doc-src/IsarImplementation/Thy/logic.thy
author wenzelm
Mon Sep 11 14:35:25 2006 +0200 (2006-09-11)
changeset 20501 de0b523b0d62
parent 20498 825a8d2335ce
child 20514 5ede702cd2ca
permissions -rw-r--r--
more rules;
wenzelm@18537
     1
wenzelm@18537
     2
(* $Id$ *)
wenzelm@18537
     3
wenzelm@18537
     4
theory logic imports base begin
wenzelm@18537
     5
wenzelm@20470
     6
chapter {* Primitive logic \label{ch:logic} *}
wenzelm@18537
     7
wenzelm@20480
     8
text {*
wenzelm@20480
     9
  The logical foundations of Isabelle/Isar are that of the Pure logic,
wenzelm@20480
    10
  which has been introduced as a natural-deduction framework in
wenzelm@20480
    11
  \cite{paulson700}.  This is essentially the same logic as ``@{text
wenzelm@20493
    12
  "\<lambda>HOL"}'' in the more abstract setting of Pure Type Systems (PTS)
wenzelm@20480
    13
  \cite{Barendregt-Geuvers:2001}, although there are some key
wenzelm@20491
    14
  differences in the specific treatment of simple types in
wenzelm@20491
    15
  Isabelle/Pure.
wenzelm@20480
    16
wenzelm@20480
    17
  Following type-theoretic parlance, the Pure logic consists of three
wenzelm@20480
    18
  levels of @{text "\<lambda>"}-calculus with corresponding arrows: @{text
wenzelm@20480
    19
  "\<Rightarrow>"} for syntactic function space (terms depending on terms), @{text
wenzelm@20480
    20
  "\<And>"} for universal quantification (proofs depending on terms), and
wenzelm@20480
    21
  @{text "\<Longrightarrow>"} for implication (proofs depending on proofs).
wenzelm@20480
    22
wenzelm@20480
    23
  Pure derivations are relative to a logical theory, which declares
wenzelm@20491
    24
  type constructors, term constants, and axioms.  Theory declarations
wenzelm@20491
    25
  support schematic polymorphism, which is strictly speaking outside
wenzelm@20491
    26
  the logic.\footnote{Incidently, this is the main logical reason, why
wenzelm@20491
    27
  the theory context @{text "\<Theta>"} is separate from the context @{text
wenzelm@20491
    28
  "\<Gamma>"} of the core calculus.}
wenzelm@20480
    29
*}
wenzelm@20480
    30
wenzelm@20480
    31
wenzelm@20451
    32
section {* Types \label{sec:types} *}
wenzelm@20437
    33
wenzelm@20437
    34
text {*
wenzelm@20480
    35
  The language of types is an uninterpreted order-sorted first-order
wenzelm@20480
    36
  algebra; types are qualified by ordered type classes.
wenzelm@20480
    37
wenzelm@20480
    38
  \medskip A \emph{type class} is an abstract syntactic entity
wenzelm@20480
    39
  declared in the theory context.  The \emph{subclass relation} @{text
wenzelm@20480
    40
  "c\<^isub>1 \<subseteq> c\<^isub>2"} is specified by stating an acyclic
wenzelm@20491
    41
  generating relation; the transitive closure is maintained
wenzelm@20491
    42
  internally.  The resulting relation is an ordering: reflexive,
wenzelm@20491
    43
  transitive, and antisymmetric.
wenzelm@20451
    44
wenzelm@20480
    45
  A \emph{sort} is a list of type classes written as @{text
wenzelm@20480
    46
  "{c\<^isub>1, \<dots>, c\<^isub>m}"}, which represents symbolic
wenzelm@20480
    47
  intersection.  Notationally, the curly braces are omitted for
wenzelm@20480
    48
  singleton intersections, i.e.\ any class @{text "c"} may be read as
wenzelm@20480
    49
  a sort @{text "{c}"}.  The ordering on type classes is extended to
wenzelm@20491
    50
  sorts according to the meaning of intersections: @{text
wenzelm@20491
    51
  "{c\<^isub>1, \<dots> c\<^isub>m} \<subseteq> {d\<^isub>1, \<dots>, d\<^isub>n}"} iff
wenzelm@20491
    52
  @{text "\<forall>j. \<exists>i. c\<^isub>i \<subseteq> d\<^isub>j"}.  The empty intersection
wenzelm@20491
    53
  @{text "{}"} refers to the universal sort, which is the largest
wenzelm@20491
    54
  element wrt.\ the sort order.  The intersections of all (finitely
wenzelm@20491
    55
  many) classes declared in the current theory are the minimal
wenzelm@20491
    56
  elements wrt.\ the sort order.
wenzelm@20480
    57
wenzelm@20491
    58
  \medskip A \emph{fixed type variable} is a pair of a basic name
wenzelm@20493
    59
  (starting with a @{text "'"} character) and a sort constraint.  For
wenzelm@20480
    60
  example, @{text "('a, s)"} which is usually printed as @{text
wenzelm@20480
    61
  "\<alpha>\<^isub>s"}.  A \emph{schematic type variable} is a pair of an
wenzelm@20480
    62
  indexname and a sort constraint.  For example, @{text "(('a, 0),
wenzelm@20491
    63
  s)"} which is usually printed as @{text "?\<alpha>\<^isub>s"}.
