reader.c

bison源代码.bison 是替代yacc的语法分析程序生成器. yacc是 Yet Another Compiler Compiler的缩写. bison又是什么呐 是一个生成可以分析文本文件结构

/* Input parser for bison
   Copyright (C) 1984, 1986, 1989, 1992 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"

#define	LTYPESTR	"\n#ifndef YYLTYPE\ntypedef\n  struct yyltype\n\
    {\n      int timestamp;\n      int first_line;\n      int first_column;\
\n      int last_line;\n      int last_column;\n      char *text;\n   }\n\
  yyltype;\n\n#define YYLTYPE yyltype\n#endif\n\n"

/* 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 int noparserflag;
extern int rawtoknumflag;
extern bucket *symval;
extern int numval;
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 char *int_to_string();
extern void fatal();
extern void fatals();
extern void warn();
extern void warni();
extern void warns();
extern void warnss();
extern void warnsss();
extern void unlex();
extern void done();

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

void reader_output_yylsp();
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 parse_thong_decl();
void copy_action();
void readgram();
void record_rule_line();
void packsymbols();
void output_token_defines();
void packgram();
int read_signed_integer();
static 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;
int *user_toknums;

/* 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 */

static bucket *errtoken;

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

extern char *version_string;


static void
skip_to_char(target)
     int target;
{
  int c;
  if (target == '\n')
    warn("   Skipping to next \\n");
  else
    warni("   Skipping to next %c", target);

  do
    c = skip_white_space();
  while (c != target && c != EOF);
  if (c != EOF) 
    ungetc(c, finput);
}


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("$undefined.")->class = STOKEN;
  /* Read the declaration section.  Copy %{ ... %} groups to ftable and fdefines file.
     Also notice any %token, %left, etc. found there.  */
  if (noparserflag) 
    fprintf(ftable, "\n/*  Bison-generated parse tables, made from %s\n",
		infile);
  else
    fprintf(ftable, "\n/*  A Bison parser, made from %s\n", infile);
  fprintf(ftable, " by  %s  */\n\n", version_string);
  fprintf(ftable, "#define YYBISON 1  /* Identify Bison output.  */\n\n");
  read_declarations();
  /* start writing the guard and action files, if they are needed.  */
  output_headers();
  /* read in the grammar, build grammar in list form.  write out guards and actions.  */
  readgram();
  /* Now we know whether we need the line-number stack.
     If we do, write its type into the .tab.h file.  */
  if (fdefines)
    reader_output_yylsp(fdefines);
  /* write closing delimiters for actions and guards.  */
  output_trailers();
  if (yylsp_needed)
    fprintf(ftable, "#define YYLSP_NEEDED\n\n");
  /* 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();
}

void
reader_output_yylsp(f)
     FILE *f;
{
  if (yylsp_needed)
    fprintf(f, LTYPESTR);
}

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

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 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 THONG:
	      parse_thong_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();
		}
	      break;

	    case PURE_PARSER:
	      pure_parser = 1;
	      break;

	    case NOOP:
	      break;

	    default:
	      warns("unrecognized: %s", token_buffer);
	      skip_to_char('%');
	  }
	}
      else if (c == EOF)
        fatal("no input grammar");
      else
	{
		char buff[100];
		sprintf(buff, "unknown character: %s", printable_version(c)); 
		warn(buff);
		skip_to_char('%');
	}
    }
}


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

void
copy_definition ()
{
  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;

  if (!nolinesflag)
    fprintf(fattrs, "#line %d \"%s\"\n", lineno, infile);

  after_percent = 0;

  c = getc(finput);

  for (;;)
    {
      switch (c)
	{
	case '\n':
	  putc(c, fattrs);
	  lineno++;
	  break;

	case '%':
          after_percent = -1;
	  break;
	      
	case '\'':
	case '"':
	  match = c;
	  putc(c, fattrs);
	  c = getc(finput);

	  while (c != match)
	    {
	      if (c == EOF)
		fatal("unterminated string at end of file");
	      if (c == '\n')
		{
		  warn("unterminated string");
		  ungetc(c, finput);
		  c = match;
		  continue;
		}

	      putc(c, fattrs);
	      
	      if (c == '\\')
		{
		  c = getc(finput);
		  if (c == EOF)
		    fatal("unterminated string at end of file");
		  putc(c, fattrs);
		  if (c == '\n')
		    lineno++;
		}

	      c = getc(finput);
	    }

	  putc(c, fattrs);
	  break;

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

	  cplus_comment = (c == '/');
	  putc(c, fattrs);
	  c = getc(finput);

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

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

	  break;

