left" valign="middle"><em>Description</em></cell>
      </row>
      <row>
        <cell align="left" valign="middle">not</cell>
        <cell align="left" valign="middle">Unary logical NOT</cell>
      </row>
      <row>
        <cell align="left" valign="middle">and</cell>
        <cell align="left" valign="middle">Logical AND</cell>
      </row>
      <row>
        <cell align="left" valign="middle">or</cell>
        <cell align="left" valign="middle">Logical OR</cell>
      </row>
      <row>
        <cell align="left" valign="middle">xor</cell>
        <cell align="left" valign="middle">Logical XOR</cell>
      </row>
      <tcaption>Logical Operators.</tcaption>
    </table>
    <p><em>Examples:</em></p>
    <pre>
1> <input>not true.</input>
false
2> <input>true and false.</input>
false
3> <input>true xor false.</input>
true
4> <input>true or garbage.</input>
** exception error: bad argument
     in operator  or/2
        called as true or garbage</pre>
  </section>

  <section>
    <title>Short-Circuit Expressions</title>
    <pre>
Expr1 orelse Expr2
Expr1 andalso Expr2</pre>
    <p><c>Expr2</c> is evaluated only if
      necessary. That is, <c>Expr2</c> is evaluated only if:</p>
    <list type="bulleted">
      <item><p><c>Expr1</c> evaluates to <c>false</c> in an
      <c>orelse</c> expression.</p>
      </item>
    </list>
      <p>or</p>
    <list type="bulleted">
      <item><p><c>Expr1</c> evaluates to <c>true</c> in an
      <c>andalso</c> expression.</p>
      </item>
    </list>
    <p>Returns either the value of <c>Expr1</c> (that is,
      <c>true</c> or <c>false</c>) or the value of <c>Expr2</c>
      (if <c>Expr2</c> is evaluated).</p>

    <p><em>Example 1:</em></p>
    <pre>
case A >= -1.0 andalso math:sqrt(A+1) > B of</pre>
    <p>This works even if <c>A</c> is less than <c>-1.0</c>,
      since in that case, <c>math:sqrt/1</c> is never evaluated.</p>
    <p><em>Example 2:</em></p>
    <pre>
OnlyOne = is_atom(L) orelse
         (is_list(L) andalso length(L) == 1),</pre>

    <p><c>Expr2</c> is not required to evaluate to a Boolean
    value. Because of that, <c>andalso</c> and <c>orelse</c> are
    tail-recursive.</p>

    <p><em>Example 3 (tail-recursive function):</em></p>
    <pre>
all(Pred, [Hd|Tail]) ->
    Pred(Hd) andalso all(Pred, Tail);
all(_, []) ->
    true.</pre>

    <change><p>Before Erlang/OTP R13A, <c>Expr2</c> was required to
    evaluate to a Boolean value, and as consequence, <c>andalso</c>
    and <c>orelse</c> were <strong>not</strong>
    tail-recursive.</p></change>
  </section>

  <section>
    <title>List Operations</title>
    <pre>
Expr1 ++ Expr2
Expr1 -- Expr2</pre>
    <p>The list concatenation operator <c>++</c> appends its second
      argument to its first and returns the resulting list.</p>
    <p>The list subtraction operator <c>--</c> produces a list that
      is a copy of the first argument. The procedure is as follows:
      for each element in the second argument, the first
      occurrence of this element (if any) is removed.</p>
    <p><em>Example:</em></p>
    <pre>
1> <input>[1,2,3]++[4,5].</input>
[1,2,3,4,5]
2> <input>[1,2,3,2,1,2]--[2,1,2].</input>
[3,1,2]</pre>

   </section>

  <section>
    <marker id="map_expressions"></marker>
	<title>Map Expressions</title>
	  <section>
	  <title>Creating Maps</title>
	  <p>
		  Constructing a new map is done by letting an expression <c>K</c> be associated with
		  another expression <c>V</c>:
	  </p>
	  <code>#{ K => V }</code>
	  <p>
		  New maps can include multiple associations at construction by listing every
		  association:
	  </p>
	  <code>#{ K1 => V1, .., Kn => Vn }</code>
	  <p>
		  An empty map is constructed by not associating any terms with each other:
	  </p>
	  <code>#{}</code>
	  <p>
		  All keys and values in the map are terms. Any expression is first evaluated and
		  then the resulting terms are used as <em>key</em> and <em>value</em> respectively.
	  </p>
	  <p>
		  Keys and values are separated by the <c>=></c> arrow and associations are
		  separated by a comma <c>,</c>.
	  </p>

	  <p>
		  <em>Examples:</em>
	  </p>
	  <code>
M0 = #{},                 % empty map
M1 = #{a => &lt;&lt;"hello"&gt;&gt;}, % single association with literals
M2 = #{1 => 2, b => b},   % multiple associations with literals
M3 = #{k => {A,B}},       % single association with variables
M4 = #{{"w", 1} => f()}.  % compound key associated with an evaluated expression</code>
	  <p>
		  Here, <c>A</c> and <c>B</c> are any expressions and <c>M0</c> through <c>M4</c>
		  are the resulting map terms.
	  </p>
	  <p>
		  If two matching keys are declared, the latter key takes precedence.
	  </p>
	  <p>
		  <em>Example:</em>
	  </p>

<pre>
1> <input>#{1 => a, 1 => b}.</input>
#{1 => b }
2> <input>#{1.0 => a, 1 => b}.</input>
#{1 => b, 1.0 => a}
</pre>
	  <p>
		  The order in which the expressions constructing the keys (and their
		  associated values) are evaluated is not defined. The syntactic order of
		  the key-value pairs in the construction is of no relevance, except in
		  the recently mentioned case of two matching keys.
	  </p>
  </section>

  <section>
	  <title>Updating Maps</title>
	  <p>
		  Updating a map has a similar syntax as constructing it.
	  </p>
	  <p>
		  An expression defining the map to be updated, is put in front of the expression
		  defining the keys to be updated and their respective values:
	  </p>
	  <code>M#{ K => V }</code>
	  <p>
		  Here <c>M</c> is a term of type map and <c>K</c> and <c>V</c> are any expression.
	  </p>
	  <p>
		  If key <c>K</c> does not match any existing key in the map, a new association
		  is created from key <c>K</c> to value <c>V</c>.
	  </p>
	  <p>     If key <c>K</c> matches an existing key in map <c>M</c>,
                  its associated value
	          is replaced by the new value <c>V</c>. In both cases, the evaluated map expression
		  returns a new map.
	  </p>
	  <p>
		  If <c>M</c> is not of type map, an exception of type <c>badmap</c> is thrown.
	  </p>
	  <p>
		  To only update an existing value, the following syntax is used:
	  </p>
	  <code>M#{ K := V } </code>
	  <p>
		  Here <c>M</c> is a term of type map, <c>V</c> is an expression and <c>K</c>
		  is an expression that evaluates to an existing key in <c>M</c>.
	  </p>
	  <p>
		  If key <c>K</c> does not match any existing keys in map <c>M</c>, an exception
		  of type <c>badkey</c> is triggered at runtime. If a matching key <c>K</c>
		  is present in map <c>M</c>, its associated value is replaced by the new
		  value <c>V</c>, and the evaluated map expression returns a new map.
	  </p>
	  <p>
		  If <c>M</c> is not of type map, an exception of type <c>badmap</c> is thrown.
	  </p>
	  <p>
		  <em>Examples:</em>
	  </p>
	  <code>
M0 = #{},
M1 = M0#{a => 0},
M2 = M1#{a => 1, b => 2},
M3 = M2#{"function" => fun() -> f() end},
M4 = M3#{a := 2, b := 3}.  % 'a' and 'b' was added in `M1` and `M2`.</code>
	  <p>
		  Here <c>M0</c> is any map. It follows that <c>M1 .. M4</c> are maps as well.
	  </p>
	  <p>
		  <em>More examples:</em>
	  </p>
<pre>
1> <input>M = #{1 => a}.</input>
#{1 => a }
2> <input>M#{1.0 => b}.</input>
#{1 => a, 1.0 => b}.
3> <input>M#{1 := b}.</input>
#{1 => b}
4> <input>M#{1.0 := b}.</input>
** exception error: bad argument
</pre>
	  <p>
		  As in construction, the order in which the key and value expressions
		  are evaluated is not defined. The
		  syntactic order of the key-value pairs in the update is of no
		  relevance, except in the case where two keys match.
		  In that case, the latter value is used.
	  </p>
  </section>