wenzelm@20451
    64
wenzelm@20480
    65
  Note that \emph{all} syntactic components contribute to the identity
wenzelm@20493
    66
  of type variables, including the sort constraint.  The core logic
wenzelm@20493
    67
  handles type variables with the same name but different sorts as
wenzelm@20493
    68
  different, although some outer layers of the system make it hard to
wenzelm@20493
    69
  produce anything like this.
wenzelm@20451
    70
wenzelm@20493
    71
  A \emph{type constructor} @{text "\<kappa>"} is a @{text "k"}-ary operator
wenzelm@20493
    72
  on types declared in the theory.  Type constructor application is
wenzelm@20494
    73
  usually written postfix as @{text "(\<alpha>\<^isub>1, \<dots>, \<alpha>\<^isub>k)\<kappa>"}.
wenzelm@20494
    74
  For @{text "k = 0"} the argument tuple is omitted, e.g.\ @{text
wenzelm@20494
    75
  "prop"} instead of @{text "()prop"}.  For @{text "k = 1"} the
wenzelm@20494
    76
  parentheses are omitted, e.g.\ @{text "\<alpha> list"} instead of @{text
wenzelm@20494
    77
  "(\<alpha>)list"}.  Further notation is provided for specific constructors,
wenzelm@20494
    78
  notably the right-associative infix @{text "\<alpha> \<Rightarrow> \<beta>"} instead of
wenzelm@20494
    79
  @{text "(\<alpha>, \<beta>)fun"}.
wenzelm@20480
    80
  
wenzelm@20480
    81
  A \emph{type} is defined inductively over type variables and type
wenzelm@20491
    82
  constructors as follows: @{text "\<tau> = \<alpha>\<^isub>s | ?\<alpha>\<^isub>s |
wenzelm@20494
    83
  (\<tau>\<^sub>1, \<dots>, \<tau>\<^sub>k)k"}.
wenzelm@20451
    84
wenzelm@20494
    85
  A \emph{type abbreviation} is a syntactic abbreviation @{text
wenzelm@20494
    86
  "(\<^vec>\<alpha>)\<kappa> = \<tau>"} of an arbitrary type expression @{text "\<tau>"} over
wenzelm@20494
    87
  variables @{text "\<^vec>\<alpha>"}.  Type abbreviations looks like type
wenzelm@20494
    88
  constructors at the surface, but are fully expanded before entering
wenzelm@20494
    89
  the logical core.
wenzelm@20480
    90
wenzelm@20480
    91
  A \emph{type arity} declares the image behavior of a type
wenzelm@20494
    92
  constructor wrt.\ the algebra of sorts: @{text "\<kappa> :: (s\<^isub>1, \<dots>,
wenzelm@20494
    93
  s\<^isub>k)s"} means that @{text "(\<tau>\<^isub>1, \<dots>, \<tau>\<^isub>k)\<kappa>"} is
wenzelm@20494
    94
  of sort @{text "s"} if every argument type @{text "\<tau>\<^isub>i"} is
wenzelm@20494
    95
  of sort @{text "s\<^isub>i"}.  Arity declarations are implicitly
wenzelm@20494
    96
  completed, i.e.\ @{text "\<kappa> :: (\<^vec>s)c"} entails @{text "\<kappa> ::
wenzelm@20491
    97
  (\<^vec>s)c'"} for any @{text "c' \<supseteq> c"}.
wenzelm@20491
    98
wenzelm@20491
    99
  \medskip The sort algebra is always maintained as \emph{coregular},
wenzelm@20491
   100
  which means that type arities are consistent with the subclass
wenzelm@20494
   101
  relation: for each type constructor @{text "\<kappa>"} and classes @{text
wenzelm@20494
   102
  "c\<^isub>1 \<subseteq> c\<^isub>2"}, any arity @{text "\<kappa> ::
wenzelm@20494
   103
  (\<^vec>s\<^isub>1)c\<^isub>1"} has a corresponding arity @{text "\<kappa>
wenzelm@20480
   104
  :: (\<^vec>s\<^isub>2)c\<^isub>2"} where @{text "\<^vec>s\<^isub>1 \<subseteq>
wenzelm@20494
   105
  \<^vec>s\<^isub>2"} holds componentwise.
wenzelm@20451
   106
wenzelm@20491
   107
  The key property of a coregular order-sorted algebra is that sort
wenzelm@20494
   108
  constraints may be always solved in a most general fashion: for each
wenzelm@20494
   109
  type constructor @{text "\<kappa>"} and sort @{text "s"} there is a most
wenzelm@20494
   110
  general vector of argument sorts @{text "(s\<^isub>1, \<dots>,
wenzelm@20491
   111
  s\<^isub>k)"} such that a type scheme @{text
wenzelm@20494
   112
  "(\<alpha>\<^bsub>s\<^isub>1\<^esub>, \<dots>, \<alpha>\<^bsub>s\<^isub>k\<^esub>)\<kappa>"} is
wenzelm@20491
   113
  of sort @{text "s"}.  Consequently, the unification problem on the
wenzelm@20491
   114
  algebra of types has most general solutions (modulo renaming and
wenzelm@20491
   115
  equivalence of sorts).  Moreover, the usual type-inference algorithm
wenzelm@20491
   116
  will produce primary types as expected \cite{nipkow-prehofer}.
