author wenzelm
Sat, 01 Jul 2000 19:49:20 +0200
changeset 9226 cbe6144f0f15
parent 104 d8205bb279a7
permissions -rw-r--r--

\chapter{Defining Logics} \label{Defining-Logics}
This chapter is intended for Isabelle experts.  It explains how to define new
logical systems, Isabelle's {\it raison d'\^etre}.  Isabelle logics are
hierarchies of theories.  A number of simple examples are contained in the
introductory manual; the full syntax of theory definitions is shown in the
{\em Reference Manual}.  The purpose of this chapter is to explain the
remaining subtleties, especially some context conditions on the class
structure and the definition of new mixfix syntax.  A full understanding of
the material requires knowledge of the internal representation of terms (data
type {\tt term}) as detailed in the {\em Reference Manual}.  Sections marked
with a * can be skipped on first reading.

\section{Classes and Types *}
\index{*arities!context conditions}

Type declarations are subject to the following two well-formedness
\item There are no two declarations $ty :: (\vec{r})c$ and $ty :: (\vec{s})c$
  with $\vec{r} \neq \vec{s}$.  For example
types ty 1
arities ty :: (\{logic\}) logic
        ty :: (\{\})logic
leads to an error message and fails.
\item If there are two declarations $ty :: (s@1,\dots,s@n)c$ and $ty ::
  (s@1',\dots,s@n')c'$ such that $c' < c$ then $s@i' \preceq s@i$ must hold
  for $i=1,\dots,n$.  The relationship $\preceq$, defined as
\[ s' \preceq s \iff \forall c\in s. \exists c'\in s'.~ c'\le c, \]
expresses that the set of types represented by $s'$ is a subset of the set of
types represented by $s$.  For example
classes term < logic
types ty 1
arities ty :: (\{logic\})logic
        ty :: (\{\})term
leads to an error message and fails.
These conditions guarantee principal types~\cite{nipkow-prehofer}.

\section{Precedence Grammars}
\index{precedence grammar|(}

The precise syntax of a logic is best defined by a context-free grammar.
These grammars obey the following conventions: identifiers denote
nonterminals, {\tt typewriter} fount denotes terminals, repetition is
indicated by \dots, and alternatives are separated by $|$.

In order to simplify the description of mathematical languages, we introduce
an extended format which permits {\bf precedences}\index{precedence}.  This
scheme generalizes precedence declarations in \ML\ and {\sc prolog}.  In this
extended grammar format, nonterminals are decorated by integers, their
precedence.  In the sequel, precedences are shown as subscripts.  A nonterminal
$A@p$ on the right-hand side of a production may only be replaced using a
production $A@q = \gamma$ where $p \le q$.

Formally, a set of context free productions $G$ induces a derivation
relation $\rew@G$ on strings as follows:
\[ \alpha A@p \beta ~\rew@G~ \alpha\gamma\beta ~~~iff~~~
   \exists q \ge p.~(A@q=\gamma) \in G
Any extended grammar of this kind can be translated into a normal context
free grammar.  However, this translation may require the introduction of a
large number of new nonterminals and productions.

The following simple grammar for arithmetic expressions demonstrates how
binding power and associativity of operators can be enforced by precedences.
$A@9$ & = & {\tt0} \\
$A@9$ & = & {\tt(} $A@0$ {\tt)} \\
$A@0$ & = & $A@0$ {\tt+} $A@1$ \\
$A@2$ & = & $A@3$ {\tt*} $A@2$ \\
$A@3$ & = & {\tt-} $A@3$
The choice of precedences determines that \verb$-$ binds tighter than
\verb$*$ which binds tighter than \verb$+$, and that \verb$+$ and \verb$*$
associate to the left and right, respectively.

To minimize the number of subscripts, we adopt the following conventions:
\item all precedences $p$ must be in the range $0 \leq p \leq max_pri$ for
  some fixed $max_pri$.
\item precedence $0$ on the right-hand side and precedence $max_pri$ on the
  left-hand side may be omitted.
In addition, we write the production $A@p = \alpha$ as $A = \alpha~(p)$.

Using these conventions and assuming $max_pri=9$, the grammar in
Example~\ref{PrecedenceEx} becomes
$A$ & = & {\tt0} & \hspace*{4em} \\
 & $|$ & {\tt(} $A$ {\tt)} \\
 & $|$ & $A$ {\tt+} $A@1$ & (0) \\
 & $|$ & $A@3$ {\tt*} $A@2$ & (2) \\
 & $|$ & {\tt-} $A@3$ & (3)

