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>2015</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><![CDATA[compile:forms/[1,2]]]></c>
    and functions in the modules
    <c><![CDATA[epp]]></c>,
    <c><![CDATA[erl_eval]]></c>,
    <c><![CDATA[erl_lint]]></c>,
    <c><![CDATA[erl_pp]]></c>,
    <c><![CDATA[erl_parse]]></c>,
    and
    <c><![CDATA[io]]></c>.
    They are also used as input and output for parse transforms (see the module
    <c><![CDATA[compile]]></c>).</p>
  <p>We use the function <c><![CDATA[Rep]]></c> to denote the mapping from an Erlang source
    construct <c><![CDATA[C]]></c> to its abstract format representation <c><![CDATA[R]]></c>, and write
    <c><![CDATA[R = Rep(C)]]></c>.
    </p>
  <p>The word <c><![CDATA[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><![CDATA[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><![CDATA[F_1]]></c>, ..., <c><![CDATA[F_k]]></c>, then
       Rep(D) = <c><![CDATA[[Rep(F_1), ..., Rep(F_k)]]]></c>.</item>
      <item>If F is an attribute <c><![CDATA[-module(Mod)]]></c>, then
       Rep(F) = <c><![CDATA[{attribute,LINE,module,Mod}]]></c>.</item>
      <item>If F is an attribute <c><![CDATA[-behavior(Behavior)]]></c>, then
       Rep(F) = <c><![CDATA[{attribute,LINE,behavior,Behavior}]]></c>.</item>
      <item>If F is an attribute <c><![CDATA[-behaviour(Behaviour)]]></c>, then
       Rep(F) = <c><![CDATA[{attribute,LINE,behaviour,Behaviour}]]></c>.</item>
      <item>If F is an attribute <c><![CDATA[-export([Fun_1/A_1, ..., Fun_k/A_k])]]></c>, then
       Rep(F) = <c><![CDATA[{attribute,LINE,export,[{Fun_1,A_1}, ..., {Fun_k,A_k}]}]]></c>.</item>
      <item>If F is an attribute <c><![CDATA[-import(Mod,[Fun_1/A_1, ..., Fun_k/A_k])]]></c>, then
       Rep(F) = <c><![CDATA[{attribute,LINE,import,{Mod,[{Fun_1,A_1}, ..., {Fun_k,A_k}]}}]]></c>.</item>
      <item>If F is an attribute <c><![CDATA[-export_type([Type_1/A_1, ..., Type_k/A_k])]]></c>, then
       Rep(F) = <c><![CDATA[{attribute,LINE,export_type,[{Type_1,A_1}, ..., {Type_k,A_k}]}]]></c>.</item>
      <item>If F is an attribute <c><![CDATA[-compile(Options)]]></c>, then
       Rep(F) = <c><![CDATA[{attribute,LINE,compile,Options}]]></c>.</item>
      <item>If F is an attribute <c><![CDATA[-file(File,Line)]]></c>, then
       Rep(F) = <c><![CDATA[{attribute,LINE,file,{File,Line}}]]></c>.</item>
      <item>If F is a record declaration
       <c>-record(Name,{V_1, ..., V_k})</c>, 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 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 wild attribute <c><![CDATA[-A(T)]]></c>, then
       Rep(F) = <c><![CDATA[{attribute,LINE,A,T}]]></c>.
