%{
/*
 *  R : A Computer Language for Statistical Data Analysis
 *  Copyright (C) 1995, 1996, 1997  Robert Gentleman and Ross Ihaka
 *  Copyright (C) 1997--2022  The R Core Team
 *  Copyright (C) 2009--2011  Romain Francois
 *
 *  This program 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 of the License, or
 *  (at your option) any later version.
 *
 *  This program 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 this program; if not, a copy is available at
 *  https://www.R-project.org/Licenses/
 */

#ifdef HAVE_CONFIG_H
#include <config.h>
#endif

#define R_USE_SIGNALS 1
#include "IOStuff.h"		/*-> Defn.h */
#include "Fileio.h"
#include "Parse.h"
#include <R_ext/Print.h>

#if !defined(__STDC_ISO_10646__) && (defined(__APPLE__) || defined(__FreeBSD__) || defined(__sun))
/* This may not be 100% true (see the comment in rlocale.h),
   but it seems true in normal locales.
 */
# define __STDC_ISO_10646__
#endif

/* #define YYDEBUG 1 */
#define YYERROR_VERBOSE 1
#define PARSE_ERROR_SIZE 256	    /* Parse error messages saved here */
#define PARSE_CONTEXT_SIZE 256	    /* Recent parse context kept in a circular buffer */

static Rboolean busy = FALSE;
static SEXP R_NullSymbol = NULL;

static int identifier ;
static void incrementId(void);
static void initData(void);
static void initId(void);
static void record_( int, int, int, int, int, int, char* ) ;

static void yyerror(const char *);
static int yylex();
int yyparse(void);

static FILE *fp_parse;
static int (*ptr_getc)(void);

static int	SavedToken;
static SEXP	SavedLval;

#define yyconst const

typedef struct yyltype
{
  int first_line;
  int first_column;
  int first_byte;

  int last_line;
  int last_column;
  int last_byte;
  
  int first_parsed;
  int last_parsed;
  
  int id;
} yyltype;


#define INIT_DATA_COUNT 16384    	/* init parser data to this size */
#define MAX_DATA_COUNT   65536		/* release it at the end if it is this size or larger*/

#define DATA_COUNT  (length( PS_DATA ) / DATA_ROWS)
#define ID_COUNT    ((length( PS_IDS ) / 2) - 1)

static void finalizeData( ) ;
static void growData( ) ;
static void growID( int ) ;

#define DATA_ROWS 8

#define _FIRST_PARSED( i ) INTEGER( PS_DATA )[ DATA_ROWS*(i)     ]
#define _FIRST_COLUMN( i ) INTEGER( PS_DATA )[ DATA_ROWS*(i) + 1 ]
#define _LAST_PARSED( i )  INTEGER( PS_DATA )[ DATA_ROWS*(i) + 2 ]
#define _LAST_COLUMN( i )  INTEGER( PS_DATA )[ DATA_ROWS*(i) + 3 ]
#define _TERMINAL( i )     INTEGER( PS_DATA )[ DATA_ROWS*(i) + 4 ]
#define _TOKEN( i )        INTEGER( PS_DATA )[ DATA_ROWS*(i) + 5 ]
#define _ID( i )           INTEGER( PS_DATA )[ DATA_ROWS*(i) + 6 ]
#define _PARENT(i)         INTEGER( PS_DATA )[ DATA_ROWS*(i) + 7 ]

#define ID_ID( i )      INTEGER(PS_IDS)[ 2*(i) ]
#define ID_PARENT( i )  INTEGER(PS_IDS)[ 2*(i) + 1 ]

static void modif_token( yyltype*, int ) ;
static void recordParents( int, yyltype*, int) ;

static int _current_token ;

/**
 * Records the current non-terminal token expression and gives it an id
 *
 * @param loc the location of the expression
 */   
static void setId(yyltype loc){
    record_( 
	    (loc).first_parsed, (loc).first_column, (loc).last_parsed, (loc).last_column, 
	    _current_token, (loc).id, 0 ) ;
}

# define YYLTYPE yyltype
# define YYLLOC_DEFAULT(Current, Rhs, N)				\
    do	{ 								\
	if (N){								\
	    (Current).first_line   = YYRHSLOC (Rhs, 1).first_line;	\
	    (Current).first_column = YYRHSLOC (Rhs, 1).first_column;	\
	    (Current).first_byte   = YYRHSLOC (Rhs, 1).first_byte;	\
	    (Current).last_line    = YYRHSLOC (Rhs, N).last_line;	\
	    (Current).last_column  = YYRHSLOC (Rhs, N).last_column;	\
	    (Current).last_byte    = YYRHSLOC (Rhs, N).last_byte;	\
	    (Current).first_parsed = YYRHSLOC (Rhs, 1).first_parsed;    \
	    (Current).last_parsed  = YYRHSLOC (Rhs, N).last_parsed;	\
	    incrementId( ) ; 						\
	    (Current).id = identifier ; 				\
	    _current_token = yyr1[yyn] ; 				\
	    if (ParseState.keepSrcRefs && ParseState.keepParseData) {	\
	        yyltype childs[N];					\
	        int ii = 0; 						\
	        for(ii=0; ii<N; ii++){					\
		      childs[ii] = YYRHSLOC (Rhs, (ii+1) ) ; 		\
	        } 							\
	        recordParents( identifier, childs, N) ; 		\
	    }								\
	} else	{							\
	  (Current).first_line   = (Current).last_line   =		\
	    YYRHSLOC (Rhs, 0).last_line;				\
	  (Current).first_parsed   = (Current).last_parsed   =		\
	    YYRHSLOC (Rhs, 0).last_parsed;				\
	  (Current).first_column = YYRHSLOC (Rhs, 0).last_column;	\
	  (Current).last_column = (Current).first_column - 1;		\
	  (Current).first_byte = YYRHSLOC (Rhs, 0).last_byte;		\
	  (Current).last_byte = (Current).first_byte - 1;		\
	  (Current).id = NA_INTEGER;                                    \
	} 								\
    } while (0)

		
# define YY_LOCATION_PRINT(File,Loc)					\
 fprintf ( File, "%d.%d.%d-%d.%d.%d (%d)",				\
 	(Loc).first_line, (Loc).first_column,	(Loc).first_byte, 	\
 	(Loc).last_line,  (Loc).last_column, 	(Loc).last_byte, 	\
	(Loc).id )

/* Useful defines so editors don't get confused ... */

#define LBRACE	'{'
#define RBRACE	'}'

/* Functions used in the parsing process */

static void	CheckFormalArgs(SEXP, SEXP, YYLTYPE *);
static SEXP	FirstArg(SEXP, SEXP); /* create list with one element */
static void 	GrowList(SEXP, SEXP); /* add element to list end */

static void	SetSingleSrcRef(SEXP);
static void	AppendToSrcRefs(SEXP);
static void	PrependToSrcRefs(SEXP);
static SEXP	SrcRefsToVectorList();

static void	IfPush(void);
static int	KeywordLookup(const char *);
static SEXP	NewList(void);
static void	NextArg(SEXP, SEXP, SEXP); /* add named element to list end */
static SEXP	TagArg(SEXP, SEXP, YYLTYPE *);
static int 	processLineDirective();
static int      checkForPlaceholder(SEXP placeholder, SEXP arg);

static int HavePlaceholder = FALSE; 
attribute_hidden SEXP R_PlaceholderToken = NULL;

static int HavePipeBind = FALSE; 
static SEXP R_PipeBindSymbol = NULL;

/* These routines allocate constants */

static SEXP	mkComplex(const char *);
SEXP		mkFalse(void);
static SEXP     mkFloat(const char *);
static SEXP 	mkInt(const char *); 
static SEXP	mkNA(void);
SEXP		mkTrue(void);

/* Internal lexer / parser state variables */

static int	EatLines = 0;
static int	GenerateCode = 0;
static int	EndOfFile = 0;
static int	Status = 1;
static int	xxgetc();
static int	xxungetc(int);
static int	xxcharcount, xxcharsave;
static int	xxlinesave, xxbytesave, xxcolsave, xxparsesave;

static SrcRefState ParseState;

#define PS_SET_SRCREFS(x)   SET_VECTOR_ELT(ParseState.sexps, 0, (x))
#define PS_SET_SRCFILE(x)   SET_VECTOR_ELT(ParseState.sexps, 1, (x))
#define PS_SET_ORIGINAL(x)  SET_VECTOR_ELT(ParseState.sexps, 2, (x))

/* direct pointer to data is kept for performance of finalizeData() */
#define PS_SET_DATA(x)      do {                \
    SEXP __x__ = (x);                           \
    SET_VECTOR_ELT(ParseState.sexps, 3, __x__); \
    ParseState.data = __x__;                    \
} while(0);

#define PS_SET_TEXT(x)      SET_VECTOR_ELT(ParseState.sexps, 4, (x))
#define PS_SET_IDS(x)       SET_VECTOR_ELT(ParseState.sexps, 5, (x))
#define PS_SET_SVS(x)       SET_VECTOR_ELT(ParseState.sexps, 6, (x))

#define PS_SRCREFS          VECTOR_ELT(ParseState.sexps, 0)
#define PS_SRCFILE          VECTOR_ELT(ParseState.sexps, 1)
#define PS_ORIGINAL         VECTOR_ELT(ParseState.sexps, 2)
#define PS_DATA             ParseState.data
#define PS_TEXT             VECTOR_ELT(ParseState.sexps, 4)
#define PS_IDS              VECTOR_ELT(ParseState.sexps, 5)
#define PS_SVS              VECTOR_ELT(ParseState.sexps, 6)

/* Memory protection in the parser

   The generated code of the parser keeps semantic values (SEXPs) on its
   semantic values stack. Values are added to the stack during shift and
   reduce operations and are removed during reduce operations or error
   handling. Values are created by the lexer before they are added to the
   stack. Values are also held in a local SEXP variable once removed from
   the stack but still needed. The stack is automatically expanded on demand.

   For memory protection, it would be natural to have that stack on the R heap
   and to use PROTECT/UNPROTECT to protect values in local SEXP variables.
   Unfortunately, bison does not seem to be customizable enough to allow this.

   Hence, semantic values, when created by the lexer or reduce operations, are
   placed on parser state precious multi-set via PRESERVE_SV. They are removed
   from the multi-set in reduce operations using RELEASE_SV, because by design
   of the bison parsers such values are subsequently removed from the stack.
   They are also automatically removed when the parsing finishes, including
   parser error (also on R error, via the context on-end action).

   Previously semantic values were protected via PROTECT/UNPROTECT_PTR with
   similar semantics but using protect stack shared with PROTECT/UNPROTECT.
   Using a separate precious multi-set is safe even with interleaving of the
   two protection schemes.
*/

#define INIT_SVS()     PS_SET_SVS(R_NewPreciousMSet(200))
#define PRESERVE_SV(x) R_PreserveInMSet((x), PS_SVS)
#define RELEASE_SV(x)  R_ReleaseFromMSet((x), PS_SVS)
#define CLEAR_SVS()    R_ReleaseMSet(PS_SVS, 500)

/* Memory leak

   yyparse(), as generated by bison, allocates extra space for the parser
   stack using malloc(). Unfortunately this means that there is a memory
   leak in case of an R error (long-jump). In principle, we could define
   yyoverflow() to relocate the parser stacks for bison and allocate say on
   the R heap, but yyoverflow() is undocumented and somewhat complicated
   (we would have to replicate some macros from the generated parser here).
   The same problem exists at least in the Rd and LaTeX parsers in tools.
*/

#include <rlocale.h>
#ifdef HAVE_LANGINFO_CODESET
# include <langinfo.h>
#endif

static int mbcs_get_next(int c, wchar_t *wc)
{
    int i, res, clen = 1; char s[9];
    mbstate_t mb_st;

    s[0] = (char) c;
    /* This assumes (probably OK) that all MBCS embed ASCII as single-byte
       lead bytes, including control chars */
    if((unsigned int) c < 0x80) {
	*wc = (wchar_t) c;
	return 1;
    }
    if(utf8locale) {
	clen = utf8clen((char) c);
	for(i = 1; i < clen; i++) {
	    c = xxgetc();
	    if(c == R_EOF) error(_("EOF whilst reading MBCS char at line %d"), ParseState.xxlineno);
	    s[i] = (char) c;
	}
	s[clen] ='\0'; /* x86 Solaris requires this */
	mbs_init(&mb_st);
	res = (int) mbrtowc(wc, s, clen, &mb_st);
	if(res == -1) error(_("invalid multibyte character in parser at line %d"), ParseState.xxlineno);
    } else {
	/* This is not necessarily correct for stateful MBCS */
	while(clen <= R_MB_CUR_MAX) {
	    mbs_init(&mb_st);
	    res = (int) mbrtowc(wc, s, clen, &mb_st);
	    if(res >= 0) break;
	    if(res == -1)
		error(_("invalid multibyte character in parser at line %d"), ParseState.xxlineno);
	    /* so res == -2 */
	    c = xxgetc();
	    if(c == R_EOF) error(_("EOF whilst reading MBCS char at line %d"), ParseState.xxlineno);
	    s[clen++] = (char) c;
	} /* we've tried enough, so must be complete or invalid by now */
    }
    for(i = clen - 1; i > 0; i--) xxungetc(s[i]);
    return clen;
}

/* Soon to be defunct entry points */

void		R_SetInput(int);
int		R_fgetc(FILE*);

/* Routines used to build the parse tree */

static SEXP	xxnullformal(void);
static SEXP	xxfirstformal0(SEXP);
static SEXP	xxfirstformal1(SEXP, SEXP);
static SEXP	xxaddformal0(SEXP, SEXP, YYLTYPE *);
static SEXP	xxaddformal1(SEXP, SEXP, SEXP, YYLTYPE *);
static SEXP	xxexprlist0();
static SEXP	xxexprlist1(SEXP, YYLTYPE *);
static SEXP	xxexprlist2(SEXP, SEXP, YYLTYPE *);
static SEXP	xxsub0(void);
static SEXP	xxsub1(SEXP, YYLTYPE *);
static SEXP	xxsymsub0(SEXP, YYLTYPE *);
static SEXP	xxsymsub1(SEXP, SEXP, YYLTYPE *);
static SEXP	xxnullsub0(YYLTYPE *);
static SEXP	xxnullsub1(SEXP, YYLTYPE *);
static SEXP	xxsublist1(SEXP);
static SEXP	xxsublist2(SEXP, SEXP);
static SEXP	xxcond(SEXP);
static SEXP	xxifcond(SEXP);
static SEXP	xxif(SEXP, SEXP, SEXP);
static SEXP	xxifelse(SEXP, SEXP, SEXP, SEXP);
static SEXP	xxforcond(SEXP, SEXP);
static SEXP	xxfor(SEXP, SEXP, SEXP);
static SEXP	xxwhile(SEXP, SEXP, SEXP);
static SEXP	xxrepeat(SEXP, SEXP);
static SEXP	xxnxtbrk(SEXP);
static SEXP	xxfuncall(SEXP, SEXP);
static SEXP	xxdefun(SEXP, SEXP, SEXP, YYLTYPE *);
static SEXP	xxpipe(SEXP, SEXP);
static SEXP	xxpipebind(SEXP, SEXP, SEXP);
static SEXP	xxunary(SEXP, SEXP);
static SEXP	xxbinary(SEXP, SEXP, SEXP);
static SEXP	xxparen(SEXP, SEXP);
static SEXP	xxsubscript(SEXP, SEXP, SEXP);
static SEXP	xxexprlist(SEXP, YYLTYPE *, SEXP);
static int	xxvalue(SEXP, int, YYLTYPE *);

#define YYSTYPE		SEXP

%}

%token-table

%token		END_OF_INPUT ERROR
%token		STR_CONST NUM_CONST NULL_CONST SYMBOL FUNCTION 
%token		INCOMPLETE_STRING
%token		LEFT_ASSIGN EQ_ASSIGN RIGHT_ASSIGN LBB
%token		FOR IN IF ELSE WHILE NEXT BREAK REPEAT
%token		GT GE LT LE EQ NE AND OR AND2 OR2
%token		NS_GET NS_GET_INT
%token		COMMENT LINE_DIRECTIVE
%token		SYMBOL_FORMALS
%token		EQ_FORMALS
%token		EQ_SUB SYMBOL_SUB
%token		SYMBOL_FUNCTION_CALL
%token		SYMBOL_PACKAGE
/* no longer used: %token COLON_ASSIGN */
%token		SLOT
%token		PIPE
%token          PLACEHOLDER
%token          PIPEBIND

/* This is the precedence table, low to high */
%left		'?'
%left		LOW WHILE FOR REPEAT
%right		IF
%left		ELSE
%right		LEFT_ASSIGN
%right		EQ_ASSIGN
%left		RIGHT_ASSIGN
%left		'~' TILDE
%left		OR OR2
%left		AND AND2
%left		UNOT NOT
%nonassoc   	GT GE LT LE EQ NE
%left		'+' '-'
%left		'*' '/'
%left		SPECIAL PIPE
%left		PIPEBIND
%left		':'
%left		UMINUS UPLUS
%right		'^'
%left		'$' '@'
%left		NS_GET NS_GET_INT
%nonassoc	'(' '[' LBB

