reader.c

C++版 词法分析、语法分析器

/* Input parser for bison
   Copyright (C) 1984, 1986, 1989 Free Software Foundation, Inc.

This file is part of Bison, the GNU Compiler Compiler.

Bison is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 2, or (at your option)
any later version.

Bison is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
GNU General Public License for more details.

You should have received a copy of the GNU General Public License
along with Bison; see the file COPYING.  If not, write to
the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.  */


/* read in the grammar specification and record it in the format described in gram.h.
  All guards are copied into the fguard file and all actions into faction,
  in each case forming the body of a C function (yyguard or yyaction)
  which contains a switch statement to decide which guard or action to execute.

The entry point is reader().  */

#include <stdio.h>
#include <ctype.h>
#include "system.h"
#include "files.h"
#include "new.h"
#include "symtab.h"
#include "lex.h"
#include "gram.h"
#include "machine.h"


/* Number of slots allocated (but not necessarily used yet) in `rline'  */
int rline_allocated;

extern char *program_name;
extern int definesflag;
extern int nolinesflag;
extern bucket *symval;
extern int numval;
extern int failure;
extern int expected_conflicts;
extern char *token_buffer;



extern void init_lex();
extern void tabinit();
extern void output_headers();
extern void output_trailers();
extern void free_symtab();
extern void open_extra_files();
extern void fatal();
extern void fatals();
extern void unlex();
extern void done();

extern int skip_white_space();
extern int parse_percent_token();
extern int lex();

void read_declarations();
void copy_definition();
void parse_token_decl();
void parse_start_decl();
void parse_type_decl();
void parse_assoc_decl();
void parse_union_decl();
void parse_expect_decl();
void copy_action();
void readgram();
void record_rule_line();
void packsymbols();
void output_token_defines();
void packgram();
int read_signed_integer();
int get_type();

typedef
  struct symbol_list
    {
      struct symbol_list *next;
      bucket *sym;
      bucket *ruleprec;
    }
  symbol_list;



int lineno;
symbol_list *grammar;
int start_flag;
bucket *startval;
char **tags;

/* Nonzero if components of semantic values are used, implying
   they must be unions.  */
static int value_components_used;

static int typed;  /* nonzero if %union has been seen.  */

static int lastprec;  /* incremented for each %left, %right or %nonassoc seen */

static int gensym_count;  /* incremented for each generated symbol */

bucket *errtoken;

/* Nonzero if any action or guard uses the @n construct.  */
int yylsp_needed;

extern char *version_string;

extern void output_before_read();
extern  void output_about_token();
void set_parser_name();
void cputc();
void hputc();
void copy_a_definition();
void copy_header_definition();
void parse_name_declaration();
void parse_define();
void read_a_name();

void
reader()
{
  start_flag = 0;
  startval = NULL;  /* start symbol not specified yet. */

#if 0
  translations = 0;  /* initially assume token number translation not needed.  */
#endif
  /* Nowadays translations is always set to 1,
     since we give `error' a user-token-number
     to satisfy the Posix demand for YYERRCODE==256.  */
  translations = 1;

  nsyms = 1;
  nvars = 0;
  nrules = 0;
  nitems = 0;
  rline_allocated = 10;
  rline = NEW2(rline_allocated, short);

  typed = 0;
  lastprec = 0;

  gensym_count = 0;

  semantic_parser = 0;
  pure_parser = 0;
  yylsp_needed = 0;

  grammar = NULL;

  init_lex();
  lineno = 1;

  /* initialize the symbol table.  */
  tabinit();
  /* construct the error token */
  errtoken = getsym("error");
  errtoken->class = STOKEN;
  errtoken->user_token_number = 256; /* Value specified by posix.  */
  /* construct a token that represents all undefined literal tokens. */
  /* it is always token number 2.  */
  getsym("$illegal.")->class = STOKEN;
  /* Read the declaration section.  Copy %{ ... %} groups to ftable and fdefines file.
     Also notice any %token, %left, etc. found there.  */
  output_before_read();

  read_declarations();
  output_headers();
  /* read in the grammar, build grammar in list form.  write out guards and actions.  */
  readgram();
  /* write closing delimiters for actions and guards.  */
  output_trailers();
  /* assign the symbols their symbol numbers.
     Write #defines for the token symbols into fdefines if requested.  */
  packsymbols();
  /* convert the grammar into the format described in gram.h.  */
  packgram();
  /* free the symbol table data structure
     since symbols are now all referred to by symbol number.  */
  free_symtab();
}