\index{precedence grammar|)}

\section{Basic syntax *}

An informal account of most of Isabelle's syntax (meta-logic, types etc) is
contained in {\em Introduction to Isabelle}.  A precise description using a
precedence grammar is shown in Figure~\ref{MetaLogicSyntax}.  This description
is the basis of all extensions by object-logics.
$prop$ &=& \ttindex{PROP} $aprop$ ~~$|$~~ {\tt(} $prop$ {\tt)} \\
     &$|$& $logic@3$ \ttindex{==} $logic@2$ & (2) \\
     &$|$& $prop@2$ \ttindex{==>} $prop@1$ & (1) \\
     &$|$& {\tt[|} $prop$ {\tt;} \dots {\tt;} $prop$ {\tt|]} {\tt==>} $prop@1$ & (1) \\
     &$|$& {\tt!!} $idts$ {\tt.} $prop$ & (0) \\\\
$logic$ &=& $prop$ ~~$|$~~ $fun$ \\\\
$aprop$ &=& $id$ ~~$|$~~ $var$
    ~~$|$~~ $fun@{max_pri}$ {\tt(} $logic$ {\tt,} \dots {\tt,} $logic$ {\tt)} \\\\
$fun$ &=& $id$ ~~$|$~~ $var$ ~~$|$~~ {\tt(} $fun$ {\tt)} \\
    &$|$& \ttindex{\%} $idts$ {\tt.} $logic$ & (0) \\\\
$idts$ &=& $idt$ ~~$|$~~ $idt@1$ $idts$ \\\\
$idt$ &=& $id$ ~~$|$~~ {\tt(} $idt$ {\tt)} \\
    &$|$& $id$ \ttindex{::} $type$ & (0) \\\\
$type$ &=& $tfree$ ~~$|$~~ $tvar$ \\
     &$|$& $tfree$ {\tt::} $sort$ ~~$|$~~ $tvar$ {\tt::} $sort$ \\
     &$|$& $id$ ~~$|$~~ $type@{max_pri}$ $id$
                ~~$|$~~ {\tt(} $type$ {\tt,} \dots {\tt,} $type$ {\tt)} $id$ \\
     &$|$& $type@1$ \ttindex{=>} $type$ & (0) \\
     &$|$& {\tt[}  $type$ {\tt,} \dots {\tt,} $type$ {\tt]} {\tt=>} $type$&(0)\\
     &$|$& {\tt(} $type$ {\tt)} \\\\
$sort$ &=& $id$ ~~$|$~~ {\tt\{\}}
                ~~$|$~~ {\tt\{} $id$ {\tt,} \dots {\tt,} $id$ {\tt\}} 
\caption{Meta-Logic Syntax}
The following main categories are defined:
\item[$prop$] Terms of type $prop$, i.e.\ formulae of the meta-logic.
\item[$aprop$] Atomic propositions.
\item[$logic$] Terms of types in class $logic$.  Initially, $logic$ contains
  merely $prop$.  As the syntax is extended by new object-logics, more
  productions for $logic$ are added (see below).
\item[$fun$] Terms potentially of function type.
\item[$type$] Types.
\item[$idts$] a list of identifiers, possibly constrained by types.  Note
  that $x::nat~y$ is parsed as $x::(nat~y)$, i.e.\ $y$ is treated like a
  type constructor applied to $nat$.

The predefined types $id$, $var$, $tfree$ and $tvar$ represent identifiers
({\tt f}), unknowns ({\tt ?f}), type variables ({\tt 'a}), and type unknowns
({\tt ?'a}) respectively.  If we think of them as nonterminals with
predefined syntax, we may assume that all their productions have precedence

\subsection{Logical types and default syntax}

Isabelle is concerned with mathematical languages which have a certain
minimal vocabulary: identifiers, variables, parentheses, and the lambda
calculus.  Logical types, i.e.\ those of class $logic$, are automatically
equipped with this basic syntax.  More precisely, for any type constructor
$ty$ with arity $(\dots)c$, where $c$ is a subclass of $logic$, the following
productions are added:
$ty$ &=& $id$ ~~$|$~~ $var$ ~~$|$~~ {\tt(} $ty$ {\tt)} \\
  &$|$& $fun@{max_pri}$ {\tt(} $logic$ {\tt,} \dots {\tt,} $logic$ {\tt)}\\
  &$|$& $ty@{max_pri}$ {\tt::} $type$\\\\
$logic$ &=& $ty$

\section{Mixfix syntax}

We distinguish between abstract and concrete syntax.  The {\em abstract}
syntax is given by the typed constants of a theory.  Abstract syntax trees are
well-typed terms, i.e.\ values of \ML\ type {\tt term}.  If none of the
constants are introduced with mixfix annotations, there is no concrete syntax
to speak of: terms can only be abstractions or applications of the form
$f(t@1,\dots,t@n)$, where $f$ is a constant or variable.  Since this notation
quickly becomes unreadable, Isabelle supports syntax definitions in the form
of unrestricted context-free grammars using mixfix annotations.

Mixfix annotations describe the {\em concrete} syntax, its translation into
the abstract syntax, and a pretty-printing scheme, all in one.  Isabelle
syntax definitions are inspired by \OBJ's~\cite{OBJ} {\em mixfix\/} syntax.
Each mixfix annotation defines a precedence grammar production and associates
an Isabelle constant with it.

