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

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

<chapter>
  <header>
    <copyright>
      <year>2001</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>The Abstract Format</title>
    <prepared>Arndt Jonasson</prepared>
    <responsible>Kenneth Lundin</responsible>
    <docno>1</docno>
    <approved>Jultomten</approved>
    <checked></checked>
    <date>00-12-01</date>
    <rev>A</rev>
    <file>absform.xml</file>
  </header>
  <p></p>
  <p>This document describes the standard representation of parse trees for Erlang
    programs as Erlang terms. This representation is known as the <em>abstract format</em>.
    Functions dealing with such parse trees are <c>compile:forms/[1,2]</c>
    and functions in the modules
    <c>epp</c>,
    <c>erl_eval</c>,
    <c>erl_lint</c>,
    <c>erl_pp</c>,
    <c>erl_parse</c>,
    and
    <c>io</c>.
    They are also used as input and output for parse transforms (see the module
    <c>compile</c>).</p>
  <p>We use the function <c>Rep</c> to denote the mapping from an Erlang source
    construct <c>C</c> to its abstract format representation <c>R</c>, and write
    <c>R = Rep(C)</c>.
    </p>
  <p>The word <c>LINE</c> below represents an integer, and denotes the
    number of the line in the source file where the construction occurred.
    Several instances of <c>LINE</c> in the same construction may denote
    different lines.</p>
  <p>Since operators are not terms in their own right, when operators are
    mentioned below, the representation of an operator should be taken to
    be the atom with a printname consisting of the same characters as the
    operator.
    </p>

  <section>
    <title>Module Declarations and Forms</title>
    <p>A module declaration consists of a sequence of forms that are either
      function declarations or attributes.</p>
    <list type="bulleted">
      <item>If D is a module declaration consisting of the forms
      <c>F_1</c>, ..., <c>F_k</c>, then
       Rep(D) = <c>[Rep(F_1), ..., Rep(F_k)]</c>.</item>
      <item>If F is an attribute <c>-export([Fun_1/A_1, ..., Fun_k/A_k])</c>, then
       Rep(F) = <c>{attribute,LINE,export,[{Fun_1,A_1}, ..., {Fun_k,A_k}]}</c>.</item>
      <item>If F is an attribute <c>-import(Mod,[Fun_1/A_1, ..., Fun_k/A_k])</c>, then
       Rep(F) = <c>{attribute,LINE,import,{Mod,[{Fun_1,A_1}, ..., {Fun_k,A_k}]}}</c>.</item>
      <item>If F is an attribute <c>-module(Mod)</c>, then
       Rep(F) = <c>{attribute,LINE,module,Mod}</c>.</item>
      <item>If F is an attribute <c>-file(File,Line)</c>, then
       Rep(F) = <c>{attribute,LINE,file,{File,Line}}</c>.</item>
      <item>If F is a function declaration
       <c>Name Fc_1 ; ... ; Name Fc_k</c>,
       where each <c>Fc_i</c> is a function clause with a
       pattern sequence of the same length <c>Arity</c>, then
       Rep(F) = <c>{function,LINE,Name,Arity,[Rep(Fc_1), ...,Rep(Fc_k)]}</c>.
      </item>
      <item>If F is a function specification
       <c>-Spec Name Ft_1; ...; Ft_k</c>,
       where <c>Spec</c> is either the atom <c>spec</c> or the atom
       <c>callback</c>, and each <c>Ft_i</c> is a possibly constrained
       function type with an argument sequence of the same length
       <c>Arity</c>, then Rep(F) =
       <c>{attribute,Line,Spec,{{Name,Arity},[Rep(Ft_1), ..., Rep(Ft_k)]}}</c>.
      </item>
      <item>If F is a function specification
       <c>-spec Mod:Name Ft_1; ...; Ft_k</c>,
       where each <c>Ft_i</c> is a possibly constrained
       function type with an argument sequence of the same length
       <c>Arity</c>, then Rep(F) =
       <c>{attribute,Line,spec,{{Mod,Name,Arity},[Rep(Ft_1), ..., Rep(Ft_k)]}}</c>.
      </item>
      <item>If F is a record declaration
       <c>-record(Name,{V_1, ..., V_k})</c>,
       where each <c>V_i</c> is a record field, then Rep(F) =
       <c>{attribute,LINE,record,{Name,[Rep(V_1), ..., Rep(V_k)]}}</c>.
       For Rep(V), see below.</item>
      <item>If F is a type declaration
       <c>-Type Name(V_1, ..., V_k) :: T</c>, where
       <c>Type</c> is either the atom <c>type</c> or the atom <c>opaque</c>,
       each <c>V_i</c> is a variable, and <c>T</c> is a type, then Rep(F) =
       <c>{attribute,LINE,Type,{Name,Rep(T),[Rep(V_1), ..., Rep(V_k)]}}</c>.
      </item>
      <item>If F is a wild attribute <c>-A(T)</c>, then
       Rep(F) = <c>{attribute,LINE,A,T}</c>.
      <br></br></item>
    </list>