  <section>
	  <title>Maps in Patterns</title>
	  <p>
		  Matching of key-value associations from maps is done as follows:
	  </p>

	  <code>#{ K := V } = M</code>
	  <p>
	    Here <c>M</c> is any map. The key <c>K</c> must be a
	    <seeguide marker="#guard_expressions">guard
	    expression</seeguide>, with all variables already bound.
	    <c>V</c> can be any pattern with either bound or unbound
	    variables.
	  </p>
	  <p>
		  If the variable <c>V</c> is unbound, it becomes bound to the value associated
		  with the key <c>K</c>, which must exist in the map <c>M</c>. If the variable
		  <c>V</c> is bound, it must match the value associated with <c>K</c> in <c>M</c>.
	  </p>
	  <change><p>Before Erlang/OTP 23, the expression defining the key
	  <c>K</c> was restricted to be either a single variable or a
	  literal.</p></change>
	  <p><em>Example:</em></p>
<pre>
1> <input>M = #{"tuple" => {1,2}}.</input>
#{"tuple" => {1,2}}
2> <input>#{"tuple" := {1,B}} = M.</input>
#{"tuple" => {1,2}}
3> <input>B.</input>
2.</pre>
	  <p>
		  This binds variable <c>B</c> to integer <c>2</c>.
	  </p>
	  <p>
		  Similarly, multiple values from the map can be matched:
	  </p>
	  <code>#{ K1 := V1, .., Kn := Vn } = M</code>
	  <p>
		  Here keys <c>K1 .. Kn</c> are any expressions with
		  literals or bound variables. If all key expressions
		  evaluate successfully and all keys exist in map
		  <c>M</c>, all variables in <c>V1 .. Vn</c> is
		  matched to the associated values of their respective
		  keys.
	  </p>
	  <p>
		  If the matching conditions are not met, the match fails, either with:
	  </p>
	  <list>
		  <item><p>A <c>badmatch</c> exception.</p>
		        <p>This is if it is used in the context of the match operator
			  as in the example.</p>
		  </item>
		  <item><p>Or resulting in the next clause being tested in function heads and
			  case expressions.</p>
		  </item>
	  </list>
	  <p>
		  Matching in maps only allows for <c>:=</c> as delimiters of associations.
	  </p>
	   <p>
		  The order in which keys are declared in matching has no relevance.
	  </p>
	  <p>
		  Duplicate keys are allowed in matching and match each pattern associated
		  to the keys:
	  </p>
	  <code>#{ K := V1, K := V2 } = M</code>
	  <p>
		  Matching an expression against an empty map literal, matches its type but
		  no variables are bound:
	  </p>
	  <code>#{} = Expr</code>
	  <p>
		  This expression matches if the expression <c>Expr</c> is of type map, otherwise
		  it fails with an exception <c>badmatch</c>.
	  </p>
	  <p>Here the key to be retrieved is constructed from an expression:</p>
	  <code>#{{tag,length(List)} := V} = Map</code>
	  <p><c>List</c> must be an already bound variable.</p>
	  <section>
		  <title>Matching Syntax</title>
		  <p>
			  Matching of literals as keys are allowed in function heads:
		  </p>
		  <code>
%% only start if not_started
handle_call(start, From, #{ state := not_started } = S) ->
...
    {reply, ok, S#{ state := start }};

%% only change if started
handle_call(change, From, #{ state := start } = S) ->
...
    {reply, ok, S#{ state := changed }};</code>
	  </section>
  </section>
  <section>
	  <title>Maps in Guards</title>
	  <p>
		  Maps are allowed in guards as long as all subexpressions are valid guard expressions.
	  </p>
	  <p>
		  The following guard BIFs handle maps:
	  </p>
	  <list>
		  <item>
			  <seemfa marker="erts:erlang#is_map/1">is_map/1</seemfa>
			   in the <c>erlang</c> module
		  </item>
		  <item>
			  <seemfa marker="erts:erlang#is_map_key/2">is_map_key/2</seemfa>
			   in the <c>erlang</c> module
		  </item>
		  <item>
			  <seemfa marker="erts:erlang#map_get/2">map_get/2</seemfa>
			   in the <c>erlang</c> module
		  </item>
		  <item>
			  <seemfa marker="erts:erlang#map_size/1">map_size/1</seemfa>
			   in the <c>erlang</c> module
		  </item>
	  </list>
	</section>
  </section>

  <section>
    <marker id="bit_syntax"></marker>
    <title>Bit Syntax Expressions</title>
    <p>
      The bit syntax operates on <em>bit strings</em>.
      A bit string is a sequence of bits ordered
      from the most significant bit to the least significant bit.
    </p>
    <code type="none"><![CDATA[<<>>  % The empty bit string, zero length
<<E1>>
<<E1,...,En>>]]></code>
    <p>
      Each element <c>Ei</c> specifies a <em>segment</em> of
      the bit string.  The segments are ordered left to right
      from the most significant bit to the least significant bit
      of the bit string.
    </p>
    <p>
      Each segment specification <c>Ei</c> is a value, followed by an
      optional <em>size expression</em>
      and an optional <em>type specifier list</em>.
    </p>
    <pre>
Ei = Value |
     Value:Size |
     Value/TypeSpecifierList |
     Value:Size/TypeSpecifierList</pre>
    <p>When used in a bit string construction, <c>Value</c> is an expression
    that is to evaluate to an integer, float, or bit string.  If the
    expression is not a single literal or variable, it
    is to be enclosed in parentheses.</p>

    <p>When used in a bit string matching, <c>Value</c> must be a variable,
    or an integer, float, or string.</p>

    <p>Notice that, for example, using a string literal as in
    <c><![CDATA[<<"abc">>]]></c> is syntactic sugar for
    <c><![CDATA[<<$a,$b,$c>>]]></c>.</p>

    <p>When used in a bit string construction, <c>Size</c> is an expression
    that is to evaluate to an integer.</p>
    
    <p>When used in a bit string matching, <c>Size</c> must be a
    <seeguide marker="#guard_expressions">guard expression</seeguide>
    that evaluates to an integer. All variables in the guard expression
    must be already bound.</p>