wenzelm@20480
   117
*}
wenzelm@20451
   118
wenzelm@20480
   119
text %mlref {*
wenzelm@20480
   120
  \begin{mldecls}
wenzelm@20480
   121
  @{index_ML_type class} \\
wenzelm@20480
   122
  @{index_ML_type sort} \\
wenzelm@20494
   123
  @{index_ML_type arity} \\
wenzelm@20480
   124
  @{index_ML_type typ} \\
wenzelm@20494
   125
  @{index_ML fold_atyps: "(typ -> 'a -> 'a) -> typ -> 'a -> 'a"} \\
wenzelm@20480
   126
  @{index_ML Sign.subsort: "theory -> sort * sort -> bool"} \\
wenzelm@20480
   127
  @{index_ML Sign.of_sort: "theory -> typ * sort -> bool"} \\
wenzelm@20480
   128
  @{index_ML Sign.add_types: "(bstring * int * mixfix) list -> theory -> theory"} \\
wenzelm@20480
   129
  @{index_ML Sign.add_tyabbrs_i: "
wenzelm@20480
   130
  (bstring * string list * typ * mixfix) list -> theory -> theory"} \\
wenzelm@20480
   131
  @{index_ML Sign.primitive_class: "string * class list -> theory -> theory"} \\
wenzelm@20480
   132
  @{index_ML Sign.primitive_classrel: "class * class -> theory -> theory"} \\
wenzelm@20480
   133
  @{index_ML Sign.primitive_arity: "arity -> theory -> theory"} \\
wenzelm@20480
   134
  \end{mldecls}
wenzelm@20480
   135
wenzelm@20480
   136
  \begin{description}
wenzelm@20480
   137
wenzelm@20480
   138
  \item @{ML_type class} represents type classes; this is an alias for
wenzelm@20480
   139
  @{ML_type string}.
wenzelm@20480
   140
wenzelm@20480
   141
  \item @{ML_type sort} represents sorts; this is an alias for
wenzelm@20480
   142
  @{ML_type "class list"}.
wenzelm@20451
   143
wenzelm@20480
   144
  \item @{ML_type arity} represents type arities; this is an alias for
wenzelm@20494
   145
  triples of the form @{text "(\<kappa>, \<^vec>s, s)"} for @{text "\<kappa> ::
wenzelm@20480
   146
  (\<^vec>s)s"} described above.
wenzelm@20480
   147
wenzelm@20480
   148
  \item @{ML_type typ} represents types; this is a datatype with
wenzelm@20480
   149
  constructors @{ML TFree}, @{ML TVar}, @{ML Type}.
wenzelm@20480
   150
wenzelm@20494
   151
  \item @{ML fold_atyps}~@{text "f \<tau>"} iterates function @{text "f"}
wenzelm@20494
   152
  over all occurrences of atoms (@{ML TFree} or @{ML TVar}) of @{text
wenzelm@20494
   153
  "\<tau>"}; the type structure is traversed from left to right.
wenzelm@20494
   154
wenzelm@20480
   155
  \item @{ML Sign.subsort}~@{text "thy (s\<^isub>1, s\<^isub>2)"}
wenzelm@20480
   156
  tests the subsort relation @{text "s\<^isub>1 \<subseteq> s\<^isub>2"}.
wenzelm@20480
   157
wenzelm@20480
   158
  \item @{ML Sign.of_sort}~@{text "thy (\<tau>, s)"} tests whether a type
wenzelm@20491
   159
  is of a given sort.
wenzelm@20480
   160
wenzelm@20494
   161
  \item @{ML Sign.add_types}~@{text "[(\<kappa>, k, mx), \<dots>]"} declares new
wenzelm@20494
   162
  type constructors @{text "\<kappa>"} with @{text "k"} arguments and
wenzelm@20480
   163
  optional mixfix syntax.
wenzelm@20451
   164
wenzelm@20494
   165
  \item @{ML Sign.add_tyabbrs_i}~@{text "[(\<kappa>, \<^vec>\<alpha>, \<tau>, mx), \<dots>]"}
wenzelm@20494
   166
  defines a new type abbreviation @{text "(\<^vec>\<alpha>)\<kappa> = \<tau>"} with
wenzelm@20491
   167
  optional mixfix syntax.
wenzelm@20480
   168
wenzelm@20480
   169
  \item @{ML Sign.primitive_class}~@{text "(c, [c\<^isub>1, \<dots>,
wenzelm@20494
   170
  c\<^isub>n])"} declares new class @{text "c"}, together with class
wenzelm@20494
   171
  relations @{text "c \<subseteq> c\<^isub>i"}, for @{text "i = 1, \<dots>, n"}.
wenzelm@20480
   172
wenzelm@20480
   173
  \item @{ML Sign.primitive_classrel}~@{text "(c\<^isub>1,
wenzelm@20480
   174
  c\<^isub>2)"} declares class relation @{text "c\<^isub>1 \<subseteq>
wenzelm@20480
   175
  c\<^isub>2"}.