A {\em mixfix declaration} {\tt consts $c$ ::\ $\tau$ ($sy$ $ps$ $p$)} is
interpreted as a grammar pro\-duction as follows:
\item $sy$ is the right-hand side of this production, specified as a {\em
    mixfix annotation}.  In general, $sy$ is of the form
  $\alpha@0\_\alpha@1\dots\alpha@{n-1}\_\alpha@n$, where each occurrence of
  ``\ttindex{_}'' denotes an argument/nonterminal and the strings
  $\alpha@i$ do not contain ``{\tt_}''.
\item $\tau$ specifies the types of the nonterminals on the left and right
  hand side. If $sy$ is of the form above, $\tau$ must be of the form
  $[\tau@1,\dots,\tau@n] \To \tau'$.  Then argument $i$ is of type $\tau@i$
  and the result, i.e.\ the left-hand side of the production, is of type
  $\tau'$.  Both the $\tau@i$ and $\tau'$ may be function types.
\item $c$ is the name of the Isabelle constant associated with this production.
  Parsing an instance of the phrase $sy$ generates the {\tt term} {\tt
    Const($c$,dummyT\footnote{Proper types are inserted later on.  See
      \S\ref{Typing}})\$$a@1$\$$\dots$\$$a@n$}\index{*dummyT}, where $a@i$ is
  the term generated by parsing the $i^{th}$ argument.
\item $ps$ must be of the form $[p@1,\dots,p@n]$, where $p@i$ is the
  minimal precedence\index{precedence} required of any phrase that may appear
  as the $i^{th}$ argument.  The null list is interpreted as a list of 0's of
  the appropriate length.
\item $p$ is the precedence of this production.
Notice that there is a close connection between abstract and concrete syntax:
each production has an associated constant, and types act as {\bf syntactic
  categories} in the concrete syntax.  To emphasize this connection, we
sometimes refer to the nonterminals on the right-hand side of a production as
its arguments and to the nonterminal on the left-hand side as its result.

The maximal legal precedence is called \ttindexbold{max_pri}, which is
currently 1000.  If you want to ignore precedences, the safest way to do so is
to use the annotation {\tt($sy$)}: this production puts no precedence
constraints on any of its arguments and has maximal precedence itself, i.e.\ 
it is always applicable and does not exclude any productions of its

In mixfix notation the grammar in Example~\ref{PrecedenceEx} can be written
as follows:
types exp 0
consts "0"  ::              "exp"  ("0" 9)
       "+"  :: "[exp,exp] => exp"  ("_ + _" [0,1] 0)
       "*"  :: "[exp,exp] => exp"  ("_ * _" [3,2] 2)
       "-"  ::       "exp => exp"  ("- _"   [3]   3)
Parsing the string \verb!"0 + - 0 + 0"! produces the term {\tt
  $p$\$($p$\$($m$\$$z$)\$$z$)\$$z$} where {\tt$p =$ Const("+",dummyT)},
{\tt$m =$ Const("-",dummyT)}, and {\tt$z =$ Const("0",dummyT)}.

The interpretation of \ttindex{_} in a mixfix annotation is always as a {\bf
  meta-character}\index{meta-character} which does not represent itself but
an argument position.  The following characters are also meta-characters:
'   (   )   /
Preceding any character with a quote (\verb$'$) turns it into an ordinary
character.  Thus you can write \verb!''! if you really want a single quote.
The purpose of the other meta-characters is explained in
\S\ref{PrettyPrinting}.  Remember that in \ML\ strings \verb$\$ is already a
(different kind of) meta-character.

\subsection{Types and syntactic categories *}

The precise mapping from types to syntactic categories is defined by the
following function:
N(\tau@1\To\tau@2) &=& fun \\
N((\tau@1,\dots,\tau@n)ty) &=& ty \\
N(\alpha) &=& logic
Only the outermost type constructor is taken into account and type variables
can range over all logical types.  This catches some ill-typed terms (like
$Cons(x,0)$, where $Cons :: [\alpha,\alpha list] \To \alpha list$ and $0 ::
nat$) but leaves the real work to the type checker.

In terms of the precedence grammar format introduced in
\S\ref{PrecedenceGrammars}, the declaration
consts \(c\) :: "[\(\tau@1\),\dots,\(\tau@n\)]\(\To\tau\)" ("\(\alpha@0\_\alpha@1\dots\alpha@{n-1}\_\alpha@n\)") [\(p@1\),\dots,\(p@n\)] \(p\))
defines the production
\[ N(\tau)@p ~~=~~ \alpha@0 ~N(\tau@1)@{p@1}~ \alpha@1~ \dots
                  ~\alpha@{n-1} ~N(\tau@n)@{p@n}~ \alpha@n