      <br></br></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>
    </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><![CDATA[A]]></c>, then
         Rep(V) = <c><![CDATA[{record_field,LINE,Rep(A)}]]></c>.</item>
        <item>If V is <c><![CDATA[A = E]]></c>,
	 where <c>E</c> is an expression, then
         Rep(V) = <c><![CDATA[{record_field,LINE,Rep(A),Rep(E)}]]></c>.</item>
        <item>If V is <c>A :: T</c>, where <c>T</c> is a
          type and it does not contain
          <c>undefined</c> syntactically, then Rep(V) =
          <c>{typed_record_field,{record_field,LINE,Rep(A)},Rep(undefined | T)}</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 integer or character literal, then
       Rep(L) = <c><![CDATA[{integer,LINE,L}]]></c>.</item>
      <item>If L is a float literal, then
       Rep(L) = <c><![CDATA[{float,LINE,L}]]></c>.</item>
      <item>If L is a string literal consisting of the characters
      <c><![CDATA[C_1]]></c>, ..., <c><![CDATA[C_k]]></c>, then
       Rep(L) = <c><![CDATA[{string,LINE,[C_1, ..., C_k]}]]></c>.</item>
      <item>If L is an atom literal, then
       Rep(L) = <c><![CDATA[{atom,LINE,L}]]></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 <c><![CDATA[Ps]]></c> is a sequence of patterns <c><![CDATA[P_1, ..., P_k]]></c>, then
      Rep(Ps) = <c><![CDATA[[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 L, then Rep(P) = Rep(L).</item>
      <item>If P is a compound pattern <c><![CDATA[P_1 = P_2]]></c>, then
       Rep(P) = <c><![CDATA[{match,LINE,Rep(P_1),Rep(P_2)}]]></c>.</item>
      <item>If P is a variable pattern <c><![CDATA[V]]></c>, then
       Rep(P) = <c><![CDATA[{var,LINE,A}]]></c>,
       where A is an atom with a printname consisting of the same characters as
      <c><![CDATA[V]]></c>.</item>
      <item>If P is a universal pattern <c><![CDATA[_]]></c>, then
       Rep(P) = <c><![CDATA[{var,LINE,'_'}]]></c>.</item>
      <item>If P is a tuple pattern <c><![CDATA[{P_1, ..., P_k}]]></c>, then
       Rep(P) = <c><![CDATA[{tuple,LINE,[Rep(P_1), ..., Rep(P_k)]}]]></c>.</item>
      <item>If P is a nil pattern <c><![CDATA[[]]]></c>, then
       Rep(P) = <c><![CDATA[{nil,LINE}]]></c>.</item>
      <item>If P is a cons pattern <c><![CDATA[[P_h | P_t]]]></c>, then
       Rep(P) = <c><![CDATA[{cons,LINE,Rep(P_h),Rep(P_t)}]]></c>.</item>
      <item>If E is a binary pattern <c><![CDATA[<<P_1:Size_1/TSL_1, ..., P_k:Size_k/TSL_k>>]]></c>, then
       Rep(E) = <c><![CDATA[{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><![CDATA[Size]]></c> is represented by <c><![CDATA[default]]></c>. An omitted <c><![CDATA[TSL]]></c>
       (type specifier list) is represented by <c><![CDATA[default]]></c>.</item>
      <item>If P is <c><![CDATA[P_1 Op P_2]]></c>, where <c><![CDATA[Op]]></c> is a binary operator (this
       is either an occurrence of <c><![CDATA[++]]></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><![CDATA[{op,LINE,Op,Rep(P_1),Rep(P_2)}]]></c>.</item>
      <item>If P is <c><![CDATA[Op P_0]]></c>, where <c><![CDATA[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><![CDATA[{op,LINE,Op,Rep(P_0)}]]></c>.</item>
      <item>If P is a record pattern <c><![CDATA[#Name{Field_1=P_1, ..., Field_k=P_k}]]></c>,
       then Rep(P) =
      <c><![CDATA[{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 <c><![CDATA[#Name.Field]]></c>, then
       Rep(P) = <c><![CDATA[{record_index,LINE,Name,Rep(Field)}]]></c>.</item>
      <item>If P is <c><![CDATA[( P_0 )]]></c>, then
       Rep(P) = <c><![CDATA[Rep(P_0)]]></c>,
       that is, patterns cannot be distinguished from their bodies.</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 sequence of expressions <c><![CDATA[E_1, ..., E_k]]></c>, and
      Rep(B) = <c><![CDATA[[Rep(E_1), ..., Rep(E_k)]]]></c>.</p>
    <p>An expression E is one of the following alternatives:</p>
    <list type="bulleted">
      <item>If P is an atomic literal <c><![CDATA[L]]></c>, then
       Rep(P) = Rep(L).</item>
      <item>If E is <c><![CDATA[P = E_0]]></c>, then
       Rep(E) = <c><![CDATA[{match,LINE,Rep(P),Rep(E_0)}]]></c>.</item>
      <item>If E is a variable <c><![CDATA[V]]></c>, then
       Rep(E) = <c><![CDATA[{var,LINE,A}]]></c>,
       where <c><![CDATA[A]]></c> is an atom with a printname consisting of the same
       characters as <c><![CDATA[V]]></c>.</item>
      <item>If E is a tuple skeleton <c><![CDATA[{E_1, ..., E_k}]]></c>, then
       Rep(E) = <c><![CDATA[{tuple,LINE,[Rep(E_1), ..., Rep(E_k)]}]]></c>.</item>
      <item>If E is <c><![CDATA[[]]]></c>, then
       Rep(E) = <c><![CDATA[{nil,LINE}]]></c>.</item>
      <item>If E is a cons skeleton <c><![CDATA[[E_h | E_t]]]></c>, then
       Rep(E) = <c><![CDATA[{cons,LINE,Rep(E_h),Rep(E_t)}]]></c>.</item>
      <item>If E is a binary constructor <c><![CDATA[<<V_1:Size_1/TSL_1, ..., V_k:Size_k/TSL_k>>]]></c>, then
       Rep(E) = <c><![CDATA[{bin,LINE,[{bin_element,LINE,Rep(V_1),Rep(Size_1),Rep(TSL_1)}, ..., {bin_element,LINE,Rep(V_k),Rep(Size_k),Rep(TSL_k)}]}]]></c>.
