Google Git
Sign in
chromium / chromium / deps / sqlite / refs/heads/chromium-version-3.38.1 / . / tool / lemon.c
blob: 8dcf65179fd03ddeda74a5601d2ecc148a45a98f [file] [log] [blame] [edit]
/*
** This file contains all sources (including headers) to the LEMON
** LALR(1) parser generator. The sources have been combined into a
** single file to make it easy to include LEMON in the source tree
** and Makefile of another program.
**
** The author of this program disclaims copyright.
*/
#include <stdio.h>
#include <stdarg.h>
#include <string.h>
#include <ctype.h>
#include <stdlib.h>
#include <assert.h>
#define ISSPACE(X) isspace((unsigned char)(X))
#define ISDIGIT(X) isdigit((unsigned char)(X))
#define ISALNUM(X) isalnum((unsigned char)(X))
#define ISALPHA(X) isalpha((unsigned char)(X))
#define ISUPPER(X) isupper((unsigned char)(X))
#define ISLOWER(X) islower((unsigned char)(X))
#ifndef __WIN32__
# if defined(_WIN32) || defined(WIN32)
# define __WIN32__
# endif
#endif
#ifdef __WIN32__
#ifdef __cplusplus
extern "C" {
#endif
extern int access(const char *path, int mode);
#ifdef __cplusplus
}
#endif
#else
#include <unistd.h>
#endif
/* #define PRIVATE static */
#define PRIVATE
#ifdef TEST
#define MAXRHS 5 /* Set low to exercise exception code */
#else
#define MAXRHS 1000
#endif
extern void memory_error();
static int showPrecedenceConflict = 0;
static char *msort(char*,char**,int(*)(const char*,const char*));
/*
** Compilers are getting increasingly pedantic about type conversions
** as C evolves ever closer to Ada.... To work around the latest problems
** we have to define the following variant of strlen().
*/
#define lemonStrlen(X) ((int)strlen(X))
/*
** Compilers are starting to complain about the use of sprintf() and strcpy(),
** saying they are unsafe. So we define our own versions of those routines too.
**
** There are three routines here: lemon_sprintf(), lemon_vsprintf(), and
** lemon_addtext(). The first two are replacements for sprintf() and vsprintf().
** The third is a helper routine for vsnprintf() that adds texts to the end of a
** buffer, making sure the buffer is always zero-terminated.
**
** The string formatter is a minimal subset of stdlib sprintf() supporting only
** a few simply conversions:
**
** %d
** %s
** %.*s
**
*/
static void lemon_addtext(
char *zBuf, /* The buffer to which text is added */
int *pnUsed, /* Slots of the buffer used so far */
const char *zIn, /* Text to add */
int nIn, /* Bytes of text to add. -1 to use strlen() */
int iWidth /* Field width. Negative to left justify */
){
if( nIn<0 ) for(nIn=0; zIn[nIn]; nIn++){}
while( iWidth>nIn ){ zBuf[(*pnUsed)++] = ' '; iWidth--; }
if( nIn==0 ) return;
memcpy(&zBuf[*pnUsed], zIn, nIn);
*pnUsed += nIn;
while( (-iWidth)>nIn ){ zBuf[(*pnUsed)++] = ' '; iWidth++; }
zBuf[*pnUsed] = 0;
}
static int lemon_vsprintf(char *str, const char *zFormat, va_list ap){
int i, j, k, c;
int nUsed = 0;
const char *z;
char zTemp[50];
str[0] = 0;
for(i=j=0; (c = zFormat[i])!=0; i++){
if( c=='%' ){
int iWidth = 0;
lemon_addtext(str, &nUsed, &zFormat[j], i-j, 0);
c = zFormat[++i];
if( ISDIGIT(c) || (c=='-' && ISDIGIT(zFormat[i+1])) ){
if( c=='-' ) i++;
while( ISDIGIT(zFormat[i]) ) iWidth = iWidth*10 + zFormat[i++] - '0';
if( c=='-' ) iWidth = -iWidth;
c = zFormat[i];
}
if( c=='d' ){
int v = va_arg(ap, int);
if( v<0 ){
lemon_addtext(str, &nUsed, "-", 1, iWidth);
v = -v;
}else if( v==0 ){
lemon_addtext(str, &nUsed, "0", 1, iWidth);
}
k = 0;
while( v>0 ){
k++;
zTemp[sizeof(zTemp)-k] = (v%10) + '0';
v /= 10;
}
lemon_addtext(str, &nUsed, &zTemp[sizeof(zTemp)-k], k, iWidth);
}else if( c=='s' ){
z = va_arg(ap, const char*);
lemon_addtext(str, &nUsed, z, -1, iWidth);
}else if( c=='.' && memcmp(&zFormat[i], ".*s", 3)==0 ){
i += 2;
k = va_arg(ap, int);
z = va_arg(ap, const char*);
lemon_addtext(str, &nUsed, z, k, iWidth);
}else if( c=='%' ){
lemon_addtext(str, &nUsed, "%", 1, 0);
}else{
fprintf(stderr, "illegal format\n");
exit(1);
}
j = i+1;
}
}
lemon_addtext(str, &nUsed, &zFormat[j], i-j, 0);
return nUsed;
}
static int lemon_sprintf(char *str, const char *format, ...){
va_list ap;
int rc;
va_start(ap, format);
rc = lemon_vsprintf(str, format, ap);
va_end(ap);
return rc;
}
static void lemon_strcpy(char *dest, const char *src){
while( (*(dest++) = *(src++))!=0 ){}
}
static void lemon_strcat(char *dest, const char *src){
while( *dest ) dest++;
lemon_strcpy(dest, src);
}
/* a few forward declarations... */
struct rule;
struct lemon;
struct action;
static struct action *Action_new(void);
static struct action *Action_sort(struct action *);
/********** From the file "build.h" ************************************/
void FindRulePrecedences(struct lemon*);
void FindFirstSets(struct lemon*);
void FindStates(struct lemon*);
void FindLinks(struct lemon*);
void FindFollowSets(struct lemon*);
void FindActions(struct lemon*);
/********* From the file "configlist.h" *********************************/
void Configlist_init(void);
struct config *Configlist_add(struct rule *, int);
struct config *Configlist_addbasis(struct rule *, int);
void Configlist_closure(struct lemon *);
void Configlist_sort(void);
void Configlist_sortbasis(void);
struct config *Configlist_return(void);
struct config *Configlist_basis(void);
void Configlist_eat(struct config *);
void Configlist_reset(void);
/********* From the file "error.h" ***************************************/
void ErrorMsg(const char *, int,const char *, ...);
/****** From the file "option.h" ******************************************/
enum option_type { OPT_FLAG=1, OPT_INT, OPT_DBL, OPT_STR,
OPT_FFLAG, OPT_FINT, OPT_FDBL, OPT_FSTR};
struct s_options {
enum option_type type;
const char *label;
char *arg;
const char *message;
};
int OptInit(char**,struct s_options*,FILE*);
int OptNArgs(void);
char *OptArg(int);
void OptErr(int);
void OptPrint(void);
/******** From the file "parse.h" *****************************************/
void Parse(struct lemon *lemp);
/********* From the file "plink.h" ***************************************/
struct plink *Plink_new(void);
void Plink_add(struct plink **, struct config *);
void Plink_copy(struct plink **, struct plink *);
void Plink_delete(struct plink *);
/********** From the file "report.h" *************************************/
void Reprint(struct lemon *);
void ReportOutput(struct lemon *);
void ReportTable(struct lemon *, int, int);
void ReportHeader(struct lemon *);
void CompressTables(struct lemon *);
void ResortStates(struct lemon *);
/********** From the file "set.h" ****************************************/
void SetSize(int); /* All sets will be of size N */
char *SetNew(void); /* A new set for element 0..N */
void SetFree(char*); /* Deallocate a set */
int SetAdd(char*,int); /* Add element to a set */
int SetUnion(char *,char *); /* A <- A U B, thru element N */
#define SetFind(X,Y) (X[Y]) /* True if Y is in set X */
/********** From the file "struct.h" *************************************/
/*
** Principal data structures for the LEMON parser generator.
*/
typedef enum {LEMON_FALSE=0, LEMON_TRUE} Boolean;
/* Symbols (terminals and nonterminals) of the grammar are stored
** in the following: */
enum symbol_type {
TERMINAL,
NONTERMINAL,
MULTITERMINAL
};
enum e_assoc {
LEFT,
RIGHT,
NONE,
UNK
};
struct symbol {
const char *name; /* Name of the symbol */
int index; /* Index number for this symbol */
enum symbol_type type; /* Symbols are all either TERMINALS or NTs */
struct rule *rule; /* Linked list of rules of this (if an NT) */
struct symbol *fallback; /* fallback token in case this token doesn't parse */
int prec; /* Precedence if defined (-1 otherwise) */
enum e_assoc assoc; /* Associativity if precedence is defined */
char *firstset; /* First-set for all rules of this symbol */
Boolean lambda; /* True if NT and can generate an empty string */
int useCnt; /* Number of times used */
char *destructor; /* Code which executes whenever this symbol is
** popped from the stack during error processing */
int destLineno; /* Line number for start of destructor. Set to
** -1 for duplicate destructors. */
char *datatype; /* The data type of information held by this
** object. Only used if type==NONTERMINAL */
int dtnum; /* The data type number. In the parser, the value
** stack is a union. The .yy%d element of this
** union is the correct data type for this object */
int bContent; /* True if this symbol ever carries content - if
** it is ever more than just syntax */
/* The following fields are used by MULTITERMINALs only */
int nsubsym; /* Number of constituent symbols in the MULTI */
struct symbol **subsym; /* Array of constituent symbols */
};
/* Each production rule in the grammar is stored in the following
** structure. */
struct rule {
struct symbol *lhs; /* Left-hand side of the rule */
const char *lhsalias; /* Alias for the LHS (NULL if none) */
int lhsStart; /* True if left-hand side is the start symbol */
int ruleline; /* Line number for the rule */
int nrhs; /* Number of RHS symbols */
struct symbol **rhs; /* The RHS symbols */
const char **rhsalias; /* An alias for each RHS symbol (NULL if none) */
int line; /* Line number at which code begins */
const char *code; /* The code executed when this rule is reduced */
const char *codePrefix; /* Setup code before code[] above */
const char *codeSuffix; /* Breakdown code after code[] above */
struct symbol *precsym; /* Precedence symbol for this rule */
int index; /* An index number for this rule */
int iRule; /* Rule number as used in the generated tables */
Boolean noCode; /* True if this rule has no associated C code */
Boolean codeEmitted; /* True if the code has been emitted already */
Boolean canReduce; /* True if this rule is ever reduced */
Boolean doesReduce; /* Reduce actions occur after optimization */
Boolean neverReduce; /* Reduce is theoretically possible, but prevented
** by actions or other outside implementation */
struct rule *nextlhs; /* Next rule with the same LHS */
struct rule *next; /* Next rule in the global list */
};