%%

prog	:	END_OF_INPUT			{ Status = 0; YYACCEPT; }
	|	'\n'				{ Status = 2; yyresult = xxvalue(NULL,2,NULL); YYACCEPT; }
	|	expr_or_assign_or_help '\n'	{ Status = 3; yyresult = xxvalue($1,3,&@1); YYACCEPT; }
	|	expr_or_assign_or_help ';'	{ Status = 4; yyresult = xxvalue($1,4,&@1); YYACCEPT; }
	|	error	 			{ Status = 1; YYABORT; }
	;

expr_or_assign_or_help  :    expr               { $$ = $1; }
                |    expr_or_assign_or_help EQ_ASSIGN expr_or_assign_or_help    { $$ = xxbinary($2,$1,$3); setId(@$); }
                |    expr_or_assign_or_help '?'  expr_or_assign_or_help		{ $$ = xxbinary($2,$1,$3); setId(@$); }
                ;

expr_or_help  :    expr				    { $$ = $1; }
	      |    expr_or_help '?' expr_or_help    { $$ = xxbinary($2,$1,$3); setId(@$); }
              ;

expr	: 	NUM_CONST			{ $$ = $1;	setId(@$); }
	|	STR_CONST			{ $$ = $1;	setId(@$); }
	|	NULL_CONST			{ $$ = $1;	setId(@$); } 
	|	PLACEHOLDER			{ $$ = $1;	setId(@$); }
	|	SYMBOL				{ $$ = $1;	setId(@$); }

	|	'{' exprlist '}'		{ $$ = xxexprlist($1,&@1,$2); setId(@$); }
	|	'(' expr_or_assign_or_help ')'	{ $$ = xxparen($1,$2);	setId(@$); }

	|	'-' expr %prec UMINUS		{ $$ = xxunary($1,$2);	setId(@$); }
	|	'+' expr %prec UMINUS		{ $$ = xxunary($1,$2);	setId(@$); }
	|	'!' expr %prec UNOT		{ $$ = xxunary($1,$2);	setId(@$); }
	|	'~' expr %prec TILDE		{ $$ = xxunary($1,$2);	setId(@$); }
	|	'?' expr_or_assign_or_help	{ $$ = xxunary($1,$2);	setId(@$); }

	|	expr ':'  expr			{ $$ = xxbinary($2,$1,$3);	setId(@$); }
	|	expr '+'  expr			{ $$ = xxbinary($2,$1,$3);	setId(@$); }
	|	expr '-' expr			{ $$ = xxbinary($2,$1,$3);	setId(@$); }
	|	expr '*' expr			{ $$ = xxbinary($2,$1,$3);	setId(@$); }
	|	expr '/' expr			{ $$ = xxbinary($2,$1,$3);	setId(@$); }
	|	expr '^' expr 			{ $$ = xxbinary($2,$1,$3);	setId(@$); }
	|	expr SPECIAL expr		{ $$ = xxbinary($2,$1,$3);	setId(@$); }
	|	expr '~' expr			{ $$ = xxbinary($2,$1,$3);	setId(@$); }
	|	expr LT expr			{ $$ = xxbinary($2,$1,$3);	setId(@$); }
	|	expr LE expr			{ $$ = xxbinary($2,$1,$3);	setId(@$); }
	|	expr EQ expr			{ $$ = xxbinary($2,$1,$3);	setId(@$); }
	|	expr NE expr			{ $$ = xxbinary($2,$1,$3);	setId(@$); }
	|	expr GE expr			{ $$ = xxbinary($2,$1,$3);	setId(@$); }
	|	expr GT expr			{ $$ = xxbinary($2,$1,$3);	setId(@$); }
	|	expr AND expr			{ $$ = xxbinary($2,$1,$3);	setId(@$); }
	|	expr OR expr			{ $$ = xxbinary($2,$1,$3);	setId(@$); }
	|	expr AND2 expr			{ $$ = xxbinary($2,$1,$3);	setId(@$); }
	|	expr OR2 expr			{ $$ = xxbinary($2,$1,$3);	setId(@$); }
	|	expr PIPE expr			{ $$ = xxpipe($1,$3);  setId(@$); }
	|	expr PIPEBIND expr		{ $$ = xxpipebind($2,$1,$3);	setId(@$); }
	|	expr LEFT_ASSIGN expr 		{ $$ = xxbinary($2,$1,$3);	setId(@$); }
	|	expr RIGHT_ASSIGN expr 		{ $$ = xxbinary($2,$3,$1);	setId(@$); }
	|	FUNCTION '(' formlist ')' cr expr_or_assign_or_help %prec LOW
						{ $$ = xxdefun($1,$3,$6,&@$); 	setId(@$); }
	|	'\\' '(' formlist ')' cr expr_or_assign_or_help %prec LOW							{ $$ = xxdefun(R_FunctionSymbol,$3,$6,&@$); 	setId(@$); }
	|	expr '(' sublist ')'		{ $$ = xxfuncall($1,$3);  setId(@$); modif_token( &@1, SYMBOL_FUNCTION_CALL ) ; }
	|	IF ifcond expr_or_assign_or_help 	{ $$ = xxif($1,$2,$3);	setId(@$); }
	|	IF ifcond expr_or_assign_or_help ELSE expr_or_assign_or_help	{ $$ = xxifelse($1,$2,$3,$5);	setId(@$); }
	|	FOR forcond expr_or_assign_or_help %prec FOR	{ $$ = xxfor($1,$2,$3);	setId(@$); }
	|	WHILE cond expr_or_assign_or_help   { $$ = xxwhile($1,$2,$3);	setId(@$); }
	|	REPEAT expr_or_assign_or_help	    { $$ = xxrepeat($1,$2);	setId(@$); }
	|	expr LBB sublist ']' ']'	{ $$ = xxsubscript($1,$2,$3);	setId(@$); }
	|	expr '[' sublist ']'		{ $$ = xxsubscript($1,$2,$3);	setId(@$); }
	|	SYMBOL NS_GET SYMBOL		{ $$ = xxbinary($2,$1,$3);      setId(@$); modif_token( &@1, SYMBOL_PACKAGE ) ; }
	|	SYMBOL NS_GET STR_CONST		{ $$ = xxbinary($2,$1,$3);      setId(@$); modif_token( &@1, SYMBOL_PACKAGE ) ; }
	|	STR_CONST NS_GET SYMBOL		{ $$ = xxbinary($2,$1,$3);	setId(@$); }
	|	STR_CONST NS_GET STR_CONST	{ $$ = xxbinary($2,$1,$3);	setId(@$); }
	|	SYMBOL NS_GET_INT SYMBOL	{ $$ = xxbinary($2,$1,$3);      setId(@$); modif_token( &@1, SYMBOL_PACKAGE ) ;}
	|	SYMBOL NS_GET_INT STR_CONST	{ $$ = xxbinary($2,$1,$3);      setId(@$); modif_token( &@1, SYMBOL_PACKAGE ) ;}
	|	STR_CONST NS_GET_INT SYMBOL	{ $$ = xxbinary($2,$1,$3);	setId(@$); }
	|	STR_CONST NS_GET_INT STR_CONST	{ $$ = xxbinary($2,$1,$3);	setId(@$); }
	|	expr '$' SYMBOL			{ $$ = xxbinary($2,$1,$3);	setId(@$); }
	|	expr '$' STR_CONST		{ $$ = xxbinary($2,$1,$3);	setId(@$); }
	|	expr '@' SYMBOL			{ $$ = xxbinary($2,$1,$3);      setId(@$); modif_token( &@3, SLOT ) ; }
	|	expr '@' STR_CONST		{ $$ = xxbinary($2,$1,$3);	setId(@$); }
	|	NEXT				{ $$ = xxnxtbrk($1);	setId(@$); }
	|	BREAK				{ $$ = xxnxtbrk($1);	setId(@$); }
	;


cond	:	'(' expr_or_help ')'			{ $$ = xxcond($2);   }
	;

ifcond	:	'(' expr_or_help ')'			{ $$ = xxifcond($2); }
	;

forcond :	'(' SYMBOL IN expr_or_help ')' 		{ $$ = xxforcond($2,$4);	setId(@$); }
	;


exprlist:					{ $$ = xxexprlist0();	setId(@$); }
	|	expr_or_assign_or_help			{ $$ = xxexprlist1($1, &@1); }
	|	exprlist ';' expr_or_assign_or_help	{ $$ = xxexprlist2($1, $3, &@3); }
	|	exprlist ';'			{ $$ = $1;		setId(@$); }
	|	exprlist '\n' expr_or_assign_or_help	{ $$ = xxexprlist2($1, $3, &@3); }
	|	exprlist '\n'			{ $$ = $1;}
	;

sublist	:	sub				{ $$ = xxsublist1($1);	  }
	|	sublist cr ',' sub		{ $$ = xxsublist2($1,$4); }
	;

sub	:					{ $$ = xxsub0();	 }
	|	expr_or_help			{ $$ = xxsub1($1, &@1);  }
	|	SYMBOL EQ_ASSIGN 		{ $$ = xxsymsub0($1, &@1); 	modif_token( &@2, EQ_SUB ) ; modif_token( &@1, SYMBOL_SUB ) ; }
	|	SYMBOL EQ_ASSIGN expr_or_help	{ $$ = xxsymsub1($1,$3, &@1); 	modif_token( &@2, EQ_SUB ) ; modif_token( &@1, SYMBOL_SUB ) ; }
	|	STR_CONST EQ_ASSIGN 		{ $$ = xxsymsub0($1, &@1); 	modif_token( &@2, EQ_SUB ) ; }
	|	STR_CONST EQ_ASSIGN expr_or_help    { $$ = xxsymsub1($1,$3, &@1); 	modif_token( &@2, EQ_SUB ) ; }
	|	NULL_CONST EQ_ASSIGN 		{ $$ = xxnullsub0(&@1); 	modif_token( &@2, EQ_SUB ) ; }
	|	NULL_CONST EQ_ASSIGN expr_or_help   { $$ = xxnullsub1($3, &@1); 	modif_token( &@2, EQ_SUB ) ; }
	;

formlist:					{ $$ = xxnullformal(); }
	|	SYMBOL				{ $$ = xxfirstformal0($1); 	modif_token( &@1, SYMBOL_FORMALS ) ; }
	|	SYMBOL EQ_ASSIGN expr_or_help	{ $$ = xxfirstformal1($1,$3); 	modif_token( &@1, SYMBOL_FORMALS ) ; modif_token( &@2, EQ_FORMALS ) ; }
	|	formlist ',' SYMBOL		{ $$ = xxaddformal0($1,$3, &@3);   modif_token( &@3, SYMBOL_FORMALS ) ; }
	|	formlist ',' SYMBOL EQ_ASSIGN expr_or_help
						{ $$ = xxaddformal1($1,$3,$5,&@3); modif_token( &@3, SYMBOL_FORMALS ) ; modif_token( &@4, EQ_FORMALS ) ;}
	;

cr	:					{ EatLines = 1; }
	;
%%


/*----------------------------------------------------------------------------*/

static int (*ptr_getc)(void);

/* Private pushback, since file ungetc only guarantees one byte.
   We need up to one MBCS-worth */
#define DECLARE_YYTEXT_BUFP(bp) char *bp = yytext ;
#define YYTEXT_PUSH(c, bp) do { \
    if ((bp) - yytext >= sizeof(yytext) - 1){ \
		error(_("input buffer overflow at line %d"), ParseState.xxlineno); \
	} \
    *(bp)++ = ((char)c);			\
} while(0) ;

#define PUSHBACK_BUFSIZE 16
static int pushback[PUSHBACK_BUFSIZE];
static unsigned int npush = 0;

static int prevpos = 0;
static int prevlines[PUSHBACK_BUFSIZE];
static int prevcols[PUSHBACK_BUFSIZE];
static int prevbytes[PUSHBACK_BUFSIZE];
static int prevparse[PUSHBACK_BUFSIZE];

static int xxgetc(void)
{
    int c;

    if(npush) c = pushback[--npush]; else  c = ptr_getc();

    prevpos = (prevpos + 1) % PUSHBACK_BUFSIZE;
    prevbytes[prevpos] = ParseState.xxbyteno;
    prevlines[prevpos] = ParseState.xxlineno;  
    prevparse[prevpos] = ParseState.xxparseno;
    prevcols[prevpos] = ParseState.xxcolno;
    	
    if (c == EOF) {
	EndOfFile = 1;
	return R_EOF;
    }
    R_ParseContextLast = (R_ParseContextLast + 1) % PARSE_CONTEXT_SIZE;
    R_ParseContext[R_ParseContextLast] = (char) c;

    if (c == '\n') {
	ParseState.xxlineno += 1;
	ParseState.xxcolno = 0;
    	ParseState.xxbyteno = 0;
    	ParseState.xxparseno += 1;
    } else {
        /* We only advance the column for the 1st byte in UTF-8, so handle later bytes specially */
	if (!known_to_be_utf8 || (unsigned char)c < 0x80 || 0xC0 <= (unsigned char)c)
            ParseState.xxcolno++;
    	ParseState.xxbyteno++;
    }

    if (c == '\t') ParseState.xxcolno = ((ParseState.xxcolno + 7) & ~7);
    
    R_ParseContextLine = ParseState.xxlineno;    

    xxcharcount++;
    return c;
}

static int xxungetc(int c)
{
    /* this assumes that c was the result of xxgetc; if not, some edits will be needed */
    ParseState.xxlineno = prevlines[prevpos];
    ParseState.xxbyteno = prevbytes[prevpos];
    ParseState.xxcolno  = prevcols[prevpos];
    ParseState.xxparseno = prevparse[prevpos];
    
    prevpos = (prevpos + PUSHBACK_BUFSIZE - 1) % PUSHBACK_BUFSIZE;

    R_ParseContextLine = ParseState.xxlineno;

    xxcharcount--;
    R_ParseContext[R_ParseContextLast] = '\0';
    /* precaution as to how % is implemented for < 0 numbers */
    R_ParseContextLast = (R_ParseContextLast + PARSE_CONTEXT_SIZE -1) % PARSE_CONTEXT_SIZE;
    if(npush >= PUSHBACK_BUFSIZE) return EOF;
    pushback[npush++] = c;
    return c;
}

/* Only used from finish_mbcs_in_parse_context. */
static int add_mbcs_byte_to_parse_context()
{
    int c;

    if (EndOfFile)
	error(_("invalid multibyte character in parser at line %d"),
	      ParseState.xxlineno);
    if(npush)
	c = pushback[--npush];
    else
	c = ptr_getc();
    if (c == EOF) 
	error(_("invalid multibyte character in parser at line %d"),
	      ParseState.xxlineno);
    
    R_ParseContextLast = (R_ParseContextLast + 1) % PARSE_CONTEXT_SIZE;
    R_ParseContext[R_ParseContextLast] = (char) c;
    return c;
}

/* On error, the parse context may end inside a multi-byte character. Add
   the missing bytes to the context to so that it contains full characters. */
static void finish_mbcs_in_parse_context()
{
    int i, c, nbytes = 0, first;
    Rboolean mbcs = FALSE;

    /* find the first byte of the context */
    for(i = R_ParseContextLast;
        R_ParseContext[i];
        i = (i + PARSE_CONTEXT_SIZE - 1) % PARSE_CONTEXT_SIZE) {

	nbytes++;
	if (nbytes == PARSE_CONTEXT_SIZE)
	    break;
    }
    if (!nbytes)
	return;
    if (!R_ParseContext[i])
	first = (i + 1) % PARSE_CONTEXT_SIZE;
    else
	/* The beginning of the context has been overwritten and for a general
	   encoding there is not way to recover it. It is possible for UTF-8,
	   though. */
	return;

    /* decode multi-byte characters */
    for(i = 0; i < nbytes; i++) {
	c = R_ParseContext[(first + i) % PARSE_CONTEXT_SIZE];
	if ((unsigned int)c < 0x80) continue; /* ASCII */

	if (utf8locale) {
	    /* UTF-8 could be handled more efficiently, searching from the end
	       of the string */
	    i += utf8clen((char) c) - 1;
	    if (i >= nbytes) {
		while (i >= nbytes) {
		    add_mbcs_byte_to_parse_context();
		    nbytes++;
		}
		return;
	    }
	} else
	    mbcs = TRUE;
    }
    if (!mbcs)
	return;

    /* copy the context to a linear buffer */
    char buf[nbytes + R_MB_CUR_MAX];

    for(i = 0; i < nbytes; i++)
	buf[i] = R_ParseContext[(first + i) % PARSE_CONTEXT_SIZE];

    for(i = 0; i < nbytes; i++) {
	wchar_t wc;
	int res;
	mbstate_t mb_st;
	
	mbs_init(&mb_st);
	res = (int) mbrtowc(&wc, buf + i, nbytes - i, &mb_st);
	while (res == -2 && nbytes < sizeof(buf)) {
	    /* This is not necessarily correct for stateful MBCS */
	    buf[nbytes++] = (char) add_mbcs_byte_to_parse_context();
	    mbs_init(&mb_st);
	    res = (int) mbrtowc(&wc, buf + i, nbytes - i, &mb_st);
	}	   
	if (res == -1)
	    error(_("invalid multibyte character in parser at line %d"),
		  ParseState.xxlineno);
	i += res - 1;
    }
}

/*
 * Increments/inits the token/grouping counter
 */
static void incrementId(void){
	identifier++; 
}

static void initId(void){
	identifier = 0 ;
}

static SEXP makeSrcref(YYLTYPE *lloc, SEXP srcfile)
{
    SEXP val;

    PROTECT(val = allocVector(INTSXP, 8));
    INTEGER(val)[0] = lloc->first_line;
    INTEGER(val)[1] = lloc->first_byte;
    INTEGER(val)[2] = lloc->last_line;
    INTEGER(val)[3] = lloc->last_byte;
    INTEGER(val)[4] = lloc->first_column;
    INTEGER(val)[5] = lloc->last_column;
    INTEGER(val)[6] = lloc->first_parsed;
    INTEGER(val)[7] = lloc->last_parsed;
    setAttrib(val, R_SrcfileSymbol, srcfile);
    setAttrib(val, R_ClassSymbol, mkString("srcref"));
    UNPROTECT(1); /* val */
    return val;
}

static void attachSrcrefs(SEXP val)
{
    SEXP srval;

