path: root/doc/misc/wisent.texi

\input texinfo  @c -*-texinfo-*-
@c %**start of header
@setfilename ../../info/wisent
@set TITLE  Wisent Parser Development
@set AUTHOR Eric M. Ludlam, David Ponce, and Richard Y. Kim
@settitle @value{TITLE}

@c *************************************************************************
@c @ Header
@c *************************************************************************

@c Merge all indexes into a single index for now.
@c We can always separate them later into two or more as needed.
@syncodeindex vr cp
@syncodeindex fn cp
@syncodeindex ky cp
@syncodeindex pg cp
@syncodeindex tp cp

@c @footnotestyle separate
@c @paragraphindent 2
@c @@smallbook
@documentencoding UTF-8
@c %**end of header

@copying
Copyright @copyright{} 1988--1993, 1995, 1998--2004, 2007, 2012--2014
Free Software Foundation, Inc.

@c Since we are both GNU manuals, we do not need to ack each other here.
@ignore
Some texts are borrowed or adapted from the manual of Bison version
1.35.  The text in section entitled ``Understanding the automaton'' is
adapted from the section ``Understanding Your Parser'' in the manual
of Bison version 1.49.
@end ignore

@quotation
Permission is granted to copy, distribute and/or modify this document
under the terms of the GNU Free Documentation License, Version 1.3 or
any later version published by the Free Software Foundation; with no
Invariant Sections, with the Front-Cover texts being ``A GNU Manual,''
and with the Back-Cover Texts as in (a) below.  A copy of the license
is included in the section entitled ``GNU Free Documentation License''.

(a) The FSF's Back-Cover Text is: ``You have the freedom to copy and
modify this GNU manual.''
@end quotation
@end copying

@dircategory Emacs misc features
@direntry
* Wisent: (wisent).             Semantic Wisent parser development.
@end direntry

@iftex
@finalout
@end iftex

@c @setchapternewpage odd
@c @setchapternewpage off

@titlepage
@sp 10
@title @value{TITLE}
@author by @value{AUTHOR}
@page
@vskip 0pt plus 1 fill
@insertcopying
@end titlepage
@page

@macro semantic{}
@i{Semantic}
@end macro

@c *************************************************************************
@c @ Document
@c *************************************************************************
@contents

@node top
@top @value{TITLE}

Wisent (the European Bison ;-) is an Emacs Lisp implementation of the
GNU Compiler Compiler Bison.

This manual describes how to use Wisent to develop grammars for
programming languages, and how to use grammars to parse language
source in Emacs buffers.

It also describes how Wisent is used with the @semantic{} tool set
described in the @ref{Top, Semantic Manual, Semantic Manual, semantic}.

@ifnottex
@insertcopying
@end ifnottex

@menu
* Wisent Overview::
* Wisent Grammar::
* Wisent Parsing::
* Wisent Semantic::
* GNU Free Documentation License::
* Index::
@end menu

@node Wisent Overview
@chapter Wisent Overview

@dfn{Wisent} (the European Bison) is an implementation in Emacs Lisp
of the GNU Compiler Compiler Bison. Its code is a port of the C code
of GNU Bison 1.28 & 1.31.

For more details on the basic concepts for understanding Wisent, it is
worthwhile to read the @ref{Top, Bison Manual, , bison}.

Wisent can generate compilers compatible with the @semantic{} tool set.
See the @ref{Top, Semantic Manual, , semantic}.

It benefits from these Bison features:

@itemize @bullet
@item
It uses a fast but not so space-efficient encoding for the parse
tables, described in Corbett's PhD thesis from Berkeley:
@quotation
@cite{Static Semantics in Compiler Error Recovery}@*
June 1985, Report No. UCB/CSD 85/251.
@end quotation

@item
For generating the lookahead sets, Wisent uses the well-known
technique of F. DeRemer and T. Pennello described in:
@quotation
@cite{Efficient Computation of LALR(1) Look-Ahead Sets}@*
October 1982, ACM TOPLAS Vol 4 No 4, 615--49,
@uref{http://dx.doi.org/10.1145/69622.357187}.
@end quotation

@item
Wisent resolves shift/reduce conflicts using operator precedence and
associativity.

@item
Parser error recovery is accomplished using rules which match the
special token @code{error}.
@end itemize

Nevertheless there are some fundamental differences between Bison and
Wisent.

@itemize
@item
Wisent is intended to be used in Emacs.  It reads and produces Emacs
Lisp data structures.  All the additional code used in grammars is
Emacs Lisp code.

@item
Contrary to Bison, Wisent does not generate a parser which combines
Emacs Lisp code and grammar constructs.  They exist separately.
Wisent reads the grammar from a Lisp data structure and then generates
grammar constructs as tables.  Afterward, the derived tables can be
included and byte-compiled in separate Emacs Lisp files, and be used
at a later time by the Wisent's parser engine.

@item
Wisent allows multiple start nonterminals and allows a call to the
parsing function to be made for a particular start nonterminal.  For
example, this is particularly useful to parse a region of an Emacs
buffer.  @semantic{} heavily depends on the availability of this feature.
@end itemize

@node Wisent Grammar
@chapter Wisent Grammar

@cindex context-free grammar
@cindex rule
In order for Wisent to parse a language, it must be described by a
@dfn{context-free grammar}.  That is a grammar specified as rules that
can be applied regardless of context.  For more information, see
@ref{Language and Grammar, , , bison}, in the Bison manual.

@cindex terminal
@cindex nonterminal
The formal grammar is formulated using @dfn{terminal} and
@dfn{nonterminal} items.  Terminals can be Emacs Lisp symbols or
characters, and nonterminals are symbols only.

@cindex token
Terminals (also known as @dfn{tokens}) represent the lexical
elements of the language like numbers, strings, etc..

For example @samp{PLUS} can represent the operator @samp{+}.

Nonterminal symbols are described by rules:

@example
@group
RESULT @equiv{} COMPONENTS@dots{}
@end group
@end example

@samp{RESULT} is a nonterminal that this rule describes and
@samp{COMPONENTS} are various terminals and nonterminals that are put
together by this rule.

For example, this rule:

@example
@group
exp @equiv{} exp PLUS exp
@end group
@end example

Says that two groupings of type @samp{exp}, with a @samp{PLUS} token
in between, can be combined into a larger grouping of type @samp{exp}.

@menu
* Grammar format::
* Example::
* Compiling a grammar::
* Conflicts::
@end menu

@node Grammar format
@section Grammar format

@cindex grammar format
To be acceptable by Wisent a context-free grammar must respect a
particular format.  That is, must be represented as an Emacs Lisp list
of the form:

@code{(@var{terminals} @var{assocs} . @var{non-terminals})}

@table @var
@item terminals
Is the list of terminal symbols used in the grammar.

@cindex associativity
@item assocs
Specify the associativity of @var{terminals}.  It is @code{nil} when
there is no associativity defined, or an alist of
@w{@code{(@var{assoc-type} . @var{assoc-value})}} elements.

@var{assoc-type} must be one of the @code{default-prec},
@code{nonassoc}, @code{left} or @code{right} symbols.  When
@var{assoc-type} is @code{default-prec}, @var{assoc-value} must be
@code{nil} or @code{t} (the default).  Otherwise it is a list of
tokens which must have been previously declared in @var{terminals}.