/* A configuration is a production rule of the grammar together with
** a mark (dot) showing how much of that rule has been processed so far.
** Configurations also contain a follow-set which is a list of terminal
** symbols which are allowed to immediately follow the end of the rule.
** Every configuration is recorded as an instance of the following: */
enum cfgstatus {
COMPLETE,
INCOMPLETE
};
struct config {
struct rule *rp; /* The rule upon which the configuration is based */
int dot; /* The parse point */
char *fws; /* Follow-set for this configuration only */
struct plink *fplp; /* Follow-set forward propagation links */
struct plink *bplp; /* Follow-set backwards propagation links */
struct state *stp; /* Pointer to state which contains this */
enum cfgstatus status; /* used during followset and shift computations */
struct config *next; /* Next configuration in the state */
struct config *bp; /* The next basis configuration */
};
enum e_action {
SHIFT,
ACCEPT,
REDUCE,
ERROR,
SSCONFLICT, /* A shift/shift conflict */
SRCONFLICT, /* Was a reduce, but part of a conflict */
RRCONFLICT, /* Was a reduce, but part of a conflict */
SH_RESOLVED, /* Was a shift. Precedence resolved conflict */
RD_RESOLVED, /* Was reduce. Precedence resolved conflict */
NOT_USED, /* Deleted by compression */
SHIFTREDUCE /* Shift first, then reduce */
};
/* Every shift or reduce operation is stored as one of the following */
struct action {
struct symbol *sp; /* The look-ahead symbol */
enum e_action type;
union {
struct state *stp; /* The new state, if a shift */
struct rule *rp; /* The rule, if a reduce */
} x;
struct symbol *spOpt; /* SHIFTREDUCE optimization to this symbol */
struct action *next; /* Next action for this state */
struct action *collide; /* Next action with the same hash */
};
/* Each state of the generated parser's finite state machine
** is encoded as an instance of the following structure. */
struct state {
struct config *bp; /* The basis configurations for this state */
struct config *cfp; /* All configurations in this set */
int statenum; /* Sequential number for this state */
struct action *ap; /* List of actions for this state */
int nTknAct, nNtAct; /* Number of actions on terminals and nonterminals */
int iTknOfst, iNtOfst; /* yy_action[] offset for terminals and nonterms */
int iDfltReduce; /* Default action is to REDUCE by this rule */
struct rule *pDfltReduce;/* The default REDUCE rule. */
int autoReduce; /* True if this is an auto-reduce state */
};
#define NO_OFFSET (-2147483647)
/* A followset propagation link indicates that the contents of one
** configuration followset should be propagated to another whenever
** the first changes. */
struct plink {
struct config *cfp; /* The configuration to which linked */
struct plink *next; /* The next propagate link */
};
/* The state vector for the entire parser generator is recorded as
** follows. (LEMON uses no global variables and makes little use of
** static variables. Fields in the following structure can be thought
** of as begin global variables in the program.) */
struct lemon {
struct state **sorted; /* Table of states sorted by state number */
struct rule *rule; /* List of all rules */
struct rule *startRule; /* First rule */
int nstate; /* Number of states */
int nxstate; /* nstate with tail degenerate states removed */
int nrule; /* Number of rules */
int nruleWithAction; /* Number of rules with actions */
int nsymbol; /* Number of terminal and nonterminal symbols */
int nterminal; /* Number of terminal symbols */
int minShiftReduce; /* Minimum shift-reduce action value */
int errAction; /* Error action value */
int accAction; /* Accept action value */
int noAction; /* No-op action value */
int minReduce; /* Minimum reduce action */
int maxAction; /* Maximum action value of any kind */
struct symbol **symbols; /* Sorted array of pointers to symbols */
int errorcnt; /* Number of errors */
struct symbol *errsym; /* The error symbol */
struct symbol *wildcard; /* Token that matches anything */
char *name; /* Name of the generated parser */
char *arg; /* Declaration of the 3th argument to parser */
char *ctx; /* Declaration of 2nd argument to constructor */
char *tokentype; /* Type of terminal symbols in the parser stack */
char *vartype; /* The default type of non-terminal symbols */
char *start; /* Name of the start symbol for the grammar */
char *stacksize; /* Size of the parser stack */
char *include; /* Code to put at the start of the C file */
char *error; /* Code to execute when an error is seen */
char *overflow; /* Code to execute on a stack overflow */
char *failure; /* Code to execute on parser failure */
char *accept; /* Code to execute when the parser excepts */
char *extracode; /* Code appended to the generated file */
char *tokendest; /* Code to execute to destroy token data */
char *vardest; /* Code for the default non-terminal destructor */
char *filename; /* Name of the input file */
char *outname; /* Name of the current output file */
char *tokenprefix; /* A prefix added to token names in the .h file */
int nconflict; /* Number of parsing conflicts */
int nactiontab; /* Number of entries in the yy_action[] table */
int nlookaheadtab; /* Number of entries in yy_lookahead[] */
int tablesize; /* Total table size of all tables in bytes */
int basisflag; /* Print only basis configurations */
int has_fallback; /* True if any %fallback is seen in the grammar */
int nolinenosflag; /* True if #line statements should not be printed */
char *argv0; /* Name of the program */
};
#define MemoryCheck(X) if((X)==0){ \
extern void memory_error(); \
memory_error(); \
}
/**************** From the file "table.h" *********************************/
/*
** All code in this file has been automatically generated
** from a specification in the file
** "table.q"
** by the associative array code building program "aagen".
** Do not edit this file! Instead, edit the specification
** file, then rerun aagen.
*/
/*
** Code for processing tables in the LEMON parser generator.
*/
/* Routines for handling a strings */
const char *Strsafe(const char *);
void Strsafe_init(void);
int Strsafe_insert(const char *);
const char *Strsafe_find(const char *);
/* Routines for handling symbols of the grammar */
struct symbol *Symbol_new(const char *);
int Symbolcmpp(const void *, const void *);
void Symbol_init(void);
int Symbol_insert(struct symbol *, const char *);
struct symbol *Symbol_find(const char *);
struct symbol *Symbol_Nth(int);
int Symbol_count(void);
struct symbol **Symbol_arrayof(void);
/* Routines to manage the state table */
int Configcmp(const char *, const char *);
struct state *State_new(void);
void State_init(void);
int State_insert(struct state *, struct config *);
struct state *State_find(struct config *);
struct state **State_arrayof(void);
/* Routines used for efficiency in Configlist_add */
void Configtable_init(void);
int Configtable_insert(struct config *);
struct config *Configtable_find(struct config *);
void Configtable_clear(int(*)(struct config *));
/****************** From the file "action.c" *******************************/
/*
** Routines processing parser actions in the LEMON parser generator.
*/
/* Allocate a new parser action */
static struct action *Action_new(void){
static struct action *actionfreelist = 0;
struct action *newaction;
if( actionfreelist==0 ){
int i;
int amt = 100;
actionfreelist = (struct action *)calloc(amt, sizeof(struct action));
if( actionfreelist==0 ){
fprintf(stderr,"Unable to allocate memory for a new parser action.");
exit(1);
}
for(i=0; i<amt-1; i++) actionfreelist[i].next = &actionfreelist[i+1];
actionfreelist[amt-1].next = 0;
}
newaction = actionfreelist;
actionfreelist = actionfreelist->next;
return newaction;
}
/* Compare two actions for sorting purposes. Return negative, zero, or
** positive if the first action is less than, equal to, or greater than
** the first
*/
static int actioncmp(
struct action *ap1,
struct action *ap2
){
int rc;
rc = ap1->sp->index - ap2->sp->index;
if( rc==0 ){
rc = (int)ap1->type - (int)ap2->type;
}
if( rc==0 && (ap1->type==REDUCE || ap1->type==SHIFTREDUCE) ){
rc = ap1->x.rp->index - ap2->x.rp->index;
}
if( rc==0 ){
rc = (int) (ap2 - ap1);
}
return rc;
}
/* Sort parser actions */
static struct action *Action_sort(
struct action *ap
){
ap = (struct action *)msort((char *)ap,(char **)&ap->next,
(int(*)(const char*,const char*))actioncmp);
return ap;
}
void Action_add(
struct action **app,
enum e_action type,
struct symbol *sp,
char *arg
){
struct action *newaction;
newaction = Action_new();
newaction->next = *app;
*app = newaction;
newaction->type = type;
newaction->sp = sp;
newaction->spOpt = 0;
if( type==SHIFT ){
newaction->x.stp = (struct state *)arg;
}else{
newaction->x.rp = (struct rule *)arg;