    <change><p>Before Erlang/OTP 23, <c>Size</c> was restricted to be an
    integer or a variable bound to an integer.</p></change>

    <p>The value of <c>Size</c> specifies the size of the segment in
    units (see below). The default value depends on the type (see
    below):</p>
    <list type="bulleted">
      <item>For <c>integer</c> it is 8.</item>
      <item>For <c>float</c> it is 64.</item>
      <item>For <c>binary</c> and <c>bitstring</c> it is
      the whole binary or bit string.</item>
    </list>
    <p>In matching, the default value for a binary or bit string
    segment is only valid for the last element. All other bit string
    or binary elements in the matching must have a size
    specification.</p>

    <marker id="binaries"></marker>
    <p><strong>Binaries</strong></p>
    <p>
      A bit string with a length that is a multiple of 8 bits
      is known as a <em>binary</em>, which is the most
      common and useful type of bit string.
    </p>
    <p>
      A binary has a canonical representation in memory.
      Here follows a sequence of bytes where each byte&apos;s
      value is its sequence number:
    </p>
    <pre>&lt;&lt;1, 2, 3, 4, 5, 6, 7, 8, 9, 10&gt;&gt;</pre>
    <p>
      Bit strings are a later generalization of binaries,
      so many texts and much information about binaries
      apply just as well for bit strings.
    </p>

    <p><strong>Example:</strong></p>
    <pre>
1> <input>&lt;&lt;A/binary, B/binary>> = &lt;&lt;"abcde">>.</input>
* 1:3: a binary field without size is only allowed at the end of a binary pattern
2> <input>&lt;&lt;A:3/binary, B/binary>> = &lt;&lt;"abcde">>.</input>
&lt;&lt;"abcde">>
3> <input>A.</input>
&lt;&lt;"abc">>
4> <input>B.</input>
&lt;&lt;"de">></pre>

    <p>For the <c>utf8</c>, <c>utf16</c>, and <c>utf32</c> types,
    <c>Size</c> must not be given. The size of the segment is implicitly
    determined by the type and value itself.</p>

    <p><c>TypeSpecifierList</c> is a list of type specifiers, in any
    order, separated by hyphens (-). Default values are used for any
    omitted type specifiers.</p>
    <taglist>
      <tag><c>Type</c>= <c>integer</c> | <c>float</c> | <c>binary</c> |
             <c>bytes</c> | <c>bitstring</c> | <c>bits</c> |
	     <c>utf8</c> | <c>utf16</c> | <c>utf32</c> </tag>
      <item>The default is <c>integer</c>. <c>bytes</c> is a shorthand for 
      <c>binary</c> and <c>bits</c> is a shorthand for <c>bitstring</c>.
      See below for more information about the <c>utf</c> types.
      </item>

      <tag><c>Signedness</c>= <c>signed</c> | <c>unsigned</c></tag>
      <item>Only matters for matching and when the type is <c>integer</c>. 
      The default is <c>unsigned</c>.</item>

      <tag><c>Endianness</c>= <c>big</c> | <c>little</c> | <c>native</c></tag>
      <item>
        Specifies byte level (octet level) endianness (byte order).
        Native-endian means that the endianness is resolved at load
        time to be either big-endian or little-endian, depending on
        what is native for the CPU that the Erlang machine is run on.
        Endianness only matters when the Type is either <c>integer</c>,
        <c>utf16</c>, <c>utf32</c>, or <c>float</c>. The default is <c>big</c>.
        <pre>&lt;&lt;16#1234:16/little>> = &lt;&lt;16#3412:16>> = &lt;&lt;16#34:8, 16#12:8>></pre>
      </item>

      <tag><c>Unit</c>= <c>unit:IntegerLiteral</c></tag>
      <item>The allowed range is 1 through 256. Defaults to 1 for <c>integer</c>,
      <c>float</c>, and <c>bitstring</c>, and to 8 for <c>binary</c>.
      For types <c>bitstring</c>, <c>bits</c>, and <c>bytes</c>, it is not allowed
      to specify a unit value different from the default value.
      No unit specifier must be given for the types <c>utf8</c>, <c>utf16</c>,
      and <c>utf32</c>.
      </item>
    </taglist>

    <section>
      <title>Integer segments</title>
      <p>The value of <c>Size</c> multiplied with the unit gives the
      size of the segment in bits.</p>

      <p>When constructing bit strings, if the size <c>N</c> of an integer
      segment is too small to contain the given integer, the most significant
      bits of the integer are silently discarded and only the <c>N</c> least
      significant bits are put into the bit string. For example, <c>&lt;&lt;16#ff:4&gt;&gt;</c>
      will result in the bit string <c>&lt;&lt;15:4&gt;&gt;</c>.</p>
    </section>

    <section>
      <title>Float segments</title>
      <p>The value of <c>Size</c> multiplied with the unit gives
      the size of the segment in bits. The size of a float segment in bits must be
      one of 16, 32, or 64.</p>

      <p>When constructing bit strings, if the size of a float segment is too small
      to contain the representation of the given float value, an exception is raised.</p>

      <p>When matching bit strings, matching of float segments fails if the bits of the segment
      does not contain the representation of a finite floating point value.</p>
    </section>

    <section>
      <title>Binary segments</title>
      <p>In this section, the phrase "binary segment" refers to any
      one of the segment types <c>binary</c>, <c>bitstring</c>,
      <c>bytes</c>, and <c>bits</c>.</p>

      <p>
        See also the paragraphs about
        <seeguide marker="#binaries">Binaries</seeguide>.
      </p>

      <p>When constructing binaries and no size is specified for a
      binary segment, the entire binary value is interpolated into the
      binary being constructed. However, the size in bits of the
      binary being interpolated must be evenly divisible by the unit
      value for the segment; otherwise an exception is raised.</p>

      <p>For example, the following examples all succeed:</p>

      <pre>
1> <input>&lt;&lt;(&lt;&lt;"abc">>)/bitstring>>.</input>
&lt;&lt;"abc">>
2> <input>&lt;&lt;(&lt;&lt;"abc">>)/binary-unit:1>>.</input>
&lt;&lt;"abc">>
3> <input>&lt;&lt;(&lt;&lt;"abc">>)/binary>>.</input>
&lt;&lt;"abc">></pre>

      <p>The first two examples have a unit value of 1 for the segment,
      while the third segment has a unit value of 8.</p>

      <p>Attempting to interpolate a bit string of size 1 into a
      binary segment with unit 8 (the default unit for <c>binary</c>)
      fails as shown in this example:</p>

      <pre>
<input>1> &lt;&lt;(&lt;&lt;1:1&gt;&gt;)/binary&gt;&gt;.</input>
** exception error: bad argument</pre>

      <p>For the construction to succeed, the unit value of the
      segment must be 1:</p>

      <pre>
2> <input>&lt;&lt;(&lt;&lt;1:1>>)/bitstring>>.</input>
&lt;&lt;1:1>>
3> <input>&lt;&lt;(&lt;&lt;1:1>>)/binary-unit:1>>.</input>
&lt;&lt;1:1>></pre>

      <p>Similarly, when matching a binary segment with no size
      specified, the match succeeds if and only if the size in bits of
      the rest of the binary is evenly divisible by the unit
      value:</p>