    <section>
      <title>Record Fields</title>
      <p>Each field in a record declaration may have an optional
        explicit default initializer expression, as well as an
        optional type.</p>
      <list type="bulleted">
        <item>If V is <c>A</c>, then
         Rep(V) = <c>{record_field,LINE,Rep(A)}</c>.</item>
        <item>If V is <c>A = E</c>,
	 where <c>E</c> is an expression, then
         Rep(V) = <c>{record_field,LINE,Rep(A),Rep(E)}</c>.</item>
        <item>If V is <c>A :: T</c>, where <c>T</c> is a type, then Rep(V) =
          <c>{typed_record_field,{record_field,LINE,Rep(A)},Rep(T)}</c>.
          </item>
        <item>If V is <c>A = E :: T</c>, where
          <c>E</c> is an expression and <c>T</c> is a type, then Rep(V) =
          <c>{typed_record_field,{record_field,LINE,Rep(A),Rep(E)},Rep(T)}</c>.
        </item>
      </list>
    </section>

    <section>
      <title>Representation of Parse Errors and End-of-file</title>
      <p>In addition to the representations of forms, the list that represents
        a module declaration (as returned by functions in <c>erl_parse</c> and
        <c>epp</c>) may contain tuples <c>{error,E}</c> and
	<c>{warning,W}</c>, denoting syntactically incorrect forms and
	warnings, and <c>{eof,LINE}</c>, denoting an end-of-stream
	encountered before a complete form had been parsed.</p>
    </section>
  </section>

  <section>
    <title>Atomic Literals</title>
    <p>There are five kinds of atomic literals, which are represented in the
      same way in patterns, expressions and guards:</p>
    <list type="bulleted">
      <item>If L is an atom literal, then
       Rep(L) = <c>{atom,LINE,L}</c>.</item>
      <item>If L is a character literal, then
       Rep(L) = <c>{char,LINE,L}</c>.</item>
      <item>If L is a float literal, then
       Rep(L) = <c>{float,LINE,L}</c>.</item>
      <item>If L is an integer literal, then
       Rep(L) = <c>{integer,LINE,L}</c>.</item>
      <item>If L is a string literal consisting of the characters
      <c>C_1</c>, ..., <c>C_k</c>, then
       Rep(L) = <c>{string,LINE,[C_1, ..., C_k]}</c>.</item>
    </list>
    <p>Note that negative integer and float literals do not occur as such; they are
      parsed as an application of the unary negation operator.</p>
  </section>

  <section>
    <title>Patterns</title>
    <p>If Ps is a sequence of patterns <c>P_1, ..., P_k</c>, then
      Rep(Ps) = <c>[Rep(P_1), ..., Rep(P_k)]</c>. Such sequences occur as the
      list of arguments to a function or fun.</p>
    <p>Individual patterns are represented as follows:</p>
    <list type="bulleted">
      <item>If P is an atomic literal <c>L</c>, then Rep(P) = Rep(L).</item>
      <item>If P is a bit string pattern
       <c>&lt;&lt;P_1:Size_1/TSL_1, ..., P_k:Size_k/TSL_k>></c>, where each
       <c>Size_i</c> is an expression that can be evaluated to an integer
       and each <c>TSL_i</c> is a type specificer list, then
       Rep(P) = <c>{bin,LINE,[{bin_element,LINE,Rep(P_1),Rep(Size_1),Rep(TSL_1)}, ..., {bin_element,LINE,Rep(P_k),Rep(Size_k),Rep(TSL_k)}]}</c>.
       For Rep(TSL), see below.
       An omitted <c>Size_i</c> is represented by <c>default</c>.
       An omitted <c>TSL_i</c> is represented by <c>default</c>.</item>
      <item>If P is a compound pattern <c>P_1 = P_2</c>, then
       Rep(P) = <c>{match,LINE,Rep(P_1),Rep(P_2)}</c>.</item>
      <item>If P is a cons pattern <c>[P_h | P_t]</c>, then
       Rep(P) = <c>{cons,LINE,Rep(P_h),Rep(P_t)}</c>.</item>
      <item>If P is a map pattern <c>#{A_1, ..., A_k}</c>, where each
       <c>A_i</c> is an association <c>P_i_1 := P_i_2</c>, then Rep(P) =
       <c>{map,LINE,[Rep(A_1), ..., Rep(A_k)]}</c>. For Rep(A), see
       below.</item>
      <item>If P is a nil pattern <c>[]</c>, then
       Rep(P) = <c>{nil,LINE}</c>.</item>
      <item>If P is an operator pattern <c>P_1 Op P_2</c>,
       where <c>Op</c> is a binary operator (this is either an occurrence
       of <c>++</c> applied to a literal string or character
       list, or an occurrence of an expression that can be evaluated to a number
       at compile time),
       then Rep(P) = <c>{op,LINE,Op,Rep(P_1),Rep(P_2)}</c>.</item>
      <item>If P is an operator pattern <c>Op P_0</c>,
       where <c>Op</c> is a unary operator (this is an occurrence of
       an expression that can be evaluated to a number at compile
       time), then Rep(P) = <c>{op,LINE,Op,Rep(P_0)}</c>.</item>
      <item>If P is a parenthesized pattern <c>( P_0 )</c>, then
       Rep(P) = <c>Rep(P_0)</c>,
       that is, parenthesized patterns cannot be distinguished from their
       bodies.</item>
      <item>If P is a record field index pattern <c>#Name.Field</c>,
       where <c>Field</c> is an atom, then
       Rep(P) = <c>{record_index,LINE,Name,Rep(Field)}</c>.</item>
      <item>If P is a record pattern
       <c>#Name{Field_1=P_1, ..., Field_k=P_k}</c>,
       where each <c>Field_i</c> is an atom or <c>_</c>, then Rep(P) =
       <c>{record,LINE,Name,[{record_field,LINE,Rep(Field_1),Rep(P_1)}, ..., {record_field,LINE,Rep(Field_k),Rep(P_k)}]}</c>.</item>
      <item>If P is a tuple pattern <c>{P_1, ..., P_k}</c>, then
       Rep(P) = <c>{tuple,LINE,[Rep(P_1), ..., Rep(P_k)]}</c>.</item>
      <item>If P is a universal pattern <c>_</c>, then
       Rep(P) = <c>{var,LINE,'_'}</c>.</item>
      <item>If P is a variable pattern <c>V</c>, then
       Rep(P) = <c>{var,LINE,A}</c>,
       where A is an atom with a printname consisting of the same characters as
      <c>V</c>.</item>
    </list>
    <p>Note that every pattern has the same source form as some expression, and is
      represented the same way as the corresponding expression.</p>
  </section>