}
}
/********************** New code to implement the "acttab" module ***********/
/*
** This module implements routines use to construct the yy_action[] table.
*/
/*
** The state of the yy_action table under construction is an instance of
** the following structure.
**
** The yy_action table maps the pair (state_number, lookahead) into an
** action_number. The table is an array of integers pairs. The state_number
** determines an initial offset into the yy_action array. The lookahead
** value is then added to this initial offset to get an index X into the
** yy_action array. If the aAction[X].lookahead equals the value of the
** of the lookahead input, then the value of the action_number output is
** aAction[X].action. If the lookaheads do not match then the
** default action for the state_number is returned.
**
** All actions associated with a single state_number are first entered
** into aLookahead[] using multiple calls to acttab_action(). Then the
** actions for that single state_number are placed into the aAction[]
** array with a single call to acttab_insert(). The acttab_insert() call
** also resets the aLookahead[] array in preparation for the next
** state number.
*/
struct lookahead_action {
int lookahead; /* Value of the lookahead token */
int action; /* Action to take on the given lookahead */
};
typedef struct acttab acttab;
struct acttab {
int nAction; /* Number of used slots in aAction[] */
int nActionAlloc; /* Slots allocated for aAction[] */
struct lookahead_action
*aAction, /* The yy_action[] table under construction */
*aLookahead; /* A single new transaction set */
int mnLookahead; /* Minimum aLookahead[].lookahead */
int mnAction; /* Action associated with mnLookahead */
int mxLookahead; /* Maximum aLookahead[].lookahead */
int nLookahead; /* Used slots in aLookahead[] */
int nLookaheadAlloc; /* Slots allocated in aLookahead[] */
int nterminal; /* Number of terminal symbols */
int nsymbol; /* total number of symbols */
};
/* Return the number of entries in the yy_action table */
#define acttab_lookahead_size(X) ((X)->nAction)
/* The value for the N-th entry in yy_action */
#define acttab_yyaction(X,N) ((X)->aAction[N].action)
/* The value for the N-th entry in yy_lookahead */
#define acttab_yylookahead(X,N) ((X)->aAction[N].lookahead)
/* Free all memory associated with the given acttab */
void acttab_free(acttab *p){
free( p->aAction );
free( p->aLookahead );
free( p );
}
/* Allocate a new acttab structure */
acttab *acttab_alloc(int nsymbol, int nterminal){
acttab *p = (acttab *) calloc( 1, sizeof(*p) );
if( p==0 ){
fprintf(stderr,"Unable to allocate memory for a new acttab.");
exit(1);
}
memset(p, 0, sizeof(*p));
p->nsymbol = nsymbol;
p->nterminal = nterminal;
return p;
}
/* Add a new action to the current transaction set.
**
** This routine is called once for each lookahead for a particular
** state.
*/
void acttab_action(acttab *p, int lookahead, int action){
if( p->nLookahead>=p->nLookaheadAlloc ){
p->nLookaheadAlloc += 25;
p->aLookahead = (struct lookahead_action *) realloc( p->aLookahead,
sizeof(p->aLookahead[0])*p->nLookaheadAlloc );
if( p->aLookahead==0 ){
fprintf(stderr,"malloc failed\n");
exit(1);
}
}
if( p->nLookahead==0 ){
p->mxLookahead = lookahead;
p->mnLookahead = lookahead;
p->mnAction = action;
}else{
if( p->mxLookahead<lookahead ) p->mxLookahead = lookahead;
if( p->mnLookahead>lookahead ){
p->mnLookahead = lookahead;
p->mnAction = action;
}
}
p->aLookahead[p->nLookahead].lookahead = lookahead;
p->aLookahead[p->nLookahead].action = action;
p->nLookahead++;
}
/*
** Add the transaction set built up with prior calls to acttab_action()
** into the current action table. Then reset the transaction set back
** to an empty set in preparation for a new round of acttab_action() calls.
**
** Return the offset into the action table of the new transaction.
**
** If the makeItSafe parameter is true, then the offset is chosen so that
** it is impossible to overread the yy_lookaside[] table regardless of
** the lookaside token. This is done for the terminal symbols, as they
** come from external inputs and can contain syntax errors. When makeItSafe
** is false, there is more flexibility in selecting offsets, resulting in
** a smaller table. For non-terminal symbols, which are never syntax errors,
** makeItSafe can be false.
*/
int acttab_insert(acttab *p, int makeItSafe){
int i, j, k, n, end;
assert( p->nLookahead>0 );
/* Make sure we have enough space to hold the expanded action table
** in the worst case. The worst case occurs if the transaction set
** must be appended to the current action table
*/
n = p->nsymbol + 1;
if( p->nAction + n >= p->nActionAlloc ){
int oldAlloc = p->nActionAlloc;
p->nActionAlloc = p->nAction + n + p->nActionAlloc + 20;
p->aAction = (struct lookahead_action *) realloc( p->aAction,
sizeof(p->aAction[0])*p->nActionAlloc);
if( p->aAction==0 ){
fprintf(stderr,"malloc failed\n");
exit(1);
}
for(i=oldAlloc; i<p->nActionAlloc; i++){
p->aAction[i].lookahead = -1;
p->aAction[i].action = -1;
}
}
/* Scan the existing action table looking for an offset that is a
** duplicate of the current transaction set. Fall out of the loop
** if and when the duplicate is found.
**
** i is the index in p->aAction[] where p->mnLookahead is inserted.
*/
end = makeItSafe ? p->mnLookahead : 0;
for(i=p->nAction-1; i>=end; i--){
if( p->aAction[i].lookahead==p->mnLookahead ){
/* All lookaheads and actions in the aLookahead[] transaction
** must match against the candidate aAction[i] entry. */
if( p->aAction[i].action!=p->mnAction ) continue;
for(j=0; j<p->nLookahead; j++){
k = p->aLookahead[j].lookahead - p->mnLookahead + i;
if( k<0 || k>=p->nAction ) break;
if( p->aLookahead[j].lookahead!=p->aAction[k].lookahead ) break;
if( p->aLookahead[j].action!=p->aAction[k].action ) break;
}
if( j<p->nLookahead ) continue;
/* No possible lookahead value that is not in the aLookahead[]
** transaction is allowed to match aAction[i] */
n = 0;
for(j=0; j<p->nAction; j++){
if( p->aAction[j].lookahead<0 ) continue;
if( p->aAction[j].lookahead==j+p->mnLookahead-i ) n++;
}
if( n==p->nLookahead ){
break; /* An exact match is found at offset i */
}
}
}
/* If no existing offsets exactly match the current transaction, find an
** an empty offset in the aAction[] table in which we can add the
** aLookahead[] transaction.
*/
if( i<end ){
/* Look for holes in the aAction[] table that fit the current
** aLookahead[] transaction. Leave i set to the offset of the hole.
** If no holes are found, i is left at p->nAction, which means the
** transaction will be appended. */
i = makeItSafe ? p->mnLookahead : 0;
for(; i<p->nActionAlloc - p->mxLookahead; i++){
if( p->aAction[i].lookahead<0 ){
for(j=0; j<p->nLookahead; j++){
k = p->aLookahead[j].lookahead - p->mnLookahead + i;
if( k<0 ) break;
if( p->aAction[k].lookahead>=0 ) break;
}
if( j<p->nLookahead ) continue;
for(j=0; j<p->nAction; j++){
if( p->aAction[j].lookahead==j+p->mnLookahead-i ) break;
}
if( j==p->nAction ){
break; /* Fits in empty slots */
}
}
}
}
/* Insert transaction set at index i. */
#if 0
printf("Acttab:");
for(j=0; j<p->nLookahead; j++){
printf(" %d", p->aLookahead[j].lookahead);
}
printf(" inserted at %d\n", i);
#endif
for(j=0; j<p->nLookahead; j++){
k = p->aLookahead[j].lookahead - p->mnLookahead + i;
p->aAction[k] = p->aLookahead[j];
if( k>=p->nAction ) p->nAction = k+1;
}
if( makeItSafe && i+p->nterminal>=p->nAction ) p->nAction = i+p->nterminal+1;
p->nLookahead = 0;
/* Return the offset that is added to the lookahead in order to get the
** index into yy_action of the action */
return i - p->mnLookahead;
}
/*
** Return the size of the action table without the trailing syntax error
** entries.
*/
int acttab_action_size(acttab *p){
int n = p->nAction;
while( n>0 && p->aAction[n-1].lookahead<0 ){ n--; }
return n;
}
/********************** From the file "build.c" *****************************/
/*
** Routines to construction the finite state machine for the LEMON
** parser generator.
*/
/* Find a precedence symbol of every rule in the grammar.
**
** Those rules which have a precedence symbol coded in the input
** grammar using the "[symbol]" construct will already have the
** rp->precsym field filled. Other rules take as their precedence
** symbol the first RHS symbol with a defined precedence. If there
** are not RHS symbols with a defined precedence, the precedence
** symbol field is left blank.
*/
void FindRulePrecedences(struct lemon *xp)
{
struct rule *rp;
for(rp=xp->rule; rp; rp=rp->next){
if( rp->precsym==0 ){
int i, j;
for(i=0; i<rp->nrhs && rp->precsym==0; i++){
struct symbol *sp = rp->rhs[i];
if( sp->type==MULTITERMINAL ){
for(j=0; j<sp->nsubsym; j++){
if( sp->subsym[j]->prec>=0 ){
rp->precsym = sp->subsym[j];
break;
}
}
}else if( sp->prec>=0 ){
rp->precsym = rp->rhs[i];
}
}
}
}
return;
}
/* Find all nonterminals which will generate the empty string.