\subsection{Copy productions *}

Productions which do not create a new node in the abstract syntax tree are
called {\bf copy productions}.  They must have exactly one nonterminal on
the right hand side.  The term generated when parsing that nonterminal is
simply passed up as the result of parsing the whole copy production.  In
Isabelle a copy production is indicated by an empty constant name, i.e.\ by
consts "" :: \(\tau\)  (\(sy\) \(ps\) \(p\))

A special kind of copy production is one where, modulo white space, $sy$ is
{\tt"_"}.  It is called a {\bf chain production}.  Chain productions should be
seen as an abbreviation mechanism.  Conceptually, they are removed from the
grammar by adding appropriate new rules.  Precedence information attached to
chain productions is ignored.  The following example demonstrates the effect:
the grammar defined by
types A,B,C 0
consts AB :: "B => A"  ("A _" [10] 517)
       "" :: "C => B"  ("_"   [0]  100)
       x  :: "C"       ("x"          5)
       y  :: "C"       ("y"         15)
admits {\tt"A y"} but not {\tt"A x"}.  Had the constant in the second
production been some non-empty string, both {\tt"A y"} and {\tt"A x"} would
be legal.


\section{Lexical conventions}

The lexical analyzer distinguishes the following kinds of tokens: delimiters,
identifiers, unknowns, type variables and type unknowns.

Delimiters are user-defined, i.e.\ they are extracted from the syntax
definition.  If $\alpha@0\_\alpha@1\dots\alpha@{n-1}\_\alpha@n$ is a mixfix
annotation, each $\alpha@i$ is decomposed into substrings
$\beta@1~\dots~\beta@k$ which are separated by and do not contain
\bfindex{white space} ( = blanks, tabs, newlines).  Each $\beta@j$ becomes a
delimiter.  Thus a delimiter can be an arbitrary string not containing white

The lexical syntax of identifiers and variables ( = unknowns) is defined in
the introductory manual.  Parsing an identifier $f$ generates {\tt
  Free($f$,dummyT)}\index{*dummyT}.  Parsing a variable {\tt?}$v$ generates
{\tt Var(($u$,$i$),dummyT)} where $i$ is the integer value of the longest
numeric suffix of $v$ (possibly $0$), and $u$ is the remaining prefix.
Parsing a variable {\tt?}$v{.}i$ generates {\tt Var(($v$,$i$),dummyT)}.  The
following table covers the four different cases that can arise:
"?v" & "?v.7" & "?v5" & "?v7.5" \\
Var(("v",0),$d$) & Var(("v",7),$d$) & Var(("v",5),$d$) & Var(("v7",5),$d$)
where $d = {\tt dummyT}$.

In mixfix annotations, \ttindexbold{id}, \ttindexbold{var},
\ttindexbold{tfree} and \ttindexbold{tvar} are the predefined categories of
identifiers, unknowns, type variables and type unknowns, respectively.

The lexical analyzer translates input strings to token lists by repeatedly
taking the maximal prefix of the input string that forms a valid token.  A
maximal prefix that is both a delimiter and an identifier or variable (like
{\tt ALL}) is treated as a delimiter.  White spaces are separators.

An important consequence of this translation scheme is that delimiters need
not be separated by white space to be recognized as separate.  If \verb$"-"$
is a delimiter but \verb$"--"$ is not, the string \verb$"--"$ is treated as
two consecutive occurrences of \verb$"-"$.  This is in contrast to \ML\ which
would treat \verb$"--"$ as a single (undeclared) identifier.  The
consequence of Isabelle's more liberal scheme is that the same string may be
parsed in a different way after extending the syntax: after adding
\verb$"--"$ as a delimiter, the input \verb$"--"$ is treated as
a single occurrence of \verb$"--"$.

\section{Infix operators}

{\tt Infixl} and {\tt infixr} declare infix operators which associate to the
left and right respectively.  As in \ML, prefixing infix operators with
\ttindexbold{op} turns them into curried functions.  Infix declarations can
be reduced to mixfix ones as follows:
"$c$" ::~$\tau$ (\ttindexbold{infixl} $p$) &
"op $c$" ::~$\tau$ ("_ $c$ _" [$p$,$p+1$] $p$) \\
"$c$" ::~$\tau$ (\ttindexbold{infixr} $p$) &
"op $c$" ::~$\tau$ ("_ $c$ _" [$p+1$,$p$] $p$)

A {\bf binder} is a variable-binding constant, such as a quantifier.
The declaration
consts \(c\) :: \(\tau\)  (binder \(Q\) \(p\))
introduces a binder $c$ of type $\tau$,
which must have the form $(\tau@1\To\tau@2)\To\tau@3$.  Its concrete syntax
is $Q~x.t$.  A binder is like a generalized quantifier where $\tau@1$ is the
type of the bound variable $x$, $\tau@2$ the type of the body $t$, and
$\tau@3$ the type of the whole term.  For example $\forall$ can be declared
like this:
consts All :: "('a => o) => o"  (binder "ALL " 10)
This allows us to write $\forall x.P$ either as {\tt ALL $x$.$P$} or {\tt
  All(\%$x$.$P$)}; the latter form is for purists only.