    PROTECT(srval = SrcRefsToVectorList());
    
    setAttrib(val, R_SrcrefSymbol, srval);
    setAttrib(val, R_SrcfileSymbol, PS_SRCFILE);
    {
	YYLTYPE wholeFile;
	wholeFile.first_line = 1;
	wholeFile.first_byte = 0;
	wholeFile.first_column = 0;
	wholeFile.last_line = ParseState.xxlineno;
	wholeFile.last_byte = ParseState.xxbyteno;
	wholeFile.last_column = ParseState.xxcolno;
	wholeFile.first_parsed = 1;
	wholeFile.last_parsed = ParseState.xxparseno;
	setAttrib(val, R_WholeSrcrefSymbol, makeSrcref(&wholeFile, PS_SRCFILE));
    }
    PS_SET_SRCREFS(R_NilValue);
    ParseState.didAttach = TRUE;
    UNPROTECT(1); /* srval */
}

static int xxvalue(SEXP v, int k, YYLTYPE *lloc)
{
    if (k > 2) {
	if (ParseState.keepSrcRefs) {
	    SEXP s = PROTECT(makeSrcref(lloc, PS_SRCFILE));
	    AppendToSrcRefs(s);
	    UNPROTECT(1); /* s */
	}
	RELEASE_SV(v);
    }
    R_CurrentExpr = v;
    return k;
}

static SEXP xxnullformal()
{
    SEXP ans;
    PRESERVE_SV(ans = R_NilValue);
    return ans;
}

static SEXP xxfirstformal0(SEXP sym)
{
    SEXP ans;
    if (GenerateCode)
	PRESERVE_SV(ans = FirstArg(R_MissingArg, sym));
    else
	PRESERVE_SV(ans = R_NilValue);
    RELEASE_SV(sym);
    return ans;
}

static SEXP xxfirstformal1(SEXP sym, SEXP expr)
{
    SEXP ans;
    if (GenerateCode)
	PRESERVE_SV(ans = FirstArg(expr, sym));
    else
	PRESERVE_SV(ans = R_NilValue);
    RELEASE_SV(expr);
    RELEASE_SV(sym);
    return ans;
}

static SEXP xxaddformal0(SEXP formlist, SEXP sym, YYLTYPE *lloc)
{
    SEXP ans;
    if (GenerateCode) {
	CheckFormalArgs(formlist, sym, lloc);
	NextArg(formlist, R_MissingArg, sym);
	ans = formlist;
    } else {
	RELEASE_SV(formlist);
	PRESERVE_SV(ans = R_NilValue);
    }
    RELEASE_SV(sym);
    return ans;
}

static SEXP xxaddformal1(SEXP formlist, SEXP sym, SEXP expr, YYLTYPE *lloc)
{
    SEXP ans;
    if (GenerateCode) {
	CheckFormalArgs(formlist, sym, lloc);
	NextArg(formlist, expr, sym);
	ans = formlist;
    } else {
	RELEASE_SV(formlist);
	PRESERVE_SV(ans = R_NilValue);
    }
    RELEASE_SV(expr);
    RELEASE_SV(sym);
    return ans;
}

static SEXP xxexprlist0(void)
{
    SEXP ans;
    if (GenerateCode) {
	PRESERVE_SV(ans = NewList());
	if (ParseState.keepSrcRefs) {
	    setAttrib(ans, R_SrcrefSymbol, PS_SRCREFS);
	    PS_SET_SRCREFS(R_NilValue);
	}
    }
    else
	PRESERVE_SV(ans = R_NilValue);
    return ans;
}

static SEXP xxexprlist1(SEXP expr, YYLTYPE *lloc)
{
    SEXP ans;
    if (GenerateCode) {
	PRESERVE_SV(ans = NewList());
	if (ParseState.keepSrcRefs) {
	    setAttrib(ans, R_SrcrefSymbol, PS_SRCREFS);
	    SEXP s = PROTECT(makeSrcref(lloc, PS_SRCFILE));
	    SetSingleSrcRef(s);
	    UNPROTECT(1); /* s */
	}
	GrowList(ans, expr);
    }
    else
	PRESERVE_SV(ans = R_NilValue);
    RELEASE_SV(expr);
    return ans;
}

static SEXP xxexprlist2(SEXP exprlist, SEXP expr, YYLTYPE *lloc)
{
    SEXP ans;
    if (GenerateCode) {
	if (ParseState.keepSrcRefs) {
	    SEXP s = PROTECT(makeSrcref(lloc, PS_SRCFILE));
	    AppendToSrcRefs(s);
	    UNPROTECT(1); /* s */
	}
	GrowList(exprlist, expr);
	ans = exprlist;
    } else {
	RELEASE_SV(exprlist);
	PRESERVE_SV(ans = R_NilValue);
    }
    RELEASE_SV(expr);
    return ans;
}

static SEXP xxsub0(void)
{
    SEXP ans;
    if (GenerateCode)
	PRESERVE_SV(ans = lang2(R_MissingArg,R_NilValue));
    else
	PRESERVE_SV(ans = R_NilValue);
    return ans;
}

static SEXP xxsub1(SEXP expr, YYLTYPE *lloc)
{
    SEXP ans;
    if (GenerateCode)
	PRESERVE_SV(ans = TagArg(expr, R_NilValue, lloc));
    else
	PRESERVE_SV(ans = R_NilValue);
    RELEASE_SV(expr);
    return ans;
}

static SEXP xxsymsub0(SEXP sym, YYLTYPE *lloc)
{
    SEXP ans;
    if (GenerateCode)
	PRESERVE_SV(ans = TagArg(R_MissingArg, sym, lloc));
    else
	PRESERVE_SV(ans = R_NilValue);
    RELEASE_SV(sym);
    return ans;
}

static SEXP xxsymsub1(SEXP sym, SEXP expr, YYLTYPE *lloc)
{
    SEXP ans;
    if (GenerateCode)
	PRESERVE_SV(ans = TagArg(expr, sym, lloc));
    else
	PRESERVE_SV(ans = R_NilValue);
    RELEASE_SV(expr);
    RELEASE_SV(sym);
    return ans;
}

static SEXP xxnullsub0(YYLTYPE *lloc)
{
    SEXP ans;
    if (GenerateCode)
	PRESERVE_SV(ans = TagArg(R_MissingArg, R_NullSymbol, lloc));
    else
	PRESERVE_SV(ans = R_NilValue);
    RELEASE_SV(R_NilValue);
    return ans;
}

static SEXP xxnullsub1(SEXP expr, YYLTYPE *lloc)
{
    SEXP ans;
    if (GenerateCode)
	PRESERVE_SV(ans = TagArg(expr, R_NullSymbol, lloc));
    else
	PRESERVE_SV(ans = R_NilValue);
    RELEASE_SV(R_NilValue);
    RELEASE_SV(expr);
    return ans;
}


static SEXP xxsublist1(SEXP sub)
{
    SEXP ans;
    if (GenerateCode)
	PRESERVE_SV(ans = FirstArg(CAR(sub),CADR(sub)));
    else
	PRESERVE_SV(ans = R_NilValue);
    RELEASE_SV(sub);
    return ans;
}

static SEXP xxsublist2(SEXP sublist, SEXP sub)
{
    SEXP ans;
    if (GenerateCode) {
	NextArg(sublist, CAR(sub), CADR(sub));
	ans = sublist;
    } else {
	RELEASE_SV(sublist);
	PRESERVE_SV(ans = R_NilValue);
    }
    RELEASE_SV(sub);
    return ans;
}

static SEXP xxcond(SEXP expr)
{
    EatLines = 1;
    return expr;
}

static SEXP xxifcond(SEXP expr)
{
    EatLines = 1;
    return expr;
}

static SEXP xxif(SEXP ifsym, SEXP cond, SEXP expr)
{
    SEXP ans;
    if (GenerateCode)
	PRESERVE_SV(ans = lang3(ifsym, cond, expr));
    else
	PRESERVE_SV(ans = R_NilValue);
    RELEASE_SV(expr);
    RELEASE_SV(cond);
    return ans;
}

static SEXP xxifelse(SEXP ifsym, SEXP cond, SEXP ifexpr, SEXP elseexpr)
{
    SEXP ans;
    if (GenerateCode)
	PRESERVE_SV(ans = lang4(ifsym, cond, ifexpr, elseexpr));
    else
	PRESERVE_SV(ans = R_NilValue);
    RELEASE_SV(elseexpr);
    RELEASE_SV(ifexpr);
    RELEASE_SV(cond);
    return ans;
}

static SEXP xxforcond(SEXP sym, SEXP expr)
{
    SEXP ans;
    EatLines = 1;
    if (GenerateCode)
	PRESERVE_SV(ans = LCONS(sym, expr));
    else
	PRESERVE_SV(ans = R_NilValue);
    RELEASE_SV(expr);
    RELEASE_SV(sym);
    return ans;
}

static SEXP xxfor(SEXP forsym, SEXP forcond, SEXP body)
{
    SEXP ans;
    if (GenerateCode)
	PRESERVE_SV(ans = lang4(forsym, CAR(forcond), CDR(forcond), body));
    else
	PRESERVE_SV(ans = R_NilValue);
    RELEASE_SV(body);
    RELEASE_SV(forcond);
    return ans;
}

static SEXP xxwhile(SEXP whilesym, SEXP cond, SEXP body)
{
    SEXP ans;
    if (GenerateCode)
	PRESERVE_SV(ans = lang3(whilesym, cond, body));
    else
	PRESERVE_SV(ans = R_NilValue);
    RELEASE_SV(body);
    RELEASE_SV(cond);
    return ans;
}

static SEXP xxrepeat(SEXP repeatsym, SEXP body)
{
    SEXP ans;
    if (GenerateCode)
	PRESERVE_SV(ans = lang2(repeatsym, body));
    else
	PRESERVE_SV(ans = R_NilValue);
    RELEASE_SV(body);
    return ans;
}

static SEXP xxnxtbrk(SEXP keyword)
{
    if (GenerateCode)
	PRESERVE_SV(keyword = lang1(keyword));
    else
	PRESERVE_SV(keyword = R_NilValue);
    return keyword;
}

static SEXP xxfuncall(SEXP expr, SEXP args)
{
    SEXP ans, sav_expr = expr;
    if (GenerateCode) {
	if (isString(expr) && expr != R_PlaceholderToken)
	    expr = installTrChar(STRING_ELT(expr, 0));
	PROTECT(expr);
	if (length(CDR(args)) == 1 && CADR(args) == R_MissingArg && TAG(CDR(args)) == R_NilValue )
	    ans = lang1(expr);
	else
	    ans = LCONS(expr, CDR(args));
	UNPROTECT(1); /* expr */
	PRESERVE_SV(ans);
    } else
	PRESERVE_SV(ans = R_NilValue);

    RELEASE_SV(args);
    RELEASE_SV(sav_expr);
    return ans;
}

static SEXP mkChar2(const char *name)
{
    cetype_t enc = CE_NATIVE;

    if(known_to_be_latin1) enc = CE_LATIN1;
    else if(known_to_be_utf8) enc = CE_UTF8;

    return mkCharLenCE(name, (int) strlen(name), enc);
}

static SEXP mkString2(const char *s, size_t len, Rboolean escaped)
{
    SEXP t;
    cetype_t enc = CE_NATIVE;

    if(known_to_be_latin1) enc = CE_LATIN1;
    else if(!escaped && known_to_be_utf8) enc = CE_UTF8;

    PROTECT(t = allocVector(STRSXP, 1));
    SET_STRING_ELT(t, 0, mkCharLenCE(s, (int) len, enc));
    UNPROTECT(1); /* t */
    return t;
}

static SEXP xxdefun(SEXP fname, SEXP formals, SEXP body, YYLTYPE *lloc)
{
    SEXP ans, srcref;

    if (GenerateCode) {
    	if (ParseState.keepSrcRefs) {
	    srcref = makeSrcref(lloc, PS_SRCFILE);
    	    ParseState.didAttach = TRUE;
    	} else
    	    srcref = R_NilValue;
	PRESERVE_SV(ans = lang4(fname, CDR(formals), body, srcref));
    } else
	PRESERVE_SV(ans = R_NilValue);
    RELEASE_SV(body);
    RELEASE_SV(formals);
    return ans;
}

static SEXP xxunary(SEXP op, SEXP arg)
{
    SEXP ans;
    if (GenerateCode)
	PRESERVE_SV(ans = lang2(op, arg));
    else
	PRESERVE_SV(ans = R_NilValue);
    RELEASE_SV(arg);
    return ans;
}

static SEXP xxbinary(SEXP n1, SEXP n2, SEXP n3)
{
    SEXP ans;
    if (GenerateCode)
	PRESERVE_SV(ans = lang3(n1, n2, n3));
    else
	PRESERVE_SV(ans = R_NilValue);
    RELEASE_SV(n2);
    RELEASE_SV(n3);
    return ans;
}

static void check_rhs(SEXP rhs)
{
    /* rule out syntactically special functions */
    /* the IS_SPECIAL_SYMBOL bit is set in names.c */
    SEXP fun = CAR(rhs);
    if (TYPEOF(fun) == SYMSXP && IS_SPECIAL_SYMBOL(fun))
	error("function '%s' not supported in RHS call of a pipe",
	      CHAR(PRINTNAME(fun)));
}

static void checkTooManyPlaceholders(SEXP rhs, SEXP args)
{
    for (SEXP rest = args; rest != R_NilValue; rest = CDR(rest))
	if (CAR(rest) == R_PlaceholderToken)
	    errorcall(rhs, _("pipe placeholder may only appear once"));
}

static SEXP xxpipe(SEXP lhs, SEXP rhs)
{
    SEXP ans;
    if (GenerateCode) {
	if (TYPEOF(rhs) != LANGSXP)
	    error(_("The pipe operator requires a function call as RHS"));

	/* allow x => log(x) on RHS */
	if (CAR(rhs) == R_PipeBindSymbol) {
	    SEXP var = CADR(rhs);
	    SEXP expr = CADDR(rhs);
	    if (TYPEOF(var) != SYMSXP)
		error(_("RHS variable must be a symbol"));
	    SEXP alist = list1(R_MissingArg);
	    SET_TAG(alist, var);
	    SEXP fun = lang4(R_FunctionSymbol, alist, expr, R_NilValue);
	    return lang2(fun, lhs);
	}

	/* check for placehilder in the RHS function */
	if (checkForPlaceholder(R_PlaceholderToken, CAR(rhs)))
	    error(_("pipe placeholder cannot be used in the RHS function"));

	/* allow top-level placeholder */
	for (SEXP a = CDR(rhs); a != R_NilValue; a = CDR(a))
	    if (CAR(a) == R_PlaceholderToken) {
		if (TAG(a) == R_NilValue)
		    error(_("pipe placeholder can only be used as a "
			    "named argument"));
		checkTooManyPlaceholders(rhs, CDR(a));
		SETCAR(a, lhs);
		return rhs;
	    }
	
	check_rhs(rhs);
	
        SEXP fun = CAR(rhs);
        SEXP args = CDR(rhs);
	PRESERVE_SV(ans = lcons(fun, lcons(lhs, args)));
    }
    else {
	PRESERVE_SV(ans = R_NilValue);
    }
    RELEASE_SV(lhs);
    RELEASE_SV(rhs);
    return ans;
}

static SEXP xxpipebind(SEXP fn, SEXP lhs, SEXP rhs)
{
    static int use_pipebind = 0;
    if (use_pipebind != 1) {
	char *lookup = getenv("_R_USE_PIPEBIND_");
	use_pipebind = ((lookup != NULL) && StringTrue(lookup)) ? 1 : 0;
    }

    if (use_pipebind)
	return xxbinary(fn, lhs, rhs);
    else
	error("'=>' is disabled; set '_R_USE_PIPEBIND_' envvar to a true value to enable it");
}

static SEXP xxparen(SEXP n1, SEXP n2)
{
    SEXP ans;
    if (GenerateCode)
	PRESERVE_SV(ans = lang2(n1, n2));
    else
	PRESERVE_SV(ans = R_NilValue);
    RELEASE_SV(n2);
    return ans;
}


/* This should probably use CONS rather than LCONS, but
   it shouldn't matter and we would rather not meddle
   See PR#7055 */

static SEXP xxsubscript(SEXP a1, SEXP a2, SEXP a3)
{
    SEXP ans;
    if (GenerateCode)
	PRESERVE_SV(ans = LCONS(a2, CONS(a1, CDR(a3))));
    else
	PRESERVE_SV(ans = R_NilValue);
    RELEASE_SV(a3);
    RELEASE_SV(a1);
    return ans;
}

static SEXP xxexprlist(SEXP a1, YYLTYPE *lloc, SEXP a2)
{
    SEXP ans;
    SEXP prevSrcrefs;

    EatLines = 0;
    if (GenerateCode) {
	SET_TYPEOF(a2, LANGSXP);
	SETCAR(a2, a1);
	if (ParseState.keepSrcRefs) {
	    PROTECT(prevSrcrefs = getAttrib(a2, R_SrcrefSymbol));
	    SEXP s = PROTECT(makeSrcref(lloc, PS_SRCFILE));
	    PrependToSrcRefs(s);
	    attachSrcrefs(a2);
	    UNPROTECT(2); /* prevSrcrefs, s */
#ifndef SWITCH_TO_REFCNT
	    /* SrcRefs got NAMED by being an attribute, preventively
	       getAttrib(), but it has not in fact been referenced. Set NAMED
	       to 0 to avoid overhead in further setAttrib calls due to cycle
	       detection. */
	    SET_NAMED(prevSrcrefs, 0);
#endif
	    PS_SET_SRCREFS(prevSrcrefs);
	}
	PRESERVE_SV(ans = a2);
    }
    else
	PRESERVE_SV(ans = R_NilValue);
    RELEASE_SV(a2);
    return ans;
}

/*--------------------------------------------------------------------------*/

static SEXP TagArg(SEXP arg, SEXP tag, YYLTYPE *lloc)
{
    switch (TYPEOF(tag)) {
    case STRSXP:
	tag = installTrChar(STRING_ELT(tag, 0));
    case NILSXP:
    case SYMSXP:
	return lang2(arg, tag);
    default:
	error(_("incorrect tag type at line %d"), lloc->first_line); return R_NilValue/* -Wall */;
    }
}


/* Stretchy List Structures : Lists are created and grown using a special */
/* dotted pair.  The CAR of the list points to the last cons-cell in the */
/* list and the CDR points to the first.  The list can be extracted from */
/* the pair by taking its CDR, while the CAR gives fast access to the end */
/* of the list. */

/* These functions must be called with arguments protected */

/* Create a stretchy-list dotted pair */
static SEXP NewList(void)
{
    SEXP s = CONS(R_NilValue, R_NilValue);
    SETCAR(s, s);
    return s;
}

/* Add a new element at the end of a stretchy list */
static void GrowList(SEXP l, SEXP s)
{
    SEXP tmp;
    tmp = CONS(s, R_NilValue);
    SETCDR(CAR(l), tmp);
    SETCAR(l, tmp);
}

/* Create a stretchy list with a single named element */
static SEXP FirstArg(SEXP s, SEXP tag)
{
    SEXP tmp;
    PROTECT(tmp = NewList());
    GrowList(tmp, s);
    SET_TAG(CAR(tmp), tag);
    UNPROTECT(1); /* tmp */
    return tmp;
}