** Then go back and compute the first sets of every nonterminal.
** The first set is the set of all terminal symbols which can begin
** a string generated by that nonterminal.
*/
void FindFirstSets(struct lemon *lemp)
{
int i, j;
struct rule *rp;
int progress;
for(i=0; i<lemp->nsymbol; i++){
lemp->symbols[i]->lambda = LEMON_FALSE;
}
for(i=lemp->nterminal; i<lemp->nsymbol; i++){
lemp->symbols[i]->firstset = SetNew();
}
/* First compute all lambdas */
do{
progress = 0;
for(rp=lemp->rule; rp; rp=rp->next){
if( rp->lhs->lambda ) continue;
for(i=0; i<rp->nrhs; i++){
struct symbol *sp = rp->rhs[i];
assert( sp->type==NONTERMINAL || sp->lambda==LEMON_FALSE );
if( sp->lambda==LEMON_FALSE ) break;
}
if( i==rp->nrhs ){
rp->lhs->lambda = LEMON_TRUE;
progress = 1;
}
}
}while( progress );
/* Now compute all first sets */
do{
struct symbol *s1, *s2;
progress = 0;
for(rp=lemp->rule; rp; rp=rp->next){
s1 = rp->lhs;
for(i=0; i<rp->nrhs; i++){
s2 = rp->rhs[i];
if( s2->type==TERMINAL ){
progress += SetAdd(s1->firstset,s2->index);
break;
}else if( s2->type==MULTITERMINAL ){
for(j=0; j<s2->nsubsym; j++){
progress += SetAdd(s1->firstset,s2->subsym[j]->index);
}
break;
}else if( s1==s2 ){
if( s1->lambda==LEMON_FALSE ) break;
}else{
progress += SetUnion(s1->firstset,s2->firstset);
if( s2->lambda==LEMON_FALSE ) break;
}
}
}
}while( progress );
return;
}
/* Compute all LR(0) states for the grammar. Links
** are added to between some states so that the LR(1) follow sets
** can be computed later.
*/
PRIVATE struct state *getstate(struct lemon *); /* forward reference */
void FindStates(struct lemon *lemp)
{
struct symbol *sp;
struct rule *rp;
Configlist_init();
/* Find the start symbol */
if( lemp->start ){
sp = Symbol_find(lemp->start);
if( sp==0 ){
ErrorMsg(lemp->filename,0,
"The specified start symbol \"%s\" is not "
"in a nonterminal of the grammar. \"%s\" will be used as the start "
"symbol instead.",lemp->start,lemp->startRule->lhs->name);
lemp->errorcnt++;
sp = lemp->startRule->lhs;
}
}else{
sp = lemp->startRule->lhs;
}
/* Make sure the start symbol doesn't occur on the right-hand side of
** any rule. Report an error if it does. (YACC would generate a new
** start symbol in this case.) */
for(rp=lemp->rule; rp; rp=rp->next){
int i;
for(i=0; i<rp->nrhs; i++){
if( rp->rhs[i]==sp ){ /* FIX ME: Deal with multiterminals */
ErrorMsg(lemp->filename,0,
"The start symbol \"%s\" occurs on the "
"right-hand side of a rule. This will result in a parser which "
"does not work properly.",sp->name);
lemp->errorcnt++;
}
}
}
/* The basis configuration set for the first state
** is all rules which have the start symbol as their
** left-hand side */
for(rp=sp->rule; rp; rp=rp->nextlhs){
struct config *newcfp;
rp->lhsStart = 1;
newcfp = Configlist_addbasis(rp,0);
SetAdd(newcfp->fws,0);
}
/* Compute the first state. All other states will be
** computed automatically during the computation of the first one.
** The returned pointer to the first state is not used. */
(void)getstate(lemp);
return;
}
/* Return a pointer to a state which is described by the configuration
** list which has been built from calls to Configlist_add.
*/
PRIVATE void buildshifts(struct lemon *, struct state *); /* Forwd ref */
PRIVATE struct state *getstate(struct lemon *lemp)
{
struct config *cfp, *bp;
struct state *stp;
/* Extract the sorted basis of the new state. The basis was constructed
** by prior calls to "Configlist_addbasis()". */
Configlist_sortbasis();
bp = Configlist_basis();
/* Get a state with the same basis */
stp = State_find(bp);
if( stp ){
/* A state with the same basis already exists! Copy all the follow-set
** propagation links from the state under construction into the
** preexisting state, then return a pointer to the preexisting state */
struct config *x, *y;
for(x=bp, y=stp->bp; x && y; x=x->bp, y=y->bp){
Plink_copy(&y->bplp,x->bplp);
Plink_delete(x->fplp);
x->fplp = x->bplp = 0;
}
cfp = Configlist_return();
Configlist_eat(cfp);
}else{
/* This really is a new state. Construct all the details */
Configlist_closure(lemp); /* Compute the configuration closure */
Configlist_sort(); /* Sort the configuration closure */
cfp = Configlist_return(); /* Get a pointer to the config list */
stp = State_new(); /* A new state structure */
MemoryCheck(stp);
stp->bp = bp; /* Remember the configuration basis */
stp->cfp = cfp; /* Remember the configuration closure */
stp->statenum = lemp->nstate++; /* Every state gets a sequence number */
stp->ap = 0; /* No actions, yet. */
State_insert(stp,stp->bp); /* Add to the state table */
buildshifts(lemp,stp); /* Recursively compute successor states */
}
return stp;
}
/*
** Return true if two symbols are the same.
*/
int same_symbol(struct symbol *a, struct symbol *b)
{
int i;
if( a==b ) return 1;
if( a->type!=MULTITERMINAL ) return 0;
if( b->type!=MULTITERMINAL ) return 0;
if( a->nsubsym!=b->nsubsym ) return 0;
for(i=0; i<a->nsubsym; i++){
if( a->subsym[i]!=b->subsym[i] ) return 0;
}
return 1;
}
/* Construct all successor states to the given state. A "successor"
** state is any state which can be reached by a shift action.
*/
PRIVATE void buildshifts(struct lemon *lemp, struct state *stp)
{
struct config *cfp; /* For looping thru the config closure of "stp" */
struct config *bcfp; /* For the inner loop on config closure of "stp" */
struct config *newcfg; /* */
struct symbol *sp; /* Symbol following the dot in configuration "cfp" */
struct symbol *bsp; /* Symbol following the dot in configuration "bcfp" */
struct state *newstp; /* A pointer to a successor state */
/* Each configuration becomes complete after it contibutes to a successor
** state. Initially, all configurations are incomplete */
for(cfp=stp->cfp; cfp; cfp=cfp->next) cfp->status = INCOMPLETE;
/* Loop through all configurations of the state "stp" */
for(cfp=stp->cfp; cfp; cfp=cfp->next){
if( cfp->status==COMPLETE ) continue; /* Already used by inner loop */
if( cfp->dot>=cfp->rp->nrhs ) continue; /* Can't shift this config */
Configlist_reset(); /* Reset the new config set */
sp = cfp->rp->rhs[cfp->dot]; /* Symbol after the dot */
/* For every configuration in the state "stp" which has the symbol "sp"
** following its dot, add the same configuration to the basis set under
** construction but with the dot shifted one symbol to the right. */
for(bcfp=cfp; bcfp; bcfp=bcfp->next){
if( bcfp->status==COMPLETE ) continue; /* Already used */
if( bcfp->dot>=bcfp->rp->nrhs ) continue; /* Can't shift this one */
bsp = bcfp->rp->rhs[bcfp->dot]; /* Get symbol after dot */
if( !same_symbol(bsp,sp) ) continue; /* Must be same as for "cfp" */
bcfp->status = COMPLETE; /* Mark this config as used */
newcfg = Configlist_addbasis(bcfp->rp,bcfp->dot+1);
Plink_add(&newcfg->bplp,bcfp);
}
/* Get a pointer to the state described by the basis configuration set
** constructed in the preceding loop */
newstp = getstate(lemp);
/* The state "newstp" is reached from the state "stp" by a shift action
** on the symbol "sp" */
if( sp->type==MULTITERMINAL ){
int i;
for(i=0; i<sp->nsubsym; i++){
Action_add(&stp->ap,SHIFT,sp->subsym[i],(char*)newstp);
}
}else{
Action_add(&stp->ap,SHIFT,sp,(char *)newstp);
}
}
}
/*
** Construct the propagation links
*/
void FindLinks(struct lemon *lemp)
{
int i;
struct config *cfp, *other;
struct state *stp;
struct plink *plp;
/* Housekeeping detail:
** Add to every propagate link a pointer back to the state to
** which the link is attached. */
for(i=0; i<lemp->nstate; i++){
stp = lemp->sorted[i];
for(cfp=stp->cfp; cfp; cfp=cfp->next){
cfp->stp = stp;
}
}
/* Convert all backlinks into forward links. Only the forward
** links are used in the follow-set computation. */
for(i=0; i<lemp->nstate; i++){
stp = lemp->sorted[i];
for(cfp=stp->cfp; cfp; cfp=cfp->next){
for(plp=cfp->bplp; plp; plp=plp->next){
other = plp->cfp;
Plink_add(&other->fplp,cfp);
}
}
}
}
/* Compute all followsets.
**
** A followset is the set of all symbols which can come immediately
** after a configuration.
*/
void FindFollowSets(struct lemon *lemp)
{
int i;
struct config *cfp;
struct plink *plp;
int progress;
int change;
for(i=0; i<lemp->nstate; i++){
for(cfp=lemp->sorted[i]->cfp; cfp; cfp=cfp->next){
cfp->status = INCOMPLETE;
}
}
do{
progress = 0;
for(i=0; i<lemp->nstate; i++){
for(cfp=lemp->sorted[i]->cfp; cfp; cfp=cfp->next){
if( cfp->status==COMPLETE ) continue;
for(plp=cfp->fplp; plp; plp=plp->next){
change = SetUnion(plp->cfp->fws,cfp->fws);
if( change ){
plp->cfp->status = INCOMPLETE;
progress = 1;
}
}
cfp->status = COMPLETE;
}
}
}while( progress );
}
static int resolve_conflict(struct action *,struct action *);
/* Compute the reduce actions, and resolve conflicts.