In case $\tau@2 = \tau@3$, nested quantifications can be written as $Q~x@1
\dots x@n.t$.  From a syntactic point of view,
consts \(c\) :: "\((\tau@1\To\tau@2)\To\tau@3\)"  (binder "\(Q\)" \(p\))
is equivalent to
consts \(c\)   :: "\((\tau@1\To\tau@2)\To\tau@3\)"
       "\(Q\)" :: "[idts,\(\tau@2\)] => \(\tau@3\)"  ("\(Q\)_.  _" \(p\))
where {\tt idts} is the syntactic category $idts$ defined in

However, there is more to binders than concrete syntax: behind the scenes the
body of the quantified expression has to be converted into a
$\lambda$-abstraction (when parsing) and back again (when printing).  This
is performed by the translation mechanism, which is discussed below.  For
binders, the definition of the required translation functions has been
automated.  Many other syntactic forms, such as set comprehension, require
special treatment.

\section{Parse translations *}
\index{parse translation|(}

So far we have pretended that there is a close enough relationship between
concrete and abstract syntax to allow an automatic translation from one to
the other using the constant name supplied with each production.  In many
cases this scheme is not powerful enough, especially for constructs involving
variable bindings.  Therefore the $ML$-section of a theory definition can
associate constant names with user-defined translation functions by including
a line
val parse_translation = \dots
where the right-hand side of this binding must be an \ML-expression of type
\verb$(string * (term list -> term))list$.

After the input string has been translated into a term according to the
syntax definition, there is a second phase in which the term is translated
using the user-supplied functions in a bottom-up manner.  Given a list $tab$
of the above type, a term $t$ is translated as follows.  If $t$ is not of the
form {\tt Const($c$,$\tau$)\$$t@1$\$\dots\$$t@n$}, then $t$ is returned
unchanged.  Otherwise all $t@i$ are translated into $t@i'$.  Let {\tt $t' =$
  Const($c$,$\tau$)\$$t@1'$\$\dots\$$t@n'$}.  If there is no pair $(c,f)$ in
$tab$, return $t'$.  Otherwise apply $f$ to $[t@1',\dots,t@n']$.  If that
raises an exception, return $t'$, otherwise return the result.
\ML-lists are constructed by {\tt[]} and {\tt::}.  For readability the
list \hbox{\tt$x$::$y$::$z$::[]} can be written \hbox{\tt[$x$,$y$,$z$]}.
In Isabelle the two forms of lists are declared as follows:
types list 1
      enum 0
arities list :: (term)term
consts "[]" :: "'a list"                   ("[]")
       ":"  :: "['a, 'a list] => 'a list"  (infixr 50)
       enum :: "enum => 'a list"           ("[_]")
       sing :: "'a => enum"                ("_")
       cons :: "['a,enum] => enum"         ("_, _")
Because \verb$::$ is already used for type constraints, it is replaced by
\verb$:$ as the infix list constructor.

In order to allow list enumeration, the new type {\tt enum} is introduced.
Its only purpose is syntactic and hence it does not need an arity, in
contrast to the logical type {\tt list}.  Although \hbox{\tt[$x$,$y$,$z$]} is
syntactically legal, it needs to be translated into a term built up from
\verb$[]$ and \verb$:$.  This is what \verb$make_list$ accomplishes:
val cons = Const("op :", dummyT);

fun make_list (Const("sing",_)$e) = cons $ e $ Const("[]", dummyT)
  | make_list (Const("cons",_)$e$es) = cons $ e $ make_list es;
To hook this translation up to Isabelle's parser, the theory definition needs
to contain the following $ML$-section:
fun enum_tr[enum] = make_list enum;
val parse_translation = [("enum",enum_tr)]
This causes \verb!Const("enum",_)$!$t$ to be replaced by

Of course the definition of \verb$make_list$ should be included in the
  Isabelle represents the set $\{ x \mid P(x) \}$ internally by $Set(\lambda
  x.P(x))$.  The internal and external forms need separate
constants:\footnote{In practice, the external form typically has a name
beginning with an {\at} sign, such as {\tt {\at}SET}.  This emphasizes that
the constant should be used only for parsing/printing.}
types set 1
arities set :: (term)term
consts Set :: "('a => o) => 'a set"
       SET :: "[id,o] => 'a set"  ("\{_ | _\}")
Parsing {\tt"\{$x$ | $P$\}"} according to this syntax yields the term {\tt
  Const("SET",dummyT) \$ Free("\(x\)",dummyT) \$ \(p\)}, where $p$ is the
result of parsing $P$.  What we need is the term {\tt
  Const("Set",dummyT)\$Abs("$x$",dummyT,$p'$)}, where $p'$ is some
``abstracted'' version of $p$.  Therefore we define a function
fun set_tr[Free(s,T), p] = Const("Set", dummyT) $
                           Abs(s, T, abstract_over(Free(s,T), p));
where \verb$abstract_over: term*term -> term$ is a predefined function such
that {\tt abstract_over($u$,$t$)} replaces every occurrence of $u$ in $t$ by
a {\tt Bound} variable of the correct index (i.e.\ 0 at top level).  Remember
that {\tt dummyT} is replaced by the correct types at a later stage (see
\S\ref{Typing}).  Function {\tt set_tr} is associated with {\tt SET} by
including the \ML-text
val parse_translation = [("SET", set_tr)];