/* Add named element to the end of a stretchy list */
static void NextArg(SEXP l, SEXP s, SEXP tag)
{
    GrowList(l, s);
    SET_TAG(CAR(l), tag);
}

/* SrcRefs (PS_SRCREFS) are represented as R_NilValue (empty) or by 
   a stretchy list (which includes another representation for empty)
   for fast append operation. */

static void SetSingleSrcRef(SEXP r)
{
    SEXP l;

    PROTECT(l = NewList());
    GrowList(l, r);
    PS_SET_SRCREFS(l);
    UNPROTECT(1); /* l */
}

static void AppendToSrcRefs(SEXP r)
{
    SEXP l = PS_SRCREFS;
    if (l == R_NilValue)
	SetSingleSrcRef(r);
    else
	GrowList(l, r);
}

static void PrependToSrcRefs(SEXP r)
{
    SEXP l = PS_SRCREFS;
    if (l == R_NilValue)
	SetSingleSrcRef(r);
    else if (CDR(l) == R_NilValue)
	/* adding to empty stretchy list */
	GrowList(l, r);
    else {
	SEXP tmp = CONS(r, CDR(l));
	SETCDR(l, tmp);
    }
}

static SEXP SrcRefsToVectorList() {
    SEXP l = PS_SRCREFS;
    if (l == R_NilValue)
	return PairToVectorList(l);
    else
	return PairToVectorList(CDR(l));
}

/*--------------------------------------------------------------------------*/

/*
 *  Parsing Entry Points:
 *
 *  The Following entry points provide language parsing facilities.
 *  Note that there are separate entry points for parsing IoBuffers
 *  (i.e. interactve use), files and R character strings.
 *
 *  The entry points provide the same functionality, they just
 *  set things up in slightly different ways.
 *
 *  The following routines parse a single expression:
 *
 *
 *	SEXP R_Parse1File(FILE *fp, int gencode, ParseStatus *status, Rboolean first)
 *   (used for R_ReplFile in main.c)
 *
 *	SEXP R_Parse1Buffer(IoBuffer *buffer, int gencode, ParseStatus *status, Rboolean first)
 *   (used for ReplIteration and R_ReplDLLdo1 in main.c)
 *
 *  The success of the parse is indicated as folllows:
 *
 *
 *	status = PARSE_NULL       - there was no statement to parse
 *		 PARSE_OK	  - complete statement
 *		 PARSE_INCOMPLETE - incomplete statement
 *		 PARSE_ERROR      - syntax error
 *		 PARSE_EOF	  - end of file
 *
 *
 *  The following routines parse several expressions and return
 *  their values in a single expression vector.
 *
 *	SEXP R_ParseFile(FILE *fp, int n, ParseStatus *status, SEXP srcfile)
 *    (used for do_edit in file edit.c)
 *
 *	SEXP R_ParseVector(SEXP *text, int n, ParseStatus *status, SEXP srcfile)
 *    (public, and used by parse(text=) in file source.c)
 *
 *	SEXP R_ParseBuffer(IoBuffer *buffer, int n, ParseStatus *status, SEXP prompt, SEXP srcfile)
 *    (used by parse(file="") in file source.c)
 *
 *      SEXP R_ParseConn(Rconnection con, int n, ParseStatus *status, SEXP srcfile)
 *    (used by parse(file=) in file source.c)
 *
 *  Here, status is 1 for a successful parse and 0 if parsing failed
 *  for some reason.
 */

#define CONTEXTSTACK_SIZE 50
static int	SavedToken;
static SEXP	SavedLval;
static char	contextstack[CONTEXTSTACK_SIZE], *contextp;

static void PutSrcRefState(SrcRefState *state);
static void UseSrcRefState(SrcRefState *state);

/* This is called once when R starts up. */
attribute_hidden
void InitParser(void)
{
    ParseState.sexps = allocVector(VECSXP, 7); /* initialized to R_NilValue */
    ParseState.data = R_NilValue;
    INIT_SVS();
    R_PreserveObject(ParseState.sexps); /* never released in an R session */
    R_NullSymbol = install("NULL");
    R_PlaceholderToken = ScalarString(mkChar("_"));
    MARK_NOT_MUTABLE(R_PlaceholderToken);
    R_PreserveObject(R_PlaceholderToken);
    R_PipeBindSymbol = install("=>");
}

static void FinalizeSrcRefStateOnError(void *dummy)
{
    R_FinalizeSrcRefState();
}

/* This is called each time a new parse sequence begins */
attribute_hidden
void R_InitSrcRefState(RCNTXT* cptr)
{
    if (busy) {
    	SrcRefState *prev = malloc(sizeof(SrcRefState));
	if (prev == NULL)
	    error(_("allocation of source reference state failed"));
    	PutSrcRefState(prev);
	ParseState.prevState = prev;
	ParseState.sexps = allocVector(VECSXP, 7);
	ParseState.data = R_NilValue;
	INIT_SVS();
	R_PreserveObject(ParseState.sexps);
	/* ParseState.sexps released in R_FinalizeSrcRefState */
    } else
	/* re-use data, text, ids arrays */
        ParseState.prevState = NULL;
    /* set up context _after_ PutSrcRefState */
    begincontext(cptr, CTXT_CCODE, R_NilValue, R_BaseEnv, R_BaseEnv,
                 R_NilValue, R_NilValue);
    cptr->cend = &FinalizeSrcRefStateOnError;
    cptr->cenddata = NULL;
    ParseState.keepSrcRefs = FALSE;
    ParseState.keepParseData = TRUE;
    ParseState.didAttach = FALSE;
    PS_SET_SRCFILE(R_NilValue);
    PS_SET_ORIGINAL(R_NilValue);
    ParseState.data_count = 0;
    ParseState.xxlineno = 1;
    ParseState.xxcolno = 0;
    ParseState.xxbyteno = 0;
    ParseState.xxparseno = 1;
    busy = TRUE;
}

attribute_hidden
void R_FinalizeSrcRefState(void)
{
    PS_SET_SRCFILE(R_NilValue);
    PS_SET_ORIGINAL(R_NilValue);
    CLEAR_SVS();

    /* Free the data, text and ids if we are restoring a previous state,
       or if they have grown too large */
    if (PS_DATA != R_NilValue) {
    	if (ParseState.prevState || DATA_COUNT > MAX_DATA_COUNT) {
	    PS_SET_DATA(R_NilValue);
	    PS_SET_TEXT(R_NilValue);
	} else /* Remove all the strings from the text vector so they don't take up memory, and clean up data */
	    for (int i=0; i < ParseState.data_count; i++) {
		SET_STRING_ELT(PS_TEXT, i, R_BlankString);
		_PARENT(i) = 0;
	    }
    } 
    if (PS_IDS != R_NilValue) {
	if (ParseState.prevState || ID_COUNT > MAX_DATA_COUNT) {
	    PS_SET_IDS(R_NilValue);
        } else {/* Remove the parent records */
            if (identifier > ID_COUNT) identifier = ID_COUNT;
            for (int i=0; i < identifier; i++) {
		ID_ID(i) = 0;
	        ID_PARENT(i) = 0;
	    }
	}
    }
    ParseState.data_count = NA_INTEGER;
    if (ParseState.prevState) {
	R_ReleaseObject(ParseState.sexps);
        SrcRefState *prev = ParseState.prevState;
    	UseSrcRefState(prev);
    	free(prev);
    } else
        busy = FALSE;
}

static void UseSrcRefState(SrcRefState *state)
{
    ParseState.keepSrcRefs = state->keepSrcRefs;
    ParseState.keepParseData = state->keepParseData;
    ParseState.sexps = state->sexps;
    ParseState.data = state->data;
    ParseState.data_count = state->data_count;
    ParseState.xxlineno = state->xxlineno;
    ParseState.xxcolno = state->xxcolno;
    ParseState.xxbyteno = state->xxbyteno;
    ParseState.xxparseno = state->xxparseno;
    ParseState.prevState = state->prevState;
    busy = TRUE;
}

static void PutSrcRefState(SrcRefState *state)
{
    state->keepSrcRefs = ParseState.keepSrcRefs;
    state->keepParseData = ParseState.keepParseData;
    state->sexps = ParseState.sexps;
    state->data = ParseState.data;
    state->data_count = ParseState.data_count;
    state->xxlineno = ParseState.xxlineno;
    state->xxcolno = ParseState.xxcolno;
    state->xxbyteno = ParseState.xxbyteno;
    state->xxparseno = ParseState.xxparseno;
    state->prevState = ParseState.prevState;
}

static void ParseInit(void)
{
    contextp = contextstack;
    *contextp = ' ';
    SavedToken = 0;
    SavedLval = R_NilValue;
    EatLines = 0;
    EndOfFile = 0;
    xxcharcount = 0;
    npush = 0;
    HavePipeBind = FALSE;
}

static void initData(void)
{
    ParseState.data_count = 0 ;
}


static void ParseContextInit(void)
{
    R_ParseContextLast = 0;
    R_ParseContext[0] = '\0';
    
    /* starts the identifier counter*/
    initId();
    initData();
}

static int checkForPipeBind(SEXP arg)
{
    if (! HavePipeBind)
    	return FALSE;
    else if (arg == R_PipeBindSymbol)
	return TRUE;
    else if (TYPEOF(arg) == LANGSXP)
	for (SEXP cur = arg; cur != R_NilValue; cur = CDR(cur))
	    if (checkForPipeBind(CAR(cur)))
		return TRUE;
    return FALSE;
}

static SEXP R_Parse1(ParseStatus *status)
{
    Status = 1; /* safety */
    switch(yyparse()) {
    case 0:
	switch(Status) {
	case 0:                     /* End of file */
	    *status = PARSE_EOF;
	    if (EndOfFile == 2) *status = PARSE_INCOMPLETE;
	    break;
	case 1:                     /* Error (currently unreachable) */
	    *status = PARSE_ERROR;
	    if (EndOfFile) *status = PARSE_INCOMPLETE;
	    break;
	case 2:                     /* Empty Line */
	    *status = PARSE_NULL;
	    break;
	case 3:                     /* Valid expr '\n' terminated */
	case 4:                     /* Valid expr ';' terminated */
	    if (checkForPlaceholder(R_PlaceholderToken, R_CurrentExpr))
		errorcall(R_CurrentExpr,
			  _("invalid use of pipe placeholder"));
	    if (checkForPipeBind(R_CurrentExpr))
		errorcall(R_CurrentExpr,
			  _("invalid use of pipe bind symbol"));
	    *status = PARSE_OK;
	    break;
	}
	break;
    case 1:                     /* Syntax error / incomplete */
	*status = PARSE_ERROR;
	if (EndOfFile) *status = PARSE_INCOMPLETE;
	break;
    case 2:
	error(_("out of memory while parsing"));
    }
    return R_CurrentExpr;
}

static FILE *fp_parse;

static int file_getc(void)
{
    return R_fgetc(fp_parse);
}

/* used in main.c */
attribute_hidden
SEXP R_Parse1File(FILE *fp, int gencode, ParseStatus *status)
{
    ParseInit();
    ParseContextInit();
    GenerateCode = gencode;
    fp_parse = fp;
    ptr_getc = file_getc;
    R_Parse1(status);
    CLEAR_SVS();
    return R_CurrentExpr;
}

static IoBuffer *iob;

static int buffer_getc(void)
{
    return R_IoBufferGetc(iob);
}

/* Used only in main.c */
attribute_hidden
SEXP R_Parse1Buffer(IoBuffer *buffer, int gencode, ParseStatus *status)
{
    Rboolean keepSource = FALSE; 
    RCNTXT cntxt;

    R_InitSrcRefState(&cntxt);
    if (gencode) {
    	keepSource = asLogical(GetOption1(install("keep.source")));
    	if (keepSource) {
    	    ParseState.keepSrcRefs = TRUE;
	    ParseState.keepParseData =
		asLogical(GetOption1(install("keep.parse.data")));
	    PS_SET_SRCFILE(NewEnvironment(R_NilValue, R_NilValue, R_EmptyEnv));
	    PS_SET_ORIGINAL(PS_SRCFILE);
	    PS_SET_SRCREFS(R_NilValue);
	}
    }
    ParseInit();
    ParseContextInit();
    GenerateCode = gencode;
    iob = buffer;
    ptr_getc = buffer_getc;
    R_Parse1(status);
    if (gencode && keepSource) {
    	if (ParseState.didAttach) {
   	    int buflen = R_IoBufferReadOffset(buffer);
   	    char buf[buflen+1];
   	    SEXP class;
   	    R_IoBufferReadReset(buffer);
   	    for (int i=0; i<buflen; i++)
   	    	buf[i] = (char) R_IoBufferGetc(buffer);

   	    buf[buflen] = 0;
	    SEXP s_filename = install("filename");
	    defineVar(s_filename, ScalarString(mkChar("")), PS_ORIGINAL);
	    SEXP s_lines = install("lines");
	    defineVar(s_lines, ScalarString(mkChar2(buf)), PS_ORIGINAL);
    	    PROTECT(class = allocVector(STRSXP, 2));
            SET_STRING_ELT(class, 0, mkChar("srcfilecopy"));
            SET_STRING_ELT(class, 1, mkChar("srcfile"));
	    setAttrib(PS_ORIGINAL, R_ClassSymbol, class);
	    UNPROTECT(1); /* class */
	}
    }
    PROTECT(R_CurrentExpr);
    endcontext(&cntxt);
    R_FinalizeSrcRefState();
    UNPROTECT(1); /* R_CurrentExpr */
    return R_CurrentExpr;
}

static TextBuffer *txtb;

static int text_getc(void)
{
    return R_TextBufferGetc(txtb);
}

static SEXP R_Parse(int n, ParseStatus *status, SEXP srcfile)
{
    int i;
    SEXP t, rval;
    RCNTXT cntxt;

    R_InitSrcRefState(&cntxt);
    ParseContextInit();

    PS_SET_SRCFILE(srcfile);
    PS_SET_ORIGINAL(srcfile);
    
    if (isEnvironment(srcfile)) {
    	ParseState.keepSrcRefs = TRUE;
	ParseState.keepParseData =
	    asLogical(GetOption1(install("keep.parse.data")));
	PS_SET_SRCREFS(R_NilValue);
    }
    
    PROTECT(t = NewList());
    for(i = 0; ; ) {
	if(n >= 0 && i >= n) break;
	ParseInit();
	rval = R_Parse1(status);
	switch(*status) {
	case PARSE_NULL:
	    break;
	case PARSE_OK:
	    PROTECT(rval);
	    GrowList(t, rval);
	    UNPROTECT(1); /* rval */
	    i++;
	    break;
	case PARSE_INCOMPLETE:
	case PARSE_ERROR:
	    UNPROTECT(1); /* t */
	    if (ParseState.keepSrcRefs && ParseState.keepParseData)
	        finalizeData();
	    endcontext(&cntxt);
	    R_FinalizeSrcRefState();	    
	    return R_NilValue;
	    break;
	case PARSE_EOF:
	    goto finish;
	    break;
	}
    }

finish:

    t = CDR(t);
    PROTECT(rval = allocVector(EXPRSXP, length(t)));
    for (n = 0 ; n < LENGTH(rval) ; n++, t = CDR(t))
	SET_VECTOR_ELT(rval, n, CAR(t));
    if (ParseState.keepSrcRefs) {
	if (ParseState.keepParseData)
	    finalizeData();
	attachSrcrefs(rval);
    }
    UNPROTECT(2); /* t, rval */
    PROTECT(rval);
    endcontext(&cntxt);
    R_FinalizeSrcRefState();
    UNPROTECT(1); /* rval */
    *status = PARSE_OK;
    return rval;
}

/* used in edit.c */
attribute_hidden
SEXP R_ParseFile(FILE *fp, int n, ParseStatus *status, SEXP srcfile)
{
    GenerateCode = 1;
    fp_parse = fp;
    ptr_getc = file_getc;
    return R_Parse(n, status, srcfile);
}

#include "Rconnections.h"
static Rconnection con_parse;

/* need to handle incomplete last line */
static int con_getc(void)
{
    int c;
    static int last=-1000;

    c = Rconn_fgetc(con_parse);
    if (c == EOF && last != '\n') c = '\n';
    return (last = c);
}

/* used in source.c */
attribute_hidden
SEXP R_ParseConn(Rconnection con, int n, ParseStatus *status, SEXP srcfile)
{
    GenerateCode = 1;
    con_parse = con;
    ptr_getc = con_getc;
    return R_Parse(n, status, srcfile);
}

/* This one is public, and used in source.c */
SEXP R_ParseVector(SEXP text, int n, ParseStatus *status, SEXP srcfile)
{
    SEXP rval;
    TextBuffer textb;
    R_TextBufferInit(&textb, text);
    txtb = &textb;
    GenerateCode = 1;
    ptr_getc = text_getc;
    rval = R_Parse(n, status, srcfile);
    R_TextBufferFree(&textb);
    return rval;
}

static const char *Prompt(SEXP prompt, int type)
{
    if(type == 1) {
	if(length(prompt) <= 0) {
	    return CHAR(STRING_ELT(GetOption1(install("prompt")), 0));
	}
	else
	    return CHAR(STRING_ELT(prompt, 0));
    }
    else {
	return CHAR(STRING_ELT(GetOption1(install("continue")), 0));
    }
}

/* used in source.c */
attribute_hidden
SEXP R_ParseBuffer(IoBuffer *buffer, int n, ParseStatus *status, SEXP prompt, 
		   SEXP srcfile)
{
    SEXP rval, t;
    char *bufp, buf[CONSOLE_BUFFER_SIZE];
    int c, i, prompt_type = 1;
    RCNTXT cntxt;

    R_IoBufferWriteReset(buffer);
    buf[0] = '\0';
    bufp = buf;
    R_InitSrcRefState(&cntxt);
    ParseContextInit();
    
    GenerateCode = 1;
    iob = buffer;
    ptr_getc = buffer_getc;

    PS_SET_SRCFILE(srcfile);
    PS_SET_ORIGINAL(srcfile);
    
    if (isEnvironment(srcfile)) {
    	ParseState.keepSrcRefs = TRUE;
	ParseState.keepParseData =
	    asLogical(GetOption1(install("keep.parse.data")));
	PS_SET_SRCREFS(R_NilValue);
    }
    