/* read from finput until %% is seen.  Discard the %%.
Handle any % declarations,
and copy the contents of any %{ ... %} groups to ftable.  */

void
read_declarations ()
{
  register int c;
  register int tok;

  for (;;)
    {
      c = skip_white_space();

      if (c == '%')
	{
	  tok = parse_percent_token();

	  switch (tok)
	    {
	    case TWO_PERCENTS:
	      return;

	    case PERCENT_LEFT_CURLY:
	      copy_definition();
	      break;
	    case PERCENT_LEFT_CURLY_HEADER:
	      copy_header_definition();
	      break;

	    case TOKEN:
	      parse_token_decl (STOKEN, SNTERM);
	      break;
	
	    case NTERM:
	      parse_token_decl (SNTERM, STOKEN);
	      break;
	
	    case TYPE:
	      parse_type_decl();
	      break;
	
	    case START:
	      parse_start_decl();
	      break;
	
	    case UNION:
	      parse_union_decl();
	      break;
	
	    case EXPECT:
	      parse_expect_decl();
	      break;
	
	    case LEFT:
	      parse_assoc_decl(LEFT_ASSOC);
	      break;

	    case RIGHT:
	      parse_assoc_decl(RIGHT_ASSOC);
	      break;

	    case NONASSOC:
	      parse_assoc_decl(NON_ASSOC);
	      break;

	    case SEMANTIC_PARSER:
	      if (semantic_parser == 0)
		{
		  semantic_parser = 1;
		  open_extra_files();
		  fprintf(stderr,
                "%semantic_parser no more supported in this version of bison !!! \n errors will be done ! use classic bison, use simple parser, or addapt this version to semantic parser\n");
		}
	      break;

	    case PURE_PARSER:
	      pure_parser = 1;
	      break;

           
	    case PARSER_NAME:
              parse_name_declaration();
	      break;
           
	    case DEFINE_SYM:
              parse_define();
	      break;
           

	    default:
	      fatal("junk after `%%' in definition section");
	    }
	}
      else if (c == EOF)
        fatal("no input grammar");
      else if (c >= 040 && c <= 0177)
	fatals ("unknown character `%c' in declaration section", c);
      else
	fatals ("unknown character with code 0x%x in declaration section", c);
    }
}


/* copy the contents of a %{ ... %} into the definitions file.
The %{ has already been read.  Return after reading the %}.  */

void
copy_definition ()
{
 if (!nolinesflag)
    fprintf(ftable, "#line %d \"%s\"\n", lineno, quoted_filename(infile));

 copy_a_definition (cputc);
}

void
copy_header_definition ()
{
  if (!nolinesflag)
    {fprintf(ftable, "#line %d \"%s\"\n", lineno, quoted_filename(infile));
     if(definesflag) 
      fprintf(fdefines, "#line %d \"%s\"\n", lineno, quoted_filename(infile));
    }
 copy_a_definition (hputc);
}
void 
hputc(c)
int c;
{
 putc(c,ftable);
 if(definesflag) putc(c,fdefines);
}
void 
cputc(c)
int c;
{
 putc(c,ftable);
}

void
copy_a_definition (do_put)
void (*do_put)();
{
  register int c;
  register int match;
  register int ended;
  register int after_percent;  /* -1 while reading a character if prev char was % */
  int cplus_comment;

  after_percent = 0;

  c = getc(finput);

  for (;;)
    {
      switch (c)
	{
	case '\n':
	  (*do_put)(c);
	  lineno++;
	  break;

	case '%':
          after_percent = -1;
	  break;
	      
	case '\'':
	case '"':
	  match = c;
	  (*do_put)(c);
	  c = getc(finput);

	  while (c != match)
	    {
	      if (c == EOF || c == '\n')
		fatal("unterminated string");