      <pre>
1> <input>&lt;&lt;_/binary-unit:16>> = &lt;&lt;"">>.</input>
&lt;&lt;>>
2> <input>&lt;&lt;_/binary-unit:16>> = &lt;&lt;"a">>.</input>
** exception error: no match of right hand side value &lt;&lt;"a">>
3> <input>&lt;&lt;_/binary-unit:16>> = &lt;&lt;"ab">>.</input>
&lt;&lt;"ab">>
4> <input>&lt;&lt;_/binary-unit:16>> = &lt;&lt;"abc">>.</input>
** exception error: no match of right hand side value &lt;&lt;"abc">>
5> <input>&lt;&lt;_/binary-unit:16>> = &lt;&lt;"abcd">>.</input>
&lt;&lt;"abcd">></pre>

      <p>When a size is explicitly specified for a binary segment,
      the segment size in bits is the value of <c>Size</c> multiplied
      by the default or explicit unit value.</p>

      <p>When constructing binaries, the size of the binary being interpolated
      into the constructed binary must be at least as large as the size of
      the binary segment.</p>

      <p><strong>Examples:</strong></p>
      <pre>
1> <input>&lt;&lt;(&lt;&lt;"abc">>):2/binary>>.</input>
&lt;&lt;"ab">>
2> <input>&lt;&lt;(&lt;&lt;"a">>):2/binary>>.</input>
** exception error: construction of binary failed
        *** segment 1 of type 'binary': the value &lt;&lt;"a">> is shorter than the size of the segment</pre>
    </section>

    <section>
      <title>Unicode segments</title>
      <p>The types <c>utf8</c>, <c>utf16</c>, and <c>utf32</c> specifies
      encoding/decoding of the <em>Unicode Transformation Format</em>s UTF-8, UTF-16,
      and UTF-32, respectively.</p>

      <p>When constructing a segment of a <c>utf</c> type,
      <c>Value</c> must be an integer in the range 0 through 16#D7FF
      or 16#E000 through 16#10FFFF. Construction fails with a
      <c>badarg</c> exception if <c>Value</c> is outside the allowed
      ranges. The sizes of the encoded values are as follows:</p>
      <list type="bulleted">
        <item>For <c>utf8</c>, <c>Value</c> is encoded in 1-4 bytes.</item>
        <item>For <c>utf16</c>, <c>Value</c> is encoded in 2 or 4 bytes.</item>
        <item>For <c>utf32</c>, <c>Value</c> is encoded in 4 bytes.</item>
      </list>

      <p>When constructing, a literal string can be given followed
      by one of the UTF types, for example: <c><![CDATA[<<"abc"/utf8>>]]></c>
      which is syntactic sugar for
      <c><![CDATA[<<$a/utf8,$b/utf8,$c/utf8>>]]></c>.</p>

      <p>A successful match of a segment of a <c>utf</c> type, results
      in an integer in the range 0 through 16#D7FF or 16#E000 through 16#10FFFF.
      The match fails if the returned value falls outside those ranges.</p>

      <p>A segment of type <c>utf8</c> matches 1-4 bytes in the bit string,
      if the bit string at the match position contains a valid UTF-8 sequence.
      (See RFC-3629 or the Unicode standard.)</p>

      <p>A segment of type <c>utf16</c> can match 2 or 4 bytes in the bit string.
      The match fails if the bit string at the match position does not contain
      a legal UTF-16 encoding of a Unicode code point. (See RFC-2781 or
      the Unicode standard.)</p>

      <p>A segment of type <c>utf32</c> can match 4 bytes in the bit string in the
      same way as an <c>integer</c> segment matches 32 bits.
      The match fails if the resulting integer is outside the legal ranges
      previously mentioned.</p>
    </section>

    <p><em>Examples:</em></p>
    <pre>
1> <input>Bin1 = &lt;&lt;1,17,42&gt;&gt;.</input>
&lt;&lt;1,17,42&gt;&gt;
2> <input>Bin2 = &lt;&lt;"abc"&gt;&gt;.</input>
&lt;&lt;97,98,99&gt;&gt;

3> <input>Bin3 = &lt;&lt;1,17,42:16&gt;&gt;.</input>
&lt;&lt;1,17,0,42&gt;&gt;
4> <input>&lt;&lt;A,B,C:16&gt;&gt; = &lt;&lt;1,17,42:16&gt;&gt;.</input>
&lt;&lt;1,17,0,42&gt;&gt;
5> <input>C.</input>
42
6> <input>&lt;&lt;D:16,E,F&gt;&gt; = &lt;&lt;1,17,42:16&gt;&gt;.</input>
&lt;&lt;1,17,0,42&gt;&gt;
7> <input>D.</input>
273
8> <input>F.</input>
42
9> <input>&lt;&lt;G,H/binary&gt;&gt; = &lt;&lt;1,17,42:16&gt;&gt;.</input>
&lt;&lt;1,17,0,42&gt;&gt;
10> <input>H.</input>
&lt;&lt;17,0,42&gt;&gt;
11> <input>&lt;&lt;G,J/bitstring&gt;&gt; = &lt;&lt;1,17,42:12&gt;&gt;.</input>
&lt;&lt;1,17,2,10:4&gt;&gt;
12> <input>J.</input>
&lt;&lt;17,2,10:4&gt;&gt;

13> <input>&lt;&lt;1024/utf8&gt;&gt;.</input>
&lt;&lt;208,128&gt;&gt;

14> <input>&lt;&lt;1:1,0:7&gt;&gt;.</input>
&lt;&lt;128&gt;&gt;
15> <input>&lt;&lt;16#123:12/little&gt;&gt; = &lt;&lt;16#231:12&gt;&gt; = &lt;&lt;2:4, 3:4, 1:4&gt;&gt;.</input>
&lt;&lt;35,1:4&gt;&gt;
</pre>
    <p>Notice that bit string patterns cannot be nested.</p>
    <p>Notice also that "<c><![CDATA[B=<<1>>]]></c>" is interpreted as
      "<c><![CDATA[B =< <1>>]]></c>" which is a syntax error. The correct way is
      to write a space after '=': "<c><![CDATA[B = <<1>>]]></c>.</p>
    <p>More examples are provided in
    <seeguide marker="system/programming_examples:bit_syntax">
    Programming Examples</seeguide>.</p>
  </section>