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

	default:
	  putc(c, fattrs);
	}

      c = getc(finput);

      if (after_percent)
	{
	  if (c == '}')
	    return;
	  putc('%', fattrs);
	}
      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 token = 0;
  register char *typename = 0;
  register struct bucket *symbol = NULL;  /* pts to symbol being defined */
  int k;

  for (;;)
    {
      if(ungetc(skip_white_space(), finput) == '%')
	return;
      token = lex();
      if (token == COMMA)
	{
	  symbol = NULL;
	  continue;
	}
      if (token == TYPENAME)
	{
	  k = strlen(token_buffer);
	  typename = NEW2(k + 1, char);
	  strcpy(typename, token_buffer);
	  value_components_used = 1;
	  symbol = NULL;
	}
      else if (token == IDENTIFIER && *symval->tag == '\"'
		&& symbol) 
	{
	  translations = 1;
	  symval->class = STOKEN;
	  symval->type_name = typename;
	  symval->user_token_number = symbol->user_token_number;
	  symbol->user_token_number = SALIAS;

	  symval->alias = symbol;	
	  symbol->alias = symval;	
	  symbol = NULL;

 	  nsyms--;   /* symbol and symval combined are only one symbol */
	}
      else if (token == IDENTIFIER)
	{
	  int oldclass = symval->class;
	  symbol = symval;

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

	  if (typename)
	    {
	      if (symbol->type_name == NULL)
		symbol->type_name = typename;
	      else if (strcmp(typename, symbol->type_name) != 0)
		warns("type redeclaration for %s", symbol->tag);
	    }
	}
      else if (symbol && token == NUMBER)
        {
	  symbol->user_token_number = numval;
	  translations = 1;
        }
      else
	{
	  warnss("`%s' is invalid in %s",
		token_buffer, 
		(what_is == STOKEN) ? "%token" : "%nterm");
	  skip_to_char('%');
	}
    }

}

/* parse what comes after %thong 
	the full syntax is
		%thong <type> token number literal
 the <type> or number may be omitted.  The number specifies the
 user_token_number.

 Two symbols are entered in the table, one for the token symbol and
 one for the literal.  Both are given the <type>, if any, from the declaration.
 The ->user_token_number of the first is SALIAS and the ->user_token_number
 of the second is set to the number, if any, from the declaration.
 The two symbols are linked via pointers in their ->alias fields.
 
 during output_defines_table, the symbol is reported
 thereafter, only the literal string is retained
 it is the literal string that is output to yytname
*/

void
parse_thong_decl ()
{
  register int token;
  register struct bucket *symbol;
  register char *typename = 0;
  int k, usrtoknum;

  translations = 1;
  token = lex();		/* fetch typename or first token */
  if (token == TYPENAME) {
    k = strlen(token_buffer);
    typename = NEW2(k + 1, char);
    strcpy(typename, token_buffer);
    value_components_used = 1;
    token = lex();		/* fetch first token */
  }

  /* process first token */

  if (token != IDENTIFIER) 
    {
      warns("unrecognized item %s, expected an identifier", 
	    token_buffer);
      skip_to_char('%');
      return;
    }
  symval->class = STOKEN;
  symval->type_name = typename;
  symval->user_token_number = SALIAS;
  symbol = symval;

  token = lex();		/* get number or literal string */
	
  if (token == NUMBER) {
    usrtoknum = numval;
    token = lex();		/* okay, did number, now get literal */
  }
  else usrtoknum = 0;

  /* process literal string token */

  if (token != IDENTIFIER || *symval->tag != '\"') 
    {
      warns("expected string constant instead of %s", 
	    token_buffer);
      skip_to_char('%');
      return;
    }
  symval->class = STOKEN;
  symval->type_name = typename;
  symval->user_token_number = usrtoknum;

  symval->alias = symbol;	
  symbol->alias = symval;	

  nsyms--;			/* symbol and symval combined are only one symbol */
}


/* parse what comes after %start */

void
parse_start_decl ()
{
  if (start_flag)
    warn("multiple %start declarations");
  if (lex() != IDENTIFIER)
    warn("invalid %start declaration");
  else
    {
      start_flag = 1;
      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;

  if (lex() != TYPENAME)
    {
      warn("%type declaration has no <typename>");
      skip_to_char('%');
      return;
    }

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

  for (;;)
    {
      register int t;

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

      t = lex();

      switch (t)
	{

	case COMMA:
	case SEMICOLON:
	  break;

	case IDENTIFIER:
	  if (symval->type_name == NULL)
	    symval->type_name = name;
	  else if (strcmp(name, symval->type_name) != 0)
	    warns("type redeclaration for %s", symval->tag);

	  break;

	default:
	  warns("invalid %%type declaration due to item: `%s'", token_buffer);
	  skip_to_char('%');
	}