*/
void FindActions(struct lemon *lemp)
{
int i,j;
struct config *cfp;
struct state *stp;
struct symbol *sp;
struct rule *rp;
/* Add all of the reduce actions
** A reduce action is added for each element of the followset of
** a configuration which has its dot at the extreme right.
*/
for(i=0; i<lemp->nstate; i++){ /* Loop over all states */
stp = lemp->sorted[i];
for(cfp=stp->cfp; cfp; cfp=cfp->next){ /* Loop over all configurations */
if( cfp->rp->nrhs==cfp->dot ){ /* Is dot at extreme right? */
for(j=0; j<lemp->nterminal; j++){
if( SetFind(cfp->fws,j) ){
/* Add a reduce action to the state "stp" which will reduce by the
** rule "cfp->rp" if the lookahead symbol is "lemp->symbols[j]" */
Action_add(&stp->ap,REDUCE,lemp->symbols[j],(char *)cfp->rp);
}
}
}
}
}
/* Add the accepting token */
if( lemp->start ){
sp = Symbol_find(lemp->start);
if( sp==0 ) sp = lemp->startRule->lhs;
}else{
sp = lemp->startRule->lhs;
}
/* Add to the first state (which is always the starting state of the
** finite state machine) an action to ACCEPT if the lookahead is the
** start nonterminal. */
Action_add(&lemp->sorted[0]->ap,ACCEPT,sp,0);
/* Resolve conflicts */
for(i=0; i<lemp->nstate; i++){
struct action *ap, *nap;
stp = lemp->sorted[i];
/* assert( stp->ap ); */
stp->ap = Action_sort(stp->ap);
for(ap=stp->ap; ap && ap->next; ap=ap->next){
for(nap=ap->next; nap && nap->sp==ap->sp; nap=nap->next){
/* The two actions "ap" and "nap" have the same lookahead.
** Figure out which one should be used */
lemp->nconflict += resolve_conflict(ap,nap);
}
}
}
/* Report an error for each rule that can never be reduced. */
for(rp=lemp->rule; rp; rp=rp->next) rp->canReduce = LEMON_FALSE;
for(i=0; i<lemp->nstate; i++){
struct action *ap;
for(ap=lemp->sorted[i]->ap; ap; ap=ap->next){
if( ap->type==REDUCE ) ap->x.rp->canReduce = LEMON_TRUE;
}
}
for(rp=lemp->rule; rp; rp=rp->next){
if( rp->canReduce ) continue;
ErrorMsg(lemp->filename,rp->ruleline,"This rule can not be reduced.\n");
lemp->errorcnt++;
}
}
/* Resolve a conflict between the two given actions. If the
** conflict can't be resolved, return non-zero.
**
** NO LONGER TRUE:
** To resolve a conflict, first look to see if either action
** is on an error rule. In that case, take the action which
** is not associated with the error rule. If neither or both
** actions are associated with an error rule, then try to
** use precedence to resolve the conflict.
**
** If either action is a SHIFT, then it must be apx. This
** function won't work if apx->type==REDUCE and apy->type==SHIFT.
*/
static int resolve_conflict(
struct action *apx,
struct action *apy
){
struct symbol *spx, *spy;
int errcnt = 0;
assert( apx->sp==apy->sp ); /* Otherwise there would be no conflict */
if( apx->type==SHIFT && apy->type==SHIFT ){
apy->type = SSCONFLICT;
errcnt++;
}
if( apx->type==SHIFT && apy->type==REDUCE ){
spx = apx->sp;
spy = apy->x.rp->precsym;
if( spy==0 || spx->prec<0 || spy->prec<0 ){
/* Not enough precedence information. */
apy->type = SRCONFLICT;
errcnt++;
}else if( spx->prec>spy->prec ){ /* higher precedence wins */
apy->type = RD_RESOLVED;
}else if( spx->prec<spy->prec ){
apx->type = SH_RESOLVED;
}else if( spx->prec==spy->prec && spx->assoc==RIGHT ){ /* Use operator */
apy->type = RD_RESOLVED; /* associativity */
}else if( spx->prec==spy->prec && spx->assoc==LEFT ){ /* to break tie */
apx->type = SH_RESOLVED;
}else{
assert( spx->prec==spy->prec && spx->assoc==NONE );
apx->type = ERROR;
}
}else if( apx->type==REDUCE && apy->type==REDUCE ){
spx = apx->x.rp->precsym;
spy = apy->x.rp->precsym;
if( spx==0 || spy==0 || spx->prec<0 ||
spy->prec<0 || spx->prec==spy->prec ){
apy->type = RRCONFLICT;
errcnt++;
}else if( spx->prec>spy->prec ){
apy->type = RD_RESOLVED;
}else if( spx->prec<spy->prec ){
apx->type = RD_RESOLVED;
}
}else{
assert(
apx->type==SH_RESOLVED ||
apx->type==RD_RESOLVED ||
apx->type==SSCONFLICT ||
apx->type==SRCONFLICT ||
apx->type==RRCONFLICT ||
apy->type==SH_RESOLVED ||
apy->type==RD_RESOLVED ||
apy->type==SSCONFLICT ||
apy->type==SRCONFLICT ||
apy->type==RRCONFLICT
);
/* The REDUCE/SHIFT case cannot happen because SHIFTs come before
** REDUCEs on the list. If we reach this point it must be because
** the parser conflict had already been resolved. */
}
return errcnt;
}
/********************* From the file "configlist.c" *************************/
/*
** Routines to processing a configuration list and building a state
** in the LEMON parser generator.
*/
static struct config *freelist = 0; /* List of free configurations */
static struct config *current = 0; /* Top of list of configurations */
static struct config **currentend = 0; /* Last on list of configs */
static struct config *basis = 0; /* Top of list of basis configs */
static struct config **basisend = 0; /* End of list of basis configs */
/* Return a pointer to a new configuration */
PRIVATE struct config *newconfig(void){
struct config *newcfg;
if( freelist==0 ){
int i;
int amt = 3;
freelist = (struct config *)calloc( amt, sizeof(struct config) );
if( freelist==0 ){
fprintf(stderr,"Unable to allocate memory for a new configuration.");
exit(1);
}
for(i=0; i<amt-1; i++) freelist[i].next = &freelist[i+1];
freelist[amt-1].next = 0;
}
newcfg = freelist;
freelist = freelist->next;
return newcfg;
}
/* The configuration "old" is no longer used */
PRIVATE void deleteconfig(struct config *old)
{
old->next = freelist;
freelist = old;
}
/* Initialized the configuration list builder */
void Configlist_init(void){
current = 0;
currentend = &current;
basis = 0;
basisend = &basis;
Configtable_init();
return;
}
/* Initialized the configuration list builder */
void Configlist_reset(void){
current = 0;
currentend = &current;
basis = 0;
basisend = &basis;
Configtable_clear(0);
return;
}
/* Add another configuration to the configuration list */
struct config *Configlist_add(
struct rule *rp, /* The rule */
int dot /* Index into the RHS of the rule where the dot goes */
){
struct config *cfp, model;
assert( currentend!=0 );
model.rp = rp;
model.dot = dot;
cfp = Configtable_find(&model);
if( cfp==0 ){
cfp = newconfig();
cfp->rp = rp;
cfp->dot = dot;
cfp->fws = SetNew();
cfp->stp = 0;
cfp->fplp = cfp->bplp = 0;
cfp->next = 0;
cfp->bp = 0;
*currentend = cfp;
currentend = &cfp->next;
Configtable_insert(cfp);
}
return cfp;
}
/* Add a basis configuration to the configuration list */
struct config *Configlist_addbasis(struct rule *rp, int dot)
{
struct config *cfp, model;
assert( basisend!=0 );
assert( currentend!=0 );
model.rp = rp;
model.dot = dot;
cfp = Configtable_find(&model);
if( cfp==0 ){
cfp = newconfig();
cfp->rp = rp;
cfp->dot = dot;
cfp->fws = SetNew();
cfp->stp = 0;
cfp->fplp = cfp->bplp = 0;
cfp->next = 0;
cfp->bp = 0;
*currentend = cfp;
currentend = &cfp->next;
*basisend = cfp;
basisend = &cfp->bp;
Configtable_insert(cfp);
}
return cfp;
}
/* Compute the closure of the configuration list */
void Configlist_closure(struct lemon *lemp)
{
struct config *cfp, *newcfp;
struct rule *rp, *newrp;
struct symbol *sp, *xsp;
int i, dot;
assert( currentend!=0 );
for(cfp=current; cfp; cfp=cfp->next){
rp = cfp->rp;
dot = cfp->dot;
if( dot>=rp->nrhs ) continue;
sp = rp->rhs[dot];
if( sp->type==NONTERMINAL ){
if( sp->rule==0 && sp!=lemp->errsym ){
ErrorMsg(lemp->filename,rp->line,"Nonterminal \"%s\" has no rules.",
sp->name);
lemp->errorcnt++;
}
for(newrp=sp->rule; newrp; newrp=newrp->nextlhs){
newcfp = Configlist_add(newrp,0);
for(i=dot+1; i<rp->nrhs; i++){
xsp = rp->rhs[i];
if( xsp->type==TERMINAL ){
SetAdd(newcfp->fws,xsp->index);
break;
}else if( xsp->type==MULTITERMINAL ){
int k;
for(k=0; k<xsp->nsubsym; k++){
SetAdd(newcfp->fws, xsp->subsym[k]->index);
}
break;
}else{
SetUnion(newcfp->fws,xsp->firstset);
if( xsp->lambda==LEMON_FALSE ) break;
}
}
if( i==rp->nrhs ) Plink_add(&cfp->fplp,newcfp);
}
}
}
return;
}
/* Sort the configuration list */
void Configlist_sort(void){
current = (struct config*)msort((char*)current,(char**)&(current->next),
Configcmp);
currentend = 0;
return;
}
/* Sort the basis configuration list */
void Configlist_sortbasis(void){
basis = (struct config*)msort((char*)current,(char**)&(current->bp),
Configcmp);
basisend = 0;
return;
}
/* Return a pointer to the head of the configuration list and
** reset the list */
struct config *Configlist_return(void){
struct config *old;
old = current;
current = 0;
currentend = 0;
return old;
}
/* Return a pointer to the head of the configuration list and
** reset the list */
struct config *Configlist_basis(void){
struct config *old;
old = basis;
basis = 0;
basisend = 0;
return old;
}
/* Free all elements of the given configuration list */
void Configlist_eat(struct config *cfp)
{
struct config *nextcfp;
for(; cfp; cfp=nextcfp){
nextcfp = cfp->next;
assert( cfp->fplp==0 );
assert( cfp->bplp==0 );
if( cfp->fws ) SetFree(cfp->fws);
deleteconfig(cfp);
}
return;