  <section>
    <marker id="funs"></marker>
    <title>Fun Expressions</title>
    <pre>
fun
    [Name](Pattern11,...,Pattern1N) [when GuardSeq1] ->
              Body1;
    ...;
    [Name](PatternK1,...,PatternKN) [when GuardSeqK] ->
              BodyK
end</pre>
    <p>A fun expression begins with the keyword <c>fun</c> and ends
      with the keyword <c>end</c>. Between them is to be a function
      declaration, similar to a
      <seeguide marker="functions#syntax">regular function declaration</seeguide>,
      except that the function name is optional and is to be a variable, if
      any.</p>
    <p>Variables in a fun head shadow the function name and both shadow
      variables in the function clause surrounding the fun expression.
      Variables bound in a fun body are local to the fun body.</p>
    <p>The return value of the expression is the resulting fun.</p>
    <p><em>Examples:</em></p>
    <pre>
1> <input>Fun1 = fun (X) -> X+1 end.</input>
#Fun&lt;erl_eval.6.39074546&gt;
2> <input>Fun1(2).</input>
3
3> <input>Fun2 = fun (X) when X>=5 -> gt; (X) -> lt end.</input>
#Fun&lt;erl_eval.6.39074546&gt;
4> <input>Fun2(7).</input>
gt
5> <input>Fun3 = fun Fact(1) -> 1; Fact(X) when X > 1 -> X * Fact(X - 1) end.</input>
#Fun&lt;erl_eval.6.39074546&gt;
6> <input>Fun3(4).</input>
24</pre>
    <p>The following fun expressions are also allowed:</p>
    <pre>
fun Name/Arity
fun Module:Name/Arity</pre>
    <p>In <c>Name/Arity</c>, <c>Name</c> is an atom and <c>Arity</c> is an integer.
      <c>Name/Arity</c> must specify an existing local function. The expression is
      syntactic sugar for:</p>
    <pre>
fun (Arg1,...,ArgN) -> Name(Arg1,...,ArgN) end</pre>
    <p>In <c>Module:Name/Arity</c>, <c>Module</c>, and <c>Name</c> are
    atoms and <c>Arity</c> is an integer. <c>Module</c>, <c>Name</c>,
    and <c>Arity</c> can also be variables. A fun defined in this way
    refers to the function <c>Name</c> with arity <c>Arity</c> in the
    <em>latest</em> version of module <c>Module</c>. A fun defined in
    this way is not dependent on the code for the module in which it
    is defined.</p>
    <change><p>Before Erlang/OTP R15, <c>Module</c>, <c>Name</c>, and <c>Arity</c> were
    not allowed to be variables.</p></change>
    <p>More examples are provided in
      <seeguide marker="system/programming_examples:funs">
      Programming Examples</seeguide>.</p>
  </section>

  <section>
    <marker id="catch"></marker>
    <title>Catch and Throw</title>
    <code type="none">
catch Expr</code>
    <p>Returns the value of <c>Expr</c> unless an exception
      occurs during the evaluation. In that case, the exception is
      caught.</p>
    <p>For exceptions of class <c>error</c>, that is,
      run-time errors,
      <c>{'EXIT',{Reason,Stack}}</c> is returned.</p>
    <p>For exceptions of class <c>exit</c>, that is,
      the code called <c>exit(Term)</c>,
      <c>{'EXIT',Term}</c> is returned.</p>
    <p>For exceptions of class <c>throw</c>, that is
      the code called <c>throw(Term)</c>,
      <c>Term</c> is returned.</p>
    <p><c>Reason</c> depends on the type of error that occurred, and
      <c>Stack</c> is the stack of recent function calls, see
      <seeguide marker="errors#exit_reasons">Exit Reasons</seeguide>.</p>
    <p><em>Examples:</em></p>
    <pre>
1> <input>catch 1+2.</input>
3
2> <input>catch 1+a.</input>
{'EXIT',{badarith,[...]}}</pre>
    <p>The BIF <c>throw(Any)</c> can be used for non-local return from
      a function. It must be evaluated within a <c>catch</c>, which
      returns the value <c>Any</c>.</p>
      <p><em>Example:</em></p>
    <pre>
3> <input>catch throw(hello).</input>
hello</pre>
    <p>If <c>throw/1</c> is not evaluated within a catch, a
      <c>nocatch</c> run-time error occurs.</p>

      <change><p>Before Erlang/OTP 24, the <c>catch</c> operator had
      the lowest precedence, making it necessary to add parentheses when
      combining it with the <c>match</c> operator:</p>
      <pre>
1> <input>A = (catch 42).</input>
42
2> <input>A.</input>
42</pre>

      <p>Starting from Erlang/OTP 24, the parentheses can be omitted:</p>
      <pre>
1> <input>A = catch 42.</input>
42
2> <input>A.</input>
42</pre>
      </change>
  </section>

  <section>
    <marker id="try"></marker>
    <title>Try</title>
    <code type="none">
try Exprs
catch
    Class1:ExceptionPattern1[:Stacktrace] [when ExceptionGuardSeq1] ->
        ExceptionBody1;
    ClassN:ExceptionPatternN[:Stacktrace] [when ExceptionGuardSeqN] ->
        ExceptionBodyN
end</code>
    <p>This is an enhancement of
      <seeguide marker="#catch">catch</seeguide>.
      It gives the possibility to:</p>
    <list type="bulleted">
      <item>Distinguish between different exception classes.</item>
      <item>Choose to handle only the desired ones.</item>
      <item>Passing the others on to an enclosing
      <c>try</c> or <c>catch</c>, or to default error handling.</item>
    </list>
    <p>Notice that although the keyword <c>catch</c> is used in
      the <c>try</c> expression, there is not a <c>catch</c> expression
      within the <c>try</c> expression.</p>
    <p>It returns the value of <c>Exprs</c> (a sequence of expressions
      <c>Expr1, ..., ExprN</c>) unless an exception occurs during
      the evaluation. In that case the exception is caught and
      the patterns <c>ExceptionPattern</c> with the right exception
      class <c>Class</c> are sequentially matched against the caught
      exception. If a match succeeds and the optional guard sequence
      <c>ExceptionGuardSeq</c> is true, the corresponding
      <c>ExceptionBody</c> is evaluated to become the return value.</p>
    <p><c>Stacktrace</c>, if specified, must be the name of a variable
      (not a pattern). The stack trace is bound to the variable when
      the corresponding <c>ExceptionPattern</c> matches.</p>
    <p>If an exception occurs during evaluation of <c>Exprs</c> but
      there is no matching <c>ExceptionPattern</c> of the right
      <c>Class</c> with a true guard sequence, the exception is passed
      on as if <c>Exprs</c> had not been enclosed in a <c>try</c>
      expression.</p>
    <p>If an exception occurs during evaluation of <c>ExceptionBody</c>,
      it is not caught.</p>
    <p>It is allowed to omit <c>Class</c> and <c>Stacktrace</c>.
      An omitted <c>Class</c> is shorthand for <c>throw</c>:</p>

    <code type="none">
try Exprs
catch
    ExceptionPattern1 [when ExceptionGuardSeq1] ->
        ExceptionBody1;
    ExceptionPatternN [when ExceptionGuardSeqN] ->
        ExceptionBodyN
end</code>