  <section>
    <title>Expressions</title>
    <p>A body B is a nonempty sequence of expressions <c>E_1, ..., E_k</c>,
      and Rep(B) = <c>[Rep(E_1), ..., Rep(E_k)]</c>.</p>
    <p>An expression E is one of the following alternatives:</p>
    <list type="bulleted">
      <item>If E is an atomic literal <c>L</c>, then Rep(E) = Rep(L).</item>
      <item>If E is a bit string comprehension
       <c>&lt;&lt;E_0 || Q_1, ..., Q_k>></c>,
       where each <c>Q_i</c> is a qualifier, then
       Rep(E) = <c>{bc,LINE,Rep(E_0),[Rep(Q_1), ..., Rep(Q_k)]}</c>.
       For Rep(Q), see below.</item>
      <item>If E is a bit string constructor
       <c>&lt;&lt;E_1:Size_1/TSL_1, ..., E_k:Size_k/TSL_k>></c>,
       where each <c>Size_i</c> is an expression and each
       <c>TSL_i</c> is a type specificer list, then Rep(E) =
       <c>{bin,LINE,[{bin_element,LINE,Rep(E_1),Rep(Size_1),Rep(TSL_1)}, ..., {bin_element,LINE,Rep(E_k),Rep(Size_k),Rep(TSL_k)}]}</c>.
       For Rep(TSL), see below.
       An omitted <c>Size_i</c> is represented by <c>default</c>.
       An omitted <c>TSL_i</c> is represented by <c>default</c>.</item>
      <item>If E is a block expression <c>begin B end</c>,
       where <c>B</c> is a body, then
       Rep(E) = <c>{block,LINE,Rep(B)}</c>.</item>
      <item>If E is a case expression <c>case E_0 of Cc_1 ; ... ; Cc_k end</c>,
       where <c>E_0</c> is an expression and each <c>Cc_i</c> is a
       case clause then Rep(E) =
      <c>{'case',LINE,Rep(E_0),[Rep(Cc_1), ..., Rep(Cc_k)]}</c>.</item>
      <item>If E is a catch expression <c>catch E_0</c>, then
       Rep(E) = <c>{'catch',LINE,Rep(E_0)}</c>.</item>
      <item>If E is a cons skeleton <c>[E_h | E_t]</c>, then
       Rep(E) = <c>{cons,LINE,Rep(E_h),Rep(E_t)}</c>.</item>
      <item>If E is a fun expression <c>fun Name/Arity</c>, then
       Rep(E) = <c>{'fun',LINE,{function,Name,Arity}}</c>.</item>
      <item>If E is a fun expression
       <c>fun Module:Name/Arity</c>, then Rep(E) =
       <c>{'fun',LINE,{function,Rep(Module),Rep(Name),Rep(Arity)}}</c>.
       (Before the R15 release: Rep(E) =
       <c>{'fun',LINE,{function,Module,Name,Arity}}</c>.)</item>
      <item>If E is a fun expression <c>fun Fc_1 ; ... ; Fc_k end</c>,
       where each <c>Fc_i</c> is a function clause then Rep(E) =
      <c>{'fun',LINE,{clauses,[Rep(Fc_1), ..., Rep(Fc_k)]}}</c>.</item>
      <item>If E is a fun expression
       <c>fun Name Fc_1 ; ... ; Name Fc_k end</c>,
       where <c>Name</c> is a variable and each
       <c>Fc_i</c> is a function clause then Rep(E) =
       <c>{named_fun,LINE,Name,[Rep(Fc_1), ..., Rep(Fc_k)]}</c>.
       </item>
      <item>If E is a function call <c>E_0(E_1, ..., E_k)</c>, then
       Rep(E) = <c>{call,LINE,Rep(E_0),[Rep(E_1), ..., Rep(E_k)]}</c>.</item>
      <item>If E is a function call <c>E_m:E_0(E_1, ..., E_k)</c>,
       then Rep(E) =
       <c>{call,LINE,{remote,LINE,Rep(E_m),Rep(E_0)},[Rep(E_1), ...,  Rep(E_k)]}</c>.
      </item>
      <item>If E is an if expression <c>if Ic_1 ; ... ; Ic_k  end</c>,
       where each <c>Ic_i</c> is an if clause then Rep(E) =
      <c>{'if',LINE,[Rep(Ic_1), ..., Rep(Ic_k)]}</c>.</item>
      <item>If E is a list comprehension <c>[E_0 || Q_1, ..., Q_k]</c>,
       where each <c>Q_i</c> is a qualifier, then Rep(E) =
       <c>{lc,LINE,Rep(E_0),[Rep(Q_1), ..., Rep(Q_k)]}</c>. For Rep(Q), see
       below.</item>
      <item>If E is a map creation <c>#{A_1, ..., A_k}</c>,
       where each <c>A_i</c> is an association <c>E_i_1 => E_i_2</c>
       or <c>E_i_1 := E_i_2</c>, then Rep(E) =
       <c>{map,LINE,[Rep(A_1), ..., Rep(A_k)]}</c>. For Rep(A), see
       below.</item>
      <item>If E is a map update <c>E_0#{A_1, ..., A_k}</c>,
       where each <c>A_i</c> is an association <c>E_i_1 => E_i_2</c>
       or <c>E_i_1 := E_i_2</c>, then Rep(E) =
       <c>{map,LINE,Rep(E_0),[Rep(A_1), ..., Rep(A_k)]}</c>.
       For Rep(A), see below.</item>
      <item>If E is a match operator expression <c>P = E_0</c>,
       where <c>P</c> is a pattern, then
       Rep(E) = <c>{match,LINE,Rep(P),Rep(E_0)}</c>.</item>
      <item>If E is nil, <c>[]</c>, then