}
/***************** From the file "error.c" *********************************/
/*
** Code for printing error message.
*/
void ErrorMsg(const char *filename, int lineno, const char *format, ...){
va_list ap;
fprintf(stderr, "%s:%d: ", filename, lineno);
va_start(ap, format);
vfprintf(stderr,format,ap);
va_end(ap);
fprintf(stderr, "\n");
}
/**************** From the file "main.c" ************************************/
/*
** Main program file for the LEMON parser generator.
*/
/* Report an out-of-memory condition and abort. This function
** is used mostly by the "MemoryCheck" macro in struct.h
*/
void memory_error(void){
fprintf(stderr,"Out of memory. Aborting...\n");
exit(1);
}
static int nDefine = 0; /* Number of -D options on the command line */
static char **azDefine = 0; /* Name of the -D macros */
/* This routine is called with the argument to each -D command-line option.
** Add the macro defined to the azDefine array.
*/
static void handle_D_option(char *z){
char **paz;
nDefine++;
azDefine = (char **) realloc(azDefine, sizeof(azDefine[0])*nDefine);
if( azDefine==0 ){
fprintf(stderr,"out of memory\n");
exit(1);
}
paz = &azDefine[nDefine-1];
*paz = (char *) malloc( lemonStrlen(z)+1 );
if( *paz==0 ){
fprintf(stderr,"out of memory\n");
exit(1);
}
lemon_strcpy(*paz, z);
for(z=*paz; *z && *z!='='; z++){}
*z = 0;
}
/* Rember the name of the output directory
*/
static char *outputDir = NULL;
static void handle_d_option(char *z){
outputDir = (char *) malloc( lemonStrlen(z)+1 );
if( outputDir==0 ){
fprintf(stderr,"out of memory\n");
exit(1);
}
lemon_strcpy(outputDir, z);
}
static char *user_templatename = NULL;
static void handle_T_option(char *z){
user_templatename = (char *) malloc( lemonStrlen(z)+1 );
if( user_templatename==0 ){
memory_error();
}
lemon_strcpy(user_templatename, z);
}
/* Merge together to lists of rules ordered by rule.iRule */
static struct rule *Rule_merge(struct rule *pA, struct rule *pB){
struct rule *pFirst = 0;
struct rule **ppPrev = &pFirst;
while( pA && pB ){
if( pA->iRule<pB->iRule ){
*ppPrev = pA;
ppPrev = &pA->next;
pA = pA->next;
}else{
*ppPrev = pB;
ppPrev = &pB->next;
pB = pB->next;
}
}
if( pA ){
*ppPrev = pA;
}else{
*ppPrev = pB;
}
return pFirst;
}
/*
** Sort a list of rules in order of increasing iRule value
*/
static struct rule *Rule_sort(struct rule *rp){
int i;
struct rule *pNext;
struct rule *x[32];
memset(x, 0, sizeof(x));
while( rp ){
pNext = rp->next;
rp->next = 0;
for(i=0; i<sizeof(x)/sizeof(x[0]) && x[i]; i++){
rp = Rule_merge(x[i], rp);
x[i] = 0;
}
x[i] = rp;
rp = pNext;
}
rp = 0;
for(i=0; i<sizeof(x)/sizeof(x[0]); i++){
rp = Rule_merge(x[i], rp);
}
return rp;
}
/* forward reference */
static const char *minimum_size_type(int lwr, int upr, int *pnByte);
/* Print a single line of the "Parser Stats" output
*/
static void stats_line(const char *zLabel, int iValue){
int nLabel = lemonStrlen(zLabel);
printf(" %s%.*s %5d\n", zLabel,
35-nLabel, "................................",
iValue);
}
/* The main program. Parse the command line and do it... */
int main(int argc, char **argv)
{
static int version = 0;
static int rpflag = 0;
static int basisflag = 0;
static int compress = 0;
static int quiet = 0;
static int statistics = 0;
static int mhflag = 0;
static int nolinenosflag = 0;
static int noResort = 0;
static int sqlFlag = 0;
static struct s_options options[] = {
{OPT_FLAG, "b", (char*)&basisflag, "Print only the basis in report."},
{OPT_FLAG, "c", (char*)&compress, "Don't compress the action table."},
{OPT_FSTR, "d", (char*)&handle_d_option, "Output directory. Default '.'"},
{OPT_FSTR, "D", (char*)handle_D_option, "Define an %ifdef macro."},
{OPT_FSTR, "f", 0, "Ignored. (Placeholder for -f compiler options.)"},
{OPT_FLAG, "g", (char*)&rpflag, "Print grammar without actions."},
{OPT_FSTR, "I", 0, "Ignored. (Placeholder for '-I' compiler options.)"},
{OPT_FLAG, "m", (char*)&mhflag, "Output a makeheaders compatible file."},
{OPT_FLAG, "l", (char*)&nolinenosflag, "Do not print #line statements."},
{OPT_FSTR, "O", 0, "Ignored. (Placeholder for '-O' compiler options.)"},
{OPT_FLAG, "p", (char*)&showPrecedenceConflict,
"Show conflicts resolved by precedence rules"},
{OPT_FLAG, "q", (char*)&quiet, "(Quiet) Don't print the report file."},
{OPT_FLAG, "r", (char*)&noResort, "Do not sort or renumber states"},
{OPT_FLAG, "s", (char*)&statistics,
"Print parser stats to standard output."},
{OPT_FLAG, "S", (char*)&sqlFlag,
"Generate the *.sql file describing the parser tables."},
{OPT_FLAG, "x", (char*)&version, "Print the version number."},
{OPT_FSTR, "T", (char*)handle_T_option, "Specify a template file."},
{OPT_FSTR, "W", 0, "Ignored. (Placeholder for '-W' compiler options.)"},
{OPT_FLAG,0,0,0}
};
int i;
int exitcode;
struct lemon lem;
struct rule *rp;
OptInit(argv,options,stderr);
if( version ){
printf("Lemon version 1.0\n");
exit(0);
}
if( OptNArgs()!=1 ){
fprintf(stderr,"Exactly one filename argument is required.\n");
exit(1);
}
memset(&lem, 0, sizeof(lem));
lem.errorcnt = 0;
/* Initialize the machine */
Strsafe_init();
Symbol_init();
State_init();
lem.argv0 = argv[0];
lem.filename = OptArg(0);
lem.basisflag = basisflag;
lem.nolinenosflag = nolinenosflag;
Symbol_new("$");
/* Parse the input file */
Parse(&lem);
if( lem.errorcnt ) exit(lem.errorcnt);
if( lem.nrule==0 ){
fprintf(stderr,"Empty grammar.\n");
exit(1);
}
lem.errsym = Symbol_find("error");
/* Count and index the symbols of the grammar */
Symbol_new("{default}");
lem.nsymbol = Symbol_count();
lem.symbols = Symbol_arrayof();
for(i=0; i<lem.nsymbol; i++) lem.symbols[i]->index = i;
qsort(lem.symbols,lem.nsymbol,sizeof(struct symbol*), Symbolcmpp);
for(i=0; i<lem.nsymbol; i++) lem.symbols[i]->index = i;
while( lem.symbols[i-1]->type==MULTITERMINAL ){ i--; }
assert( strcmp(lem.symbols[i-1]->name,"{default}")==0 );
lem.nsymbol = i - 1;
for(i=1; ISUPPER(lem.symbols[i]->name[0]); i++);
lem.nterminal = i;