    <p>The <c>try</c> expression can have an <c>of</c>
      section:
      </p>
    <code type="none">
try Exprs of
    Pattern1 [when GuardSeq1] ->
        Body1;
    ...;
    PatternN [when GuardSeqN] ->
        BodyN
catch
    Class1:ExceptionPattern1[:Stacktrace] [when ExceptionGuardSeq1] ->
        ExceptionBody1;
    ...;
    ClassN:ExceptionPatternN[:Stacktrace] [when ExceptionGuardSeqN] ->
        ExceptionBodyN
end</code>
    <p>If the evaluation of <c>Exprs</c> succeeds without an exception,
      the patterns <c>Pattern</c> are sequentially matched against
      the result in the same way as for a
      <seeguide marker="#case">case</seeguide> expression, except that if
      the matching fails, a <c>try_clause</c> run-time error occurs instead of a
      <c>case_clause</c>.</p>
    <p>Only exceptions occurring during the evaluation of <c>Exprs</c> can be
      caught by the <c>catch</c> section. Exceptions occurring in a <c>Body</c>
      or due to a failed match are not caught.</p>
    <p>The <c>try</c> expression can also be augmented with an
      <c>after</c> section, intended to be used for cleanup with side
      effects:</p>
    <code type="none">
try Exprs of
    Pattern1 [when GuardSeq1] ->
        Body1;
    ...;
    PatternN [when GuardSeqN] ->
        BodyN
catch
    Class1:ExceptionPattern1[:Stacktrace] [when ExceptionGuardSeq1] ->
        ExceptionBody1;
    ...;
    ClassN:ExceptionPatternN[:Stacktrace] [when ExceptionGuardSeqN] ->
        ExceptionBodyN
after
    AfterBody
end</code>
    <p><c>AfterBody</c> is evaluated after either <c>Body</c> or
      <c>ExceptionBody</c>, no matter which one. The evaluated value of
      <c>AfterBody</c> is lost; the return value of the <c>try</c>
      expression is the same with an <c>after</c> section as without.</p>
    <p>Even if an exception occurs during evaluation of <c>Body</c> or
      <c>ExceptionBody</c>, <c>AfterBody</c> is evaluated. In this case
      the exception is passed on after <c>AfterBody</c> has been
      evaluated, so the exception from the <c>try</c> expression is
      the same with an <c>after</c> section as without.</p>
    <p>If an exception occurs during evaluation of <c>AfterBody</c>
      itself, it is not caught. So if <c>AfterBody</c> is evaluated after
      an exception in <c>Exprs</c>, <c>Body</c>, or <c>ExceptionBody</c>,
      that exception is lost and masked by the exception in
      <c>AfterBody</c>.</p>
    <p>The <c>of</c>, <c>catch</c>, and <c>after</c> sections are all
      optional, as long as there is at least a <c>catch</c> or an
      <c>after</c> section. So the following are valid <c>try</c>
      expressions:</p>
    <code type="none">
try Exprs of 
    Pattern when GuardSeq -> 
        Body 
after 
    AfterBody 
end

try Exprs
catch 
    ExpressionPattern -> 
        ExpressionBody
after
    AfterBody
end

try Exprs after AfterBody end</code>
    <p>Next is an example of using <c>after</c>. This closes the file,
      even in the event of exceptions in <c>file:read/2</c> or in
      <c>binary_to_term/1</c>. The exceptions are the same as
      without the <c>try</c>...<c>after</c>...<c>end</c> expression:</p>
    <code type="none">
termize_file(Name) ->
    {ok,F} = file:open(Name, [read,binary]),
    try
        {ok,Bin} = file:read(F, 1024*1024),
        binary_to_term(Bin)
    after
        file:close(F)
    end.</code>
    <p>Next is an example of using <c>try</c> to emulate <c>catch Expr</c>:</p>
    <code type="none">
try Expr
catch
    throw:Term -> Term;
    exit:Reason -> {'EXIT',Reason}
    error:Reason:Stk -> {'EXIT',{Reason,Stk}}
end</code>
  </section>

  <section>
    <title>Parenthesized Expressions</title>
    <pre>
(Expr)</pre>
    <p>Parenthesized expressions are useful to override
      <seeguide marker="#prec">operator precedences</seeguide>,
      for example, in arithmetic expressions:</p>
    <pre>
1> <input>1 + 2 * 3.</input>
7
2> <input>(1 + 2) * 3.</input>
9</pre>
  </section>

  <section>
    <marker id="block_expressions"></marker>
    <title>Block Expressions</title>
    <pre>
begin
   Expr1,
   ...,
   ExprN
end</pre>
    <p>Block expressions provide a way to group a sequence of
      expressions, similar to a clause body. The return value is
      the value of the last expression <c>ExprN</c>.</p>
  </section>

  <section>
    <marker id="lcs"></marker>
    <title>Comprehensions</title>
    <p>Comprehensions provide a succinct notation for iterating over
    one or more terms and constructing a new term. Comprehensions come
    in three different flavors, depending on the type of term they
    build.</p>
    <p>List comprehensions construct lists. They have the following syntax:</p>
    <pre>
[Expr || Qualifier1, . . ., QualifierN]</pre>
    <p>Here, <c>Expr</c> is an arbitrary expression, and each
    <c>Qualifier</c> is either a <strong>generator</strong> or a
    <strong>filter</strong>.</p>

    <p>Bit string comprehensions construct bit strings or binaries.
    They have the following syntax:</p>
    <pre>
&lt;&lt; BitStringExpr || Qualifier1, . . ., QualifierN &gt;&gt;</pre>

    <p><c>BitStringExpr</c> is an expression that evaluates to a bit string.
    If <c>BitStringExpr</c> is a function call, it must be
    enclosed in parentheses.  Each <c>Qualifier</c> is either a
    <strong>generator</strong> or a <strong>filter</strong>.</p>

    <p>Map comprehensions construct maps. They have the following syntax:</p>
    <pre>
#{KeyExpr => ValueExpr || Qualifier1, . . ., QualifierN}</pre>
    <p>Here, <c>KeyExpr</c> and <c>ValueExpr</c> are arbitrary
    expressions, and each <c>Qualifier</c> is either a
    <strong>generator</strong> or a <strong>filter</strong>.</p>

    <change><p>Map comprehensions and map generators were introduced
    in Erlang/OTP 26.</p></change>

    <p>There are three kinds of generators.</p>

    <p>A <em>list generator</em> has the following syntax:</p>

<pre>
Pattern &lt;- ListExpr</pre>

    <p>where <c>ListExpr</c> is an expression that evaluates to a
    list of terms.</p>

    <p>A <em>bit string generator</em> has the following syntax:</p>

    <pre>
BitstringPattern &lt;= BitStringExpr</pre>

    <p>where <c>BitStringExpr</c> is an expression that evaluates to a
    bit string.</p>

    <p>A <em>map generator</em> has the following syntax:</p>

    <pre>
KeyPattern := ValuePattern &lt;- MapExpression</pre>

    <p>where <c>MapExpr</c> is an expression that evaluates to a map,
    or a map iterator obtained by calling <seemfa
    marker="stdlib:maps#iterator/1">maps:iterator/1</seemfa> or
    <seemfa
    marker="stdlib:maps#iterator/2">maps:iterator/2</seemfa>.</p>

    <p>A <em>filter</em> is an expression that evaluates to
    <c>true</c> or <c>false</c>.</p>

    <p>The variables in the generator patterns shadow previously bound variables,
    including variables bound in a previous generator pattern.</p>

    <p>Variables bound in a generator expression are not visible outside the expression:</p>

    <pre>
1> <input>[{E,L} || E &lt;- L=[1,2,3]].</input>
* 1:5: variable 'L' is unbound</pre>

    <p>A <strong>list comprehension</strong> returns a list, where the list elements are the
    result of evaluating <c>Expr</c> for each combination of generator
    elements for which all filters are true.</p>

    <p>A <strong>bit string comprehension</strong> returns a bit string, which is
    created by concatenating the results of evaluating <c>BitStringExpr</c> for
    each combination of bit string generator elements for which all
    filters are true.</p>

    <p>A <strong>map comprehension</strong> returns a map, where the
    map elements are the result of evaluating <c>KeyExpr</c> and
    <c>ValueExpr</c> for each combination of generator elements for
    which all filters are true. If the key expressions are not unique,
    the last occurrence is stored in the map.</p>