    PROTECT(t = NewList());
    for(i = 0; ; ) {
	if(n >= 0 && i >= n) break;
	if (!*bufp) {
	    if(R_ReadConsole((char *) Prompt(prompt, prompt_type),
			     (unsigned char *)buf, CONSOLE_BUFFER_SIZE, 1) == 0)
		goto finish;
	    bufp = buf;
	}
	while ((c = *bufp++)) {
	    R_IoBufferPutc(c, buffer);
	    if (c == ';' || c == '\n') break;
	}

	/* Was a call to R_Parse1Buffer, but we don't want to reset
	   xxlineno and xxcolno */
	ParseInit();
	/* Not calling ParseContextInit() as it resets parse data, and
	   to be consistent with R_Parse */
	R_Parse1(status);
	rval = R_CurrentExpr;

	switch(*status) {
	case PARSE_NULL:
	    break;
	case PARSE_OK:
	    PROTECT(rval);
	    GrowList(t, rval);
	    UNPROTECT(1); /* rval */
	    i++;
	    break;
	case PARSE_INCOMPLETE:
	case PARSE_ERROR:
	    UNPROTECT(1); /* t */
	    R_IoBufferWriteReset(buffer);
	    endcontext(&cntxt);
	    R_FinalizeSrcRefState();
	    return R_NilValue;
	    break;
	case PARSE_EOF:
	    goto finish;
	    break;
	}
    }
finish:
    R_IoBufferWriteReset(buffer);
    t = CDR(t);
    PROTECT(rval = allocVector(EXPRSXP, length(t)));
    for (n = 0 ; n < LENGTH(rval) ; n++, t = CDR(t))
	SET_VECTOR_ELT(rval, n, CAR(t));
    if (ParseState.keepSrcRefs) {
	if (ParseState.keepParseData)
	    finalizeData();
	attachSrcrefs(rval);
    }
    UNPROTECT(2); /* t, rval */
    PROTECT(rval);
    endcontext(&cntxt);
    R_FinalizeSrcRefState();
    UNPROTECT(1); /* rval */
    *status = PARSE_OK;
    return rval;
}


/*----------------------------------------------------------------------------
 *
 *  The Lexical Analyzer:
 *
 *  Basic lexical analysis is performed by the following
 *  routines.  Input is read a line at a time, and, if the
 *  program is in batch mode, each input line is echoed to
 *  standard output after it is read.
 *
 *  The function yylex() scans the input, breaking it into
 *  tokens which are then passed to the parser.  The lexical
 *  analyser maintains a symbol table (in a very messy fashion).
 *
 *  The fact that if statements need to parse differently
 *  depending on whether the statement is being interpreted or
 *  part of the body of a function causes the need for ifpop
 *  and IfPush.  When an if statement is encountered an 'i' is
 *  pushed on a stack (provided there are parentheses active).
 *  At later points this 'i' needs to be popped off of the if
 *  stack.
 *
 */

static void IfPush(void)
{
    if (*contextp==LBRACE ||
	*contextp=='['    ||
	*contextp=='('    ||
	*contextp == 'i') {
	if(contextp - contextstack >= CONTEXTSTACK_SIZE)
	    error(_("contextstack overflow"));
	*++contextp = 'i';
    }

}

static void ifpop(void)
{
    if (*contextp=='i')
	*contextp-- = 0;
}

/* This is only called following ., so we only care if it is
   an ANSI digit or not */
static int typeofnext(void)
{
    int k, c;

    c = xxgetc();
    if (isdigit(c)) k = 1; else k = 2;
    xxungetc(c);
    return k;
}

static int nextchar(int expect)
{
    int c = xxgetc();
    if (c == expect)
	return 1;
    else
	xxungetc(c);
    return 0;
}

/* Special Symbols */
/* Syntactic Keywords + Symbolic Constants */

struct {
    char *name;
    int token;
}
static keywords[] = {
    { "NULL",	    NULL_CONST },
    { "NA",	    NUM_CONST  },
    { "TRUE",	    NUM_CONST  },
    { "FALSE",	    NUM_CONST  },
    { "Inf",	    NUM_CONST  },
    { "NaN",	    NUM_CONST  },
    { "NA_integer_", NUM_CONST  },
    { "NA_real_",    NUM_CONST  },
    { "NA_character_", NUM_CONST  },
    { "NA_complex_", NUM_CONST  },
    { "function",   FUNCTION   },
    { "while",	    WHILE      },
    { "repeat",	    REPEAT     },
    { "for",	    FOR	       },
    { "if",	    IF	       },
    { "in",	    IN	       },
    { "else",	    ELSE       },
    { "next",	    NEXT       },
    { "break",	    BREAK      },
    { "...",	    SYMBOL     },
    { 0,	    0	       }
};

/* KeywordLookup has side effects, it sets yylval */

static int KeywordLookup(const char *s)
{
    int i;
    for (i = 0; keywords[i].name; i++) {
	if (strcmp(keywords[i].name, s) == 0) {
	    switch (keywords[i].token) {
	    case NULL_CONST:
		PRESERVE_SV(yylval = R_NilValue);
		break;
	    case NUM_CONST:
		if(GenerateCode) {
		    switch(i) {
		    case 1:
			PRESERVE_SV(yylval = mkNA());
			break;
		    case 2:
			PRESERVE_SV(yylval = mkTrue());
			break;
		    case 3:
			PRESERVE_SV(yylval = mkFalse());
			break;
		    case 4:
			PRESERVE_SV(yylval = allocVector(REALSXP, 1));
			REAL(yylval)[0] = R_PosInf;
			break;
		    case 5:
			PRESERVE_SV(yylval = allocVector(REALSXP, 1));
			REAL(yylval)[0] = R_NaN;
			break;
		    case 6:
			PRESERVE_SV(yylval = allocVector(INTSXP, 1));
			INTEGER(yylval)[0] = NA_INTEGER;
			break;
		    case 7:
			PRESERVE_SV(yylval = allocVector(REALSXP, 1));
			REAL(yylval)[0] = NA_REAL;
			break;
		    case 8:
			PRESERVE_SV(yylval = allocVector(STRSXP, 1));
			SET_STRING_ELT(yylval, 0, NA_STRING);
			break;
		    case 9:
			PRESERVE_SV(yylval = allocVector(CPLXSXP, 1));
			COMPLEX(yylval)[0].r = COMPLEX(yylval)[0].i = NA_REAL;
			break;
		    }
		} else
		    PRESERVE_SV(yylval = R_NilValue);
		break;
	    case FUNCTION:
	    case WHILE:
	    case REPEAT:
	    case FOR:
	    case IF:
	    case NEXT:
	    case BREAK:
		yylval = install(s);
		break;
	    case IN:
	    case ELSE:
		break;
	    case SYMBOL:
		PRESERVE_SV(yylval = install(s));
		break;
	    }
	    return keywords[i].token;
	}
    }
    return 0;
}

static SEXP mkFloat(const char *s)
{
    return ScalarReal(R_atof(s));
}

static SEXP mkInt(const char *s)
{
    double f = R_atof(s);  /* or R_strtol? */
    return ScalarInteger((int) f);
}

static SEXP mkComplex(const char *s)
{
    SEXP t = R_NilValue;
    double f;
    f = R_atof(s); /* FIXME: make certain the value is legitimate. */
    t = allocVector(CPLXSXP, 1);
    COMPLEX(t)[0].r = 0;
    COMPLEX(t)[0].i = f;
    return t;
}

static SEXP mkNA(void)
{
    SEXP t = allocVector(LGLSXP, 1);
    LOGICAL(t)[0] = NA_LOGICAL;
    return t;
}

attribute_hidden
SEXP mkTrue(void)
{
    SEXP s = allocVector(LGLSXP, 1);
    LOGICAL(s)[0] = 1;
    return s;
}

SEXP mkFalse(void)
{
    SEXP s = allocVector(LGLSXP, 1);
    LOGICAL(s)[0] = 0;
    return s;
}

static void yyerror(const char *s)
{
    static const char *const yytname_translations[] =
    {
    /* the left column are strings coming from bison, the right
       column are translations for users.
       The first YYENGLISH from the right column are English to be translated,
       the rest are to be copied literally.  The #if 0 block below allows xgettext
       to see these.
    */
#define YYENGLISH 8
	"$undefined",	"input",
	"END_OF_INPUT",	"end of input",
	"ERROR",	"input",
	"STR_CONST",	"string constant",
	"NUM_CONST",	"numeric constant",
	"SYMBOL",	"symbol",
	"LEFT_ASSIGN",	"assignment",
	"'\\n'",	"end of line",
	"NULL_CONST",	"'NULL'",
	"FUNCTION",	"'function'",
	"EQ_ASSIGN",	"'='",
	"RIGHT_ASSIGN",	"'->'",
	"LBB",		"'[['",
	"FOR",		"'for'",
	"IN",		"'in'",
	"IF",		"'if'",
	"ELSE",		"'else'",
	"WHILE",	"'while'",
	"NEXT",		"'next'",
	"BREAK",	"'break'",
	"REPEAT",	"'repeat'",
	"GT",		"'>'",
	"GE",		"'>='",
	"LT",		"'<'",
	"LE",		"'<='",
	"EQ",		"'=='",
	"NE",		"'!='",
	"AND",		"'&'",
	"OR",		"'|'",
	"AND2",		"'&&'",
	"OR2",		"'||'",
	"NS_GET",	"'::'",
	"NS_GET_INT",	"':::'",
	"PIPE",         "'|>'",
	"PIPEBIND",     "'=>'",
	"PLACEHOLDER",  "'_'",
	0
    };
    static char const yyunexpected[] = "syntax error, unexpected ";
    static char const yyexpecting[] = ", expecting ";
    char *expecting;
    
    if (!EndOfFile)
	/* On EndOfFile, there are no more bytes to add, but trying to do
	   so may have non-trivial performance overhead and this can be
	   reached also in non-error situations, e.g. from repl.
	*/
	finish_mbcs_in_parse_context();

    R_ParseError     = yylloc.first_line;
    R_ParseErrorCol  = yylloc.first_column;
    R_ParseErrorFile = PS_SRCFILE;

    if (!strncmp(s, yyunexpected, sizeof yyunexpected -1)) {
	int i;
	/* Edit the error message */
	expecting = strstr(s + sizeof yyunexpected -1, yyexpecting);
	if (expecting) *expecting = '\0';
	for (i = 0; yytname_translations[i]; i += 2) {
	    if (!strcmp(s + sizeof yyunexpected - 1, yytname_translations[i])) {
                switch(i/2)
                {
                case 0:
                        snprintf(R_ParseErrorMsg, PARSE_ERROR_SIZE, _("unexpected input"));
                                break;
                case 1:
                        snprintf(R_ParseErrorMsg, PARSE_ERROR_SIZE, _("unexpected end of input"));
                                break;
                case 2:
                        snprintf(R_ParseErrorMsg, PARSE_ERROR_SIZE, _("unexpected input"));
                                break;
                case 3:
                        snprintf(R_ParseErrorMsg, PARSE_ERROR_SIZE, _("unexpected string constant"));
                                break;
                case 4:
                        snprintf(R_ParseErrorMsg, PARSE_ERROR_SIZE, _("unexpected numeric constant"));
                                break;
                case 5:
                        snprintf(R_ParseErrorMsg, PARSE_ERROR_SIZE, _("unexpected symbol"));
                                break;
                case 6:
                        snprintf(R_ParseErrorMsg, PARSE_ERROR_SIZE, _("unexpected assignment"));
                                break;
                case 7:
                        snprintf(R_ParseErrorMsg, PARSE_ERROR_SIZE, _("unexpected end of line"));
                                break;
                default:
		  if (!strcmp(s + sizeof yyunexpected - 1, "PLACEHOLDER")) {
		      /* cheat to avoid changing the parse error
			 message for mis-use of _ */
		      snprintf(R_ParseErrorMsg, PARSE_ERROR_SIZE,
			       _("unexpected input"));
		      break;
		  }
                  snprintf(R_ParseErrorMsg, PARSE_ERROR_SIZE, _("unexpected %s"),
                           yytname_translations[i+1]);
                                break;
                }
                
		return;
	    }
	}
	snprintf(R_ParseErrorMsg, PARSE_ERROR_SIZE - 1, _("unexpected %s"),
                 s + sizeof yyunexpected - 1);
    } else {
	strncpy(R_ParseErrorMsg, s, PARSE_ERROR_SIZE - 1);
        R_ParseErrorMsg[PARSE_ERROR_SIZE - 1] = '\0';
    }
}

static void CheckFormalArgs(SEXP formlist, SEXP _new, YYLTYPE *lloc)
{
    while (formlist != R_NilValue) {
	if (TAG(formlist) == _new) {
	    error(_("repeated formal argument '%s' on line %d"), EncodeChar(PRINTNAME(_new)),
								 lloc->first_line);
	}
	formlist = CDR(formlist);
    }
}

/* This is used as the buffer for NumericValue, SpecialValue and
   SymbolValue.  None of these could conceivably need 8192 bytes.

   It has not been used as the buffer for input character strings
   since Oct 2007 (released as 2.7.0), and for comments since 2.8.0
 */
static char yytext[MAXELTSIZE];

static int SkipSpace(void)
{
    int c;

#if defined(USE_RI18N_FNS) // includes Win32
    static wctype_t blankwct = 0;
    if (!blankwct)
	blankwct = Ri18n_wctype("blank");
#endif

#ifdef Win32
    if(!mbcslocale) { /* 0xa0 is NBSP in all 8-bit Windows locales */
	while ((c = xxgetc()) == ' ' || c == '\t' || c == '\f' ||
	       (unsigned int) c == 0xa0) ;
	return c;
    } else {
	int i, clen;
	wchar_t wc;
	while (1) {
	    c = xxgetc();
	    if (c == ' ' || c == '\t' || c == '\f') continue;
	    if (c == '\n' || c == R_EOF) break;
	    if ((unsigned int) c < 0x80) break;
	    clen = mbcs_get_next(c, &wc);  /* always 2 */
	    if(! Ri18n_iswctype(wc, blankwct) ) break;
	    for(i = 1; i < clen; i++) c = xxgetc();
	}
	return c;
    }
#endif
#if defined(__STDC_ISO_10646__)
    if(mbcslocale) { /* wctype functions need Unicode wchar_t */
	int i, clen;
	wchar_t wc;
	while (1) {
	    c = xxgetc();
	    if (c == ' ' || c == '\t' || c == '\f') continue;
	    if (c == '\n' || c == R_EOF) break;
	    if ((unsigned int) c < 0x80) break;
	    clen = mbcs_get_next(c, &wc);
#if defined(USE_RI18N_FNS)
	    if(! Ri18n_iswctype(wc, blankwct) ) break;
#else
	    if(! iswblank(wc) ) break;
#endif
	    for(i = 1; i < clen; i++) c = xxgetc();
	}
    } else
#endif
	// does not support non-ASCII spaces, unlike Windows
	while ((c = xxgetc()) == ' ' || c == '\t' || c == '\f') ;
    return c;
}

/* Note that with interactive use, EOF cannot occur inside */
/* a comment.  However, semicolons inside comments make it */
/* appear that this does happen.  For this reason we use the */
/* special assignment EndOfFile=2 to indicate that this is */
/* going on.  This is detected and dealt with in Parse1Buffer. */

static int SkipComment(void)
{
    int c='#', i;
    
    /* locations before the # character was read */
    int _first_column = ParseState.xxcolno ;
    int _first_parsed = ParseState.xxparseno ;
    int type = COMMENT ;

    Rboolean maybeLine = (ParseState.xxcolno == 1);
    Rboolean doSave;

    DECLARE_YYTEXT_BUFP(yyp);
    
    if (maybeLine) {
    	char lineDirective[] = "#line";
    	YYTEXT_PUSH(c, yyp);
    	for (i=1; i<5; i++) {
    	    c = xxgetc();
  	    if (c != (int)(lineDirective[i])) {
  	    	maybeLine = FALSE;
  	    	break;
  	    }
            YYTEXT_PUSH(c, yyp);
  	}
  	if (maybeLine)     
	    c = processLineDirective(&type);
    }
    // we want to track down the character
    // __before__ the new line character
    int _last_column  = ParseState.xxcolno ;
    int _last_parsed  = ParseState.xxparseno ;
    
    if (c == '\n') {
        _last_column = prevcols[prevpos];
        _last_parsed = prevparse[prevpos];
    }
    
    doSave = !maybeLine;
    
    while (c != '\n' && c != R_EOF) {
        // Comments can be any length; we only record the ones that fit in yytext.
        if (doSave) {
            YYTEXT_PUSH(c, yyp);
            doSave = (yyp - yytext) < sizeof(yytext) - 2;
        }
 	_last_column = ParseState.xxcolno ;
	_last_parsed = ParseState.xxparseno ;
	c = xxgetc();
    }
    if (c == R_EOF) EndOfFile = 2;
    incrementId( ) ;
    YYTEXT_PUSH('\0', yyp);
    record_( _first_parsed, _first_column, _last_parsed, _last_column,
	     type, identifier, doSave ? yytext : 0 ) ;
    return c;
}

static int NumericValue(int c)
{
    int seendot = (c == '.');
    int seenexp = 0;
    int last = c;
    int nd = 0;
    int asNumeric = 0;
    int count = 1; /* The number of characters seen */