For details, see @ref{Contextual Precedence, , , bison}, in the
Bison manual.

@item non-terminals
Is the list of nonterminal definitions.  Each definition has the form:

@code{(@var{nonterm} . @var{rules})}

Where @var{nonterm} is the nonterminal symbol defined and
@var{rules} the list of rules that describe this nonterminal.  Each
rule is a list:

@code{(@var{components} [@var{precedence}] [@var{action}])}

Where:

@table @var
@item components
Is a list of various terminals and nonterminals that are put together
by this rule.

For example,

@example
@group
(exp ((exp ?+ exp))          ;; exp: exp '+' exp
     )                       ;;    ;
@end group
@end example

Says that two groupings of type @samp{exp}, with a @samp{+} token in
between, can be combined into a larger grouping of type @samp{exp}.

@cindex grammar coding conventions
By convention, a nonterminal symbol should be in lower case, such as
@samp{exp}, @samp{stmt} or @samp{declaration}.  Terminal symbols
should be upper case to distinguish them from nonterminals: for
example, @samp{INTEGER}, @samp{IDENTIFIER}, @samp{IF} or
@samp{RETURN}.  A terminal symbol that represents a particular keyword
in the language is conventionally the same as that keyword converted
to upper case.  The terminal symbol @code{error} is reserved for error
recovery.

@cindex middle-rule actions
Scattered among the components can be @dfn{middle-rule} actions.
Usually only @var{action} is provided (@pxref{action}).

If @var{components} in a rule is @code{nil}, it means that the rule
can match the empty string.  For example, here is how to define a
comma-separated sequence of zero or more @samp{exp} groupings:

@smallexample
@group
(expseq  (nil)               ;; expseq: ;; empty
         ((expseq1))         ;;       | expseq1
         )                   ;;       ;

(expseq1 ((exp))             ;; expseq1: exp
         ((expseq1 ?, exp))  ;;        | expseq1 ',' exp
         )                   ;;        ;
@end group
@end smallexample

@cindex precedence level
@item precedence
Assign the rule the precedence of the given terminal item, overriding
the precedence that would be deduced for it, that is the one of the
last terminal in it.  Notice that only terminals declared in
@var{assocs} have a precedence level.  The altered rule precedence
then affects how conflicts involving that rule are resolved.

@var{precedence} is an optional vector of one terminal item.

Here is how @var{precedence} solves the problem of unary minus.
First, declare a precedence for a fictitious terminal symbol named
@code{UMINUS}.  There are no tokens of this type, but the symbol
serves to stand for its precedence:

@example
@dots{}
((default-prec t) ;; This is the default
 (left '+' '-')
 (left '*')
 (left UMINUS))
@end example

Now the precedence of @code{UMINUS} can be used in specific rules:

@smallexample
@group
(exp    @dots{}                  ;; exp:    @dots{}
         ((exp ?- exp))      ;;         | exp '-' exp
        @dots{}                  ;;         @dots{}
         ((?- exp) [UMINUS]) ;;         | '-' exp %prec UMINUS
        @dots{}                  ;;         @dots{}
        )                    ;;         ;
@end group
@end smallexample

If you forget to append @code{[UMINUS]} to the rule for unary minus,
Wisent silently assumes that minus has its usual precedence.  This
kind of problem can be tricky to debug, since one typically discovers
the mistake only by testing the code.

Using @code{(default-prec nil)} declaration makes it easier to
discover this kind of problem systematically.  It causes rules that
lack a @var{precedence} modifier to have no precedence, even if the
last terminal symbol mentioned in their components has a declared
precedence.

If @code{(default-prec nil)} is in effect, you must specify
@var{precedence} for all rules that participate in precedence conflict
resolution.  Then you will see any shift/reduce conflict until you
tell Wisent how to resolve it, either by changing your grammar or by
adding an explicit precedence.  This will probably add declarations to
the grammar, but it helps to protect against incorrect rule
precedences.

The effect of @code{(default-prec nil)} can be reversed by giving
@code{(default-prec t)}, which is the default.

For more details, see @ref{Contextual Precedence, , , bison}, in the
Bison manual.

It is important to understand that @var{assocs} declarations defines
associativity but also assign a precedence level to terminals.  All
terminals declared in the same @code{left}, @code{right} or
@code{nonassoc} association get the same precedence level.  The
precedence level is increased at each new association.

On the other hand, @var{precedence} explicitly assign the precedence
level of the given terminal to a rule.

@cindex semantic actions
@item @anchor{action}action
An action is an optional Emacs Lisp function call, like this:

@code{(identity $1)}

The result of an action determines the semantic value of a rule.

From an implementation standpoint, the function call will be embedded
in a lambda expression, and several useful local variables will be
defined:

@table @code
@vindex $N
@item $@var{n}
Where @var{n} is a positive integer.  Like in Bison, the value of
@code{$@var{n}} is the semantic value of the @var{n}th element of
@var{components}, starting from 1.  It can be of any Lisp data
type.

@vindex $region@var{n}
@item $regionN
Where @var{n} is a positive integer.  For each @code{$@var{n}}
variable defined there is a corresponding @code{$region@var{n}}
variable.  Its value is a pair @code{(@var{start-pos} .
@var{end-pos})} that represent the start and end positions (in the
lexical input stream) of the @code{$@var{n}} value.  It can be
@code{nil} when the component positions are not available, like for an
empty string component for example.

@vindex $region
@item $region
Its value is the leftmost and rightmost positions of input data
matched by all @var{components} in the rule.  This is a pair
@code{(@var{leftmost-pos} .  @var{rightmost-pos})}.  It can be
@code{nil} when components positions are not available.

@vindex $nterm
@item $nterm
This variable is initialized with the nonterminal symbol
(@var{nonterm}) the rule belongs to.  It could be useful to improve
error reporting or debugging.  It is also used to automatically
provide incremental re-parse entry points for @semantic{} tags
(@pxref{Wisent Semantic}).

@vindex $action
@item $action
The value of @code{$action} is the symbolic name of the current
semantic action (@pxref{Debugging actions}).
@end table

When an action is not specified a default value is supplied, it is
@code{(identity $1)}.  This means that the default semantic value of a
rule is the value of its first component.  Excepted for a rule
matching the empty string, for which the default action is to return
@code{nil}.
@end table
@end table

@node Example
@section Example

@cindex grammar example
Here is an example to parse simple infix arithmetic expressions.  See
@ref{Infix Calc, , , bison}, in the Bison manual for details.