       For Rep(TSL), see below.
       An omitted <c><![CDATA[Size]]></c> is represented by <c><![CDATA[default]]></c>. An omitted <c><![CDATA[TSL]]></c>
       (type specifier list) is represented by <c><![CDATA[default]]></c>.</item>
      <item>If E is <c><![CDATA[E_1 Op E_2]]></c>, where <c><![CDATA[Op]]></c> is a binary operator,
       then Rep(E) = <c><![CDATA[{op,LINE,Op,Rep(E_1),Rep(E_2)}]]></c>.</item>
      <item>If E is <c><![CDATA[Op E_0]]></c>, where <c><![CDATA[Op]]></c> is a unary operator, then
       Rep(E) = <c><![CDATA[{op,LINE,Op,Rep(E_0)}]]></c>.</item>
      <item>If E is <c><![CDATA[#Name{Field_1=E_1, ..., Field_k=E_k}]]></c>, then
       Rep(E) =
      <c><![CDATA[{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 <c><![CDATA[E_0#Name{Field_1=E_1, ..., Field_k=E_k}]]></c>, then
       Rep(E) =
      <c><![CDATA[{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 <c><![CDATA[#Name.Field]]></c>, then
       Rep(E) = <c><![CDATA[{record_index,LINE,Name,Rep(Field)}]]></c>.</item>
      <item>If E is <c><![CDATA[E_0#Name.Field]]></c>, then
       Rep(E) = <c><![CDATA[{record_field,LINE,Rep(E_0),Name,Rep(Field)}]]></c>.</item>
      <item>If E is <c><![CDATA[#{W_1, ..., W_k}]]></c> where each
       <c><![CDATA[W_i]]></c> is a map assoc or exact field, then Rep(E) =
       <c><![CDATA[{map,LINE,[Rep(W_1), ..., Rep(W_k)]}]]></c>. For Rep(W), see
       below.</item>
      <item>If E is <c><![CDATA[E_0#{W_1, ..., W_k}]]></c> where
       <c><![CDATA[W_i]]></c> is a map assoc or exact field, then Rep(E) =
       <c><![CDATA[{map,LINE,Rep(E_0),[Rep(W_1), ..., Rep(W_k)]}]]></c>. For
       Rep(W), see below.</item>
      <item>If E is <c><![CDATA[catch E_0]]></c>, then
       Rep(E) = <c><![CDATA[{'catch',LINE,Rep(E_0)}]]></c>.</item>
      <item>If E is <c><![CDATA[E_0(E_1, ..., E_k)]]></c>, then
       Rep(E) = <c><![CDATA[{call,LINE,Rep(E_0),[Rep(E_1), ..., Rep(E_k)]}]]></c>.</item>
      <item>If E is <c><![CDATA[E_m:E_0(E_1, ..., E_k)]]></c>, then
       Rep(E) =
      <c><![CDATA[{call,LINE,{remote,LINE,Rep(E_m),Rep(E_0)},[Rep(E_1), ...,  Rep(E_k)]}]]></c>.</item>
      <item>If E is a list comprehension <c><![CDATA[[E_0 || W_1, ..., W_k]]]></c>,
       where each <c><![CDATA[W_i]]></c> is a generator or a filter, then
       Rep(E) = <c><![CDATA[{lc,LINE,Rep(E_0),[Rep(W_1), ..., Rep(W_k)]}]]></c>. For Rep(W), see
       below.</item>
      <item>If E is a binary comprehension <c><![CDATA[<<E_0 || W_1, ..., W_k>>]]></c>,
       where each <c><![CDATA[W_i]]></c> is a generator or a filter, then
       Rep(E) = <c><![CDATA[{bc,LINE,Rep(E_0),[Rep(W_1), ..., Rep(W_k)]}]]></c>. For Rep(W), see
       below.</item>
      <item>If E is <c><![CDATA[begin B end]]></c>, where <c><![CDATA[B]]></c> is a body, then