    DECLARE_YYTEXT_BUFP(yyp);
    YYTEXT_PUSH(c, yyp);
    /* We don't care about other than ASCII digits */
    while (isdigit(c = xxgetc()) || c == '.' || c == 'e' || c == 'E'
	   || c == 'x' || c == 'X' || c == 'L')
    {
	count++;
	if (c == 'L') /* must be at the end.  Won't allow 1Le3 (at present). */
	{   YYTEXT_PUSH(c, yyp);
	    break;
	}
	
	if (c == 'x' || c == 'X') {
	    if (count > 2 || last != '0') break;  /* 0x must be first */
	    YYTEXT_PUSH(c, yyp);
	    while(isdigit(c = xxgetc()) || ('a' <= c && c <= 'f') ||
		  ('A' <= c && c <= 'F') || c == '.') {
		if (c == '.') {
		    if (seendot) return ERROR;
		    seendot = 1;
		}
		YYTEXT_PUSH(c, yyp);
		nd++;
	    }
	    if (nd == 0) return ERROR;
	    if (c == 'p' || c == 'P') {
	        seenexp = 1;
		YYTEXT_PUSH(c, yyp);
		c = xxgetc();
		if (!isdigit(c) && c != '+' && c != '-') return ERROR;
		if (c == '+' || c == '-') {
		    YYTEXT_PUSH(c, yyp);
		    c = xxgetc();
		}
		for(nd = 0; isdigit(c); c = xxgetc(), nd++)
		    YYTEXT_PUSH(c, yyp);
		if (nd == 0) return ERROR;
	    }
            if (seendot && !seenexp) return ERROR;
	    if (c == 'L') /* for getParseData */
	    {
		// seenexp will be checked later
		YYTEXT_PUSH(c, yyp);
		break;
	    }
	    break;
	}
	if (c == 'E' || c == 'e') {
	    if (seenexp)
		break;
	    seenexp = 1;
	    seendot = seendot == 1 ? seendot : 2;
	    YYTEXT_PUSH(c, yyp);
	    c = xxgetc();
	    if (!isdigit(c) && c != '+' && c != '-') return ERROR;
	    if (c == '+' || c == '-') {
		YYTEXT_PUSH(c, yyp);
		c = xxgetc();
		if (!isdigit(c)) return ERROR;
	    }
	}
	if (c == '.') {
	    if (seendot)
		break;
	    seendot = 1;
	}
	YYTEXT_PUSH(c, yyp);
	last = c;
    }
    
    if(c == 'i')
	YYTEXT_PUSH(c, yyp); /* for getParseData */
	
    YYTEXT_PUSH('\0', yyp);    
    /* Make certain that things are okay. */
    if(c == 'L') {
	double a = R_atof(yytext);
	int b = (int) a;
	/* We are asked to create an integer via the L, so we check that the
	   double and int values are the same. If not, this is a problem and we
	   will not lose information and so use the numeric value.
	*/
	if(a != (double) b) {
	    if(GenerateCode) {
		if(seendot == 1 && seenexp == 0)
		    warning(_("integer literal %s contains decimal; using numeric value"), yytext);
		else {
		    /* hide the L for the warning message */
		    warning(_("non-integer value %s qualified with L; using numeric value"), yytext);
		}
	    }
	    asNumeric = 1;
	    seenexp = 1;
	}
    }

    if(c == 'i') {
	yylval = GenerateCode ? mkComplex(yytext) : R_NilValue;
    } else if(c == 'L' && asNumeric == 0) {
	if(GenerateCode && seendot == 1 && seenexp == 0)
	    warning(_("integer literal %s contains unnecessary decimal point"), yytext);
	yylval = GenerateCode ? mkInt(yytext) : R_NilValue;
#if 0  /* do this to make 123 integer not double */
    } else if(!(seendot || seenexp)) {
	if(c != 'L') xxungetc(c);
	if (GenerateCode) {
	    double a = R_atof(yytext);
	    int b = (int) a;
	    yylval = (a != (double) b) ? mkFloat(yytext) : mkInt(yytext);
	} else yylval = R_NilValue;
#endif
    } else {
	if(c != 'L')
	    xxungetc(c);
	yylval = GenerateCode ? mkFloat(yytext) : R_NilValue;
    }

    PRESERVE_SV(yylval);
    return NUM_CONST;
}

/* Strings may contain the standard ANSI escapes and octal */
/* specifications of the form \o, \oo or \ooo, where 'o' */
/* is an octal digit. */

/* The buffer is reallocated on the R heap if needed; not by malloc */
/* to avoid memory leak in case of R error (long jump) */
#define STEXT_PUSH(c) do {                  \
	size_t nc = bp - stext;             \
	if (nc >= nstext - 1) {             \
	    char *old = stext;              \
	    SEXP st1;		            \
	    nstext *= 2;                    \
	    PROTECT(st1 = allocVector(RAWSXP, nstext)); \
	    stext = (char *)RAW(st1);       \
	    memmove(stext, old, nc);        \
	    REPROTECT(st1, sti);	    \
	    UNPROTECT(1); /* st1 */         \
	    bp = stext+nc; }		    \
	*bp++ = ((char) c);		    \
} while(0)


/* The idea here is that if a string contains \u escapes that are not
   valid in the current locale, we should switch to UTF-8 for that
   string.  Needs Unicode wide-char support or out substitutes.

   Defining __STDC_ISO_10646__ is done by the OS (or not) in wchar.t.
   Some (e.g. macOS, Solaris, FreeBSD) have Unicode wchar_t but do not
   define it: we override macOS and FreeBSD earlier in this file.
*/

#if defined(Win32) || defined(__STDC_ISO_10646__)
typedef wchar_t ucs_t;
# define mbcs_get_next2 mbcs_get_next
#else
typedef unsigned int ucs_t;
# define WC_NOT_UNICODE
// which is used to select our mbtoucs rather than system mbrtowc
static int mbcs_get_next2(int c, ucs_t *wc)
{
    int i, res, clen = 1; char s[9];

    s[0] = c;
    /* This assumes (probably OK) that all MBCS embed ASCII as single-byte
       lead bytes, including control chars */
    if((unsigned int) c < 0x80) {
	*wc = (wchar_t) c;
	return 1;
    }
    if(utf8locale) {
	clen = utf8clen(c);
	for(i = 1; i < clen; i++) {
	    c = xxgetc();
	    if(c == R_EOF) error(_("EOF whilst reading MBCS char at line %d"), ParseState.xxlineno);
	    s[i] = (char) c;
	}
	s[clen] ='\0'; /* x86 Solaris requires this */
	res = mbtoucs(wc, s, clen);
	if(res == -1) error(_("invalid multibyte character in parser at line %d"), ParseState.xxlineno);
    } else {
	/* This is not necessarily correct for stateful MBCS */
	while(clen <= R_MB_CUR_MAX) {
	    res = mbtoucs(wc, s, clen);
	    if(res >= 0) break;
	    if(res == -1)
		error(_("invalid multibyte character in parser at line %d"), ParseState.xxlineno);
	    /* so res == -2 */
	    c = xxgetc();
	    if(c == R_EOF) error(_("EOF whilst reading MBCS char at line %d"), ParseState.xxlineno);
	    s[clen++] = c;
	} /* we've tried enough, so must be complete or invalid by now */
    }
    for(i = clen - 1; i > 0; i--) xxungetc(s[i]);
    return clen;
}
#endif

#define WTEXT_PUSH(c) do { if(wcnt < 10000) wcs[wcnt++] = c; } while(0)

static SEXP mkStringUTF8(const ucs_t *wcs, int cnt)
{
    SEXP t;
    int nb;

/* NB: cnt includes the terminator */
#ifdef Win32
    nb = cnt*4; /* UCS-2/UTF-16 so max 4 bytes per wchar_t */
#else
    nb = cnt*6;
#endif
    R_CheckStack2(nb);
    char s[nb];
    memset(s, 0, nb); /* safety */
    // This used to differentiate WC_NOT_UNICODE but not needed
    wcstoutf8(s, (const wchar_t *)wcs, sizeof(s));
    PROTECT(t = allocVector(STRSXP, 1));
    SET_STRING_ELT(t, 0, mkCharCE(s, CE_UTF8));
    UNPROTECT(1); /* t */
    return t;
}
/*
 * Skip at Least `min` Bytes in Complete Character Steps
 *
 * min: minimum number bytes of prefix of "c" to skip
 * returns: min or more as needed to skip complete characters
 *
 * Assumptions:
 * - sizeof(buffer) >= min + R_MB_CUR_MAX, i.e. at least 1 full char in buffer.
 * - MBCS encodings are valid (they've been read already so should be).
 * - Stateless encodings.
 */

static int skipBytesByChar(char *c, int min) {
    int res = 0;
    
    if(!mbcslocale) 
	res = min;
    else {
	if(utf8locale) {
	    /* Find first non continuation byte; we assume UTF-8 is valid. */
	    char *cc = c + min;
	    while(((unsigned char)*cc & 0xc0) == 0x80) ++cc;
	    res = (int) (cc - c);
	} else {
	    mbstate_t mb_st;
	    mbs_init(&mb_st);
	    while(res < min)
		res += (int) mbrtowc(NULL, c + res, R_MB_CUR_MAX, &mb_st);
	}
    }
    return res;
}

#define CTEXT_PUSH(c) do {                                             \
	if (ct - currtext >= 1000) {                                   \
	    int skip = skipBytesByChar(currtext, 100 + 4);             \
	    memmove(currtext, "... ", 4);                              \
	    memmove(currtext + 4, currtext + skip, 1000 - skip + 1);   \
	    ct -= skip - 4;                                            \
	    currtext_truncated = TRUE;                                 \
	}                                                              \
	*ct++ = ((char) c);                                            \
} while(0)
#define CTEXT_POP() ct--

/* Bidi formatting could confuse the code */
#define BIDI_CHECK(wc) do {                                           \
	if((wc) >= 0x202A && (wc) <= 0x2069 &&                        \
	  !((wc) > 0x202E && (wc) < 0x2066))                          \
	    error(_("bidi formatting not allowed (line %d), use escapes instead (\\u%04x)"),  \
		  ParseState.xxlineno, (wc));                         \
} while(0)

/* forSymbol is true when parsing backticked symbols */
static int StringValue(int c, Rboolean forSymbol)
{
    int quote = c;
    char currtext[1010], *ct = currtext;
    char st0[MAXELTSIZE];
    unsigned int nstext = MAXELTSIZE;
    char *stext = st0, *bp = st0;
    PROTECT_INDEX sti;
    int wcnt = 0;
    ucs_t wcs[10001];
    Rboolean oct_or_hex = FALSE, use_wcs = FALSE, currtext_truncated = FALSE;

    PROTECT_WITH_INDEX(R_NilValue, &sti);
    CTEXT_PUSH(c);
    while ((c = xxgetc()) != R_EOF && c != quote) {
	CTEXT_PUSH(c);
	if (c == '\n') {
	    xxungetc(c); CTEXT_POP();
	    /* Fix suggested by Mark Bravington to allow multiline strings
	     * by pretending we've seen a backslash. Was:
	     * return ERROR;
	     */
	    c = '\\';
	}
	if (c == '\\') {
	    c = xxgetc(); CTEXT_PUSH(c);
	    if ('0' <= c && c <= '7') {
		int octal = c - '0';
		if ('0' <= (c = xxgetc()) && c <= '7') {
		    CTEXT_PUSH(c);
		    octal = 8 * octal + c - '0';
		    if ('0' <= (c = xxgetc()) && c <= '7') {
			CTEXT_PUSH(c);
			octal = 8 * octal + c - '0';
		    } else {
			xxungetc(c);
		    }
		} else {
		    xxungetc(c);
		}
		if (!octal)
		    error(_("nul character not allowed (line %d)"), ParseState.xxlineno);
		if(octal > 0xff)
		    error(_("exceeded maximum allowed octal value \\377 (line %d)"), ParseState.xxlineno);
		c = octal;
		oct_or_hex = TRUE;
	    }
	    else if(c == 'x') {
		int val = 0; int i, ext;
		for(i = 0; i < 2; i++) {
		    c = xxgetc(); CTEXT_PUSH(c);
		    if(c >= '0' && c <= '9') ext = c - '0';
		    else if (c >= 'A' && c <= 'F') ext = c - 'A' + 10;
		    else if (c >= 'a' && c <= 'f') ext = c - 'a' + 10;
		    else {
			xxungetc(c);
			CTEXT_POP();
			if (i == 0) { /* was just \x */
			    *ct = '\0';
			    errorcall(R_NilValue, _("'\\x' used without hex digits in character string starting \"%s\""), currtext);
			}
			break;
		    }
		    val = 16*val + ext;
		}
		if (!val)
		    error(_("nul character not allowed (line %d)"), ParseState.xxlineno);
		c = val;
		oct_or_hex = TRUE;
	    }
	    else if(c == 'u') {
		unsigned int val = 0; int i, ext; 
		Rboolean delim = FALSE;

		if(forSymbol) 
		    error(_("\\uxxxx sequences not supported inside backticks (line %d)"), ParseState.xxlineno);
		if((c = xxgetc()) == '{') {
		    delim = TRUE;
		    CTEXT_PUSH(c);
		} else xxungetc(c);
		for(i = 0; i < 4; i++) {
		    c = xxgetc(); CTEXT_PUSH(c);
		    if(c >= '0' && c <= '9') ext = c - '0';
		    else if (c >= 'A' && c <= 'F') ext = c - 'A' + 10;
		    else if (c >= 'a' && c <= 'f') ext = c - 'a' + 10;
		    else {
			xxungetc(c);
			CTEXT_POP();
			if (i == 0) { /* was just \u */
			    *ct = '\0';
			    errorcall(R_NilValue, _("'\\u' used without hex digits in character string starting \"%s\""), currtext);
			}
			break;
		    }
		    val = 16*val + ext;
		}
		if(delim) {
		    if((c = xxgetc()) != '}')
			error(_("invalid \\u{xxxx} sequence (line %d)"),
			      ParseState.xxlineno);
		    else CTEXT_PUSH(c);
		}
		if (!val)
		    error(_("nul character not allowed (line %d)"), ParseState.xxlineno);
		WTEXT_PUSH(val); /* this assumes wchar_t is Unicode */
		use_wcs = TRUE;
		continue;
	    }
	    else if(c == 'U') {
		unsigned int val = 0; int i, ext;
		Rboolean delim = FALSE;
		if(forSymbol) 
		    error(_("\\Uxxxxxxxx sequences not supported inside backticks (line %d)"), ParseState.xxlineno);
		if((c = xxgetc()) == '{') {
		    delim = TRUE;
		    CTEXT_PUSH(c);
		} else xxungetc(c);
		for(i = 0; i < 8; i++) {
		    c = xxgetc(); CTEXT_PUSH(c);
		    if(c >= '0' && c <= '9') ext = c - '0';
		    else if (c >= 'A' && c <= 'F') ext = c - 'A' + 10;
		    else if (c >= 'a' && c <= 'f') ext = c - 'a' + 10;
		    else {
			xxungetc(c);
			CTEXT_POP();
			if (i == 0) { /* was just \U */
			    *ct = '\0';
			    errorcall(R_NilValue, _("'\\U' used without hex digits in character string starting \"%s\""), currtext);
			}
			break;
		    }
		    val = 16*val + ext;
		}
		if(delim) {
		    if((c = xxgetc()) != '}')
			error(_("invalid \\U{xxxxxxxx} sequence (line %d)"),
			      ParseState.xxlineno);
		    else CTEXT_PUSH(c);
		}
		if (!val)
		    error(_("nul character not allowed (line %d)"),
			  ParseState.xxlineno);
		if (val > 0x10FFFF) {
		    if(delim)
			error(_("invalid \\U{xxxxxxxx} value %6x (line %d)"),
			      val, ParseState.xxlineno);
		    else
			error(_("invalid \\Uxxxxxxxx value %6x (line %d)"),
			      val, ParseState.xxlineno);
		}
#ifdef Win32
		if (0x010000 <= val && val <= 0x10FFFF) {   /* Need surrogate pair in Windows */
		    val = val - 0x010000;
		    WTEXT_PUSH( 0xD800 | (val >> 10) );
		    val = 0xDC00 | (val & 0x03FF);
		}
#endif
		WTEXT_PUSH(val);
		use_wcs = TRUE;
		continue;
	    }
	    else {
		switch (c) {
		case 'a':
		    c = '\a';
		    break;
		case 'b':
		    c = '\b';
		    break;
		case 'f':
		    c = '\f';
		    break;
		case 'n':
		    c = '\n';
		    break;
		case 'r':
		    c = '\r';
		    break;
		case 't':
		    c = '\t';
		    break;
		case 'v':
		    c = '\v';
		    break;
		case '\\':
		    c = '\\';
		    break;
		case '"':
		case '\'':
		case '`':
		case ' ':
		case '\n':
		    break;
		default:
		    *ct = '\0';
		    errorcall(R_NilValue, _("'\\%c' is an unrecognized escape in character string starting \"%s\""), c, currtext);
		}
	    }
	} else if(mbcslocale) {
	    int i, clen;
	    ucs_t wc;
	    clen = mbcs_get_next2(c, &wc);
	    BIDI_CHECK(wc);
	    WTEXT_PUSH(wc);
	    ParseState.xxbyteno += clen-1;
	    
	    for(i = 0; i < clen - 1; i++){
		STEXT_PUSH(c);
		c = xxgetc();
		if (c == R_EOF) break;
		CTEXT_PUSH(c);
		if (c == '\n') {
		    xxungetc(c); CTEXT_POP();
		    c = '\\';
		}
	    }
	    if (c == R_EOF) break;
	    STEXT_PUSH(c);
	    continue;
	}
	STEXT_PUSH(c);
	if ((unsigned int) c < 0x80) WTEXT_PUSH(c);
	else { 
	    /* have an 8-bit char in the current encoding */
	    /* FIXME: `wc` values will be wrong when native encoding differs
	       from that indicated by `known_to_be*` */
	    int res = 0;
#ifdef WC_NOT_UNICODE
	    ucs_t wc;
	    char s[2] = " ";
	    s[0] = (char) c;
	    res = (int) mbtoucs(&wc, s, 2);
#else
	    wchar_t wc;
	    char s[2] = " ";
	    s[0] = (char) c;
	    /* This is not necessarily correct for stateful SBCS */
	    mbstate_t mb_st;
	    mbs_init(&mb_st);
	    res = (int) mbrtowc(&wc, s, 2, &mb_st);
#endif
	    if(res < 0) wc = 0xFFFD; /* placeholder for invalid encoding */
	    WTEXT_PUSH(wc);
	}
    }
    STEXT_PUSH('\0');
    WTEXT_PUSH(0);
    yytext[0] = '\0';
    if (c == R_EOF) {
	PRESERVE_SV(yylval = R_NilValue);
	UNPROTECT(1); /* release stext */
    	return INCOMPLETE_STRING;
    } else {
    	CTEXT_PUSH(c);
    	CTEXT_PUSH('\0');
    }
    if (!currtext_truncated)
    	strcpy(yytext, currtext);
    else if (forSymbol || !use_wcs) {
        size_t total = strlen(stext);
        snprintf(yytext, MAXELTSIZE, "[%u chars quoted with '%c']", (unsigned int)total, quote);
    } else 
        snprintf(yytext, MAXELTSIZE, "[%d wide chars quoted with '%c']", wcnt, quote);
    if(forSymbol) {
	PRESERVE_SV(yylval = install(stext));
	UNPROTECT(1); /* release stext */
	return SYMBOL;
    } else {
	if(use_wcs) {
	    if(oct_or_hex)
		error(_("mixing Unicode and octal/hex escapes in a string is not allowed"));
	    if(wcnt < 10000)
		PRESERVE_SV(yylval = mkStringUTF8(wcs, wcnt)); /* include terminator */
	    else
		error(_("string at line %d containing Unicode escapes not in this locale\nis too long (max 10000 chars)"), ParseState.xxlineno);
	} else
	    PRESERVE_SV(yylval = mkString2(stext,  bp - stext - 1, oct_or_hex));
	UNPROTECT(1); /* release stext */
	return STR_CONST;
    }
}

static int RawStringValue(int c0, int c)
{
    int quote = c;
    int delim = ')';
    char currtext[1010], *ct = currtext;
    char st0[MAXELTSIZE];
    unsigned int nstext = MAXELTSIZE;
    char *stext = st0, *bp = st0;
    PROTECT_INDEX sti;
    int wcnt = 0;
    ucs_t wcs[10001];
    Rboolean oct_or_hex = FALSE, use_wcs = FALSE, currtext_truncated = FALSE;