wenzelm@20480
   176
wenzelm@20494
   177
  \item @{ML Sign.primitive_arity}~@{text "(\<kappa>, \<^vec>s, s)"} declares
wenzelm@20494
   178
  arity @{text "\<kappa> :: (\<^vec>s)s"}.
wenzelm@20480
   179
wenzelm@20480
   180
  \end{description}
wenzelm@20437
   181
*}
wenzelm@20437
   182
wenzelm@20437
   183
wenzelm@20480
   184
wenzelm@20451
   185
section {* Terms \label{sec:terms} *}
wenzelm@18537
   186
wenzelm@18537
   187
text {*
wenzelm@20451
   188
  \glossary{Term}{FIXME}
wenzelm@18537
   189
wenzelm@20491
   190
  The language of terms is that of simply-typed @{text "\<lambda>"}-calculus
wenzelm@20491
   191
  with de-Bruijn indices for bound variables, and named free
wenzelm@20491
   192
  variables, and constants.  Terms with loose bound variables are
wenzelm@20491
   193
  usually considered malformed.  The types of variables and constants
wenzelm@20491
   194
  is stored explicitly at each occurrence in the term (which is a
wenzelm@20491
   195
  known performance issue).
wenzelm@20491
   196
wenzelm@20451
   197
  FIXME de-Bruijn representation of lambda terms
wenzelm@20480
   198
wenzelm@20480
   199
  Term syntax provides explicit abstraction @{text "\<lambda>x :: \<alpha>. b(x)"}
wenzelm@20480
   200
  and application @{text "t u"}, while types are usually implicit
wenzelm@20480
   201
  thanks to type-inference.
wenzelm@20480
   202
wenzelm@20498
   203
wenzelm@20498
   204
  \[
wenzelm@20501
   205
  \infer{@{text "a\<^isub>\<tau> :: \<tau>"}}{}
wenzelm@20498
   206
  \qquad
wenzelm@20501
   207
  \infer{@{text "(\<lambda>x\<^sub>\<tau>. t) :: \<tau> \<Rightarrow> \<sigma>"}}{@{text "t :: \<sigma>"}}
wenzelm@20501
   208
  \qquad
wenzelm@20501
   209
  \infer{@{text "t u :: \<sigma>"}}{@{text "t :: \<tau> \<Rightarrow> \<sigma>"} & @{text "u :: \<tau>"}}
wenzelm@20498
   210
  \]
wenzelm@20498
   211
wenzelm@18537
   212
*}
wenzelm@18537
   213
wenzelm@18537
   214
wenzelm@18537
   215
text {*
wenzelm@18537
   216
wenzelm@18537
   217
FIXME
wenzelm@18537
   218
wenzelm@18537
   219
\glossary{Schematic polymorphism}{FIXME}
wenzelm@18537
   220
wenzelm@18537
   221
\glossary{Type variable}{FIXME}
wenzelm@18537
   222
wenzelm@18537
   223
*}
wenzelm@18537
   224
wenzelm@18537
   225
wenzelm@20451
   226
section {* Theorems \label{sec:thms} *}
wenzelm@18537
   227
wenzelm@18537
   228
text {*
wenzelm@20501
   229
  \glossary{Proposition}{A \seeglossary{term} of \seeglossary{type}
wenzelm@20501
   230
  @{text "prop"}.  Internally, there is nothing special about
wenzelm@20501
   231
  propositions apart from their type, but the concrete syntax enforces
wenzelm@20501
   232
  a clear distinction.  Propositions are structured via implication
wenzelm@20501
   233
  @{text "A \<Longrightarrow> B"} or universal quantification @{text "\<And>x. B x"} ---
wenzelm@20501
   234
  anything else is considered atomic.  The canonical form for
wenzelm@20501
   235
  propositions is that of a \seeglossary{Hereditary Harrop Formula}. FIXME}
wenzelm@20480
   236
wenzelm@20501
   237
  \glossary{Theorem}{A proven proposition within a certain theory and
wenzelm@20501
   238
  proof context, formally @{text "\<Gamma> \<turnstile>\<^sub>\<Theta> \<phi>"}; both contexts are
wenzelm@20501
   239
  rarely spelled out explicitly.  Theorems are usually normalized
wenzelm@20501
   240
  according to the \seeglossary{HHF} format. FIXME}
wenzelm@20480
   241
wenzelm@20501
   242
  \glossary{Fact}{Sometimes used interchangably for
wenzelm@20501
   243
  \seeglossary{theorem}.  Strictly speaking, a list of theorems,
wenzelm@20501
   244
  essentially an extra-logical conjunction.  Facts emerge either as
wenzelm@20501
   245
  local assumptions, or as results of local goal statements --- both
wenzelm@20501
   246
  may be simultaneous, hence the list representation. FIXME}
wenzelm@18537
   247
wenzelm@20501
   248
  \glossary{Schematic variable}{FIXME}
wenzelm@20501
   249
wenzelm@20501
   250
  \glossary{Fixed variable}{A variable that is bound within a certain
wenzelm@20501
   251
  proof context; an arbitrary-but-fixed entity within a portion of
wenzelm@20501
   252
  proof text. FIXME}
wenzelm@18537
   253
wenzelm@20501
   254
  \glossary{Free variable}{Synonymous for \seeglossary{fixed
wenzelm@20501
   255
  variable}. FIXME}
wenzelm@20501
   256
wenzelm@20501
   257
  \glossary{Bound variable}{FIXME}
wenzelm@18537
   258
wenzelm@20501
   259
  \glossary{Variable}{See \seeglossary{schematic variable},
wenzelm@20501
   260
  \seeglossary{fixed variable}, \seeglossary{bound variable}, or
wenzelm@20501
   261
  \seeglossary{type variable}.  The distinguishing feature of
wenzelm@20501
   262
  different variables is their binding scope. FIXME}
wenzelm@18537
   263
wenzelm@20501
   264
  A \emph{proposition} is a well-formed term of type @{text "prop"}.