    <p><strong>Examples:</strong></p>

    <p>Multiplying each element in a list by two:</p>
    <pre>
1> <input>[X*2 || X &lt;- [1,2,3]].</input>
[2,4,6]</pre>

    <p>Multiplying each byte in a binary by two, returning a list:</p>
    <pre>
1> <input>[X*2 || &lt;&lt;X&gt;&gt; &lt;= &lt;&lt;1,2,3&gt;&gt; &gt;&gt;.</input>
[2,4,6]</pre>

    <p>Multiplying each byte in a binary by two:</p>

    <pre>
1> <input>&lt;&lt; &lt;&lt;(X*2)&gt;&gt; || &lt;&lt;X&gt;&gt; &lt;= &lt;&lt;1,2,3&gt;&gt; &gt;&gt;.</input>
&lt;&lt;2,4,6&gt;&gt;</pre>

    <p>Multiplying each element in a list by two, returning a binary:</p>

    <pre>
1> <input>&lt;&lt; &lt;&lt;(X*2)&gt;&gt; || X &lt;- [1,2,3]].</input>
&lt;&lt;2,4,6&gt;&gt;</pre>

    <p>Creating a mapping from an integer to its square:</p>
    <pre>
1> <input>#{X => X*X || X &lt;- [1,2,3]}.</input>
#{1 => 1,2 => 4,3 => 9}</pre>

    <p>Multiplying the value of each element in a map by two:</p>
    <pre>
1> <input>#{K => 2*V || K := V &lt;- #{a => 1,b => 2,c => 3}}.</input>
#{a => 2,b => 4,c => 6}</pre>

    <p>Filtering a list, keeping odd numbers:</p>
    <pre>
1> <input>[X || X &lt;- [1,2,3,4,5], X rem 2 =:= 1].</input>
[1,3,5]</pre>

    <p>Filtering a list, keeping only elements that match:</p>
    <pre>
1> <input>[X || {_,_}=X &lt;- [{a,b}, [a], {x,y,z}, {1,2}]].</input>
[{a,b},{1,2}]</pre>

    <p>Combining elements from two list generators:</p>
    <pre>
1> <input>[{P,Q} || P &lt;- [a,b,c], Q &lt;- [1,2]].</input>
[{a,1},{a,2},{b,1},{b,2},{c,1},{c,2}]
</pre>

    <p>More examples are provided in
    <seeguide marker="system/programming_examples:list_comprehensions">
    Programming Examples.</seeguide></p>

    <p>When there are no generators, a comprehension returns either a
    term constructed from a single element (the result of evaluating
    <c>Expr</c>) if all filters are true, or a term constructed from
    no elements (that is, <c>[]</c> for list comprehension,
    <c>&lt;&lt;&gt;&gt;</c> for a bit string comprehension, and
    <c>#{}</c> for a map comprehension).</p>
    <p><em>Example:</em></p>
    <pre>
1> <input>[2 || is_integer(2)].</input>
[2]
2> <input>[x || is_integer(x)].</input>
[]</pre>

    <p>What happens when the filter expression does not evaluate to
    a boolean value depends on the expression:</p>

    <list>
      <item><p>If the expression is a <seeguide
      marker="#guard_expressions">guard expression</seeguide>, failure
      to evaluate or evaluating to a non-boolean value is equivalent
      to evaluating to <c>false</c>.</p></item>

      <item><p>If the expression is not a guard expression and
      evaluates to a non-Boolean value <c>Val</c>, an exception
      <c>{bad_filter, Val}</c> is triggered at runtime. If the
      evaluation of the expression raises an exception, it is not
      caught by the comprehension.</p></item>
    </list>

    <p><strong>Examples</strong> (using a guard expression as filter):</p>

    <pre>
1> <input>List = [1,2,a,b,c,3,4].</input>
[1,2,a,b,c,3,4]
2> <input>[E || E &lt;- List, E rem 2].</input>
[]
3> <input>[E || E &lt;- List, E rem 2 =:= 0].</input>
[2,4]</pre>

    <p><strong>Examples</strong> (using a non-guard expression as filter):</p>

    <pre>
1> <input>List = [1,2,a,b,c,3,4].</input>
[1,2,a,b,c,3,4]
2> <input>FaultyIsEven = fun(E) -> E rem 2 end.</input>
#Fun&lt;erl_eval.42.17316486&gt;
3> <input>[E || E &lt;- List, FaultyIsEven(E)].</input>
** exception error: bad filter 1
4> <input>IsEven = fun(E) -> E rem 2 =:= 0 end.</input>
#Fun&lt;erl_eval.42.17316486&gt;
5> <input>[E || E &lt;- List, IsEven(E)].</input>
** exception error: an error occurred when evaluating an arithmetic expression
     in operator  rem/2
        called as a rem 2
6> <input>[E || E &lt;- List, is_integer(E), IsEven(E)].</input>
[2,4]</pre>
  </section>

  <section>
    <marker id="guards"></marker>
    <title>Guard Sequences</title>

    <p>A <em>guard sequence</em> is a sequence of guards, separated
      by semicolon (;). The guard sequence is true if at least one of
      the guards is true. (The remaining guards, if any, are not
      evaluated.)</p>
    <p><c>Guard1;...;GuardK</c></p>
    <p>A <em>guard</em> is a sequence of guard expressions, separated
      by comma (,). The guard is true if all guard expressions
      evaluate to <c>true</c>.</p>
    <p><c>GuardExpr1,...,GuardExprN</c></p>
  </section>