If you want to run the above examples in Isabelle, you should note that an
$ML$-section needs to contain not just a definition of
\verb$parse_translation$ but also of a variable \verb$print_translation$.  The
purpose of the latter is to reverse the effect of the former during printing;
details are found in \S\ref{Print-translations}.  Hence you need to include
the line
val print_translation = [];
This is instructive because the terms are then printed out in their internal
form.  For example the input \hbox{\tt[$x$,$y$,$z$]} is echoed as
\hbox{\tt$x$:$y$:$z$:[]}.  This helps to check that your parse translation is
working correctly.

%Explicit type constraints disappear with type checking but are still
%visible to the parse translation functions.

\index{parse translation|)}


Syntax definitions provide printing information in three distinct ways:
\item the syntax of the language (as used for parsing),
\item pretty printing information, and
\item print translation functions.
The bare mixfix declarations enable Isabelle to print terms, but the result
will not necessarily be pretty and may look different from what you expected.
To produce a pleasing layout, you need to read the following sections.

\subsection{Printing with mixfix declarations}

Let {\tt$t =$ Const($c$,_)\$$t@1$\$\dots\$$t@n$} be a term and let
consts \(c\) :: \(\tau\) (\(sy\))
be a mixfix declaration where $sy$ is of the form
$\alpha@0\_\alpha@1\dots\alpha@{n-1}\_\alpha@n$.  Printing $t$ according to
$sy$ means printing the string
$\alpha@0\beta@1\alpha@1\ldots\alpha@{n-1}\beta@n\alpha@n$, where $\beta@i$
is the result of printing $t@i$.

Note that the system does {\em not\/} insert blanks.  They should be part of
the mixfix syntax if they are required to separate tokens or achieve a
certain layout.

\subsection{Pretty printing}
\index{pretty printing}

In order to format the output, it is possible to embed pretty printing
directives in mixfix annotations.  These directives are ignored during parsing
and affect only printing.  The characters {\tt(}, {\tt)} and {\tt/} are
interpreted as meta-characters\index{meta-character} when found in a mixfix
annotation.  Their meaning is
\item[~{\tt(}~] Open a block.  A sequence of digits following it is
  interpreted as the \bfindex{indentation} of this block.  It causes the
  output to be indented by $n$ positions if a line break occurs within the
  block.  If {\tt(} is not followed by a digit, the indentation defaults to
\item[~{\tt)}~] Close a block.
\item[~\ttindex{/}~] Allow a \bfindex{line break}.  White space immediately
  following {\tt/} is not printed if the line is broken at this point.

\subsection{Print translations *}
\index{print translation|(}

Since terms are translated after parsing (see \S\ref{Parse-translations}),
there is a similar mechanism to translate them back before printing.
Therefore the $ML$-section of a theory definition can associate constant
names with user-defined translation functions by including a line
val print_translation = \dots
where the right-hand side of this binding is again an \ML-expression of type
\verb$(string * (term list -> term))list$.
Including a pair $(c,f)$ in this list causes the printer to print
$f[t@1,\dots,t@n]$ whenever it finds {\tt Const($c$,_)\$$t@1$\$\dots\$$t@n$}.
Reversing the effect of the parse translation in Example~\ref{list-enum} is
accomplished by the following function:
fun make_enum (Const("op :",_) $ e $ es) = case es of
        Const("[]",_)  =>  Const("sing",dummyT) $ e
      | _  =>  Const("enum",dummyT) $ e $ make_enum es;
It translates \hbox{\tt$x$:$y$:$z$:[]} to \hbox{\tt[$x$,$y$,$z$]}.  However,
if the input does not terminate with an empty list, e.g.\ \hbox{\tt$x$:$xs$},
\verb$make_enum$ raises exception {\tt Match}.  This signals that the
attempted translation has failed and the term should be printed as is.
The connection with Isabelle's pretty printer is established as follows:
fun enum_tr'[x,xs] = Const("enum",dummyT) $
                     make_enum(Const("op :",dummyT)$x$xs);
val print_translation = [("op :", enum_tr')];
This declaration causes the printer to print \hbox{\tt enum_tr'[$x$,$y$]}
whenever it finds \verb!Const("op :",_)$!$x$\verb!$!$y$.
  In Example~\ref{SET} we showed how to translate the concrete syntax for set
  comprehension into the proper internal form.  The string {\tt"\{$x$ |
    $P$\}"} now becomes {\tt Const("Set",_)\$Abs("$x$",_,$p$)}.  If, however,
  the latter term were printed without translating it back, it would result
  in {\tt"Set(\%$x$.$P$)"}.  Therefore the abstraction has to be turned back
  into a term that matches the concrete mixfix syntax:
fun set_tr'[Abs(x,T,P)] =
      let val (x',P') = variant_abs(x,T,P)
      in Const("SET",dummyT) $ Free(x',T) $ P' end;
The function \verb$variant_abs$, a basic term manipulation function, replaces
the bound variable $x$ by a {\tt Free} variable $x'$ having a unique name.  A
term produced by {\tt set_tr'} can now be printed according to the concrete
syntax defined in Example~\ref{SET} above.