       Rep(E) = <c>{nil,LINE}</c>.</item>
      <item>If E is an operator expression <c>E_1 Op E_2</c>,
       where <c>Op</c> is a binary operator other than the match
       operator <c>=</c>, then
       Rep(E) = <c>{op,LINE,Op,Rep(E_1),Rep(E_2)}</c>.</item>
      <item>If E is an operator expression <c>Op E_0</c>,
       where <c>Op</c> is a unary operator, then
       Rep(E) = <c>{op,LINE,Op,Rep(E_0)}</c>.</item>
      <item>If E is a parenthesized expression <c>( E_0 )</c>, then
       Rep(E) = <c>Rep(E_0)</c>, that is, parenthesized
       expressions cannot be distinguished from their bodies.</item>
      <item>If E is a receive expression <c>receive Cc_1 ; ... ; Cc_k end</c>,
       where each <c>Cc_i</c> is a case clause then Rep(E) =
      <c>{'receive',LINE,[Rep(Cc_1), ..., Rep(Cc_k)]}</c>.</item>
      <item>If E is a receive expression
       <c>receive Cc_1 ; ... ; Cc_k after E_0 -> B_t end</c>,
       where each <c>Cc_i</c> is a case clause,
      <c>E_0</c> is an expression and <c>B_t</c> is a body, then Rep(E) =
      <c>{'receive',LINE,[Rep(Cc_1), ..., Rep(Cc_k)],Rep(E_0),Rep(B_t)}</c>.</item>
      <item>If E is a record creation
       <c>#Name{Field_1=E_1, ..., Field_k=E_k}</c>,
       where each <c>Field_i</c> is an atom or <c>_</c>, then Rep(E) =
      <c>{record,LINE,Name,[{record_field,LINE,Rep(Field_1),Rep(E_1)}, ..., {record_field,LINE,Rep(Field_k),Rep(E_k)}]}</c>.</item>
      <item>If E is a record field access <c>E_0#Name.Field</c>,
       where <c>Field</c> is an atom, then
       Rep(E) = <c>{record_field,LINE,Rep(E_0),Name,Rep(Field)}</c>.</item>
      <item>If E is a record field index <c>#Name.Field</c>,
       where <c>Field</c> is an atom, then
       Rep(E) = <c>{record_index,LINE,Name,Rep(Field)}</c>.</item>
      <item>If E is a record update
       <c>E_0#Name{Field_1=E_1, ..., Field_k=E_k}</c>,
       where each <c>Field_i</c> is an atom, then Rep(E) =
      <c>{record,LINE,Rep(E_0),Name,[{record_field,LINE,Rep(Field_1),Rep(E_1)}, ..., {record_field,LINE,Rep(Field_k),Rep(E_k)}]}</c>.</item>
      <item>If E is a tuple skeleton <c>{E_1, ..., E_k}</c>, then
       Rep(E) = <c>{tuple,LINE,[Rep(E_1), ..., Rep(E_k)]}</c>.</item>
      <item>If E is a try expression <c>try B catch Tc_1 ; ... ; Tc_k end</c>,
       where <c>B</c> is a body and each <c>Tc_i</c> is a catch clause then
       Rep(E) =
      <c>{'try',LINE,Rep(B),[],[Rep(Tc_1), ..., Rep(Tc_k)],[]}</c>.</item>
      <item>If E is a try expression
       <c>try B of Cc_1 ; ... ; Cc_k catch Tc_1 ; ... ; Tc_n end</c>,
       where <c>B</c> is a body,
       each <c>Cc_i</c> is a case clause and
       each <c>Tc_j</c> is a catch clause then Rep(E) =
      <c>{'try',LINE,Rep(B),[Rep(Cc_1), ..., Rep(Cc_k)],[Rep(Tc_1), ..., Rep(Tc_n)],[]}</c>.</item>
      <item>If E is a try expression <c>try B after A end</c>,
       where <c>B</c> and <c>A</c> are bodies then Rep(E) =
      <c>{'try',LINE,Rep(B),[],[],Rep(A)}</c>.</item>
      <item>If E is a try expression
       <c>try B of Cc_1 ; ... ; Cc_k after A end</c>,
       where <c>B</c> and <c>A</c> are a bodies and
       each <c>Cc_i</c> is a case clause then Rep(E) =
      <c>{'try',LINE,Rep(B),[Rep(Cc_1), ..., Rep(Cc_k)],[],Rep(A)}</c>.</item>
      <item>If E is a try expression
       <c>try B catch Tc_1 ; ... ; Tc_k after A end</c>,
       where <c>B</c> and <c>A</c> are bodies and
       each <c>Tc_i</c> is a catch clause then Rep(E) =
      <c>{'try',LINE,Rep(B),[],[Rep(Tc_1), ..., Rep(Tc_k)],Rep(A)}</c>.</item>
      <item>If E is a try expression
       <c>try B of Cc_1 ; ... ; Cc_k  catch Tc_1 ; ... ; Tc_n after A end</c>,
       where <c>B</c> and <c>A</c> are a bodies,
       each <c>Cc_i</c> is a case clause, and
       each <c>Tc_j</c> is a catch clause then
       Rep(E) =
      <c>{'try',LINE,Rep(B),[Rep(Cc_1), ..., Rep(Cc_k)],[Rep(Tc_1),  ..., Rep(Tc_n)],Rep(A)}</c>.</item>
      <item>If E is a variable <c>V</c>, then Rep(E) = <c>{var,LINE,A}</c>,
       where <c>A</c> is an atom with a printname consisting of the same
       characters as <c>V</c>.</item>
    </list>