@lisp
@group
'(
  ;; Terminals
  (NUM)

  ;; Terminal associativity & precedence
  ((nonassoc ?=)
   (left ?- ?+)
   (left ?* ?/)
   (left NEG)
   (right ?^))

  ;; Rules
  (input
   ((line))
   ((input line)
    (format "%s %s" $1 $2))
   )

  (line
   ((?;)
    (progn ";"))
   ((exp ?;)
    (format "%s;" $1))
   ((error ?;)
    (progn "Error;")))
   )

  (exp
   ((NUM)
    (string-to-number $1))
   ((exp ?= exp)
    (= $1 $3))
   ((exp ?+ exp)
    (+ $1 $3))
   ((exp ?- exp)
    (- $1 $3))
   ((exp ?* exp)
    (* $1 $3))
   ((exp ?/ exp)
    (/ $1 $3))
   ((?- exp) [NEG]
    (- $2))
   ((exp ?^ exp)
    (expt $1 $3))
   ((?\( exp ?\))
    (progn $2))
   )
  )
@end group
@end lisp

In the bison-like @dfn{WY} format (@pxref{Wisent Semantic}) the
grammar looks like this:

@example
@group
%token <number> NUM

%nonassoc '=' ;; comparison
%left '-' '+'
%left '*' '/'
%left NEG     ;; negation--unary minus
%right '^'    ;; exponentiation

%%

input:
    line
  | input line
    (format "%s %s" $1 $2)
  ;

line:
    ';'
    @{";"@}
  | exp ';'
    (format "%s;" $1)
  | error ';'
    @{"Error;"@}
  ;

exp:
    NUM
    (string-to-number $1)
  | exp '=' exp
    (= $1 $3)
  | exp '+' exp
    (+ $1 $3)
  | exp '-' exp
    (- $1 $3)
  | exp '*' exp
    (* $1 $3)
  | exp '/' exp
    (/ $1 $3)
  | '-' exp %prec NEG
    (- $2)
  | exp '^' exp
    (expt $1 $3)
  | '(' exp ')'
    @{$2@}
  ;

%%
@end group
@end example

@node Compiling a grammar
@section Compiling a grammar

@cindex automaton
After providing a context-free grammar in a suitable format, it must
be translated into a set of tables (an @dfn{automaton}) that will be
used to derive the parser.  Like Bison, Wisent translates grammars that
must be @dfn{LALR(1)}.

@cindex LALR(1) grammar
@cindex look-ahead token
A grammar is @acronym{LALR(1)} if it is possible to tell how to parse
any portion of an input string with just a single token of look-ahead:
the @dfn{look-ahead token}.  See @ref{Language and Grammar, , ,
bison}, in the Bison manual for more information.

@cindex grammar compilation
Grammar translation (compilation) is achieved by the function:

@cindex compiling a grammar
@vindex wisent-single-start-flag
@findex wisent-compile-grammar
@defun wisent-compile-grammar grammar &optional start-list
Compile @var{grammar} and return an @acronym{LALR(1)} automaton.

Optional argument @var{start-list} is a list of start symbols
(nonterminals).  If @code{nil} the first nonterminal defined in the
grammar is the default start symbol.  If @var{start-list} contains
only one element, it defines the start symbol.  If @var{start-list}
contains more than one element, all are defined as potential start
symbols, unless @code{wisent-single-start-flag} is non-@code{nil}.  In
that case the first element of @var{start-list} defines the start
symbol and others are ignored.

The @acronym{LALR(1)} automaton is a vector of the form:

@code{[@var{actions gotos starts functions}]}

@table @var
@item actions
A state/token matrix telling the parser what to do at every state
based on the current look-ahead token.  That is shift, reduce, accept
or error.  See also @ref{Wisent Parsing}.

@item gotos
A state/nonterminal matrix telling the parser the next state to go to
after reducing with each rule.

@item starts
An alist which maps the allowed start symbols (nonterminals) to
lexical tokens that will be first shifted into the parser stack.

@item functions
An obarray of semantic action symbols.  A semantic action is actually
an Emacs Lisp function (lambda expression).
@end table
@end defun

@node Conflicts
@section Conflicts

Normally, a grammar should produce an automaton where at each state
the parser has only one action to do (@pxref{Wisent Parsing}).

@cindex ambiguous grammar
In certain cases, a grammar can produce an automaton where, at some
states, there are more than one action possible.  Such a grammar is
@dfn{ambiguous}, and generates @dfn{conflicts}.

@cindex deterministic automaton
The parser can't be driven by an automaton which isn't completely
@dfn{deterministic}, that is which contains conflicts.  It is
necessary to resolve the conflicts to eliminate them.  Wisent resolves
conflicts like Bison does.

@cindex grammar conflicts
@cindex conflicts resolution
There are two sorts of conflicts:

@table @dfn
@cindex shift/reduce conflicts
@item shift/reduce conflicts
When either a shift or a reduction would be valid at the same state.

Such conflicts are resolved by choosing to shift, unless otherwise
directed by operator precedence declarations.
See @ref{Shift/Reduce , , , bison}, in the Bison manual for more
information.

@cindex reduce/reduce conflicts
@item reduce/reduce conflicts
That occurs if there are two or more rules that apply to the same
sequence of input.  This usually indicates a serious error in the
grammar.

Such conflicts are resolved by choosing to use the rule that appears
first in the grammar, but it is very risky to rely on this.  Every
reduce/reduce conflict must be studied and usually eliminated.  See
@ref{Reduce/Reduce , , , bison}, in the Bison manual for more
information.
@end table

@menu
* Grammar Debugging::
* Understanding the automaton::
@end menu

@node Grammar Debugging
@subsection Grammar debugging

@cindex grammar debugging
@cindex grammar verbose description
To help writing a new grammar, @code{wisent-compile-grammar} can
produce a verbose report containing a detailed description of the
grammar and parser (equivalent to what Bison reports with the
@option{--verbose} option).

To enable the verbose report you can set to non-@code{nil} the
variable:

@vindex wisent-verbose-flag
@deffn Option wisent-verbose-flag
non-@code{nil} means to report verbose information on generated parser.
@end deffn

Or interactively use the command:

@findex wisent-toggle-verbose-flag
@deffn Command wisent-toggle-verbose-flag
Toggle whether to report verbose information on generated parser.
@end deffn

The verbose report is printed in the temporary buffer
@code{*wisent-log*} when running interactively, or in file
@file{wisent.output} when running in batch mode.  Different
reports are separated from each other by a line like this:

@example
@group
*** Wisent @var{source-file} - 2002-06-27 17:33
@end group
@end example

where @var{source-file} is the name of the Emacs Lisp file from which
the grammar was read.  See @ref{Understanding the automaton}, for
details on the verbose report.

@table @strong
@item Please Note
To help debugging the grammar compiler itself, you can set this
variable to print the content of some internal data structures:

@vindex wisent-debug-flag
@defvar wisent-debug-flag
non-@code{nil} means enable some debug stuff.
@end defvar
@end table

@node Understanding the automaton
@subsection Understanding the automaton

@cindex understanding the automaton
This section (took from the manual of Bison 1.49) describes how to use
the verbose report printed by @code{wisent-compile-grammar} to
understand the generated automaton, to tune or fix a grammar.