Notice that the application of {\tt set_tr'} fails if the second component of
the argument is not an abstraction, but for example just a {\tt Free}
variable.  This is intentional because it signals to the caller that the
translation is inapplicable.  As a result {\tt Const("Set",_)\$Free("P",_)}
prints as {\tt"Set(P)"}.

The full theory extension, including concrete syntax and both translation
functions, has the following form:
types set 1
arities set :: (term)term
consts Set :: "('a => o) => 'a set"
       SET :: "[id,o] => 'a set"  ("\{_ | _\}")
fun set_tr[Free(s,T), p] = \dots;
val parse_translation = [("SET", set_tr)];
fun set_tr'[Abs(x,T,P)] = \dots;
val print_translation = [("Set", set_tr')];
As during the parse translation process, the types attached to constants
during print translation are ignored.  Thus {\tt Const("SET",dummyT)} in
{\tt set_tr'} above is acceptable.  The types of {\tt Free}s and {\tt Var}s
however must be preserved because they may get printed (see {\tt

\index{print translation|)}

\subsection{Printing a term}

Let $tab$ be the set of all string-function pairs of print translations in the
current syntax.

Terms are printed recursively; print translations are applied top down:
\item {\tt Free($x$,_)} is printed as $x$.
\item {\tt Var(($x$,$i$),_)} is printed as $x$, if $i = 0$ and $x$ does not
  end with a digit, as $x$ followed by $i$, if $i \neq 0$ and $x$ does not
  end with a digit, and as {\tt $x$.$i$}, if $x$ ends with a digit.  Thus the
  following cases can arise:
\verb$Var(("v",0),_)$ & \verb$Var(("v",7),_)$ & \verb$Var(("v5",0),_)$ \\
"?v" & "?v7" & "?v5.0"
\item {\tt Abs($x@1$,_,Abs($x@2$,_,\dots Abs($x@n$,_,$p$)\dots))}, where $p$
  is not an abstraction, is printed as {\tt \%$y@1\dots y@n$.$P$}, where $P$
  is the result of printing $p$, and the $x@i$ are replaced by $y@i$.  The
  latter are (possibly new) unique names.
\item {\tt Bound($i$)} is printed as {\tt B.$i$} \footnote{The occurrence of
    such ``loose'' bound variables indicates that either you are trying to
    print a subterm of an abstraction, or there is something wrong with your
    print translations.}.
\item The application {\tt$t =$ Const($c$,_)\$$t@1$\$\dots\$$t@n$} (where
  $n$ may be $0$!) is printed as follows:

  If there is a pair $(c,f)$ in $tab$, print $f[t@1,\dots,t@n]$.  If this
  application raises exception {\tt Match} or there is no pair $(c,f)$ in
  $tab$, let $sy$ be the mixfix annotation associated with $c$.  If there is
  no such $sy$, or if $sy$ does not have exactly $n$ argument positions, $t$
  is printed as an application; otherwise $t$ is printed according to $sy$.

  All other applications are printed as applications.
Printing a term {\tt $c$\$$t@1$\$\dots\$$t@n$} as an application means
printing it as {\tt $s$($s@1$,\dots,$s@n$)}, where $s@i$ is the result of
printing $t@i$.  If $c$ is a {\tt Const}, $s$ is its first argument;
otherwise $s$ is the result of printing $c$ as described above.

The printer also inserts parentheses where they are necessary for reasons
of precedence.

\section{Identifiers, constants, and type inference *}

There is one final step in the translation from strings to terms that we have
not covered yet.  It explains how constants are distinguished from {\tt Free}s
and how {\tt Free}s and {\tt Var}s are typed.  Both issues arise because {\tt
  Free}s and {\tt Var}s are not declared.

An identifier $f$ that does not appear as a delimiter in the concrete syntax
can be either a free variable or a constant.  Since the parser knows only
about those constants which appear in mixfix annotations, it parses $f$ as
{\tt Free("$f$",dummyT)}, where \ttindex{dummyT} is the predefined dummy {\tt
  typ}.  Although the parser produces these very raw terms, most user
interface level functions like {\tt goal} type terms according to the given
theory, say $T$.  In a first step, every occurrence of {\tt Free($f$,_)} or
{\tt Const($f$,_)} is replaced by {\tt Const($f$,$\tau$)}, provided there is
a constant $f$ of {\tt typ} $\tau$ in $T$.  This means that identifiers are
treated as {\tt Free}s iff they are not declared in the theory.  The types of
the remaining {\tt Free}s (and {\tt Var}s) are inferred as in \ML.  Type
constraints can be used to remove ambiguities.