    <section>
      <title>Qualifiers</title>
      <p>A qualifier Q is one of the following alternatives:</p>
      <list type="bulleted">
        <item>If Q is a filter <c>E</c>, where <c>E</c> is an expression, then
         Rep(Q) = <c>Rep(E)</c>.</item>
        <item>If Q is a generator <c>P &lt;- E</c>, where <c>P</c> is
	 a pattern and <c>E</c> is an expression, then
         Rep(Q) = <c>{generate,LINE,Rep(P),Rep(E)}</c>.</item>
        <item>If Q is a bit string generator
	 <c>P &lt;= E</c>, where <c>P</c> is
	 a pattern and <c>E</c> is an expression, then
         Rep(Q) = <c>{b_generate,LINE,Rep(P),Rep(E)}</c>.</item>
      </list>
    </section>

    <section>
      <title>Bit String Element Type Specifiers</title>
      <p>A type specifier list TSL for a bit string element is a sequence
        of type specifiers <c>TS_1 - ... - TS_k</c>, and
        Rep(TSL) = <c>[Rep(TS_1), ..., Rep(TS_k)]</c>.</p>
      <list type="bulleted">
        <item>If TS is a type specifier <c>A</c>, where <c>A</c> is an atom,
	 then Rep(TS) = <c>A</c>.</item>
        <item>If TS is a type specifier <c>A:Value</c>,
	 where <c>A</c> is an atom and <c>Value</c> is an integer,
	 then Rep(TS) = <c>{A,Value}</c>.</item>
      </list>
    </section>

    <section>
      <title>Associations</title>
      <p>An association A is one of the following alternatives:</p>
      <list type="bulleted">
	<item>If A is an association <c>K => V</c>,
	 then Rep(A) = <c>{map_field_assoc,LINE,Rep(K),Rep(V)}</c>.
         </item>
	<item>If A is an association <c>K := V</c>,
         then Rep(A) = <c>{map_field_exact,LINE,Rep(K),Rep(V)}</c>.
         </item>
      </list>
    </section>
  </section>