/* Assign sequential rule numbers. Start with 0. Put rules that have no
** reduce action C-code associated with them last, so that the switch()
** statement that selects reduction actions will have a smaller jump table.
*/
for(i=0, rp=lem.rule; rp; rp=rp->next){
rp->iRule = rp->code ? i++ : -1;
}
lem.nruleWithAction = i;
for(rp=lem.rule; rp; rp=rp->next){
if( rp->iRule<0 ) rp->iRule = i++;
}
lem.startRule = lem.rule;
lem.rule = Rule_sort(lem.rule);
/* Generate a reprint of the grammar, if requested on the command line */
if( rpflag ){
Reprint(&lem);
}else{
/* Initialize the size for all follow and first sets */
SetSize(lem.nterminal+1);
/* Find the precedence for every production rule (that has one) */
FindRulePrecedences(&lem);
/* Compute the lambda-nonterminals and the first-sets for every
** nonterminal */
FindFirstSets(&lem);
/* Compute all LR(0) states. Also record follow-set propagation
** links so that the follow-set can be computed later */
lem.nstate = 0;
FindStates(&lem);
lem.sorted = State_arrayof();
/* Tie up loose ends on the propagation links */
FindLinks(&lem);
/* Compute the follow set of every reducible configuration */
FindFollowSets(&lem);
/* Compute the action tables */
FindActions(&lem);
/* Compress the action tables */
if( compress==0 ) CompressTables(&lem);
/* Reorder and renumber the states so that states with fewer choices
** occur at the end. This is an optimization that helps make the
** generated parser tables smaller. */
if( noResort==0 ) ResortStates(&lem);
/* Generate a report of the parser generated. (the "y.output" file) */
if( !quiet ) ReportOutput(&lem);
/* Generate the source code for the parser */
ReportTable(&lem, mhflag, sqlFlag);
/* Produce a header file for use by the scanner. (This step is
** omitted if the "-m" option is used because makeheaders will
** generate the file for us.) */
if( !mhflag ) ReportHeader(&lem);
}
if( statistics ){
printf("Parser statistics:\n");
stats_line("terminal symbols", lem.nterminal);
stats_line("non-terminal symbols", lem.nsymbol - lem.nterminal);
stats_line("total symbols", lem.nsymbol);
stats_line("rules", lem.nrule);
stats_line("states", lem.nxstate);
stats_line("conflicts", lem.nconflict);
stats_line("action table entries", lem.nactiontab);
stats_line("lookahead table entries", lem.nlookaheadtab);
stats_line("total table size (bytes)", lem.tablesize);
}
if( lem.nconflict > 0 ){
fprintf(stderr,"%d parsing conflicts.\n",lem.nconflict);
}
/* return 0 on success, 1 on failure. */
exitcode = ((lem.errorcnt > 0) || (lem.nconflict > 0)) ? 1 : 0;
exit(exitcode);
return (exitcode);
}
/******************** From the file "msort.c" *******************************/
/*
** A generic merge-sort program.
**
** USAGE:
** Let "ptr" be a pointer to some structure which is at the head of
** a null-terminated list. Then to sort the list call:
**
** ptr = msort(ptr,&(ptr->next),cmpfnc);
**
** In the above, "cmpfnc" is a pointer to a function which compares
** two instances of the structure and returns an integer, as in
** strcmp. The second argument is a pointer to the pointer to the
** second element of the linked list. This address is used to compute
** the offset to the "next" field within the structure. The offset to
** the "next" field must be constant for all structures in the list.
**
** The function returns a new pointer which is the head of the list
** after sorting.
**
** ALGORITHM:
** Merge-sort.
*/
/*
** Return a pointer to the next structure in the linked list.
*/
#define NEXT(A) (*(char**)(((char*)A)+offset))
/*
** Inputs:
** a: A sorted, null-terminated linked list. (May be null).
** b: A sorted, null-terminated linked list. (May be null).
** cmp: A pointer to the comparison function.
** offset: Offset in the structure to the "next" field.
**
** Return Value:
** A pointer to the head of a sorted list containing the elements
** of both a and b.
**
** Side effects:
** The "next" pointers for elements in the lists a and b are
** changed.
*/
static char *merge(
char *a,
char *b,
int (*cmp)(const char*,const char*),
int offset
){
char *ptr, *head;
if( a==0 ){
head = b;
}else if( b==0 ){
head = a;
}else{
if( (*cmp)(a,b)<=0 ){
ptr = a;
a = NEXT(a);
}else{
ptr = b;
b = NEXT(b);
}
head = ptr;
while( a && b ){
if( (*cmp)(a,b)<=0 ){
NEXT(ptr) = a;
ptr = a;
a = NEXT(a);
}else{
NEXT(ptr) = b;
ptr = b;
b = NEXT(b);
}
}
if( a ) NEXT(ptr) = a;
else NEXT(ptr) = b;
}
return head;
}
/*
** Inputs:
** list: Pointer to a singly-linked list of structures.
** next: Pointer to pointer to the second element of the list.
** cmp: A comparison function.
**
** Return Value:
** A pointer to the head of a sorted list containing the elements
** orginally in list.
**
** Side effects:
** The "next" pointers for elements in list are changed.
*/
#define LISTSIZE 30
static char *msort(
char *list,
char **next,
int (*cmp)(const char*,const char*)
){
unsigned long offset;
char *ep;
char *set[LISTSIZE];
int i;
offset = (unsigned long)((char*)next - (char*)list);
for(i=0; i<LISTSIZE; i++) set[i] = 0;
while( list ){
ep = list;
list = NEXT(list);
NEXT(ep) = 0;
for(i=0; i<LISTSIZE-1 && set[i]!=0; i++){
ep = merge(ep,set[i],cmp,offset);
set[i] = 0;
}
set[i] = ep;
}
ep = 0;
for(i=0; i<LISTSIZE; i++) if( set[i] ) ep = merge(set[i],ep,cmp,offset);
return ep;
}
/************************ From the file "option.c" **************************/
static char **g_argv;
static struct s_options *op;
static FILE *errstream;
#define ISOPT(X) ((X)[0]=='-'||(X)[0]=='+'||strchr((X),'=')!=0)
/*
** Print the command line with a carrot pointing to the k-th character
** of the n-th field.
*/
static void errline(int n, int k, FILE *err)
{
int spcnt, i;
if( g_argv[0] ) fprintf(err,"%s",g_argv[0]);
spcnt = lemonStrlen(g_argv[0]) + 1;
for(i=1; i<n && g_argv[i]; i++){
fprintf(err," %s",g_argv[i]);
spcnt += lemonStrlen(g_argv[i])+1;
}
spcnt += k;
for(; g_argv[i]; i++) fprintf(err," %s",g_argv[i]);
if( spcnt<20 ){
fprintf(err,"\n%*s^-- here\n",spcnt,"");
}else{
fprintf(err,"\n%*shere --^\n",spcnt-7,"");
}
}
/*
** Return the index of the N-th non-switch argument. Return -1
** if N is out of range.
*/
static int argindex(int n)
{
int i;
int dashdash = 0;
if( g_argv!=0 && *g_argv!=0 ){
for(i=1; g_argv[i]; i++){
if( dashdash || !ISOPT(g_argv[i]) ){
if( n==0 ) return i;
n--;
}
if( strcmp(g_argv[i],"--")==0 ) dashdash = 1;
}
}
return -1;
}
static char emsg[] = "Command line syntax error: ";
/*
** Process a flag command line argument.
*/
static int handleflags(int i, FILE *err)
{
int v;
int errcnt = 0;
int j;
for(j=0; op[j].label; j++){
if( strncmp(&g_argv[i][1],op[j].label,lemonStrlen(op[j].label))==0 ) break;
}
v = g_argv[i][0]=='-' ? 1 : 0;
if( op[j].label==0 ){
if( err ){
fprintf(err,"%sundefined option.\n",emsg);
errline(i,1,err);
}
errcnt++;
}else if( op[j].arg==0 ){
/* Ignore this option */
}else if( op[j].type==OPT_FLAG ){
*((int*)op[j].arg) = v;
}else if( op[j].type==OPT_FFLAG ){
(*(void(*)(int))(op[j].arg))(v);
}else if( op[j].type==OPT_FSTR ){
(*(void(*)(char *))(op[j].arg))(&g_argv[i][2]);
}else{
if( err ){
fprintf(err,"%smissing argument on switch.\n",emsg);
errline(i,1,err);
}
errcnt++;
}
return errcnt;
}
/*
** Process a command line switch which has an argument.
*/
static int handleswitch(int i, FILE *err)
{
int lv = 0;
double dv = 0.0;
char *sv = 0, *end;
char *cp;
int j;
int errcnt = 0;
cp = strchr(g_argv[i],'=');
assert( cp!=0 );
*cp = 0;
for(j=0; op[j].label; j++){
if( strcmp(g_argv[i],op[j].label)==0 ) break;
}
*cp = '=';
if( op[j].label==0 ){
if( err ){
fprintf(err,"%sundefined option.\n",emsg);
errline(i,0,err);
}
errcnt++;
}else{
cp++;
switch( op[j].type ){
case OPT_FLAG:
case OPT_FFLAG:
if( err ){
fprintf(err,"%soption requires an argument.\n",emsg);
errline(i,0,err);
}
errcnt++;
break;
case OPT_DBL:
case OPT_FDBL:
dv = strtod(cp,&end);
if( *end ){
if( err ){
fprintf(err,
"%sillegal character in floating-point argument.\n",emsg);
errline(i,(int)((char*)end-(char*)g_argv[i]),err);
}
errcnt++;
}
break;
case OPT_INT:
case OPT_FINT:
lv = strtol(cp,&end,0);
if( *end ){
if( err ){
fprintf(err,"%sillegal character in integer argument.\n",emsg);
errline(i,(int)((char*)end-(char*)g_argv[i]),err);
}
errcnt++;
}
break;
case OPT_STR:
case OPT_FSTR:
sv = cp;
break;
}
switch( op[j].type ){
case OPT_FLAG:
case OPT_FFLAG:
break;
case OPT_DBL:
*(double*)(op[j].arg) = dv;
break;
case OPT_FDBL:
(*(void(*)(double))(op[j].arg))(dv);
break;
case OPT_INT:
*(int*)(op[j].arg) = lv;
break;
case OPT_FINT:
(*(void(*)(int))(op[j].arg))((int)lv);
break;
case OPT_STR:
*(char**)(op[j].arg) = sv;
break;
case OPT_FSTR:
(*(void(*)(char *))(op[j].arg))(sv);
break;
}
}
return errcnt;
}
int OptInit(char **a, struct s_options *o, FILE *err)
{
int errcnt = 0;
g_argv = a;
op = o;
errstream = err;
if( g_argv && *g_argv && op ){
int i;
for(i=1; g_argv[i]; i++){
if( g_argv[i][0]=='+' || g_argv[i][0]=='-' ){
errcnt += handleflags(i,err);
}else if( strchr(g_argv[i],'=') ){
errcnt += handleswitch(i,err);
}
}
}
if( errcnt>0 ){
fprintf(err,"Valid command line options for \"%s\" are:\n",*a);
OptPrint();
exit(1);
}
return 0;
}
int OptNArgs(void){
int cnt = 0;
int dashdash = 0;
int i;
if( g_argv!=0 && g_argv[0]!=0 ){
for(i=1; g_argv[i]; i++){
if( dashdash || !ISOPT(g_argv[i]) ) cnt++;
if( strcmp(g_argv[i],"--")==0 ) dashdash = 1;
}
}
return cnt;
}
char *OptArg(int n)
{
int i;
i = argindex(n);
return i>=0 ? g_argv[i] : 0;
}
void OptErr(int n)
{
int i;
i = argindex(n);
if( i>=0 ) errline(i,0,errstream);
}
void OptPrint(void){
int i;
int max, len;
max = 0;
for(i=0; op[i].label; i++){
len = lemonStrlen(op[i].label) + 1;
switch( op[i].type ){
case OPT_FLAG:
case OPT_FFLAG:
break;
case OPT_INT:
case OPT_FINT:
len += 9; /* length of "<integer>" */
break;
case OPT_DBL:
case OPT_FDBL:
len += 6; /* length of "<real>" */
break;
case OPT_STR:
case OPT_FSTR:
len += 8; /* length of "<string>" */
break;
}
if( len>max ) max = len;
}
for(i=0; op[i].label; i++){
switch( op[i].type ){
case OPT_FLAG:
case OPT_FFLAG:
fprintf(errstream," -%-*s %s\n",max,op[i].label,op[i].message);
break;
case OPT_INT:
case OPT_FINT:
fprintf(errstream," -%s<integer>%*s %s\n",op[i].label,
(int)(max-lemonStrlen(op[i].label)-9),"",op[i].message);
break;
case OPT_DBL:
case OPT_FDBL:
fprintf(errstream," -%s<real>%*s %s\n",op[i].label,
(int)(max-lemonStrlen(op[i].label)-6),"",op[i].message);
break;
case OPT_STR:
case OPT_FSTR:
fprintf(errstream," -%s<string>%*s %s\n",op[i].label,
(int)(max-lemonStrlen(op[i].label)-8),"",op[i].message);
break;