One peculiarity of the current type inference algorithm is that variables
with the same name must have the same type, irrespective of whether they are
schematic, free or bound.  For example, take the first-order formula $f(x) = x
\land (\forall f.~ f=f)$ where ${=} :: [\alpha{::}term,\alpha]\To o$ and
$\forall :: (\alpha{::}term\To o)\To o$.  The first conjunct forces
$x::\alpha{::}term$ and $f::\alpha\To\alpha$, the second one
$f::\beta{::}term$.  Although the two $f$'s are distinct, they are required to
have the same type.  Unifying $\alpha\To\alpha$ and $\beta{::}term$ fails
because, in first-order logic, function types are not in class $term$.

\section{Putting it all together}

Having discussed the individual building blocks of a logic definition, it
remains to be shown how they fit together.  In particular we need to say how
an object-logic syntax is hooked up to the meta-logic.  Since all theorems
must conform to the syntax for $prop$ (see Figure~\ref{MetaLogicSyntax}),
that syntax has to be extended with the object-level syntax.  Assume that the
syntax of your object-logic defines a category $o$ of formulae.  These
formulae can now appear in axioms and theorems wherever $prop$ does if you
add the production
\[ prop ~=~ form.  \]
More precisely, you need a coercion from formulae to propositions:
Base = Pure +
types o 0
arities o :: logic
consts Trueprop :: "o => prop"  ("_"  5)
The constant {\tt Trueprop} (the name is arbitrary) acts as an invisible
coercion function.  Assuming this definition resides in a file {\tt base.thy},
you have to load it with the command {\tt use_thy"base"}.

One of the simplest nontrivial logics is {\em minimal logic} of
implication.  Its definition in Isabelle needs no advanced features but
illustrates the overall mechanism quite nicely:
Hilbert = Base +
consts "-->" :: "[o,o] => o"  (infixr 10)
K   "P --> Q --> P"
S   "(P --> Q --> R) --> (P --> Q) --> P --> R"
MP  "[| P --> Q; P |] ==> Q"
After loading this definition you can start to prove theorems in this logic:
goal Hilbert.thy "P --> P";
{\out Level 0}
{\out P --> P}
{\out  1.  P --> P}
by (resolve_tac [Hilbert.MP] 1);
{\out Level 1}
{\out P --> P}
{\out  1.  ?P --> P --> P}
{\out  2.  ?P}
by (resolve_tac [Hilbert.MP] 1);
{\out Level 2}
{\out P --> P}
{\out  1.  ?P1 --> ?P --> P --> P}
{\out  2.  ?P1}
{\out  3.  ?P}
by (resolve_tac [Hilbert.S] 1);
{\out Level 3}
{\out P --> P}
{\out  1.  P --> ?Q2 --> P}
{\out  2.  P --> ?Q2}
by (resolve_tac [Hilbert.K] 1);
{\out Level 4}
{\out P --> P}
{\out  1.  P --> ?Q2}
by (resolve_tac [Hilbert.K] 1);
{\out Level 5}
{\out P --> P}
{\out No subgoals!}
As you can see, this Hilbert-style formulation of minimal logic is easy to
define but difficult to use.  The following natural deduction formulation is
far preferable:
MinI = Base +
consts "-->" :: "[o,o] => o"  (infixr 10)
impI  "(P ==> Q) ==> P --> Q"
impE  "[| P --> Q; P |] ==> Q"
Note, however, that although the two systems are equivalent, this fact cannot
be proved within Isabelle: {\tt S} and {\tt K} can be derived in \verb$MinI$
(exercise!), but {\tt impI} cannot be derived in \verb!Hilbert!.  The reason
is that {\tt impI} is only an {\em admissible} rule in \verb!Hilbert!,
something that can only be shown by induction over all possible proofs in

It is a very simple matter to extend minimal logic with falsity:
MinIF = MinI +
consts False :: "o"
FalseE  "False ==> P"
On the other hand, we may wish to introduce conjunction only:
MinC = Base +
consts "&" :: "[o,o] => o"  (infixr 30)
conjI  "[| P; Q |] ==> P & Q"
conjE1 "P & Q ==> P"
conjE2 "P & Q ==> Q"
And if we want to have all three connectives together, we define:
MinIFC = MinIF + MinC
Now we can prove mixed theorems like
goal MinIFC.thy "P & False --> Q";
by (resolve_tac [MinI.impI] 1);
by (dresolve_tac [MinC.conjE2] 1);
by (eresolve_tac [MinIF.FalseE] 1);
Try this as an exercise!