	      (*do_put)(c);
	      
	      if (c == '\\')
		{
		  c = getc(finput);
		  if (c == EOF)
		    fatal("unterminated string");
		  (*do_put)(c);
		  if (c == '\n')
		    lineno++;
		}

	      c = getc(finput);
	    }

	  (*do_put)(c);
	  break;

	case '/':
	  (*do_put)(c);
	  c = getc(finput);
	  if (c != '*' && c != '/')
	    continue;

	  cplus_comment = (c == '/');
	  (*do_put)(c);
	  c = getc(finput);

	  ended = 0;
	  while (!ended)
	    {
	      if (!cplus_comment && c == '*')
		{
		  while (c == '*')
		    {
		      (*do_put)(c);
		      c = getc(finput);
		    }

		  if (c == '/')
		    {
		      (*do_put)(c);
		      ended = 1;
		    }
		}
	      else if (c == '\n')
		{
		  lineno++;
		  (*do_put)(c);
		  if (cplus_comment)
		    ended = 1;
		  else
		    c = getc(finput);
		}
	      else if (c == EOF)
		fatal("unterminated comment in `%{' definition");
	      else
		{
		  (*do_put)(c);
		  c = getc(finput);
		}
	    }

	  break;

	case EOF:
	  fatal("unterminated `%{' definition");

	default:
	  (*do_put)(c);
	}

      c = getc(finput);

      if (after_percent)
	{
	  if (c == '}')
	    return;
          (*do_put)('%');
	}
      after_percent = 0;

    }

}



/* parse what comes after %token or %nterm.
For %token, what_is is STOKEN and what_is_not is SNTERM.
For %nterm, the arguments are reversed.  */

void
parse_token_decl (what_is, what_is_not)
     int what_is, what_is_not;
{
/*   register int start_lineno; JF */
  register int token = 0;
  register int prev;
  register char *typename = 0;
  int k;

/*   start_lineno = lineno; JF */

  for (;;)
    {
      if(ungetc(skip_white_space(), finput) == '%')
	return;

/*      if (lineno != start_lineno)
	return; JF */

      /* we have not passed a newline, so the token now starting is in this declaration */
      prev = token;

      token = lex();
      if (token == COMMA)
	continue;
      if (token == TYPENAME)
	{
	  k = strlen(token_buffer);
	  typename = NEW2(k + 1, char);
	  strcpy(typename, token_buffer);
	  value_components_used = 1;
	}
      else if (token == IDENTIFIER)
	{
	  int oldclass = symval->class;

	  if (symval->class == what_is_not)
	    fatals("symbol %s redefined", symval->tag);
	  symval->class = what_is;
	  if (what_is == SNTERM && oldclass != SNTERM)
	    symval->value = nvars++;

	  if (typename)
	    {
	      if (symval->type_name == NULL)
		symval->type_name = typename;
	      else
		fatals("type redeclaration for %s", symval->tag);
	    }
	}
      else if (prev == IDENTIFIER && token == NUMBER)
        {
	  symval->user_token_number = numval;
	  translations = 1;
        }
      else
	fatal("invalid text in %token or %nterm declaration");
    }

}



/* parse what comes after %start */

void
parse_start_decl ()
{
  if (start_flag)
    fatal("multiple %start declarations");
  start_flag = 1;
  if (lex() != IDENTIFIER)
    fatal("invalid %start declaration");
  startval = symval;
}



/* read in a %type declaration and record its information for get_type_name to access */

void
parse_type_decl ()
{
  register int k;
  register char *name;
/*   register int start_lineno; JF */

  if (lex() != TYPENAME)
    fatal("ill-formed %type declaration");

  k = strlen(token_buffer);
  name = NEW2(k + 1, char);
  strcpy(name, token_buffer);

/*   start_lineno = lineno; */

  for (;;)
    {
      register int t;

      if(ungetc(skip_white_space(), finput) == '%')
	return;

/*       if (lineno != start_lineno)
	return; JF */

      /* we have not passed a newline, so the token now starting is in this declaration */

      t = lex();

      switch (t)
	{

	case COMMA:
	case SEMICOLON:
	  break;

	case IDENTIFIER:
	  if (symval->type_name == NULL)
	    symval->type_name = name;
	  else
	    fatals("type redeclaration for %s", symval->tag);

	  break;

	default:
	  fatal("invalid %type declaration");
	}
    }
}



/* read in a %left, %right or %nonassoc declaration and record its information.  */
/* assoc is either LEFT_ASSOC, RIGHT_ASSOC or NON_ASSOC.  */

void
parse_assoc_decl (assoc)
int assoc;
{
  register int k;
  register char *name = NULL;
/*  register int start_lineno; JF */
  register int prev = 0;	/* JF added = 0 to keep lint happy */

  lastprec++;  /* Assign a new precedence level, never 0.  */

/*   start_lineno = lineno; */

  for (;;)
    {
      register int t;

      if(ungetc(skip_white_space(), finput) == '%')
	return;

      /* if (lineno != start_lineno)
	return; JF */

      /* we have not passed a newline, so the token now starting is in this declaration */

      t = lex();