We will use the following example:

@example
@group
(let ((wisent-verbose-flag t)) ;; Print a verbose report!
  (wisent-compile-grammar
   '((NUM STR)                          ; %token NUM STR

     ((left ?+ ?-)                      ; %left '+' '-';
      (left ?*))                        ; %left '*'

     (exp                               ; exp:
      ((exp ?+ exp))                    ;    exp '+' exp
      ((exp ?- exp))                    ;  | exp '-' exp
      ((exp ?* exp))                    ;  | exp '*' exp
      ((exp ?/ exp))                    ;  | exp '/' exp
      ((NUM))                           ;  | NUM
      )                                 ;  ;

     (useless                           ; useless:
      ((STR))                           ;    STR
      )                                 ;  ;
     )
   'nil)                                ; no %start declarations
  )
@end group
@end example

When evaluating the above expression, grammar compilation first issues
the following two clear messages:

@example
@group
Grammar contains 1 useless nonterminals and 1 useless rules
Grammar contains 7 shift/reduce conflicts
@end group
@end example

The @samp{*wisent-log*} buffer details things!

The first section reports conflicts that were solved using precedence
and/or associativity:

@example
@group
Conflict in state 7 between rule 1 and token '+' resolved as reduce.
Conflict in state 7 between rule 1 and token '-' resolved as reduce.
Conflict in state 7 between rule 1 and token '*' resolved as shift.
Conflict in state 8 between rule 2 and token '+' resolved as reduce.
Conflict in state 8 between rule 2 and token '-' resolved as reduce.
Conflict in state 8 between rule 2 and token '*' resolved as shift.
Conflict in state 9 between rule 3 and token '+' resolved as reduce.
Conflict in state 9 between rule 3 and token '-' resolved as reduce.
Conflict in state 9 between rule 3 and token '*' resolved as reduce.
@end group
@end example

The next section reports useless tokens, nonterminal and rules (note
that useless tokens might be used by the scanner):

@example
@group
Useless nonterminals:

   useless


Terminals which are not used:

   STR


Useless rules:

#6     useless: STR;
@end group
@end example

The next section lists states that still have conflicts:

@example
@group
State 7 contains 1 shift/reduce conflict.
State 8 contains 1 shift/reduce conflict.
State 9 contains 1 shift/reduce conflict.
State 10 contains 4 shift/reduce conflicts.
@end group
@end example

The next section reproduces the grammar used:

@example
@group
Grammar

  Number, Rule
  1       exp -> exp '+' exp
  2       exp -> exp '-' exp
  3       exp -> exp '*' exp
  4       exp -> exp '/' exp
  5       exp -> NUM
@end group
@end example

And reports the uses of the symbols:

@example
@group
Terminals, with rules where they appear

$EOI (-1)
error (1)
NUM (2) 5
STR (3) 6
'+' (4) 1
'-' (5) 2
'*' (6) 3
'/' (7) 4


Nonterminals, with rules where they appear

exp (8)
    on left: 1 2 3 4 5, on right: 1 2 3 4
@end group
@end example

The report then details the automaton itself, describing each state
with it set of @dfn{items}, also known as @dfn{pointed rules}.  Each
item is a production rule together with a point (marked by @samp{.})
that the input cursor.

@example
@group
state 0

    NUM shift, and go to state 1

    exp go to state 2
@end group
@end example

State 0 corresponds to being at the very beginning of the parsing, in
the initial rule, right before the start symbol (@samp{exp}).  When
the parser returns to this state right after having reduced a rule
that produced an @samp{exp}, it jumps to state 2.  If there is no such
transition on a nonterminal symbol, and the lookahead is a @samp{NUM},
then this token is shifted on the parse stack, and the control flow
jumps to state 1.  Any other lookahead triggers a parse error.

In the state 1...

@example
@group
state 1

    exp  ->  NUM .   (rule 5)

    $default    reduce using rule 5 (exp)
@end group
@end example

the rule 5, @samp{exp: NUM;}, is completed.  Whatever the lookahead
(@samp{$default}), the parser will reduce it.  If it was coming from
state 0, then, after this reduction it will return to state 0, and
will jump to state 2 (@samp{exp: go to state 2}).

@example
@group
state 2

    exp  ->  exp . '+' exp   (rule 1)
    exp  ->  exp . '-' exp   (rule 2)
    exp  ->  exp . '*' exp   (rule 3)
    exp  ->  exp . '/' exp   (rule 4)

    $EOI        shift, and go to state 11
    '+' shift, and go to state 3
    '-' shift, and go to state 4
    '*' shift, and go to state 5
    '/' shift, and go to state 6
@end group
@end example

In state 2, the automaton can only shift a symbol.  For instance,
because of the item @samp{exp -> exp . '+' exp}, if the lookahead if
@samp{+}, it will be shifted on the parse stack, and the automaton
control will jump to state 3, corresponding to the item
@samp{exp -> exp . '+' exp}:

@example
@group
state 3

    exp  ->  exp '+' . exp   (rule 1)

    NUM shift, and go to state 1

    exp go to state 7
@end group
@end example

Since there is no default action, any other token than those listed
above will trigger a parse error.

The interpretation of states 4 to 6 is straightforward:

@example
@group
state 4

    exp  ->  exp '-' . exp   (rule 2)

    NUM shift, and go to state 1

    exp go to state 8



state 5

    exp  ->  exp '*' . exp   (rule 3)

    NUM shift, and go to state 1

    exp go to state 9



state 6

    exp  ->  exp '/' . exp   (rule 4)

    NUM shift, and go to state 1

    exp go to state 10
@end group
@end example

As was announced in beginning of the report, @samp{State 7 contains 1
shift/reduce conflict.}:

@example
@group
state 7

    exp  ->  exp . '+' exp   (rule 1)
    exp  ->  exp '+' exp .   (rule 1)
    exp  ->  exp . '-' exp   (rule 2)
    exp  ->  exp . '*' exp   (rule 3)
    exp  ->  exp . '/' exp   (rule 4)

    '*' shift, and go to state 5
    '/' shift, and go to state 6

    '/' [reduce using rule 1 (exp)]
    $default    reduce using rule 1 (exp)
@end group
@end example

Indeed, there are two actions associated to the lookahead @samp{/}:
either shifting (and going to state 6), or reducing rule 1.  The
conflict means that either the grammar is ambiguous, or the parser
lacks information to make the right decision.  Indeed the grammar is
ambiguous, as, since we did not specify the precedence of @samp{/},
the sentence @samp{NUM + NUM / NUM} can be parsed as @samp{NUM + (NUM
/ NUM)}, which corresponds to shifting @samp{/}, or as @samp{(NUM +
NUM) / NUM}, which corresponds to reducing rule 1.

Because in @acronym{LALR(1)} parsing a single decision can be made,
Wisent arbitrarily chose to disable the reduction, see
@ref{Conflicts}.  Discarded actions are reported in between square
brackets.

Note that all the previous states had a single possible action: either
shifting the next token and going to the corresponding state, or
reducing a single rule.  In the other cases, i.e., when shifting
@emph{and} reducing is possible or when @emph{several} reductions are
possible, the lookahead is required to select the action.  State 7 is
one such state: if the lookahead is @samp{*} or @samp{/} then the
action is shifting, otherwise the action is reducing rule 1.  In other
words, the first two items, corresponding to rule 1, are not eligible
when the lookahead is @samp{*}, since we specified that @samp{*} has
higher precedence that @samp{+}.  More generally, some items are
eligible only with some set of possible lookaheads.