       Rep(E) = <c><![CDATA[{block,LINE,Rep(B)}]]></c>.</item>
      <item>If E is <c><![CDATA[if Ic_1 ; ... ; Ic_k  end]]></c>,
       where each <c><![CDATA[Ic_i]]></c> is an if clause then
       Rep(E) =
      <c><![CDATA[{'if',LINE,[Rep(Ic_1), ..., Rep(Ic_k)]}]]></c>.</item>
      <item>If E is <c><![CDATA[case E_0 of Cc_1 ; ... ; Cc_k end]]></c>,
       where <c><![CDATA[E_0]]></c> is an expression and each <c><![CDATA[Cc_i]]></c> is a
       case clause then
       Rep(E) =
      <c><![CDATA[{'case',LINE,Rep(E_0),[Rep(Cc_1), ..., Rep(Cc_k)]}]]></c>.</item>
      <item>If E is <c><![CDATA[try B catch Tc_1 ; ... ; Tc_k end]]></c>,
       where <c><![CDATA[B]]></c> is a body and each <c><![CDATA[Tc_i]]></c> is a catch clause then 
       Rep(E) =
      <c><![CDATA[{'try',LINE,Rep(B),[],[Rep(Tc_1), ..., Rep(Tc_k)],[]}]]></c>.</item>
      <item>If E is <c><![CDATA[try B of Cc_1 ; ... ; Cc_k catch Tc_1 ; ... ; Tc_n end]]></c>,
       where <c><![CDATA[B]]></c> is a body, 
       each <c><![CDATA[Cc_i]]></c> is a case clause and 
       each <c><![CDATA[Tc_j]]></c> is a catch clause then 
       Rep(E) =
      <c><![CDATA[{'try',LINE,Rep(B),[Rep(Cc_1), ..., Rep(Cc_k)],[Rep(Tc_1), ..., Rep(Tc_n)],[]}]]></c>.</item>
      <item>If E is <c><![CDATA[try B after A end]]></c>,
       where <c><![CDATA[B]]></c> and <c><![CDATA[A]]></c> are bodies then
       Rep(E) =
      <c><![CDATA[{'try',LINE,Rep(B),[],[],Rep(A)}]]></c>.</item>
      <item>If E is <c><![CDATA[try B of Cc_1 ; ... ; Cc_k after A end]]></c>,
       where <c><![CDATA[B]]></c> and <c><![CDATA[A]]></c> are a bodies and
       each <c><![CDATA[Cc_i]]></c> is a case clause then 
       Rep(E) =
      <c><![CDATA[{'try',LINE,Rep(B),[Rep(Cc_1), ..., Rep(Cc_k)],[],Rep(A)}]]></c>.</item>
      <item>If E is <c><![CDATA[try B catch Tc_1 ; ... ; Tc_k after A end]]></c>,
       where <c><![CDATA[B]]></c> and <c><![CDATA[A]]></c> are bodies and
       each <c><![CDATA[Tc_i]]></c> is a catch clause then 
       Rep(E) =
      <c><![CDATA[{'try',LINE,Rep(B),[],[Rep(Tc_1), ..., Rep(Tc_k)],Rep(A)}]]></c>.</item>
      <item>If E is <c><![CDATA[try B of Cc_1 ; ... ; Cc_k  catch Tc_1 ; ... ; Tc_n after A end]]></c>,
       where <c><![CDATA[B]]></c> and <c><![CDATA[A]]></c> are a bodies, 
       each <c><![CDATA[Cc_i]]></c> is a case clause and
       each <c><![CDATA[Tc_j]]></c> is a catch clause then 
       Rep(E) =
      <c><![CDATA[{'try',LINE,Rep(B),[Rep(Cc_1), ..., Rep(Cc_k)],[Rep(Tc_1),  ..., Rep(Tc_n)],Rep(A)}]]></c>.</item>
      <item>If E is <c><![CDATA[receive Cc_1 ; ... ; Cc_k end]]></c>,
       where each <c><![CDATA[Cc_i]]></c> is a case clause then
       Rep(E) =
      <c><![CDATA[{'receive',LINE,[Rep(Cc_1), ..., Rep(Cc_k)]}]]></c>.</item>