  <section>
    <title>Clauses</title>
    <p>There are function clauses, if clauses, case clauses
      and catch clauses.</p>
    <p>A clause <c>C</c> is one of the following alternatives:</p>
    <list type="bulleted">
      <item>If C is a case clause <c>P -> B</c>,
       where <c>P</c> is a pattern and <c>B</c> is a body, then
       Rep(C) = <c>{clause,LINE,[Rep(P)],[],Rep(B)}</c>.</item>
      <item>If C is a case clause <c>P when Gs -> B</c>,
       where <c>P</c> is a pattern,
      <c>Gs</c> is a guard sequence and <c>B</c> is a body, then
       Rep(C) = <c>{clause,LINE,[Rep(P)],Rep(Gs),Rep(B)}</c>.</item>
      <item>If C is a catch clause <c>P -> B</c>,
       where <c>P</c> is a pattern and <c>B</c> is a body, then
       Rep(C) = <c>{clause,LINE,[Rep({throw,P,_})],[],Rep(B)}</c>.</item>
      <item>If C is a catch clause <c>X : P -> B</c>,
       where <c>X</c> is an atomic literal or a variable pattern,
      <c>P</c> is a pattern, and <c>B</c> is a body, then
       Rep(C) = <c>{clause,LINE,[Rep({X,P,_})],[],Rep(B)}</c>.</item>
      <item>If C is a catch clause <c>P when Gs -> B</c>,
       where <c>P</c> is a pattern, <c>Gs</c> is a guard sequence,
       and <c>B</c> is a body, then
       Rep(C) = <c>{clause,LINE,[Rep({throw,P,_})],Rep(Gs),Rep(B)}</c>.</item>
      <item>If C is a catch clause <c>X : P when Gs -> B</c>,
       where <c>X</c> is an atomic literal or a variable pattern,
      <c>P</c> is a pattern, <c>Gs</c> is a guard sequence,
       and <c>B</c> is a body, then
       Rep(C) = <c>{clause,LINE,[Rep({X,P,_})],Rep(Gs),Rep(B)}</c>.</item>
      <item>If C is a function clause <c>( Ps ) -> B</c>,
       where <c>Ps</c> is a pattern sequence and <c>B</c> is a body, then
       Rep(C) = <c>{clause,LINE,Rep(Ps),[],Rep(B)}</c>.</item>
      <item>If C is a function clause <c>( Ps ) when Gs -> B</c>,
       where <c>Ps</c> is a pattern sequence,
      <c>Gs</c> is a guard sequence and <c>B</c> is a body, then
       Rep(C) = <c>{clause,LINE,Rep(Ps),Rep(Gs),Rep(B)}</c>.</item>
      <item>If C is an if clause <c>Gs -> B</c>,
       where <c>Gs</c> is a guard sequence and <c>B</c> is a body, then
       Rep(C) = <c>{clause,LINE,[],Rep(Gs),Rep(B)}</c>.</item>
    </list>
  </section>

  <section>
    <title>Guards</title>
    <p>A guard sequence Gs is a sequence of guards <c>G_1; ...; G_k</c>, and
      Rep(Gs) = <c>[Rep(G_1), ..., Rep(G_k)]</c>. If the guard sequence is
      empty, Rep(Gs) = <c>[]</c>.</p>
    <p>A guard G is a nonempty sequence of guard tests
      <c>Gt_1, ..., Gt_k</c>, and Rep(G) =
      <c>[Rep(Gt_1), ..., Rep(Gt_k)]</c>.</p>
    <p>A guard test <c>Gt</c> is one of the following alternatives:</p>
    <list type="bulleted">
      <item>If Gt is an atomic literal <c>L</c>, then Rep(Gt) = Rep(L).</item>
      <item>If Gt is a bit string constructor
       <c>&lt;&lt;Gt_1:Size_1/TSL_1, ..., Gt_k:Size_k/TSL_k>></c>,
       where each <c>Size_i</c> is a guard test and each
       <c>TSL_i</c> is a type specificer list, then
       Rep(Gt) = <c>{bin,LINE,[{bin_element,LINE,Rep(Gt_1),Rep(Size_1),Rep(TSL_1)}, ..., {bin_element,LINE,Rep(Gt_k),Rep(Size_k),Rep(TSL_k)}]}</c>.
       For Rep(TSL), see above.
       An omitted <c>Size_i</c> is represented by <c>default</c>.
       An omitted <c>TSL_i</c> is represented by <c>default</c>.</item>
      <item>If Gt is a cons skeleton <c>[Gt_h | Gt_t]</c>, then
       Rep(Gt) = <c>{cons,LINE,Rep(Gt_h),Rep(Gt_t)}</c>.</item>
      <item>If Gt is a function call <c>A(Gt_1, ..., Gt_k)</c>,
       where <c>A</c> is an atom, then Rep(Gt) =
       <c>{call,LINE,Rep(A),[Rep(Gt_1), ..., Rep(Gt_k)]}</c>.</item>
      <item>If Gt is a function call <c>A_m:A(Gt_1, ..., Gt_k)</c>,
       where <c>A_m</c> is the atom <c>erlang</c> and <c>A</c> is
       an atom or an operator, then Rep(Gt) =
       <c>{call,LINE,{remote,LINE,Rep(A_m),Rep(A)},[Rep(Gt_1), ..., Rep(Gt_k)]}</c>.</item>
      <item>If Gt is a map creation <c>#{A_1, ..., A_k}</c>,
       where each <c>A_i</c> is an association <c>Gt_i_1 => Gt_i_2</c>
       or <c>Gt_i_1 := Gt_i_2</c>, then Rep(Gt) =
       <c>{map,LINE,[Rep(A_1), ..., Rep(A_k)]}</c>. For Rep(A), see
       above.</item>
      <item>If Gt is a map update <c>Gt_0#{A_1, ..., A_k}</c>, where each
       <c>A_i</c> is an association <c>Gt_i_1 => Gt_i_2</c>
       or <c>Gt_i_1 := Gt_i_2</c>, then Rep(Gt) =
       <c>{map,LINE,Rep(Gt_0),[Rep(A_1), ..., Rep(A_k)]}</c>.
       For Rep(A), see above.</item>
      <item>If Gt is nil, <c>[]</c>,
       then Rep(Gt) = <c>{nil,LINE}</c>.</item>
      <item>If Gt is an operator guard test <c>Gt_1 Op Gt_2</c>,
       where <c>Op</c> is a binary operator other than the match
       operator <c>=</c>, then
       Rep(Gt) = <c>{op,LINE,Op,Rep(Gt_1),Rep(Gt_2)}</c>.</item>
      <item>If Gt is an operator guard test <c>Op Gt_0</c>,
       where <c>Op</c> is a unary operator, then
       Rep(Gt) = <c>{op,LINE,Op,Rep(Gt_0)}</c>.</item>
      <item>If Gt is a parenthesized guard test <c>( Gt_0 )</c>, then
       Rep(Gt) = <c>Rep(Gt_0)</c>, that is, parenthesized
       guard tests cannot be distinguished from their bodies.</item>
      <item>If Gt is a record creation
       <c>#Name{Field_1=Gt_1, ..., Field_k=Gt_k}</c>,
       where each <c>Field_i</c> is an atom or <c>_</c>, then Rep(Gt) =
      <c>{record,LINE,Name,[{record_field,LINE,Rep(Field_1),Rep(Gt_1)}, ..., {record_field,LINE,Rep(Field_k),Rep(Gt_k)}]}</c>.</item>
      <item>If Gt is  a record field access <c>Gt_0#Name.Field</c>,
       where <c>Field</c> is an atom, then
       Rep(Gt) = <c>{record_field,LINE,Rep(Gt_0),Name,Rep(Field)}</c>.</item>
      <item>If Gt is a record field index <c>#Name.Field</c>,
       where <c>Field</c> is an atom, then
       Rep(Gt) = <c>{record_index,LINE,Name,Rep(Field)}</c>.</item>
      <item>If Gt is a tuple skeleton <c>{Gt_1, ..., Gt_k}</c>, then
       Rep(Gt) = <c>{tuple,LINE,[Rep(Gt_1), ..., Rep(Gt_k)]}</c>.</item>
      <item>If Gt is a variable pattern <c>V</c>, then
       Rep(Gt) = <c>{var,LINE,A}</c>, where A is an atom with
       a printname consisting of the same characters as <c>V</c>.</item>
    </list>
    <p>Note that every guard test has the same source form as some expression,
      and is represented the same way as the corresponding expression.</p>
  </section>