States 8 to 10 are similar:

@example
@group
state 8

    exp  ->  exp . '+' exp   (rule 1)
    exp  ->  exp . '-' exp   (rule 2)
    exp  ->  exp '-' exp .   (rule 2)
    exp  ->  exp . '*' exp   (rule 3)
    exp  ->  exp . '/' exp   (rule 4)

    '*' shift, and go to state 5
    '/' shift, and go to state 6

    '/' [reduce using rule 2 (exp)]
    $default    reduce using rule 2 (exp)


state 9

    exp  ->  exp . '+' exp   (rule 1)
    exp  ->  exp . '-' exp   (rule 2)
    exp  ->  exp . '*' exp   (rule 3)
    exp  ->  exp '*' exp .   (rule 3)
    exp  ->  exp . '/' exp   (rule 4)

    '/' shift, and go to state 6

    '/' [reduce using rule 3 (exp)]
    $default    reduce using rule 3 (exp)


state 10

    exp  ->  exp . '+' exp   (rule 1)
    exp  ->  exp . '-' exp   (rule 2)
    exp  ->  exp . '*' exp   (rule 3)
    exp  ->  exp . '/' exp   (rule 4)
    exp  ->  exp '/' exp .   (rule 4)

    '+' shift, and go to state 3
    '-' shift, and go to state 4
    '*' shift, and go to state 5
    '/' shift, and go to state 6

    '+' [reduce using rule 4 (exp)]
    '-' [reduce using rule 4 (exp)]
    '*' [reduce using rule 4 (exp)]
    '/' [reduce using rule 4 (exp)]
    $default    reduce using rule 4 (exp)
@end group
@end example

Observe that state 10 contains conflicts due to the lack of precedence
of @samp{/} wrt @samp{+}, @samp{-}, and @samp{*}, but also because the
associativity of @samp{/} is not specified.

Finally, the state 11 (plus 12) is named the @dfn{final state}, or the
@dfn{accepting state}:

@example
@group
state 11

    $EOI        shift, and go to state 12



state 12

    $default    accept
@end group
@end example

The end of input is shifted @samp{$EOI shift,} and the parser exits
successfully (@samp{go to state 12}, that terminates).

@node Wisent Parsing
@chapter Wisent Parsing

@cindex bottom-up parser
@cindex shift-reduce parser
The Wisent's parser is what is called a @dfn{bottom-up} or
@dfn{shift-reduce} parser which repeatedly:

@table @dfn
@cindex shift
@item shift
That is pushes the value of the last lexical token read (the
look-ahead token) into a value stack, and reads a new one.

@cindex reduce
@item reduce
That is replaces a nonterminal by its semantic value.  The values of
the components which form the right hand side of a rule are popped
from the value stack and reduced by the semantic action of this rule.
The result is pushed back on top of value stack.
@end table

The parser will stop on:

@table @dfn
@cindex accept
@item accept
When all input has been successfully parsed.  The semantic value of
the start nonterminal is on top of the value stack.

@cindex syntax error
@item error
When a syntax error (an unexpected token in input) has been detected.
At this point the parser issues an error message and either stops or
calls a recovery routine to try to resume parsing.
@end table

@cindex table-driven parser
The above elementary actions are driven by the @acronym{LALR(1)}
automaton built by @code{wisent-compile-grammar} from a context-free
grammar.

The Wisent's parser is entered by calling the function:

@findex wisent-parse
@defun wisent-parse automaton lexer &optional error start
Parse input using the automaton specified in @var{automaton}.

@table @var
@item automaton
Is an @acronym{LALR(1)} automaton generated by
@code{wisent-compile-grammar} (@pxref{Wisent Grammar}).

@item lexer
Is a function with no argument called by the parser to obtain the next
terminal (token) in input (@pxref{Writing a lexer}).

@item error
Is an optional reporting function called when a parse error occurs.
It receives a message string to report.  It defaults to the function
@code{wisent-message} (@pxref{Report errors}).

@item start
Specify the start symbol (nonterminal) used by the parser as its goal.
It defaults to the start symbol defined in the grammar
(@pxref{Wisent Grammar}).
@end table
@end defun

The following two normal hooks permit to do some useful processing
respectively before to start parsing, and after the parser terminated.

@vindex wisent-pre-parse-hook
@defvar wisent-pre-parse-hook
Normal hook run just before entering the @var{LR} parser engine.
@end defvar

@vindex wisent-post-parse-hook
@defvar wisent-post-parse-hook
Normal hook run just after the @var{LR} parser engine terminated.
@end defvar

@menu
* Writing a lexer::
* Actions goodies::
* Report errors::
* Error recovery::
* Debugging actions::
@end menu

@node Writing a lexer
@section What the parser must receive

It is important to understand that the parser does not parse
characters, but lexical tokens, and does not know anything about
characters in text streams!

@cindex lexical analysis
@cindex lexer
@cindex scanner
Reading input data to produce lexical tokens is performed by a lexer
(also called a scanner) in a lexical analysis step, before the syntax
analysis step performed by the parser.  The parser automatically calls
the lexer when it needs the next token to parse.

@cindex lexical tokens
A Wisent's lexer is an Emacs Lisp function with no argument.  It must
return a valid lexical token of the form:

@code{(@var{token-class value} [@var{start} . @var{end}])}

@table @var
@item token-class
Is a category of lexical token identifying a terminal as specified in
the grammar (@pxref{Wisent Grammar}).  It can be a symbol or a character
literal.

@item value
Is the value of the lexical token.  It can be of any valid Emacs Lisp
data type.

@item start
@itemx end
Are the optional beginning and ending positions of @var{value} in the
input stream.
@end table

When there are no more tokens to read the lexer must return the token
@code{(list wisent-eoi-term)} to each request.

@vindex wisent-eoi-term
@defvar wisent-eoi-term
Predefined constant, End-Of-Input terminal symbol.
@end defvar

@code{wisent-lex} is an example of a lexer that reads lexical tokens
produced by a @semantic{} lexer, and translates them into lexical tokens
suitable to the Wisent parser.  See also @ref{Wisent Lex}.

To call the lexer in a semantic action use the function
@code{wisent-lexer}.  See also @ref{Actions goodies}.

@node Actions goodies
@section Variables and macros useful in grammar actions.

@vindex wisent-input
@defvar wisent-input
The last token read.
This variable only has meaning in the scope of @code{wisent-parse}.
@end defvar

@findex wisent-lexer
@defun wisent-lexer
Obtain the next terminal in input.
@end defun

@findex wisent-region
@defun wisent-region &rest positions
Return the start/end positions of the region including
@var{positions}.  Each element of @var{positions} is a pair
@w{@code{(@var{start-pos} .  @var{end-pos})}} or @code{nil}.  The
returned value is the pair @w{@code{(@var{min-start-pos} .
@var{max-end-pos})}} or @code{nil} if no @var{positions} are
available.
@end defun

@node Report errors
@section The error reporting function

@cindex error reporting
When the parser encounters a syntax error it calls a user-defined
function.  It must be an Emacs Lisp function with one argument: a
string containing the message to report.