      <item>If E is <c><![CDATA[receive Cc_1 ; ... ; Cc_k after E_0 -> B_t end]]></c>,
       where each <c><![CDATA[Cc_i]]></c> is a case clause, 
      <c><![CDATA[E_0]]></c> is an expression and <c><![CDATA[B_t]]></c> is a body, then
       Rep(E) =
      <c><![CDATA[{'receive',LINE,[Rep(Cc_1), ..., Rep(Cc_k)],Rep(E_0),Rep(B_t)}]]></c>.</item>
      <item>If E is <c><![CDATA[fun Name / Arity]]></c>, then
       Rep(E) = <c><![CDATA[{'fun',LINE,{function,Name,Arity}}]]></c>.</item>
      <item>If E is <c><![CDATA[fun Module:Name/Arity]]></c>, then
       Rep(E) = <c><![CDATA[{'fun',LINE,{function,Rep(Module),Rep(Name),Rep(Arity)}}]]></c>.
      (Before the R15 release: Rep(E) = <c><![CDATA[{'fun',LINE,{function,Module,Name,Arity}}]]></c>.)</item>
      <item>If E is <c><![CDATA[fun Fc_1 ; ... ; Fc_k end]]></c> 
       where each <c><![CDATA[Fc_i]]></c> is a function clause then Rep(E) =
      <c><![CDATA[{'fun',LINE,{clauses,[Rep(Fc_1), ..., Rep(Fc_k)]}}]]></c>.</item>
      <item>If E is <c><![CDATA[fun Name Fc_1 ; ... ; Name Fc_k end]]></c>
       where <c><![CDATA[Name]]></c> is a variable and each
       <c><![CDATA[Fc_i]]></c> is a function clause then Rep(E) =
       <c><![CDATA[{named_fun,LINE,Name,[Rep(Fc_1), ..., Rep(Fc_k)]}]]></c>.
       </item>
      <item>If E is <c><![CDATA[( E_0 )]]></c>, then
       Rep(E) = <c>Rep(E_0)</c>, that is, parenthesized
       expressions cannot be distinguished from their bodies.</item>
    </list>

    <section>
      <title>Generators and Filters</title>
      <p>When W is a generator or a filter (in the body of a list or binary comprehension), then:</p>
      <list type="bulleted">
        <item>If W is a generator <c><![CDATA[P <- E]]></c>, where <c><![CDATA[P]]></c> is a pattern and <c><![CDATA[E]]></c>
         is an expression, then
         Rep(W) = <c><![CDATA[{generate,LINE,Rep(P),Rep(E)}]]></c>.</item>
        <item>If W is a generator <c><![CDATA[P <= E]]></c>, where <c><![CDATA[P]]></c> is a pattern and <c><![CDATA[E]]></c>
         is an expression, then
         Rep(W) = <c><![CDATA[{b_generate,LINE,Rep(P),Rep(E)}]]></c>.</item>
        <item>If W is a filter <c><![CDATA[E]]></c>, which is an expression, then
         Rep(W) = <c><![CDATA[Rep(E)]]></c>.</item>
      </list>
    </section>

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

    <section>
      <title>Map Assoc and Exact Fields</title>
      <p>When W is an assoc or exact field (in the body of a map), then:</p>
      <list type="bulleted">
        <item>If W is an assoc field <c><![CDATA[K => V]]></c>, where
         <c><![CDATA[K]]></c> and <c><![CDATA[V]]></c> are both expressions,
         then Rep(W) = <c><![CDATA[{map_field_assoc,LINE,Rep(K),Rep(V)}]]></c>.