  <section>
    <title>Types</title>
    <list type="bulleted">
      <item>If T is an annotated type <c>A :: T_0</c>,
       where <c>A</c> is a variable, then Rep(T) =
       <c>{ann_type,LINE,[Rep(A),Rep(T_0)]}</c>.</item>
      <item>If T is an atom or integer literal L, then Rep(T) = Rep(L).
      </item>
      <item>If T is a bit string type <c>&lt;&lt;_:M,_:_*N>></c>,
       where <c>M</c> and <c>N</c> are singleton integer types, then Rep(T) =
       <c>{type,LINE,binary,[Rep(M),Rep(N)]}</c>.</item>
      <item>If T is the empty list type <c>[]</c>, then Rep(T) =
       <c>{type,Line,nil,[]}</c>.</item>
      <item>If T is a fun type <c>fun()</c>, then Rep(T) =
       <c>{type,LINE,'fun',[]}</c>.</item>
      <item>If T is a fun type <c>fun((...) -> T_0)</c>, then
       Rep(T) = <c>{type,LINE,'fun',[{type,LINE,any},Rep(T_0)]}</c>.
       </item>
      <item>If T is a fun type <c>fun(Ft)</c>, where
       <c>Ft</c> is a function type,
       then Rep(T) = <c>Rep(Ft)</c>. For Rep(Ft), see below.</item>
      <item>If T is an integer range type <c>L .. H</c>,
       where <c>L</c> and <c>H</c> are singleton integer types, then
       Rep(T) = <c>{type,LINE,range,[Rep(L),Rep(H)]}</c>.</item>
      <item>If T is a map type <c>map()</c>, then Rep(T) =
       <c>{type,LINE,map,any}</c>.</item>
      <item>If T is a map type <c>#{A_1, ..., A_k}</c>, where each
       <c>A_i</c> is an association type, then Rep(T) =
       <c>{type,LINE,map,[Rep(A_1), ..., Rep(A_k)]}</c>.
       For Rep(A), see below.</item>
      <item>If T is an operator type <c>T_1 Op T_2</c>,
       where <c>Op</c> is a binary operator (this is an occurrence of
       an expression that can be evaluated to an integer at compile
       time), then
       Rep(T) = <c>{op,LINE,Op,Rep(T_1),Rep(T_2)}</c>.</item>
      <item>If T is an operator type <c>Op T_0</c>, where <c>Op</c> is a
       unary operator (this is an occurrence of
       an expression that can be evaluated to an integer at compile time),
       then Rep(T) = <c>{op,LINE,Op,Rep(T_0)}</c>.</item>
      <item>If T is <c>( T_0 )</c>, then Rep(T) = <c>Rep(T_0)</c>,
       that is, parenthesized types cannot be distinguished from their
       bodies.</item>
      <item>If T is a predefined (or built-in) type <c>N(T_1, ..., T_k)</c>,
       then Rep(T) =
       <c>{type,LINE,N,[Rep(T_1), ..., Rep(T_k)]}</c>.</item>
      <item>If T is a record type <c>#Name{F_1, ..., F_k}</c>,
       where each <c>F_i</c> is a record field type, then Rep(T) =
       <c>{type,LINE,record,[Rep(Name),Rep(F_1), ..., Rep(F_k)]}</c>.
       For Rep(F), see below.</item>
      <item>If T is a remote type <c>M:N(T_1, ..., T_k)</c>, then Rep(T) =
       <c>{remote_type,LINE,[Rep(M),Rep(N),[Rep(T_1), ..., Rep(T_k)]]}</c>.
       </item>
      <item>If T is a tuple type <c>tuple()</c>, then Rep(T) =
       <c>{type,LINE,tuple,any}</c>.</item>
      <item>If T is a tuple type <c>{T_1, ..., T_k}</c>, then Rep(T) =
       <c>{type,LINE,tuple,[Rep(T_1), ..., Rep(T_k)]}</c>.</item>
      <item>If T is a type union <c>T_1 | ... | T_k</c>, then Rep(T) =
       <c>{type,LINE,union,[Rep(T_1), ..., Rep(T_k)]}</c>.</item>
      <item>If T is a type variable <c>V</c>, then Rep(T) =
       <c>{var,LINE,A}</c>, where <c>A</c> is an atom with a printname
       consisting of the same characters as <c>V</c>. A type variable
       is any variable except underscore (<c>_</c>).</item>
      <item>If T is a user-defined type <c>N(T_1, ..., T_k)</c>,
       then Rep(T) =
       <c>{user_type,LINE,N,[Rep(T_1), ..., Rep(T_k)]}</c>.</item>
    </list>