    CTEXT_PUSH(c0); /* 'r' or 'R' */
    CTEXT_PUSH(c);  /* opening quote */

    /* count dashes between the opening quote and opening delimiter */
    int ndash = 0;
    while (nextchar('-')) { CTEXT_PUSH('-'); ndash++; }

    c = xxgetc();
    CTEXT_PUSH(c);
    switch(c) {
    case '(': delim = ')'; break;
    case '[': delim = ']'; break;
    case '{': delim = '}'; break;
    case '|': delim = '|'; break;
    default:
	error(_("malformed raw string literal at line %d"),
	      ParseState.xxlineno);
    }

    PROTECT_WITH_INDEX(R_NilValue, &sti);
    while ((c = xxgetc()) != R_EOF) {
	if (c == delim) {
	    /* count the dashes after the closing delimiter */
	    int nd = 0;
	    while (nd < ndash && nextchar('-')) nd++;
	    
	    if (nd == ndash && nextchar(quote))
		/* right number of dashes, right quote: were done! */
		break;
	    else {
		/* not done: emit closing delimiter, dashes, and continue */
		CTEXT_PUSH(delim);
		STEXT_PUSH(delim);
		WTEXT_PUSH(delim);
		for (int i = 0; i < nd; i++) {
		    CTEXT_PUSH('-');
		    STEXT_PUSH('-');
		    WTEXT_PUSH('-');
		}
		continue;
	    }
	}
	CTEXT_PUSH(c);
	if(mbcslocale) {
	    int i, clen;
	    ucs_t wc;
	    clen = mbcs_get_next2(c, &wc);
	    BIDI_CHECK(wc);
	    WTEXT_PUSH(wc);
	    ParseState.xxbyteno += clen-1;
	    
	    for(i = 0; i < clen - 1; i++){
		STEXT_PUSH(c);
		c = xxgetc();
		if (c == R_EOF) break;
		CTEXT_PUSH(c);
	    }
	    if (c == R_EOF) break;
	    STEXT_PUSH(c);
	    continue;
	}
	STEXT_PUSH(c);
	if ((unsigned int) c < 0x80) WTEXT_PUSH(c);
	else { /* have an 8-bit char in the current encoding */
	    int res = 0;
#ifdef WC_NOT_UNICODE
	    ucs_t wc;
	    char s[2] = " ";
	    s[0] = (char) c;
	    res = (int) mbtoucs(&wc, s, 2);
#else
	    wchar_t wc;
	    char s[2] = " ";
	    s[0] = (char) c;
	    /* This is not necessarily correct for stateful SBCS */
	    mbstate_t mb_st;
	    mbs_init(&mb_st);
	    res = (int) mbrtowc(&wc, s, 2, &mb_st);
#endif
	    if(res < 0) wc = 0xFFFD; /* placeholder for invalid encoding */
	    WTEXT_PUSH(wc);
	}
    }
    STEXT_PUSH('\0');
    WTEXT_PUSH(0);
    yytext[0] = '\0';
    if (c == R_EOF) {
	PRESERVE_SV(yylval = R_NilValue);
	UNPROTECT(1); /* release stext */
    	return INCOMPLETE_STRING;
    } else {
	/* record delim, dashes, and quote, and terminate string */
	CTEXT_PUSH(delim);
	for (int i = 0; i < ndash; i++)
	    CTEXT_PUSH('-');
	CTEXT_PUSH(quote);
    	CTEXT_PUSH('\0');
    }
    if (!currtext_truncated)
    	strcpy(yytext, currtext);
    else if (!use_wcs) {
        size_t total = strlen(stext);
        snprintf(yytext, MAXELTSIZE, "[%u chars quoted with '%c']", (unsigned int)total, quote);
    } else 
        snprintf(yytext, MAXELTSIZE, "[%d wide chars quoted with '%c']", wcnt, quote);
    if(use_wcs) {
	if(oct_or_hex)
	    error(_("mixing Unicode and octal/hex escapes in a string is not allowed"));
	if(wcnt < 10000)
	    PRESERVE_SV(yylval = mkStringUTF8(wcs, wcnt)); /* include terminator */
	else
	    error(_("string at line %d containing Unicode escapes not in this locale\nis too long (max 10000 chars)"), ParseState.xxlineno);
    } else
	PRESERVE_SV(yylval = mkString2(stext,  bp - stext - 1, oct_or_hex));
    UNPROTECT(1); /* release stext */
    return STR_CONST;
}

static int SpecialValue(int c)
{
    DECLARE_YYTEXT_BUFP(yyp);
    YYTEXT_PUSH(c, yyp);
    while ((c = xxgetc()) != R_EOF && c != '%') {
	if (c == '\n') {
	    xxungetc(c);
	    return ERROR;
	}
	YYTEXT_PUSH(c, yyp);
    }
    if (c == '%')
	YYTEXT_PUSH(c, yyp);
    YYTEXT_PUSH('\0', yyp);
    yylval = install(yytext);
    return SPECIAL;
}

/* return 1 if name is a valid name 0 otherwise */
attribute_hidden
int isValidName(const char *name)
{
    const char *p = name;
    int i;

    if(mbcslocale) {
	/* the only way to establish which chars are alpha etc is to
	   use the wchar variants */
	size_t n = strlen(name), used;
	wchar_t wc;
	/* This is not necessarily correct for stateful MBCS */
	mbstate_t mb_st;
	mbs_init(&mb_st);
	used = Mbrtowc(&wc, p, n, &mb_st); p += used; n -= used;
	if(used == 0) return 0;
	if (wc != L'.' && !iswalpha(wc) ) return 0;
	if (wc == L'.') {
	    /* We don't care about other than ASCII digits */
	    if(isdigit(0xff & (int)*p)) return 0;
	    /* Mbrtowc(&wc, p, n, NULL); if(iswdigit(wc)) return 0; */
	}
	while((used = Mbrtowc(&wc, p, n, &mb_st))) {
	    if (!(iswalnum(wc) || wc == L'.' || wc == L'_')) break;
	    p += used; n -= used;
	}
	if (*p != '\0') return 0;
    } else {
	int c = 0xff & *p++;
	if (c != '.' && !isalpha(c) ) return 0;
	if (c == '.' && isdigit(0xff & (int)*p)) return 0;
	while ( c = 0xff & *p++, (isalnum(c) || c == '.' || c == '_') ) ;
	if (c != '\0') return 0;
    }

    if (strcmp(name, "...") == 0) return 1;

    for (i = 0; keywords[i].name != NULL; i++)
	if (strcmp(keywords[i].name, name) == 0) return 0;

    return 1;
}


static int SymbolValue(int c)
{
    int kw;
    DECLARE_YYTEXT_BUFP(yyp);
    if(mbcslocale) {
	// FIXME potentially need R_wchar_t with UTF-8 Windows.
	wchar_t wc; int i, clen;
	clen = mbcs_get_next(c, &wc);
	while(1) {
	    /* at this point we have seen one char, so push its bytes
	       and get one more */
	    for(i = 0; i < clen; i++) {
		YYTEXT_PUSH(c, yyp);
		c = xxgetc();
	    }
	    if(c == R_EOF) break;
	    if(c == '.' || c == '_') {
		clen = 1;
		continue;
	    }
	    clen = mbcs_get_next(c, &wc);
	    if(!iswalnum(wc)) break;
	}
    } else
	do {
	    YYTEXT_PUSH(c, yyp);
	} while ((c = xxgetc()) != R_EOF &&
		 (isalnum(c) || c == '.' || c == '_'));
    xxungetc(c);
    YYTEXT_PUSH('\0', yyp);
    if ((kw = KeywordLookup(yytext))) 
	return kw;
    
    PRESERVE_SV(yylval = install(yytext));
    return SYMBOL;
}

static int Placeholder(int c)
{
    DECLARE_YYTEXT_BUFP(yyp);
    YYTEXT_PUSH(c, yyp);
    YYTEXT_PUSH('\0', yyp);
    HavePlaceholder = TRUE;
    PRESERVE_SV(yylval = R_PlaceholderToken);
    return PLACEHOLDER;
}

static void setParseFilename(SEXP newname) {
    SEXP class;
    
    if (isEnvironment(PS_SRCFILE)) {
	SEXP oldname = findVar(install("filename"), PS_SRCFILE);
    	if (isString(oldname) && length(oldname) > 0 &&
    	    strcmp(CHAR(STRING_ELT(oldname, 0)),
    	           CHAR(STRING_ELT(newname, 0))) == 0) return;
	PS_SET_SRCFILE(NewEnvironment(R_NilValue, R_NilValue, R_EmptyEnv));
	defineVar(install("filename"), newname, PS_SRCFILE);
	defineVar(install("original"), PS_ORIGINAL, PS_SRCFILE);

	PROTECT(class = allocVector(STRSXP, 2));
	SET_STRING_ELT(class, 0, mkChar("srcfilealias"));
	SET_STRING_ELT(class, 1, mkChar("srcfile"));
	setAttrib(PS_SRCFILE, R_ClassSymbol, class);
	UNPROTECT(1); /* class */
    } else 
	PS_SET_SRCFILE(duplicate(newname));
    RELEASE_SV(newname);
}

static int processLineDirective(int *type)
{
    int c, tok, linenumber;
    c = SkipSpace();
    if (!isdigit(c)) return(c);
    tok = NumericValue(c);
    linenumber = atoi(yytext);
    c = SkipSpace();
    if (c == '"') 
	tok = StringValue(c, FALSE);
    else
    	xxungetc(c);
    if (tok == STR_CONST) 
	setParseFilename(yylval);
    while ((c = xxgetc()) != '\n' && c != R_EOF) /* skip */ ;
    ParseState.xxlineno = linenumber;
    *type = LINE_DIRECTIVE;
    /* we don't change xxparseno here:  it counts parsed lines, not official lines */
    R_ParseContext[R_ParseContextLast] = '\0';  /* Context report shouldn't show the directive */
    return(c);
}

/* Get the R symbol, and set yytext at the same time */
static SEXP install_and_save(char * text)
{
    strcpy(yytext, text);
    return install(text);
}

/* Get an R symbol, and set different yytext.  Used for translation of -> to <-. ->> to <<- */
static SEXP install_and_save2(char * text, char * savetext)
{
    strcpy(yytext, savetext);
    return install(text);
}


/* Split the input stream into tokens. */
/* This is the lowest of the parsing levels. */

static int token(void)
{
    int c;
    wchar_t wc;

    if (SavedToken) {
	c = SavedToken;
	yylval = SavedLval;
	SavedLval = R_NilValue;
	SavedToken = 0;
	yylloc.first_line = xxlinesave;
	yylloc.first_column = xxcolsave;
	yylloc.first_byte = xxbytesave;
	yylloc.first_parsed = xxparsesave;
	return c;
    }
    xxcharsave = xxcharcount; /* want to be able to go back one token */

    c = SkipSpace();
    if (c == '#') c = SkipComment();

    yylloc.first_line = ParseState.xxlineno;
    yylloc.first_column = ParseState.xxcolno;
    yylloc.first_byte = ParseState.xxbyteno;
    yylloc.first_parsed = ParseState.xxparseno;

    if (c == R_EOF) return END_OF_INPUT;

    /* Either digits or symbols can start with a "." */
    /* so we need to decide which it is and jump to  */
    /* the correct spot. */

    if (c == '.' && typeofnext() >= 2) goto symbol;

    /* literal numbers */

    if (c == '.') return NumericValue(c);
    /* We don't care about other than ASCII digits */
    if (isdigit(c)) return NumericValue(c);

    /* raw string literal */

    if (c == 'r' || c == 'R') {
	if (nextchar('"'))
	    return RawStringValue(c, '"');
	else if (nextchar('\''))
	    return RawStringValue(c, '\'');
    }

    /* literal strings */

    if (c == '\"' || c == '\'')
	return StringValue(c, FALSE);

    /* special functions */

    if (c == '%')
	return SpecialValue(c);

    /* functions, constants and variables */

    if (c == '`')
	return StringValue(c, TRUE);
 symbol:

    if (c == '.') return SymbolValue(c);
    if (c == '_') return Placeholder(c);
    if(mbcslocale) {
	// FIXME potentially need R_wchar_t with UTF-8 Windows.
	mbcs_get_next(c, &wc);
	if (iswalpha(wc)) return SymbolValue(c);
    } else
	if (isalpha(c)) return SymbolValue(c);

    /* compound tokens */

    switch (c) {
    case '<':
	if (nextchar('=')) {
	    yylval = install_and_save("<=");
	    return LE;
	}
	if (nextchar('-')) {
	    yylval = install_and_save("<-");
	    return LEFT_ASSIGN;
	}
	if (nextchar('<')) {
	    if (nextchar('-')) {
		yylval = install_and_save("<<-");
		return LEFT_ASSIGN;
	    }
	    else
		return ERROR;
	}
	yylval = install_and_save("<");
	return LT;
    case '-':
	if (nextchar('>')) {
	    if (nextchar('>')) {
		yylval = install_and_save2("<<-", "->>");
		return RIGHT_ASSIGN;
	    }
	    else {
		yylval = install_and_save2("<-", "->");
		return RIGHT_ASSIGN;
	    }
	}
	yylval = install_and_save("-");
	return '-';
    case '>':
	if (nextchar('=')) {
	    yylval = install_and_save(">=");
	    return GE;
	}
	yylval = install_and_save(">");
	return GT;
    case '!':
	if (nextchar('=')) {
	    yylval = install_and_save("!=");
	    return NE;
	}
	yylval = install_and_save("!");
	return '!';
    case '=':
	if (nextchar('=')) {
	    yylval = install_and_save("==");
	    return EQ;
	}
	else if (nextchar('>')) {
	    yylval = install_and_save("=>");
	    HavePipeBind = TRUE;
	    return PIPEBIND;
	}		 
	yylval = install_and_save("=");
	return EQ_ASSIGN;
    case ':':
	if (nextchar(':')) {
	    if (nextchar(':')) {
		yylval = install_and_save(":::");
		return NS_GET_INT;
	    }
	    else {
		yylval = install_and_save("::");
		return NS_GET;
	    }
	}
	if (nextchar('=')) {
	    yylval = install_and_save(":=");
	    return LEFT_ASSIGN;
	}
	yylval = install_and_save(":");
	return ':';
    case '&':
	if (nextchar('&')) {
	    yylval = install_and_save("&&");
	    return AND2;
	}
	yylval = install_and_save("&");
	return AND;
    case '|':
	if (nextchar('|')) {
	    yylval = install_and_save("||");
	    return OR2;
	}
	else if (nextchar('>')) {
	    yylval = install_and_save("|>");
	    return PIPE;
	}
	yylval = install_and_save("|");
	return OR;
    case LBRACE:
	yylval = install_and_save("{");
	return c;
    case RBRACE:
        strcpy(yytext, "}");
	return c;
    case '(':
	yylval = install_and_save("(");
	return c;
    case ')':
        strcpy(yytext, ")");
	return c;
    case '[':
	if (nextchar('[')) {
	    yylval = install_and_save("[[");
	    return LBB;
	}
	yylval = install_and_save("[");
	return c;
    case ']':
        strcpy(yytext, "]");
	return c;
    case '?':
	yylval = install_and_save("?");
	return c;
    case '*':
	/* Replace ** by ^.  This has been here since 1998, but is
	   undocumented (at least in the obvious places).  It is in
	   the index of the Blue Book with a reference to p. 431, the
	   help for 'Deprecated'.  S-PLUS 6.2 still allowed this, so
	   presumably it was for compatibility with S. */
	if (nextchar('*')) {
	    yylval = install_and_save2("^", "**");
	    return '^';
	} else
	    yylval = install_and_save("*");
	return c;
    case '+':
    case '/':
    case '^':
    case '~':
    case '$':
    case '@':
    case '\\':
	yytext[0] = (char) c;
	yytext[1] = '\0';
	yylval = install(yytext);
	return c;
    default:
        yytext[0] = (char) c;
        yytext[1] = '\0';
	return c;
    }
}

/**
 * Sets the first elements of the yyloc structure with current 
 * information
 */
static void setfirstloc(void)
{
    yylloc.first_line   = ParseState.xxlineno;
    yylloc.first_column = ParseState.xxcolno;
    yylloc.first_byte   = ParseState.xxbyteno;
    yylloc.first_parsed = ParseState.xxparseno;
}

static void setlastloc(void)
{
    yylloc.last_line = ParseState.xxlineno;
    yylloc.last_column = ParseState.xxcolno;
    yylloc.last_byte = ParseState.xxbyteno;
    yylloc.last_parsed = ParseState.xxparseno;
}

/**
 * Wrap around the token function. Returns the same result
 * but increments the identifier, after a call to token_, 
 * the identifier variable contains the id of the token
 * just returned
 *
 * @return the same as token
 */

static int token_(void){
    // capture the position before retrieving the token
    setfirstloc( ) ;

    // get the token
    int res = token( ) ;

    // capture the position after
    int _last_col  = ParseState.xxcolno ;
    int _last_parsed = ParseState.xxparseno ;