         </item>
        <item>If W is an exact field <c><![CDATA[K := V]]></c>, where
         <c><![CDATA[K]]></c> and <c><![CDATA[V]]></c> are both expressions,
         then Rep(W) = <c><![CDATA[{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><![CDATA[C]]></c> is one of the following alternatives:</p>
    <list type="bulleted">
      <item>If C is a function clause <c><![CDATA[( Ps ) -> B]]></c> 
       where <c><![CDATA[Ps]]></c> is a pattern sequence and <c><![CDATA[B]]></c> is a body, then 
       Rep(C) = <c><![CDATA[{clause,LINE,Rep(Ps),[],Rep(B)}]]></c>.</item>
      <item>If C is a function clause <c><![CDATA[( Ps ) when Gs -> B]]></c> 
       where <c><![CDATA[Ps]]></c> is a pattern sequence, 
      <c><![CDATA[Gs]]></c> is a guard sequence and <c><![CDATA[B]]></c> is a body, then 
       Rep(C) = <c><![CDATA[{clause,LINE,Rep(Ps),Rep(Gs),Rep(B)}]]></c>.</item>
      <item>If C is an if clause <c><![CDATA[Gs -> B]]></c> 
       where <c><![CDATA[Gs]]></c> is a guard sequence and <c><![CDATA[B]]></c> is a body, then 
       Rep(C) = <c><![CDATA[{clause,LINE,[],Rep(Gs),Rep(B)}]]></c>.</item>
      <item>If C is a case clause <c><![CDATA[P -> B]]></c> 
       where <c><![CDATA[P]]></c> is a pattern and <c><![CDATA[B]]></c> is a body, then 
       Rep(C) = <c><![CDATA[{clause,LINE,[Rep(P)],[],Rep(B)}]]></c>.</item>
      <item>If C is a case clause <c><![CDATA[P when Gs -> B]]></c> 
       where <c><![CDATA[P]]></c> is a pattern, 
      <c><![CDATA[Gs]]></c> is a guard sequence and <c><![CDATA[B]]></c> is a body, then 
       Rep(C) = <c><![CDATA[{clause,LINE,[Rep(P)],Rep(Gs),Rep(B)}]]></c>.</item>
      <item>If C is a catch clause <c><![CDATA[P -> B]]></c> 
       where <c><![CDATA[P]]></c> is a pattern and <c><![CDATA[B]]></c> is a body, then 
       Rep(C) = <c><![CDATA[{clause,LINE,[Rep({throw,P,_})],[],Rep(B)}]]></c>.</item>
      <item>If C is a catch clause <c><![CDATA[X : P -> B]]></c> 
       where <c><![CDATA[X]]></c> is an atomic literal or a variable pattern, 
      <c><![CDATA[P]]></c> is a pattern and <c><![CDATA[B]]></c> is a body, then 
       Rep(C) = <c><![CDATA[{clause,LINE,[Rep({X,P,_})],[],Rep(B)}]]></c>.</item>
      <item>If C is a catch clause <c><![CDATA[P when Gs -> B]]></c> 
       where <c><![CDATA[P]]></c> is a pattern, <c><![CDATA[Gs]]></c> is a guard sequence 
       and <c><![CDATA[B]]></c> is a body, then 
       Rep(C) = <c><![CDATA[{clause,LINE,[Rep({throw,P,_})],Rep(Gs),Rep(B)}]]></c>.</item>
      <item>If C is a catch clause <c><![CDATA[X : P when Gs -> B]]></c> 
       where <c><![CDATA[X]]></c> is an atomic literal or a variable pattern, 
      <c><![CDATA[P]]></c> is a pattern, <c><![CDATA[Gs]]></c> is a guard sequence 
       and <c><![CDATA[B]]></c> is a body, then 
       Rep(C) = <c><![CDATA[{clause,LINE,[Rep({X,P,_})],Rep(Gs),Rep(B)}]]></c>.</item>
    </list>
  </section>

  <section>
    <title>Guards</title>
    <p>A guard sequence Gs is a sequence of guards <c><![CDATA[G_1; ...; G_k]]></c>, and
      Rep(Gs) = <c><![CDATA[[Rep(G_1), ..., Rep(G_k)]]]></c>. If the guard sequence is
      empty, Rep(Gs) = <c><![CDATA[[]]]></c>.</p>
    <p>A guard G is a nonempty sequence of guard tests <c><![CDATA[Gt_1, ..., Gt_k]]></c>, and
      Rep(G) = <c><![CDATA[[Rep(Gt_1), ..., Rep(Gt_k)]]]></c>.</p>