By default the parser uses this function to report error messages:

@findex wisent-message
@defun wisent-message string &rest args
Print a one-line message if @code{wisent-parse-verbose-flag} is set.
Pass @var{string} and @var{args} arguments to @dfn{message}.
@end defun

@table @strong
@item Please Note:
@code{wisent-message} uses the following function to print lexical
tokens:

@defun wisent-token-to-string token
Return a printed representation of lexical token @var{token}.
@end defun

The general printed form of a lexical token is:

@w{@code{@var{token}(@var{value})@@@var{location}}}
@end table

To control the verbosity of the parser you can set to non-@code{nil}
this variable:

@vindex wisent-parse-verbose-flag
@deffn Option wisent-parse-verbose-flag
non-@code{nil} means to issue more messages while parsing.
@end deffn

Or interactively use the command:

@findex wisent-parse-toggle-verbose-flag
@deffn Command wisent-parse-toggle-verbose-flag
Toggle whether to issue more messages while parsing.
@end deffn

When the error reporting function is entered the variable
@code{wisent-input} contains the unexpected token as returned by the
lexer.

The error reporting function can be called from a semantic action too
using the special macro @code{wisent-error}.  When called from a
semantic action entered by error recovery (@pxref{Error recovery}) the
value of the variable @code{wisent-recovering} is non-@code{nil}.

@node Error recovery
@section Error recovery

@cindex error recovery
The error recovery mechanism of the Wisent's parser conforms to the
one Bison uses.  See @ref{Error Recovery, , , bison}, in the Bison
manual for details.

@cindex error token
To recover from a syntax error you must write rules to recognize the
special token @code{error}.  This is a terminal symbol that is
automatically defined and reserved for error handling.

When the parser encounters a syntax error, it pops the state stack
until it finds a state that allows shifting the @code{error} token.
After it has been shifted, if the old look-ahead token is not
acceptable to be shifted next, the parser reads tokens and discards
them until it finds a token which is acceptable.

@cindex error recovery strategy
Strategies for error recovery depend on the choice of error rules in
the grammar.  A simple and useful strategy is simply to skip the rest
of the current statement if an error is detected:

@example
@group
(statement (( error ?; )) ;; on error, skip until ';' is read
           )
@end group
@end example

It is also useful to recover to the matching close-delimiter of an
opening-delimiter that has already been parsed:

@example
@group
(primary (( ?@{ expr  ?@} ))
         (( ?@{ error ?@} ))
         @dots{}
         )
@end group
@end example

@cindex error recovery actions
Note that error recovery rules may have actions, just as any other
rules can.  Here are some predefined hooks, variables, functions or
macros, useful in such actions:

@vindex wisent-nerrs
@defvar wisent-nerrs
The number of parse errors encountered so far.
@end defvar

@vindex wisent-recovering
@defvar wisent-recovering
non-@code{nil} means that the parser is recovering.
This variable only has meaning in the scope of @code{wisent-parse}.
@end defvar

@findex wisent-error
@defun wisent-error msg
Call the user supplied error reporting function with message
@var{msg} (@pxref{Report errors}).

For an example of use, @xref{wisent-skip-token}.
@end defun

@findex wisent-errok
@defun wisent-errok
Resume generating error messages immediately for subsequent syntax
errors.

The parser suppress error message for syntax errors that happens
shortly after the first, until three consecutive input tokens have
been successfully shifted.

Calling @code{wisent-errok} in an action, make error messages resume
immediately.  No error messages will be suppressed if you call it in
an error rule's action.

For an example of use, @xref{wisent-skip-token}.
@end defun

@findex wisent-clearin
@defun wisent-clearin
Discard the current lookahead token.
This will cause a new lexical token to be read.

In an error rule's action the previous lookahead token is reanalyzed
immediately.  @code{wisent-clearin} may be called to clear this token.

For example, suppose that on a parse error, an error handling routine
is called that advances the input stream to some point where parsing
should once again commence.  The next symbol returned by the lexical
scanner is probably correct.  The previous lookahead token ought to
be discarded with @code{wisent-clearin}.

For an example of use, @xref{wisent-skip-token}.
@end defun

@findex wisent-abort
@defun wisent-abort
Abort parsing and save the lookahead token.
@end defun

@findex wisent-set-region
@defun wisent-set-region start end
Change the region of text matched by the current nonterminal.
@var{start} and @var{end} are respectively the beginning and end
positions of the region occupied by the group of components associated
to this nonterminal.  If @var{start} or @var{end} values are not a
valid positions the region is set to @code{nil}.

For an example of use, @xref{wisent-skip-token}.
@end defun

@vindex wisent-discarding-token-functions
@defvar wisent-discarding-token-functions
List of functions to be called when discarding a lexical token.
These functions receive the lexical token discarded.
When the parser encounters unexpected tokens, it can discards them,
based on what directed by error recovery rules.  Either when the
parser reads tokens until one is found that can be shifted, or when an
semantic action calls the function @code{wisent-skip-token} or
@code{wisent-skip-block}.
For language specific hooks, make sure you define this as a local
hook.

For example, in @semantic{}, this hook is set to the function
@code{wisent-collect-unmatched-syntax} to collect unmatched lexical
tokens (@pxref{Useful functions}).
@end defvar

@findex wisent-skip-token
@defun wisent-skip-token
@anchor{wisent-skip-token}
Skip the lookahead token in order to resume parsing.
Return nil.
Must be used in error recovery semantic actions.

It typically looks like this:

@lisp
@group
(wisent-message "%s: skip %s" $action
                (wisent-token-to-string wisent-input))
(run-hook-with-args
 'wisent-discarding-token-functions wisent-input)
(wisent-clearin)
(wisent-errok)))
@end group
@end lisp
@end defun

@findex wisent-skip-block
@defun wisent-skip-block
Safely skip a block in order to resume parsing.
Return nil.
Must be used in error recovery semantic actions.

A block is data between an open-delimiter (syntax class @code{(}) and
a matching close-delimiter (syntax class @code{)}):

@example
@group
(a parenthesized block)
[a block between brackets]
@{a block between braces@}
@end group
@end example

The following example uses @code{wisent-skip-block} to safely skip a
block delimited by @samp{LBRACE} (@code{@{}) and @samp{RBRACE}
(@code{@}}) tokens, when a syntax error occurs in
@samp{other-components}:

@example
@group
(block ((LBRACE other-components RBRACE))
       ((LBRACE RBRACE))
       ((LBRACE error)
        (wisent-skip-block))
       )
@end group
@end example
@end defun

@node Debugging actions
@section Debugging semantic actions

@cindex semantic action symbols
Each semantic action is represented by a symbol interned in an
@dfn{obarray} that is part of the @acronym{LALR(1)} automaton
(@pxref{Compiling a grammar}).  @code{symbol-function} on a semantic
action symbol return the semantic action lambda expression.

A semantic action symbol name has the form
@code{@var{nonterminal}:@var{index}}, where @var{nonterminal} is the
name of the nonterminal symbol the action belongs to, and @var{index}
is an action sequence number within the scope of @var{nonterminal}.
For example, this nonterminal definition:

@example
@group
input:
   line                     [@code{input:0}]
 | input line
   (format "%s %s" $1 $2)   [@code{input:1}]
 ;
@end group
@end example

Will produce two semantic actions, and associated symbols:

@table @code
@item input:0
A default action that returns @code{$1}.