wenzelm@20501
   265
  The connectives of minimal logic are declared as constants of the
wenzelm@20501
   266
  basic theory:
wenzelm@18537
   267
wenzelm@20501
   268
  \smallskip
wenzelm@20501
   269
  \begin{tabular}{ll}
wenzelm@20501
   270
  @{text "all :: (\<alpha> \<Rightarrow> prop) \<Rightarrow> prop"} & universal quantification (binder @{text "\<And>"}) \\
wenzelm@20501
   271
  @{text "\<Longrightarrow> :: prop \<Rightarrow> prop \<Rightarrow> prop"} & implication (right associative infix) \\
wenzelm@20501
   272
  \end{tabular}
wenzelm@18537
   273
wenzelm@20501
   274
  \medskip A \emph{theorem} is a proven proposition, depending on a
wenzelm@20501
   275
  collection of assumptions, and axioms from the theory context.  The
wenzelm@20501
   276
  judgment @{text "A\<^isub>1, \<dots>, A\<^isub>n \<turnstile> B"} is defined
wenzelm@20501
   277
  inductively by the primitive inferences given in
wenzelm@20501
   278
  \figref{fig:prim-rules}; there is a global syntactic restriction
wenzelm@20501
   279
  that the hypotheses may not contain schematic variables.
wenzelm@18537
   280
wenzelm@20501
   281
  \begin{figure}[htb]
wenzelm@20501
   282
  \begin{center}
wenzelm@20498
   283
  \[
wenzelm@20498
   284
  \infer[@{text "(axiom)"}]{@{text "\<turnstile> A"}}{@{text "A \<in> \<Theta>"}}
wenzelm@20498
   285
  \qquad
wenzelm@20498
   286
  \infer[@{text "(assume)"}]{@{text "A \<turnstile> A"}}{}
wenzelm@20498
   287
  \]
wenzelm@20498
   288
  \[
wenzelm@20498
   289
  \infer[@{text "(\<And>_intro)"}]{@{text "\<Gamma> \<turnstile> \<And>x. b x"}}{@{text "\<Gamma> \<turnstile> b x"} & @{text "x \<notin> \<Gamma>"}}
wenzelm@20498
   290
  \qquad
wenzelm@20498
   291
  \infer[@{text "(\<And>_elim)"}]{@{text "\<Gamma> \<turnstile> b a"}}{@{text "\<Gamma> \<turnstile> \<And>x. b x"}}
wenzelm@20498
   292
  \]
wenzelm@20498
   293
  \[
wenzelm@20498
   294
  \infer[@{text "(\<Longrightarrow>_intro)"}]{@{text "\<Gamma> - A \<turnstile> A \<Longrightarrow> B"}}{@{text "\<Gamma> \<turnstile> B"}}
wenzelm@20498
   295
  \qquad
wenzelm@20498
   296
  \infer[@{text "(\<Longrightarrow>_elim)"}]{@{text "\<Gamma>\<^sub>1 \<union> \<Gamma>\<^sub>2 \<turnstile> B"}}{@{text "\<Gamma>\<^sub>1 \<turnstile> A \<Longrightarrow> B"} & @{text "\<Gamma>\<^sub>2 \<turnstile> A"}}
wenzelm@20498
   297
  \]
wenzelm@20501
   298
  \caption{Primitive inferences of the Pure logic}\label{fig:prim-rules}
wenzelm@20501
   299
  \end{center}
wenzelm@20501
   300
  \end{figure}
wenzelm@18537
   301
wenzelm@20501
   302
  The introduction and elimination rules for @{text "\<And>"} and @{text
wenzelm@20501
   303
  "\<Longrightarrow>"} are analogous to formation of (dependently typed) @{text
wenzelm@20501
   304
  "\<lambda>"}-terms representing the underlying proof objects.  Proof terms
wenzelm@20501
   305
  are \emph{irrelevant} in the Pure logic, they may never occur within
wenzelm@20501
   306
  propositions, i.e.\ the @{text "\<Longrightarrow>"} arrow of the framework is a
wenzelm@20501
   307
  non-dependent one.