    <p>A guard test <c><![CDATA[Gt]]></c> is one of the following alternatives:</p>
    <list type="bulleted">
      <item>If Gt is an atomic literal L, then Rep(Gt) = Rep(L).</item>
      <item>If Gt is a variable pattern <c><![CDATA[V]]></c>, then
       Rep(Gt) = <c><![CDATA[{var,LINE,A}]]></c>,
       where A is an atom with a printname consisting of the same characters as
      <c><![CDATA[V]]></c>.</item>
      <item>If Gt is a tuple skeleton <c><![CDATA[{Gt_1, ..., Gt_k}]]></c>, then
       Rep(Gt) = <c><![CDATA[{tuple,LINE,[Rep(Gt_1), ..., Rep(Gt_k)]}]]></c>.</item>
      <item>If Gt is <c><![CDATA[[]]]></c>, then
       Rep(Gt) = <c><![CDATA[{nil,LINE}]]></c>.</item>
      <item>If Gt is a cons skeleton <c><![CDATA[[Gt_h | Gt_t]]]></c>, then
       Rep(Gt) = <c><![CDATA[{cons,LINE,Rep(Gt_h),Rep(Gt_t)}]]></c>.</item>
      <item>If Gt is a binary constructor <c><![CDATA[<<Gt_1:Size_1/TSL_1, ..., Gt_k:Size_k/TSL_k>>]]></c>, then
       Rep(Gt) = <c><![CDATA[{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><![CDATA[Size]]></c> is represented by <c><![CDATA[default]]></c>. An omitted <c><![CDATA[TSL]]></c>
       (type specifier list) is represented by <c><![CDATA[default]]></c>.</item>
      <item>If Gt is <c><![CDATA[Gt_1 Op Gt_2]]></c>, where <c><![CDATA[Op]]></c>
       is a binary operator, then Rep(Gt) = <c><![CDATA[{op,LINE,Op,Rep(Gt_1),Rep(Gt_2)}]]></c>.</item>
      <item>If Gt is <c><![CDATA[Op Gt_0]]></c>, where <c><![CDATA[Op]]></c> is a unary operator, then
       Rep(Gt) = <c><![CDATA[{op,LINE,Op,Rep(Gt_0)}]]></c>.</item>
      <item>If Gt is <c><![CDATA[#Name{Field_1=Gt_1, ..., Field_k=Gt_k}]]></c>, then
       Rep(E) =
      <c><![CDATA[{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 <c><![CDATA[#Name.Field]]></c>, then
       Rep(Gt) = <c><![CDATA[{record_index,LINE,Name,Rep(Field)}]]></c>.</item>
      <item>If Gt is <c><![CDATA[Gt_0#Name.Field]]></c>, then
       Rep(Gt) = <c><![CDATA[{record_field,LINE,Rep(Gt_0),Name,Rep(Field)}]]></c>.</item>
      <item>If Gt is <c><![CDATA[A(Gt_1, ..., Gt_k)]]></c>, where <c><![CDATA[A]]></c> is an atom, then
       Rep(Gt) = <c><![CDATA[{call,LINE,Rep(A),[Rep(Gt_1), ..., Rep(Gt_k)]}]]></c>.</item>
      <item>If Gt is <c><![CDATA[A_m:A(Gt_1, ..., Gt_k)]]></c>, where <c><![CDATA[A_m]]></c> is 
       the atom <c><![CDATA[erlang]]></c> and <c><![CDATA[A]]></c> is an atom or an operator, then
       Rep(Gt) = <c><![CDATA[{call,LINE,{remote,LINE,Rep(A_m),Rep(A)},[Rep(Gt_1), ..., Rep(Gt_k)]}]]></c>.</item>
      <item>If Gt is <c><![CDATA[{A_m,A}(Gt_1, ..., Gt_k)]]></c>, where <c><![CDATA[A_m]]></c> is 
       the atom <c><![CDATA[erlang]]></c> and <c><![CDATA[A]]></c> is an atom or an operator, then
       Rep(Gt) = <c><![CDATA[{call,LINE,Rep({A_m,A}),[Rep(Gt_1), ..., Rep(Gt_k)]}]]></c>.</item>
      <item>If Gt is <c><![CDATA[( Gt_0 )]]></c>, then
       Rep(Gt) = <c><![CDATA[Rep(Gt_0)]]></c>, that is, parenthesized
       guard tests cannot be distinguished from their bodies.</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>Anno :: Type</c>,
       where <c>Anno</c> is a variable and
       <c>Type</c> is a type, then Rep(T) =
       <c>{ann_type,LINE,[Rep(Anno),Rep(Type)]}</c>.</item>
      <item>If T is an atom or integer literal L, then Rep(T) = Rep(L).