    <section>
      <title>Function Types</title>
      <p>A function type Ft is one of the following alternatives:</p>
      <list type="bulleted">
       <item>If Ft is a constrained function type <c>Ft_1 when Fc</c>,
	where <c>Ft_1</c> is a function type and
	<c>Fc</c> is a function constraint, then Rep(T) =
	<c>{type,LINE,bounded_fun,[Rep(Ft_1),Rep(Fc)]}</c>.
        For Rep(Fc), see below.</item>
       <item>If Ft is a function type <c>(T_1, ..., T_n) -> T_0</c>,
        where each <c>T_i</c> is a type, then
	Rep(Ft) = <c>{type,LINE,'fun',[{type,LINE,product,[Rep(T_1),
	..., Rep(T_n)]},Rep(T_0)]}</c>.</item>
      </list>
    </section>

    <section>
      <title>Function Constraints</title>
      <p>A function constraint Fc is a nonempty sequence of constraints
	<c>C_1, ..., C_k</c>, and
	Rep(Fc) = <c>[Rep(C_1), ..., Rep(C_k)]</c>.</p>
      <list type="bulleted">
       <item>If C is a constraint <c>is_subtype(V, T)</c> or <c>V :: T</c>,
        where <c>V</c> is a type variable and <c>T</c> is a type, then
        Rep(C) = <c>{type,LINE,constraint,[{atom,LINE,is_subtype},[Rep(V),Rep(T)]]}</c>.
       </item>
      </list>
    </section>

    <section>
      <title>Association Types</title>
      <list type="bulleted">
       <item>If A is an association type <c>K => V</c>, where
        <c>K</c> and <c>V</c> are types, then Rep(A) =
        <c>{type,LINE,map_field_assoc,[Rep(K),Rep(V)]}</c>.</item>
       <item>If A is an association type <c>K := V</c>, where
        <c>K</c> and <c>V</c> are types, then Rep(A) =
        <c>{type,LINE,map_field_exact,[Rep(K),Rep(V)]}</c>.</item>
      </list>
    </section>

    <section>
      <title>Record Field Types</title>
      <list type="bulleted">
       <item>If F is a record field type <c>Name :: Type</c>,
	where <c>Type</c> is a type, then Rep(F) =
	<c>{type,LINE,field_type,[Rep(Name),Rep(Type)]}</c>.</item>
      </list>
    </section>
  </section>

  <section>
    <title>The Abstract Format After Preprocessing</title>
    <p>The compilation option <c>debug_info</c> can be given to the
      compiler to have the abstract code stored in
      the <c>abstract_code</c> chunk in the BEAM file
      (for debugging purposes).</p>
    <p>In OTP R9C and later, the <c>abstract_code</c> chunk will
      contain</p>
    <p><c>{raw_abstract_v1,AbstractCode}</c></p>
    <p>where <c>AbstractCode</c> is the abstract code as described
      in this document.</p>
    <p>In releases of OTP prior to R9C, the abstract code after some more
      processing was stored in the BEAM file. The first element of the
      tuple would be either <c>abstract_v1</c> (R7B) or <c>abstract_v2</c>
      (R8B).</p>
  </section>
</chapter>