@item input:1
That returns @code{(format "%s %s" $1 $2)}.
@end table

@cindex debugging semantic actions
Debugging uses the Lisp debugger to investigate what is happening
during execution of semantic actions.
Three commands are available to debug semantic actions.  They receive
two arguments:

@itemize @bullet
@item The automaton that contains the semantic action.

@item The semantic action symbol.
@end itemize

@findex wisent-debug-on-entry
@deffn Command wisent-debug-on-entry automaton function
Request @var{automaton}'s @var{function} to invoke debugger each time it is called.
@var{function} must be a semantic action symbol that exists in @var{automaton}.
@end deffn

@findex wisent-cancel-debug-on-entry
@deffn Command wisent-cancel-debug-on-entry automaton function
Undo effect of @code{wisent-debug-on-entry} on @var{automaton}'s @var{function}.
@var{function} must be a semantic action symbol that exists in @var{automaton}.
@end deffn

@findex wisent-debug-show-entry
@deffn Command wisent-debug-show-entry automaton function
Show the source of @var{automaton}'s semantic action @var{function}.
@var{function} must be a semantic action symbol that exists in @var{automaton}.
@end deffn

@node Wisent Semantic
@chapter How to use Wisent with Semantic

@cindex tags
This section presents how the Wisent's parser can be used to produce
@dfn{tags} for the @semantic{} tool set.

@semantic{} tags form a hierarchy of Emacs Lisp data structures that
describes a program in a way independent of programming languages.
Tags map program declarations, like functions, methods, variables,
data types, classes, includes, grammar rules, etc..

@cindex WY grammar format
To use the Wisent parser with @semantic{} you have to define
your grammar in @dfn{WY} form, a grammar format very close
to the one used by Bison.

Please @inforef{top, Semantic Grammar Framework Manual, grammar-fw}
for more information on @semantic{} grammars.

@menu
* Grammar styles::
* Wisent Lex::
@end menu

@node Grammar styles
@section Grammar styles

@cindex grammar styles
@semantic{} parsing heavily depends on how you wrote the grammar.
There are mainly two styles to write a Wisent's grammar intended to be
used with the @semantic{} tool set: the @dfn{Iterative style} and the
@dfn{Bison style}.  Each one has pros and cons, and in certain cases
it can be worth a mix of the two styles!

@menu
* Iterative style::
* Bison style::
* Mixed style::
* Start nonterminals::
* Useful functions::
@end menu

@node Iterative style
@subsection Iterative style

@cindex grammar iterative style
The @dfn{iterative style} is the preferred style to use with @semantic{}.
It relies on an iterative parser back-end mechanism which parses start
nonterminals one at a time and automagically skips unexpected lexical
tokens in input.

Compared to rule-based iterative functions (@pxref{Bison style}),
iterative parsers are better in that they can handle obscure errors
more cleanly.

@cindex raw tag
Each start nonterminal must produces a @dfn{raw tag} by calling a
@code{TAG}-like grammar macro with appropriate parameters.  See also
@ref{Start nonterminals}.

@cindex expanded tag
Then, each parsing iteration automatically translates a raw tag into
@dfn{expanded tags}, updating the raw tag structure with internal
properties and buffer related data.

After parsing completes, it results in a tree of expanded tags.

The following example is a snippet of the iterative style Java grammar
provided in the @semantic{} distribution in the file
@file{semantic/wisent/java-tags.wy}.

@example
@group
@dots{}
;; Alternate entry points
;;    - Needed by partial re-parse
%start formal_parameter
@dots{}
;;    - Needed by EXPANDFULL clauses
%start formal_parameters
@dots{}

formal_parameter_list
  : PAREN_BLOCK
    (EXPANDFULL $1 formal_parameters)
  ;

formal_parameters
  : LPAREN
    ()
  | RPAREN
    ()
  | formal_parameter COMMA
  | formal_parameter RPAREN
  ;

formal_parameter
  : formal_parameter_modifier_opt type variable_declarator_id
    (VARIABLE-TAG $3 $2 nil :typemodifiers $1)
  ;
@end group
@end example

@findex EXPANDFULL
It shows the use of the @code{EXPANDFULL} grammar macro to parse a
@samp{PAREN_BLOCK} which contains a @samp{formal_parameter_list}.
@code{EXPANDFULL} tells to recursively parse @samp{formal_parameters}
inside @samp{PAREN_BLOCK}.  The parser iterates until it digested all
available input data inside the @samp{PAREN_BLOCK}, trying to match
any of the @samp{formal_parameters} rules:

@itemize
@item @samp{LPAREN}

@item @samp{RPAREN}

@item @samp{formal_parameter COMMA}

@item @samp{formal_parameter RPAREN}
@end itemize

At each iteration it will return a @samp{formal_parameter} raw tag,
or @code{nil} to skip unwanted (single @samp{LPAREN} or @samp{RPAREN}
for example) or unexpected input data.  Those raw tags will be
automatically expanded by the iterative back-end parser.

@node Bison style
@subsection Bison style

@cindex grammar bison style
What we call the @dfn{Bison style} is the traditional style of Bison's
grammars.  Compared to iterative style, it is not straightforward to
use grammars written in Bison style in @semantic{}.  Mainly because such
grammars are designed to parse the whole input data in one pass, and
don't use the iterative parser back-end mechanism (@pxref{Iterative
style}).  With Bison style the parser is called once to parse the
grammar start nonterminal.

The following example is a snippet of the Bison style Java grammar
provided in the @semantic{} distribution in the file
@file{semantic/wisent/java.wy}.

@example
@group
%start formal_parameter
@dots{}

formal_parameter_list
  : formal_parameter_list COMMA formal_parameter
    (cons $3 $1)
  | formal_parameter
    (list $1)
  ;

formal_parameter
  : formal_parameter_modifier_opt type variable_declarator_id
    (EXPANDTAG
     (VARIABLE-TAG $3 $2 :typemodifiers $1)
     )
  ;
@end group
@end example

The first consequence is that syntax errors are not automatically
handled by @semantic{}.  Thus, it is necessary to explicitly handle
them at the grammar level, providing error recovery rules to skip
unexpected input data.

The second consequence is that the iterative parser can't do automatic
tag expansion, except for the start nonterminal value.  It is
necessary to explicitly expand tags from concerned semantic actions by
calling the grammar macro @code{EXPANDTAG} with a raw tag as
parameter.  See also @ref{Start nonterminals}, for incremental
re-parse considerations.

@node Mixed style
@subsection Mixed style

@cindex grammar mixed style
@example
@group
%start grammar
;; Reparse
%start prologue epilogue declaration nonterminal rule
@dots{}

%%

grammar:
    prologue
  | epilogue
  | declaration
  | nonterminal
  | PERCENT_PERCENT
  ;
@dots{}

nonterminal:
    SYMBOL COLON rules SEMI
    (TAG $1 'nonterminal :children $3)
  ;

rules:
    lifo_rules
    (apply 'nconc (nreverse $1))
  ;

lifo_rules:
    lifo_rules OR rule
    (cons $3 $1)
  | rule
    (list $1)
  ;

rule:
    rhs
    (let* ((rhs $1)
           name type comps prec action elt)
      @dots{}
      (EXPANDTAG
       (TAG name 'rule :type type :value comps :prec prec :expr action)
       ))
  ;
@end group
@end example

This example shows how iterative and Bison styles can be combined in
the same grammar to obtain a good compromise between grammar
complexity and an efficient parsing strategy in an interactive
environment.