      </item>
      <item>If T is <c>L Op R</c>,
       where <c>Op</c> is a binary operator and <c>L</c> and <c>R</c>
       are types  (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(L),Rep(R)}</c>.</item>
      <item>If T is <c>Op A</c>, where <c>Op</c> is a
       unary operator and <c>A</c> is a type (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(A)}</c>.</item>
      <item>If T is a bitstring 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((...) -> B)</c>,
       where <c>B</c> is a type, then
       Rep(T) = <c>{type,LINE,'fun',[{type,LINE,any},Rep(B)]}</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>.</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>#{P_1, ..., P_k}</c>, where each
       <c>P_i</c> is a map pair type, then Rep(T) =
       <c>{type,LINE,map,[Rep(P_1), ..., Rep(P_k)]}</c>.</item>
      <item>If T is a map pair type <c>K => V</c>, where
       <c>K</c> and <c>V</c> are types, then Rep(T) =
       <c>{type,LINE,map_field_assoc,[Rep(K),Rep(V)]}</c>.</item>
      <item>If T is a predefined (or built-in) type <c>N(A_1, ..., A_k)</c>,
       where each <c>A_i</c> is a type, then Rep(T) =
       <c>{type,LINE,N,[Rep(A_1), ..., Rep(A_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>.
       </item>
      <item>If T is a record field type <c>Name :: Type</c>,
       where <c>Type</c> is a type, then Rep(T) =
       <c>{type,LINE,field_type,[Rep(Name),Rep(Type)]}</c>.</item>
      <item>If T is a remote type <c>M:N(A_1, ..., A_k)</c>, where
       each <c>A_i</c> is a type, then Rep(T) =
       <c>{remote_type,LINE,[Rep(M),Rep(N),[Rep(A_1), ..., Rep(A_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>{A_1, ..., A_k}</c>, where
       each <c>A_i</c> is a type, then Rep(T) =
       <c>{type,LINE,tuple,[Rep(A_1), ..., Rep(A_k)]}</c>.</item>
      <item>If T is a type union <c>T_1 | ... | T_k</c>,
       where each <c>T_i</c> is a type, 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(A_1, ..., A_k)</c>,
       where each <c>A_i</c> is a type, then Rep(T) =
       <c>{user_type,LINE,N,[Rep(A_1), ..., Rep(A_k)]}</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>
    </list>

    <section>
      <title>Function Types</title>
      <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>.</item>
       <item>If Ft is a function type <c>(A_1, ..., A_n) -> B</c>,
        where each <c>A_i</c> and <c>B</c> are types, then
	Rep(Ft) = <c>{type,LINE,'fun',[{type,LINE,product,[Rep(A_1),
	..., Rep(A_n)]},Rep(B)]}</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,[Rep(F),[Rep(V),Rep(T)]]}</c>.
       </item>
      </list>
    </section>
  </section>

  <section>
    <title>The Abstract Format After Preprocessing</title>
    <p>The compilation option <c><![CDATA[debug_info]]></c> can be given to the
      compiler to have the abstract code stored in 
      the <c><![CDATA[abstract_code]]></c> chunk in the BEAM file
      (for debugging purposes).</p>
    <p>In OTP R9C and later, the <c><![CDATA[abstract_code]]></c> chunk will
      contain</p>
    <p><c><![CDATA[{raw_abstract_v1,AbstractCode}]]></c></p>
    <p>where <c><![CDATA[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><![CDATA[abstract_v1]]></c> (R7B) or <c><![CDATA[abstract_v2]]></c>
      (R8B).</p>
  </section>
</chapter>