wenzelm@20491
   308
wenzelm@20501
   309
  Also note that fixed parameters as in @{text "\<And>_intro"} need not be
wenzelm@20501
   310
  recorded in the context @{text "\<Gamma>"}, since syntactic types are
wenzelm@20501
   311
  always inhabitable.  An ``assumption'' @{text "x :: \<tau>"} is logically
wenzelm@20501
   312
  vacuous, because @{text "\<tau>"} is always non-empty.  This is the deeper
wenzelm@20501
   313
  reason why @{text "\<Gamma>"} only consists of hypothetical proofs, but no
wenzelm@20501
   314
  hypothetical terms.
wenzelm@20501
   315
wenzelm@20501
   316
  The corresponding proof terms are left implicit in the classic
wenzelm@20501
   317
  ``LCF-approach'', although they could be exploited separately
wenzelm@20501
   318
  \cite{Berghofer-Nipkow:2000}.  The implementation provides a runtime
wenzelm@20501
   319
  option to control the generation of full proof terms.
wenzelm@20501
   320
wenzelm@20501
   321
  \medskip The axiomatization of a theory is implicitly closed by
wenzelm@20501
   322
  forming all instances of type and term variables: @{text "\<turnstile> A\<theta>"} for
wenzelm@20501
   323
  any substirution instance of axiom @{text "\<turnstile> A"}.  By pushing
wenzelm@20501
   324
  substitution through derivations inductively, we get admissible
wenzelm@20501
   325
  substitution rules for theorems shown in \figref{fig:subst-rules}.
wenzelm@20501
   326
wenzelm@20501
   327
  \begin{figure}[htb]
wenzelm@20501
   328
  \begin{center}
wenzelm@20498
   329
  \[
wenzelm@20501
   330
  \infer{@{text "\<Gamma> \<turnstile> B[?\<alpha>]"}}{@{text "\<Gamma> \<turnstile> B[\<alpha>]"} & @{text "\<alpha> \<notin> \<Gamma>"}}
wenzelm@20501
   331
  \quad
wenzelm@20501
   332
  \infer[\quad@{text "(generalize)"}]{@{text "\<Gamma> \<turnstile> B[?x]"}}{@{text "\<Gamma> \<turnstile> B[x]"} & @{text "x \<notin> \<Gamma>"}}
wenzelm@20498
   333
  \]
wenzelm@20498
   334
  \[
wenzelm@20501
   335
  \infer{@{text "\<Gamma> \<turnstile> B[\<tau>]"}}{@{text "\<Gamma> \<turnstile> B[?\<alpha>]"}}
wenzelm@20501
   336
  \quad
wenzelm@20501
   337
  \infer[\quad@{text "(instantiate)"}]{@{text "\<Gamma> \<turnstile> B[t]"}}{@{text "\<Gamma> \<turnstile> B[?x]"}}
wenzelm@20498
   338
  \]
wenzelm@20501
   339
  \caption{Admissible substitution rules}\label{fig:subst-rules}
wenzelm@20501
   340
  \end{center}
wenzelm@20501
   341
  \end{figure}
wenzelm@18537
   342
wenzelm@20498
   343
  Note that @{text "instantiate_term"} could be derived using @{text
wenzelm@20498
   344
  "\<And>_intro/elim"}, but this is not how it is implemented.  The type
wenzelm@20501
   345
  instantiation rule is a genuine admissible one, due to the lack of
wenzelm@20501
   346
  true polymorphism in the logic.
wenzelm@20498
   347
wenzelm@20501
   348
  Since @{text "\<Gamma>"} may never contain any schematic variables, the
wenzelm@20501
   349
  @{text "instantiate"} do not require an explicit side-condition.  In
wenzelm@20501
   350
  principle, variables could be substituted in hypotheses as well, but
wenzelm@20501
   351
  this could disrupt monotonicity of the basic calculus: derivations
wenzelm@20501
   352
  could leave the current proof context.
wenzelm@20498
   353
wenzelm@20501
   354
  \medskip The framework also provides builtin equality @{text "\<equiv>"},
wenzelm@20501
   355
  which is conceptually axiomatized shown in \figref{fig:equality},
wenzelm@20501
   356
  although the implementation provides derived rules directly:
wenzelm@20501
   357
wenzelm@20501
   358
  \begin{figure}[htb]
wenzelm@20501
   359
  \begin{center}
wenzelm@20498
   360
  \begin{tabular}{ll}
wenzelm@20498
   361
  @{text "\<equiv> :: \<alpha> \<Rightarrow> \<alpha> \<Rightarrow> prop"} & equality relation (infix) \\
wenzelm@20501
   362
  @{text "\<turnstile> (\<lambda>x. b x) a \<equiv> b a"} & @{text "\<beta>"}-conversion \\
wenzelm@20498
   363
  @{text "\<turnstile> x \<equiv> x"} & reflexivity law \\
wenzelm@20498
   364
  @{text "\<turnstile> x \<equiv> y \<Longrightarrow> P x \<Longrightarrow> P y"} & substitution law \\
wenzelm@20498
   365
  @{text "\<turnstile> (\<And>x. f x \<equiv> g x) \<Longrightarrow> f \<equiv> g"} & extensionality \\
wenzelm@20498
   366
  @{text "\<turnstile> (A \<Longrightarrow> B) \<Longrightarrow> (B \<Longrightarrow> A) \<Longrightarrow> A \<equiv> B"} & coincidence with equivalence \\
wenzelm@20498
   367
  \end{tabular}
wenzelm@20501
   368
  \caption{Conceptual axiomatization of equality.}\label{fig:equality}
wenzelm@20501
   369
  \end{center}
wenzelm@20501
   370
  \end{figure}
wenzelm@20501
   371
wenzelm@20501
   372
  Since the basic representation of terms already accounts for @{text
wenzelm@20501
   373
  "\<alpha>"}-conversion, Pure equality essentially acts like @{text
wenzelm@20501
   374
  "\<alpha>\<beta>\<eta>"}-equivalence on terms, while coinciding with bi-implication.