@samp{nonterminal} is parsed using iterative style via the main
@samp{grammar} rule.  The semantic action uses the @code{TAG} macro to
produce a raw tag, automagically expanded by @semantic{}.

But @samp{rules} part is parsed in Bison style! Why?

Rule delimiters are the colon (@code{:}), that follows the nonterminal
name, and a final semicolon (@code{;}).  Unfortunately these
delimiters are not @code{open-paren}/@code{close-paren} type, and the
Emacs' syntactic analyzer can't easily isolate data between them to
produce a @samp{RULES_PART} parenthesis-block-like lexical token.
Consequently it is not possible to use @code{EXPANDFULL} to iterate in
@samp{RULES_PART}, like this:

@example
@group
nonterminal:
    SYMBOL COLON rules SEMI
    (TAG $1 'nonterminal :children $3)
  ;

rules:
    RULES_PART  ;; @strong{Map a parenthesis-block-like lexical token}
    (EXPANDFULL $1 'rules)
  ;

rules:
    COLON
    ()
    OR
    ()
    SEMI
    ()
    rhs
    rhs
    (let* ((rhs $1)
           name type comps prec action elt)
      @dots{}
      (TAG name 'rule :type type :value comps :prec prec :expr action)
      )
  ;
@end group
@end example

In such cases, when it is difficult for Emacs to obtain
parenthesis-block-like lexical tokens, the best solution is to use the
traditional Bison style with error recovery!

In some extreme cases, it can also be convenient to extend the lexer,
to deliver new lexical tokens, to simplify the grammar.

@node Start nonterminals
@subsection Start nonterminals

@cindex start nonterminals
@cindex @code{reparse-symbol} property
When you write a grammar for @semantic{}, it is important to carefully
indicate the start nonterminals.  Each one defines an entry point in
the grammar, and after parsing its semantic value is returned to the
back-end iterative engine.  Consequently:

@strong{The semantic value of a start nonterminal must be a produced
by a TAG like grammar macro}.

Start nonterminals are declared by @code{%start} statements.  When
nothing is specified the first nonterminal that appears in the grammar
is the start nonterminal.

Generally, the following nonterminals must be declared as start
symbols:

@itemize @bullet
@item The main grammar entry point
@quotation
Of course!
@end quotation

@item nonterminals passed to @code{EXPAND}/@code{EXPANDFULL}
@quotation
These grammar macros recursively parse a part of input data, based on
rules of the given nonterminal.

For example, the following will parse @samp{PAREN_BLOCK} data using
the @samp{formal_parameters} rules:

@example
@group
formal_parameter_list
  : PAREN_BLOCK
    (EXPANDFULL $1 formal_parameters)
  ;
@end group
@end example

The semantic value of @samp{formal_parameters} becomes the value of
the @code{EXPANDFULL} expression.  It is a list of @semantic{} tags
spliced in the tags tree.

Because the automaton must know that @samp{formal_parameters} is a
start symbol, you must declare it like this:

@example
@group
%start formal_parameters
@end group
@end example
@end quotation
@end itemize

@cindex incremental re-parse
@cindex reparse-symbol
The @code{EXPANDFULL} macro has a side effect it is important to know,
related to the incremental re-parse mechanism of @semantic{}: the
nonterminal symbol parameter passed to @code{EXPANDFULL} also becomes
the @code{reparse-symbol} property of the tag returned by the
@code{EXPANDFULL} expression.

When buffer's data mapped by a tag is modified, @semantic{}
schedules an incremental re-parse of that data, using the tag's
@code{reparse-symbol} property as start nonterminal.

@strong{The rules associated to such start symbols must be carefully
reviewed to ensure that the incremental parser will work!}

Things are a little bit different when the grammar is written in Bison
style.

@strong{The @code{reparse-symbol} property is set to the nonterminal
symbol the rule that explicitly uses @code{EXPANDTAG} belongs to.}

For example:

@example
@group
rule:
    rhs
    (let* ((rhs $1)
           name type comps prec action elt)
      @dots{}
      (EXPANDTAG
       (TAG name 'rule :type type :value comps :prec prec :expr action)
       ))
  ;
@end group
@end example

Set the @code{reparse-symbol} property of the expanded tag to
@samp{rule}.  A important consequence is that:

@strong{Every nonterminal having any rule that calls @code{EXPANDTAG}
in a semantic action, should be declared as a start symbol!}

@node Useful functions
@subsection Useful functions

Here is a description of some predefined functions it might be useful
to know when writing new code to use Wisent in @semantic{}:

@findex wisent-collect-unmatched-syntax
@defun wisent-collect-unmatched-syntax input
Add @var{input} lexical token to the cache of unmatched tokens, in
variable @code{semantic-unmatched-syntax-cache}.

See implementation of the function @code{wisent-skip-token} in
@ref{Error recovery}, for an example of use.
@end defun

@node Wisent Lex
@section The Wisent Lex lexer

@findex semantic-lex
The lexical analysis step of @semantic{} is performed by the general
function @code{semantic-lex}.  For more information, @inforef{Writing
Lexers, ,semantic-langdev}.

@code{semantic-lex} produces lexical tokens of the form:

@example
@group
@code{(@var{token-class start} . @var{end})}
@end group
@end example

@table @var
@item token-class
Is a symbol that identifies a lexical token class, like @code{symbol},
@code{string}, @code{number}, or @code{PAREN_BLOCK}.

@item start
@itemx end
Are the start and end positions of mapped data in the input buffer.
@end table

The Wisent's parser doesn't depend on the nature of analyzed input
stream (buffer, string, etc.), and requires that lexical tokens have a
different form (@pxref{Writing a lexer}):

@example
@group
@code{(@var{token-class value} [@var{start} . @var{end}])}
@end group
@end example

@cindex lexical token mapping
@code{wisent-lex} is the default Wisent's lexer used in @semantic{}.

@vindex wisent-lex-istream
@findex wisent-lex
@defun wisent-lex
Return the next available lexical token in Wisent's form.

The variable @code{wisent-lex-istream} contains the list of lexical
tokens produced by @code{semantic-lex}.  Pop the next token available
and convert it to a form suitable for the Wisent's parser.
@end defun

Mapping of lexical tokens as produced by @code{semantic-lex} into
equivalent Wisent lexical tokens is straightforward:

@example
@group
(@var{token-class start} . @var{end})
     @result{} (@var{token-class value start} . @var{end})
@end group
@end example

@var{value} is the input @code{buffer-substring} from @var{start} to
@var{end}.

@node GNU Free Documentation License
@appendix GNU Free Documentation License

@include doclicense.texi

@node Index
@unnumbered Index
@printindex cp

@iftex
@contents
@summarycontents
@end iftex

@bye

@c Following comments are for the benefit of ispell.

@c  LocalWords:  Wisent automagically wisent Wisent's LALR obarray