    _current_token = res ;
    incrementId( ) ;
    yylloc.id = identifier ;

    // record the position
    if( res != '\n' && res != END_OF_INPUT)
	record_( yylloc.first_parsed, yylloc.first_column, 
	         _last_parsed, _last_col,
		res, identifier, yytext );

    return res; 
}


static int yylex(void)
{
    int tok;

 again:

    tok = token_();

    /* Newlines must be handled in a context */
    /* sensitive way.  The following block of */
    /* deals directly with newlines in the */
    /* body of "if" statements. */

    if (tok == '\n') {

	if (EatLines || *contextp == '[' || *contextp == '(')
	    goto again;

	/* The essence of this is that in the body of */
	/* an "if", any newline must be checked to */
	/* see if it is followed by an "else". */
	/* such newlines are discarded. */

	if (*contextp == 'i') {

	    /* Find the next non-newline token */

	    while(tok == '\n')
		tok = token_();

	    /* If we encounter "}", ")" or "]" then */
	    /* we know that all immediately preceding */
	    /* "if" bodies have been terminated. */
	    /* The corresponding "i" values are */
	    /* popped off the context stack. */

	    if (tok == RBRACE || tok == ')' || tok == ']' ) {
		while (*contextp == 'i')
		    ifpop();
		*contextp-- = 0;
		setlastloc();
		return tok;
	    }

	    /* When a "," is encountered, it terminates */
	    /* just the immediately preceding "if" body */
	    /* so we pop just a single "i" of the */
	    /* context stack. */

	    if (tok == ',') {
		ifpop();
		setlastloc();
		return tok;
	    }

	    /* Tricky! If we find an "else" we must */
	    /* ignore the preceding newline.  Any other */
	    /* token means that we must return the newline */
	    /* to terminate the "if" and "push back" that */
	    /* token so that we will obtain it on the next */
	    /* call to token.  In either case sensitivity */
	    /* is lost, so we pop the "i" from the context */
	    /* stack. */

	    if(tok == ELSE) {
		EatLines = 1;
		ifpop();
		setlastloc();
		return ELSE;
	    }
	    else {
		ifpop();
		SavedToken = tok;
		xxlinesave = yylloc.first_line;
		xxcolsave  = yylloc.first_column;
		xxbytesave = yylloc.first_byte;
		xxparsesave = yylloc.first_parsed;
		SavedLval = yylval;
		setlastloc();
		if (ParseState.keepSrcRefs && ParseState.keepParseData &&
		    yytext[0])

		    /* unrecord the pushed back token if not null */
		    ParseState.data_count--;
		return '\n';
	    }
	}
	else {
	    setlastloc();
	    return '\n';
	}
    }

    /* Additional context sensitivities */

    switch(tok) {

	/* Any newlines immediately following the */
	/* the following tokens are discarded. The */
	/* expressions are clearly incomplete. */

    case '+':
    case '-':
    case '*':
    case '/':
    case '^':
    case LT:
    case LE:
    case GE:
    case GT:
    case EQ:
    case NE:
    case OR:
    case AND:
    case OR2:
    case AND2:
    case PIPE:
    case PIPEBIND:
    case SPECIAL:
    case FUNCTION:
    case WHILE:
    case REPEAT:
    case FOR:
    case IN:
    case '?':
    case '!':
    case '=':
    case ':':
    case '~':
    case '$':
    case '@':
    case LEFT_ASSIGN:
    case RIGHT_ASSIGN:
    case EQ_ASSIGN:
	EatLines = 1;
	break;

	/* Push any "if" statements found and */
	/* discard any immediately following newlines. */

    case IF:
	IfPush();
	EatLines = 1;
	break;

	/* Terminate any immediately preceding "if" */
	/* statements and discard any immediately */
	/* following newlines. */

    case ELSE:
	ifpop();
	EatLines = 1;
	break;

	/* These tokens terminate any immediately */
	/* preceding "if" statements. */

    case ';':
    case ',':
	ifpop();
	break;

	/* Any newlines following these tokens can */
	/* indicate the end of an expression. */

    case SYMBOL:
    case PLACEHOLDER:
    case STR_CONST:
    case NUM_CONST:
    case NULL_CONST:
    case NEXT:
    case BREAK:
	EatLines = 0;
	break;

	/* Handle brackets, braces and parentheses */

    case LBB:
	if(contextp - contextstack >= CONTEXTSTACK_SIZE - 1)
	    error(_("contextstack overflow at line %d"), ParseState.xxlineno);
	*++contextp = '[';
	*++contextp = '[';
	break;

    case '[':
	if(contextp - contextstack >= CONTEXTSTACK_SIZE)
	    error(_("contextstack overflow at line %d"), ParseState.xxlineno);
	*++contextp = (char) tok;
	break;

    case LBRACE:
	if(contextp - contextstack >= CONTEXTSTACK_SIZE)
	    error(_("contextstack overflow at line %d"), ParseState.xxlineno);
	*++contextp = (char) tok;
	EatLines = 1;
	break;

    case '(':
	if(contextp - contextstack >= CONTEXTSTACK_SIZE)
	    error(_("contextstack overflow at line %d"), ParseState.xxlineno);
	*++contextp = (char) tok;
	break;

    case ']':
	while (*contextp == 'i')
	    ifpop();
	*contextp-- = 0;
	EatLines = 0;
	break;

    case RBRACE:
	while (*contextp == 'i')
	    ifpop();
	*contextp-- = 0;
	break;

    case ')':
	while (*contextp == 'i')
	    ifpop();
	*contextp-- = 0;
	EatLines = 0;
	break;

    }
    setlastloc();
    return tok;
}
/**
 * Records location information about a symbol. The information is
 * used to fill the data 
 * 
 */
static void record_( int first_parsed, int first_column, int last_parsed, int last_column,
	int token, int id, char* text_in ){
	
	if (!ParseState.keepSrcRefs || !ParseState.keepParseData
	    || id == NA_INTEGER) return;
	
	// don't care about zero sized things
	if( !yytext[0] ) return ;
	
	if (ParseState.data_count == DATA_COUNT)
	    growData();
	
	_FIRST_COLUMN( ParseState.data_count ) = first_column; 
	_FIRST_PARSED( ParseState.data_count ) = first_parsed;
	_LAST_COLUMN( ParseState.data_count )  = last_column;  
	_LAST_PARSED( ParseState.data_count )  = last_parsed; 
	_TOKEN( ParseState.data_count )        = token;        
	_ID( ParseState.data_count )           = id ;          
	_PARENT(ParseState.data_count)         = 0 ; 
	if ( text_in )
	    SET_STRING_ELT(PS_TEXT, ParseState.data_count, mkChar2(text_in));
	else
	    SET_STRING_ELT(PS_TEXT, ParseState.data_count, mkChar(""));
	
	if( id > ID_COUNT )
	    growID(id) ;

	ID_ID( id ) = ParseState.data_count ; 
	
	ParseState.data_count++ ;
}

/**
 * records parent as the parent of all its childs. This grows the 
 * parents list with a new vector. The first element of the new 
 * vector is the parent id, and other elements are childs id
 *
 * @param parent id of the parent expression
 * @param childs array of location information for all child symbols
 * @param nchilds number of childs
 */
static void recordParents( int parent, yyltype * childs, int nchilds){
	
	if( parent > ID_COUNT ){
		growID(parent) ;
	}
	
	/* some of the childs might be an empty token (like cr)
	   which we do not want to track */
	int ii;    /* loop index */
	yyltype loc ;
	for( ii=0; ii<nchilds; ii++){
		loc = childs[ii] ;
		if( loc.id == NA_INTEGER || (loc.first_line == loc.last_line && loc.first_byte > loc.last_byte) )
			continue ;
		/*  This shouldn't happen... */
		if (loc.id < 0 || loc.id > identifier) {
		    error(_("internal parser error at line %d"),  ParseState.xxlineno);
		}
		ID_PARENT( loc.id ) = parent;
	}
	
}

/**
 * The token pointed by the location has the wrong token type, 
 * This updates the type
 *
 * @param loc location information for the token to track
 */ 
static void modif_token( yyltype* loc, int tok ){
	
	int id = loc->id ;
	
	if (!ParseState.keepSrcRefs || !ParseState.keepParseData
	    || id < 0 || id > ID_COUNT) return;
	    
	if( tok == SYMBOL_FUNCTION_CALL ){
		// looking for first child of id
		int j = ID_ID( id ) ;
		int parent = id ;
		
		if (j < 0 || j > ID_COUNT)
	            return;
	            
		while( ID_PARENT( _ID(j) ) != parent ){
		    j-- ; 
		    if (j < 0)
	        	return;
		}
			
		if( _TOKEN(j) == SYMBOL ){
		    _TOKEN(j) = SYMBOL_FUNCTION_CALL ;
		}
		
	} else{
		_TOKEN( ID_ID(id) ) = tok ;
	}
	
}

/* this local version of lengthgets() always copies and doesn't fill with NA */
static SEXP lengthgets2(SEXP x, int len) {
    SEXP result;
    PROTECT(result = allocVector( TYPEOF(x), len ));
    
    len = (len < length(x)) ? len : length(x);
    switch(TYPEOF(x)) {
    	case INTSXP: 
    	    for (int i = 0; i < len; i++)
    	    	INTEGER(result)[i] = INTEGER(x)[i];
	    for (int i = len; i < length(result); i++)
		INTEGER(result)[i] = 0;
    	    break;
    	case STRSXP:
    	    for (int i = 0; i < len; i++)
    	    	SET_STRING_ELT(result, i, STRING_ELT(x, i));
    	    break;
    	default:
	    UNIMPLEMENTED_TYPE("lengthgets2", x);
    }
    UNPROTECT(1); /* result */
    return result;
}

static void finalizeData( ){
	
    int nloc = ParseState.data_count ;

    int i, j, id ;
    int parent ;

    /* store parents in the data */
    for( i=0; i<nloc; i++){
	id = _ID(i);
	parent = ID_PARENT( id ) ;
	while( parent != 0 && ID_ID(parent) == 0 )
	    parent = ID_PARENT( parent ) ;
	_PARENT(i) = parent ;

#define FD_FAST_UPDATE_PARENTS
#ifdef FD_FAST_UPDATE_PARENTS
	/*
	   With long generated expressions, updating the parents can take
	   a lot of time due to long chains of nodes not represented in the
	   parse data. To reduce the overhead somewhat, we create shortcuts
	   in the IDS array to point directly to the parent that is in the
	   parse data.
	*/
	int data_parent = parent;
	parent = ID_PARENT( id ) ;
	while( parent != data_parent ){
	    ID_PARENT( id ) = data_parent; /* set shortcut */
	    id = parent;
	    parent = ID_PARENT( parent );
	}
#endif
    }

    /* attach comments to closest enclosing symbol */
    /* not updating ID_PARENT anymore */

#define FD_FAST_ASSIGN_COMMENTS
#ifdef FD_FAST_ASSIGN_COMMENTS
    /*
       All terminals (tokens) are ordered by start and end location, including
       the comments, in the data.

       All non-terminals, including to be found parents of the comments, are
       ordered by their end location. When they have the same end location
       in the code, they are ordered by their decreasing start location
       (children before parents).

       All terminals and non-terminals are also before their parents (if any),
       so a comment is also befor its parent in the data.

       Consequently: the first non-terminal after a comment that encloses the
       comment is its (immediate) parent. The original algorithm for every
       comment linearly searches for the first enclosing non-terminal and
       returns it, but it has quadratic complexity and dominates the whole
       parsing for long inputs (used when FD_FAST_ASSIGN_COMMENTS is not
       defined).

       This algorithm uses the parental information available on nodes that
       follow the comments. That information has been filled by the parser
       during reductions (but not for comments, because those are not in the
       grammar). A node following a comment is either the parent of the
       comment, or some of its parents are, or is an orphan.

       Note that a non-terminal may end before a terminal (e.g. comment) in the
       code but be after the terminal in the data (due to look-ahead). It seems
       that the parent of the comment has to be within parents of the
       non-terminal as well, but I am not sure how to prove it, so the algorithm
       just skips non-terminals preceding the comment in the code (so is not
       strictly linear).
      */

    for(i = nloc-1; i >= 0; i--) {
	if (_TOKEN(i) == COMMENT) {
	    int orphan = 1;
	    int istartl = _FIRST_PARSED(i);
	    int istartc = _FIRST_COLUMN(i);

	    /* look for first node j that does not end before the comment i */
	    for(j = i + 1; j < nloc && _LAST_PARSED(j) <= istartl; j++);

	    if (j < nloc) {
		for(;;) {
		    int jstartl = _FIRST_PARSED(j);
		    int jstartc = _FIRST_COLUMN(j);

		    if (jstartl < istartl || (jstartl == istartl
		                              && jstartc <= istartc)) {
			/* j starts before or at the comment */
			_PARENT(i) = _ID(j);
			orphan = 0;
			break;
		    }
		    /* find parent of j */
		    int jparent = _PARENT(j);
		    if (jparent == 0)
			break; /* orphan */
		    j = ID_ID(jparent);
		}
	    }
	    if (orphan)
		_PARENT(i) = 0;
	}
    }
#else
    /* the original algorithm, which is slow for large inputs */

    int comment_line, comment_first_col;
    int this_first_parsed, this_last_parsed, this_first_col ;
    int orphan ;

    for( i=0; i<nloc; i++){
	if( _TOKEN(i) == COMMENT ){
	    comment_line = _FIRST_PARSED( i ) ;
	    comment_first_col = _FIRST_COLUMN( i ) ;

	    orphan = 1 ;
	    for( j=i+1; j<nloc; j++){
		this_first_parsed = _FIRST_PARSED( j ) ;
		this_first_col = _FIRST_COLUMN( j ) ;
		this_last_parsed  = _LAST_PARSED( j ) ;

		/* the comment needs to start after the current symbol */
		if( comment_line < this_first_parsed ) continue ;
		if( (comment_line == this_first_parsed) & (comment_first_col < this_first_col) ) continue ;

		/* the current symbol must finish after the comment */
		if( this_last_parsed <= comment_line ) continue ; 

		/* we have a match, record the parent and stop looking */
		_PARENT(i) = _ID(j);
		orphan = 0;
		break ;
	    }
	    if(orphan){
		_PARENT(i) = 0 ;
	    }
	}
    }
#endif


    /* now rework the parents of comments, we try to attach 
    comments that are not already attached (parent=0) to the next
    enclosing top-level expression */ 

    for( i=0; i<nloc; i++){
	int token = _TOKEN(i); 
	if( token == COMMENT && _PARENT(i) == 0 ){
	    for( j=i; j<nloc; j++){
		int token_j = _TOKEN(j); 
		if( token_j == COMMENT ) continue ;
		if( _PARENT(j) != 0 ) continue ;
		_PARENT(i) = - _ID(j) ;
		break ;
	    }
	}
    }

    /* attach the token names as an attribute so we don't need to switch to a dataframe, and decide on terminals */
    SEXP tokens;
    PROTECT(tokens = allocVector( STRSXP, nloc ) );
    for (int i=0; i<nloc; i++) {
        int token = _TOKEN(i);
        int xlat = yytranslate[token];
        if (xlat == 2) /* "unknown" */
            xlat = token;
        if (xlat < YYNTOKENS + YYNNTS)
    	    SET_STRING_ELT(tokens, i, mkChar(yytname[xlat]));
    	else { /* we have a token which doesn't have a name, e.g. an illegal character as in PR#15518 */
    	    char name[2];
    	    name[0] = (char) xlat;
    	    name[1] = 0;
    	    SET_STRING_ELT(tokens, i, mkChar(name));
    	}
    	_TERMINAL(i) = xlat < YYNTOKENS;
    }
    SEXP dims, newdata, newtext;
    if (nloc) {
	PROTECT( newdata = lengthgets2(PS_DATA, nloc * DATA_ROWS));
	PROTECT( newtext = lengthgets2(PS_TEXT, nloc));
    } else {
	PROTECT( newdata = allocVector( INTSXP, 0));
	PROTECT( newtext = allocVector( STRSXP, 0));
    }
    PROTECT( dims = allocVector( INTSXP, 2 ) ) ;
    INTEGER(dims)[0] = DATA_ROWS ;
    INTEGER(dims)[1] = nloc ;
    setAttrib( newdata, install( "dim" ), dims ) ;
    setAttrib( newdata, install("tokens"), tokens );
    setAttrib( newdata, install("text"), newtext );
    
    setAttrib(newdata, R_ClassSymbol, mkString("parseData"));
    
    /* Put it into the srcfile environment */
    if (isEnvironment(PS_SRCFILE))
	defineVar(install("parseData"), newdata, PS_SRCFILE);
    UNPROTECT(4); /* tokens, newdata, newtext, dims */
}

/**
 * Grows the data
 */
static void growData(){
	
    int new_data_count;	
    if (PS_DATA == R_NilValue) {
        new_data_count = INIT_DATA_COUNT;
	PS_SET_DATA(allocVector(INTSXP, 0));
	PS_SET_TEXT(allocVector(STRSXP, 0));
    } else
        new_data_count = 2*DATA_COUNT;
	
    PS_SET_DATA(lengthgets2(PS_DATA, new_data_count * DATA_ROWS));
    PS_SET_TEXT(lengthgets2(PS_TEXT, new_data_count));
}

/**
 * Grows the ids vector so that ID_ID(target) can be called
 */
static void growID( int target ){
	
    int new_count;
    if (PS_IDS == R_NilValue) {
        new_count = INIT_DATA_COUNT/2 - 1;
        PS_SET_IDS(allocVector(INTSXP, 0));
    } else
    	new_count = ID_COUNT;
    	
    while (target > new_count)
    	new_count = 2*new_count + 1;
    	
    if (new_count <= ID_COUNT)
    	return;
    
    int new_size = (1 + new_count)*2;
    PS_SET_IDS(lengthgets2(PS_IDS, new_size));
}

static int checkForPlaceholder(SEXP placeholder, SEXP arg)
{
    if (! HavePlaceholder)
    	return FALSE;
    else if (arg == placeholder)
	return TRUE;
    else if (TYPEOF(arg) == LANGSXP)
	for (SEXP cur = arg; cur != R_NilValue; cur = CDR(cur))
	    if (checkForPlaceholder(placeholder, CAR(cur)))
		return TRUE;
    return FALSE;
}