wenzelm@20498
   375
wenzelm@20498
   376
wenzelm@20501
   377
  \medskip Conjunction is defined in Pure as a derived connective, see
wenzelm@20501
   378
  \figref{fig:conjunction}.  This is occasionally useful to represent
wenzelm@20501
   379
  simultaneous statements behind the scenes --- framework conjunction
wenzelm@20501
   380
  is usually not exposed to the user.
wenzelm@20498
   381
wenzelm@20501
   382
  \begin{figure}[htb]
wenzelm@20501
   383
  \begin{center}
wenzelm@20501
   384
  \begin{tabular}{ll}
wenzelm@20501
   385
  @{text "& :: prop \<Rightarrow> prop \<Rightarrow> prop"} & conjunction (hidden) \\
wenzelm@20501
   386
  @{text "\<turnstile> A & B \<equiv> (\<And>C. (A \<Longrightarrow> B \<Longrightarrow> C) \<Longrightarrow> C)"} \\
wenzelm@20501
   387
  \end{tabular}
wenzelm@20501
   388
  \caption{Definition of conjunction.}\label{fig:equality}
wenzelm@20501
   389
  \end{center}
wenzelm@20501
   390
  \end{figure}
wenzelm@20501
   391
wenzelm@20501
   392
  The definition allows to derive the usual introduction @{text "\<turnstile> A \<Longrightarrow>
wenzelm@20501
   393
  B \<Longrightarrow> A & B"}, and destructions @{text "A & B \<Longrightarrow> A"} and @{text "A & B
wenzelm@20501
   394
  \<Longrightarrow> B"}.
wenzelm@20491
   395
*}
wenzelm@18537
   396
wenzelm@20480
   397
wenzelm@20491
   398
section {* Rules \label{sec:rules} *}
wenzelm@18537
   399
wenzelm@18537
   400
text {*
wenzelm@18537
   401
wenzelm@18537
   402
FIXME
wenzelm@18537
   403
wenzelm@20491
   404
  A \emph{rule} is any Pure theorem in HHF normal form; there is a
wenzelm@20491
   405
  separate calculus for rule composition, which is modeled after
wenzelm@20491
   406
  Gentzen's Natural Deduction \cite{Gentzen:1935}, but allows
wenzelm@20491
   407
  rules to be nested arbitrarily, similar to \cite{extensions91}.
wenzelm@20491
   408
wenzelm@20491
   409
  Normally, all theorems accessible to the user are proper rules.
wenzelm@20491
   410
  Low-level inferences are occasional required internally, but the
wenzelm@20491
   411
  result should be always presented in canonical form.  The higher
wenzelm@20491
   412
  interfaces of Isabelle/Isar will always produce proper rules.  It is
wenzelm@20491
   413
  important to maintain this invariant in add-on applications!
wenzelm@20491
   414
wenzelm@20491
   415
  There are two main principles of rule composition: @{text
wenzelm@20491
   416
  "resolution"} (i.e.\ backchaining of rules) and @{text
wenzelm@20491
   417
  "by-assumption"} (i.e.\ closing a branch); both principles are
wenzelm@20491
   418
  combined in the variants of @{text "elim-resosultion"} and @{text
wenzelm@20491
   419
  "dest-resolution"}.  Raw @{text "composition"} is occasionally
wenzelm@20491
   420
  useful as well, also it is strictly speaking outside of the proper
wenzelm@20491
   421
  rule calculus.
wenzelm@20491
   422
wenzelm@20491
   423
  Rules are treated modulo general higher-order unification, which is
wenzelm@20491
   424
  unification modulo the equational theory of @{text "\<alpha>\<beta>\<eta>"}-conversion
wenzelm@20491
   425
  on @{text "\<lambda>"}-terms.  Moreover, propositions are understood modulo
wenzelm@20491
   426
  the (derived) equivalence @{text "(A \<Longrightarrow> (\<And>x. B x)) \<equiv> (\<And>x. A \<Longrightarrow> B x)"}.
wenzelm@20491
   427
wenzelm@20491
   428
  This means that any operations within the rule calculus may be
wenzelm@20491
   429
  subject to spontaneous @{text "\<alpha>\<beta>\<eta>"}-HHF conversions.  It is common
wenzelm@20491
   430
  practice not to contract or expand unnecessarily.  Some mechanisms
wenzelm@20491
   431
  prefer an one form, others the opposite, so there is a potential
wenzelm@20491
   432
  danger to produce some oscillation!