  <section>
    <marker id="guard_expressions"></marker>
    <title>Guard Expressions</title>
    <p>The set of valid <em>guard expressions</em> is a subset of the
    set of valid Erlang expressions.  The reason for restricting the
    set of valid expressions is that evaluation of a guard expression
    must be guaranteed to be free of side effects. Valid guard
    expressions are the following:</p>
    <list type="bulleted">
      <item>Variables</item>
      <item>Constants (atoms, integer, floats, lists, tuples, records,
      binaries, and maps)</item>
      <item>Expressions that construct atoms, integer, floats, lists,
      tuples, records, binaries, and maps</item>
      <item>Expressions that update a map</item>
      <item>The record expressions <c>Expr#Name.Field</c> and <c>#Name.Field</c></item>
      <item>Calls to the BIFs specified in tables <em>Type Test BIFs</em> and
      <em>Other BIFs Allowed in Guard Expressions</em></item>
      <item>Term comparisons</item>
      <item>Arithmetic expressions</item>
      <item>Boolean expressions</item>
      <item>Short-circuit expressions (<c>andalso</c>/<c>orelse</c>)</item>
    </list>
    <table>
      <row>
        <cell align="left" valign="middle"><c>is_atom/1</c></cell>
      </row>
      <row>
        <cell align="left" valign="middle"><c>is_binary/1</c></cell>
      </row>
      <row>
        <cell align="left" valign="middle"><c>is_bitstring/1</c></cell>
      </row>
      <row>
        <cell align="left" valign="middle"><c>is_boolean/1</c></cell>
      </row>
      <row>
        <cell align="left" valign="middle"><c>is_float/1</c></cell>
      </row>
      <row>
        <cell align="left" valign="middle"><c>is_function/1</c></cell>
      </row>
      <row>
        <cell align="left" valign="middle"><c>is_function/2</c></cell>
      </row>
      <row>
        <cell align="left" valign="middle"><c>is_integer/1</c></cell>
      </row>
      <row>
        <cell align="left" valign="middle"><c>is_list/1</c></cell>
      </row>
      <row>
        <cell align="left" valign="middle"><c>is_map/1</c></cell>
      </row>
      <row>
        <cell align="left" valign="middle"><c>is_number/1</c></cell>
      </row>
      <row>
        <cell align="left" valign="middle"><c>is_pid/1</c></cell>
      </row>
      <row>
        <cell align="left" valign="middle"><c>is_port/1</c></cell>
      </row>
      <row>
        <cell align="left" valign="middle"><c>is_record/2</c></cell>
      </row>
      <row>
        <cell align="left" valign="middle"><c>is_record/3</c></cell>
      </row>
      <row>
        <cell align="left" valign="middle"><c>is_reference/1</c></cell>
      </row>
      <row>
        <cell align="left" valign="middle"><c>is_tuple/1</c></cell>
      </row>
      <tcaption>Type Test BIFs</tcaption>
    </table>
    <p>Notice that most type test BIFs have older equivalents, without
      the <c>is_</c> prefix. These old BIFs are retained for backwards
      compatibility only and are not to be used in new code. They are
      also only allowed at top level. For example, they are not allowed
      in Boolean expressions in guards.</p>
    <table>
      <row>
        <cell align="left" valign="middle"><c>abs(Number)</c></cell>
      </row>
      <row>
        <cell align="left" valign="middle"><c>bit_size(Bitstring)</c></cell>
      </row>
      <row>
        <cell align="left" valign="middle"><c>byte_size(Bitstring)</c></cell>
      </row>
      <row>
        <cell align="left" valign="middle"><c>element(N, Tuple)</c></cell>
      </row>
      <row>
        <cell align="left" valign="middle"><c>float(Term)</c></cell>
      </row>
      <row>
        <cell align="left" valign="middle"><c>hd(List)</c></cell>
      </row>
      <row>
        <cell align="left" valign="middle"><c>is_map_key(Key, Map)</c></cell>
      </row>
      <row>
        <cell align="left" valign="middle"><c>length(List)</c></cell>
      </row>
      <row>
        <cell align="left" valign="middle"><c>map_get(Key, Map)</c></cell>
      </row>
      <row>
        <cell align="left" valign="middle"><c>map_size(Map)</c></cell>
      </row>
      <row>
        <cell align="left" valign="middle"><c>max(A, B)</c></cell>
      </row>
      <row>
        <cell align="left" valign="middle"><c>min(A, B)</c></cell>
      </row>
      <row>
        <cell align="left" valign="middle"><c>node()</c></cell>
      </row>
      <row>
        <cell align="left" valign="middle"><c>node(Pid|Ref|Port)</c></cell>
      </row>
      <row>
        <cell align="left" valign="middle"><c>round(Number)</c></cell>
      </row>
      <row>
        <cell align="left" valign="middle"><c>self()</c></cell>
      </row>
      <row>
        <cell align="left" valign="middle"><c>size(Tuple|Bitstring)</c></cell>
      </row>
      <row>
        <cell align="left" valign="middle"><c>tl(List)</c></cell>
      </row>
      <row>
        <cell align="left" valign="middle"><c>trunc(Number)</c></cell>
      </row>
      <row>
        <cell align="left" valign="middle"><c>tuple_size(Tuple)</c></cell>
      </row>
      <tcaption>Other BIFs Allowed in Guard Expressions</tcaption>
    </table>

    <change><p>The <c>min/2</c> and <c>max/2</c> BIFs are allowed to be
    used in guards from Erlang/OTP 26.</p></change>

    <p>If an arithmetic expression, a Boolean expression, a
    short-circuit expression, or a call to a guard BIF fails (because
    of invalid arguments), the entire guard fails. If the guard was
    part of a guard sequence, the next guard in the sequence (that is,
    the guard following the next semicolon) is evaluated.</p>

  </section>

  <section>
    <marker id="prec"></marker>
    <title>Operator Precedence</title>
    <p>Operator precedence in descending order:</p>
    <table>
      <row>
        <cell align="left" valign="middle">:</cell>
        <cell align="left" valign="middle">&nbsp;</cell>
      </row>
      <row>
        <cell align="left" valign="middle">#</cell>
        <cell align="left" valign="middle">&nbsp;</cell>
      </row>
      <row>
        <cell align="left" valign="middle">Unary + - bnot not</cell>
        <cell align="left" valign="middle">&nbsp;</cell>
      </row>
      <row>
        <cell align="left" valign="middle">/ * div rem band and</cell>
        <cell align="left" valign="middle">Left-associative</cell>
      </row>
      <row>
        <cell align="left" valign="middle">+ - bor bxor bsl bsr or xor</cell>
        <cell align="left" valign="middle">Left-associative</cell>
      </row>
      <row>
        <cell align="left" valign="middle">++ --</cell>
        <cell align="left" valign="middle">Right-associative</cell>
      </row>
      <row>
        <cell align="left" valign="middle">== /= =&lt; &lt; >= > =:= =/=</cell>
        <cell align="left" valign="middle">Non-associative</cell>
      </row>
      <row>
        <cell align="left" valign="middle">andalso</cell>
        <cell align="left" valign="middle">Left-associative</cell>
      </row>
      <row>
        <cell align="left" valign="middle">orelse</cell>
        <cell align="left" valign="middle">Left-associative</cell>
      </row>
      <row>
        <cell align="left" valign="middle">catch</cell>
        <cell align="left" valign="middle">&nbsp;</cell>
      </row>
      <row>
        <cell align="left" valign="middle">= !</cell>
        <cell align="left" valign="middle">Right-associative</cell>
      </row>
      <row>
        <cell align="left" valign="middle">?=</cell>
        <cell align="left" valign="middle">Non-associative</cell>
      </row>
      <tcaption>Operator Precedence</tcaption>
    </table>
    <change><p>Before Erlang/OTP 24, the <c>catch</c> operator had the lowest
    precedence.</p></change>
    <note><p>The <c>=</c> operator in the table is the
    <seeguide marker="#match_operator">match operator</seeguide>.
    The character <c>=</c> can also denote the
    <seeguide marker="#compound_pattern_operator">compound pattern operator</seeguide>,
    which can only be used in patterns.</p>
    <p><c>?=</c> is restricted in that it can only be used at
    the top-level inside a <c>maybe</c> block.</p>
    </note>
    <p>When evaluating an expression, the operator with the highest
      precedence is evaluated first. Operators with the same precedence
      are evaluated according to their associativity. Non-associative
      operators cannot be combined with operators of the same precedence.</p>
      <p><em>Examples:</em></p>
      <p>The left-associative arithmetic operators are evaluated left to
      right:</p>
    <pre>
<input>6 + 5 * 4 - 3 / 2</input> evaluates to
<input>6 + 20 - 1.5</input> evaluates to
<input>26 - 1.5</input> evaluates to
<input>24.5</input></pre>

<p>The non-associative operators cannot be combined:</p>
    <pre>
1> <input>1 &lt; X &lt; 10.</input>
* 1:7: syntax error before: '&lt;'</pre>
  </section>
</chapter>