wenzelm@20491
   433
wenzelm@20491
   434
  Only few operations really work \emph{modulo} HHF conversion, but
wenzelm@20491
   435
  expect a normal form: quantifiers @{text "\<And>"} before implications
wenzelm@20491
   436
  @{text "\<Longrightarrow>"} at each level of nesting.
wenzelm@20491
   437
wenzelm@18537
   438
\glossary{Hereditary Harrop Formula}{The set of propositions in HHF
wenzelm@18537
   439
format is defined inductively as @{text "H = (\<And>x\<^sup>*. H\<^sup>* \<Longrightarrow>
wenzelm@18537
   440
A)"}, for variables @{text "x"} and atomic propositions @{text "A"}.
wenzelm@18537
   441
Any proposition may be put into HHF form by normalizing with the rule
wenzelm@18537
   442
@{text "(A \<Longrightarrow> (\<And>x. B x)) \<equiv> (\<And>x. A \<Longrightarrow> B x)"}.  In Isabelle, the outermost
wenzelm@18537
   443
quantifier prefix is represented via \seeglossary{schematic
wenzelm@18537
   444
variables}, such that the top-level structure is merely that of a
wenzelm@18537
   445
\seeglossary{Horn Clause}}.
wenzelm@18537
   446
wenzelm@18537
   447
\glossary{HHF}{See \seeglossary{Hereditary Harrop Formula}.}
wenzelm@18537
   448
wenzelm@20498
   449
wenzelm@20498
   450
  \[
wenzelm@20498
   451
  \infer[@{text "(assumption)"}]{@{text "C\<vartheta>"}}
wenzelm@20498
   452
  {@{text "(\<And>\<^vec>x. \<^vec>H \<^vec>x \<Longrightarrow> A \<^vec>x) \<Longrightarrow> C"} & @{text "A\<vartheta> = H\<^sub>i\<vartheta>"}~~\text{(for some~@{text i})}}
wenzelm@20498
   453
  \]
wenzelm@20498
   454
wenzelm@20498
   455
wenzelm@20498
   456
  \[
wenzelm@20498
   457
  \infer[@{text "(compose)"}]{@{text "\<^vec>A\<vartheta> \<Longrightarrow> C\<vartheta>"}}
wenzelm@20498
   458
  {@{text "\<^vec>A \<Longrightarrow> B"} & @{text "B' \<Longrightarrow> C"} & @{text "B\<vartheta> = B'\<vartheta>"}}
wenzelm@20498
   459
  \]
wenzelm@20498
   460
wenzelm@20498
   461
wenzelm@20498
   462
  \[
wenzelm@20498
   463
  \infer[@{text "(\<And>_lift)"}]{@{text "(\<And>\<^vec>x. \<^vec>A (?\<^vec>a \<^vec>x)) \<Longrightarrow> (\<And>\<^vec>x. B (?\<^vec>a \<^vec>x))"}}{@{text "\<^vec>A ?\<^vec>a \<Longrightarrow> B ?\<^vec>a"}}
wenzelm@20498
   464
  \]
wenzelm@20498
   465
  \[
wenzelm@20498
   466
  \infer[@{text "(\<Longrightarrow>_lift)"}]{@{text "(\<^vec>H \<Longrightarrow> \<^vec>A) \<Longrightarrow> (\<^vec>H \<Longrightarrow> B)"}}{@{text "\<^vec>A \<Longrightarrow> B"}}
wenzelm@20498
   467
  \]
wenzelm@20498
   468
wenzelm@20498
   469
  The @{text resolve} scheme is now acquired from @{text "\<And>_lift"},
wenzelm@20498
   470
  @{text "\<Longrightarrow>_lift"}, and @{text compose}.
wenzelm@20498
   471
wenzelm@20498
   472
  \[
wenzelm@20498
   473
  \infer[@{text "(resolution)"}]
wenzelm@20498
   474
  {@{text "(\<And>\<^vec>x. \<^vec>H \<^vec>x \<Longrightarrow> \<^vec>A (?\<^vec>a \<^vec>x))\<vartheta> \<Longrightarrow> C\<vartheta>"}}
wenzelm@20498
   475
  {\begin{tabular}{l}
wenzelm@20498
   476
    @{text "\<^vec>A ?\<^vec>a \<Longrightarrow> B ?\<^vec>a"} \\
wenzelm@20498
   477
    @{text "(\<And>\<^vec>x. \<^vec>H \<^vec>x \<Longrightarrow> B' \<^vec>x) \<Longrightarrow> C"} \\
wenzelm@20498
   478
    @{text "(\<lambda>\<^vec>x. B (?\<^vec>a \<^vec>x))\<vartheta> = B'\<vartheta>"} \\
wenzelm@20498
   479
   \end{tabular}}
wenzelm@20498
   480
  \]
wenzelm@20498
   481
wenzelm@20498
   482
wenzelm@20498
   483
  FIXME @{text "elim_resolution"}, @{text "dest_resolution"}
wenzelm@18537
   484
*}
wenzelm@18537
   485
wenzelm@20498
   486
wenzelm@18537
   487
end