}
}
}
/*********************** From the file "parse.c" ****************************/
/*
** Input file parser for the LEMON parser generator.
*/
/* The state of the parser */
enum e_state {
INITIALIZE,
WAITING_FOR_DECL_OR_RULE,
WAITING_FOR_DECL_KEYWORD,
WAITING_FOR_DECL_ARG,
WAITING_FOR_PRECEDENCE_SYMBOL,
WAITING_FOR_ARROW,
IN_RHS,
LHS_ALIAS_1,
LHS_ALIAS_2,
LHS_ALIAS_3,
RHS_ALIAS_1,
RHS_ALIAS_2,
PRECEDENCE_MARK_1,
PRECEDENCE_MARK_2,
RESYNC_AFTER_RULE_ERROR,
RESYNC_AFTER_DECL_ERROR,
WAITING_FOR_DESTRUCTOR_SYMBOL,
WAITING_FOR_DATATYPE_SYMBOL,
WAITING_FOR_FALLBACK_ID,
WAITING_FOR_WILDCARD_ID,
WAITING_FOR_CLASS_ID,
WAITING_FOR_CLASS_TOKEN,
WAITING_FOR_TOKEN_NAME
};
struct pstate {
char *filename; /* Name of the input file */
int tokenlineno; /* Linenumber at which current token starts */
int errorcnt; /* Number of errors so far */
char *tokenstart; /* Text of current token */
struct lemon *gp; /* Global state vector */
enum e_state state; /* The state of the parser */
struct symbol *fallback; /* The fallback token */
struct symbol *tkclass; /* Token class symbol */
struct symbol *lhs; /* Left-hand side of current rule */
const char *lhsalias; /* Alias for the LHS */
int nrhs; /* Number of right-hand side symbols seen */
struct symbol *rhs[MAXRHS]; /* RHS symbols */
const char *alias[MAXRHS]; /* Aliases for each RHS symbol (or NULL) */
struct rule *prevrule; /* Previous rule parsed */
const char *declkeyword; /* Keyword of a declaration */
char **declargslot; /* Where the declaration argument should be put */
int insertLineMacro; /* Add #line before declaration insert */
int *decllinenoslot; /* Where to write declaration line number */
enum e_assoc declassoc; /* Assign this association to decl arguments */
int preccounter; /* Assign this precedence to decl arguments */
struct rule *firstrule; /* Pointer to first rule in the grammar */
struct rule *lastrule; /* Pointer to the most recently parsed rule */
};
/* Parse a single token */
static void parseonetoken(struct pstate *psp)
{
const char *x;
x = Strsafe(psp->tokenstart); /* Save the token permanently */
#if 0
printf("%s:%d: Token=[%s] state=%d\n",psp->filename,psp->tokenlineno,
x,psp->state);
#endif
switch( psp->state ){
case INITIALIZE:
psp->prevrule = 0;
psp->preccounter = 0;
psp->firstrule = psp->lastrule = 0;
psp->gp->nrule = 0;
/* Fall thru to next case */
case WAITING_FOR_DECL_OR_RULE:
if( x[0]=='%' ){
psp->state = WAITING_FOR_DECL_KEYWORD;
}else if( ISLOWER(x[0]) ){
psp->lhs = Symbol_new(x);
psp->nrhs = 0;
psp->lhsalias = 0;
psp->state = WAITING_FOR_ARROW;
}else if( x[0]=='{' ){
if( psp->prevrule==0 ){
ErrorMsg(psp->filename,psp->tokenlineno,
"There is no prior rule upon which to attach the code "
"fragment which begins on this line.");
psp->errorcnt++;
}else if( psp->prevrule->code!=0 ){
ErrorMsg(psp->filename,psp->tokenlineno,
"Code fragment beginning on this line is not the first "
"to follow the previous rule.");
psp->errorcnt++;
}else if( strcmp(x, "{NEVER-REDUCE")==0 ){
psp->prevrule->neverReduce = 1;
}else{
psp->prevrule->line = psp->tokenlineno;
psp->prevrule->code = &x[1];
psp->prevrule->noCode = 0;
}
}else if( x[0]=='[' ){
psp->state = PRECEDENCE_MARK_1;
}else{
ErrorMsg(psp->filename,psp->tokenlineno,
"Token \"%s\" should be either \"%%\" or a nonterminal name.",
x);
psp->errorcnt++;
}
break;
case PRECEDENCE_MARK_1:
if( !ISUPPER(x[0]) ){
ErrorMsg(psp->filename,psp->tokenlineno,
"The precedence symbol must be a terminal.");
psp->errorcnt++;
}else if( psp->prevrule==0 ){
ErrorMsg(psp->filename,psp->tokenlineno,
"There is no prior rule to assign precedence \"[%s]\".",x);
psp->errorcnt++;
}else if( psp->prevrule->precsym!=0 ){
ErrorMsg(psp->filename,psp->tokenlineno,
"Precedence mark on this line is not the first "
"to follow the previous rule.");
psp->errorcnt++;
}else{
psp->prevrule->precsym = Symbol_new(x);
}
psp->state = PRECEDENCE_MARK_2;
break;
case PRECEDENCE_MARK_2:
if( x[0]!=']' ){
ErrorMsg(psp->filename,psp->tokenlineno,
"Missing \"]\" on precedence mark.");
psp->errorcnt++;
}
psp->state = WAITING_FOR_DECL_OR_RULE;
break;
case WAITING_FOR_ARROW:
if( x[0]==':' && x[1]==':' && x[2]=='=' ){
psp->state = IN_RHS;
}else if( x[0]=='(' ){
psp->state = LHS_ALIAS_1;
}else{
ErrorMsg(psp->filename,psp->tokenlineno,
"Expected to see a \":\" following the LHS symbol \"%s\".",
psp->lhs->name);
psp->errorcnt++;
psp->state = RESYNC_AFTER_RULE_ERROR;
}
break;
case LHS_ALIAS_1:
if( ISALPHA(x[0]) ){
psp->lhsalias = x;
psp->state = LHS_ALIAS_2;
}else{
ErrorMsg(psp->filename,psp->tokenlineno,
"\"%s\" is not a valid alias for the LHS \"%s\"\n",
x,psp->lhs->name);
psp->errorcnt++;
psp->state = RESYNC_AFTER_RULE_ERROR;
}
break;
case LHS_ALIAS_2:
if( x[0]==')' ){
psp->state = LHS_ALIAS_3;
}else{
ErrorMsg(psp->filename,psp->tokenlineno,
"Missing \")\" following LHS alias name \"%s\".",psp->lhsalias);
psp->errorcnt++;
psp->state = RESYNC_AFTER_RULE_ERROR;
}
break;
case LHS_ALIAS_3:
if( x[0]==':' && x[1]==':' && x[2]=='=' ){
psp->state = IN_RHS;
}else{
ErrorMsg(psp->filename,psp->tokenlineno,
"Missing \"->\" following: \"%s(%s)\".",
psp->lhs->name,psp->lhsalias);
psp->errorcnt++;
psp->state = RESYNC_AFTER_RULE_ERROR;
}
break;
case IN_RHS:
if( x[0]=='.' ){
struct rule *rp;
rp = (struct rule *)calloc( sizeof(struct rule) +
sizeof(struct symbol*)*psp->nrhs + sizeof(char*)*psp->nrhs, 1);
if( rp==0 ){
ErrorMsg(psp->filename,psp->tokenlineno,
"Can't allocate enough memory for this rule.");
psp->errorcnt++;
psp->prevrule = 0;
}else{
int i;
rp->ruleline = psp->tokenlineno;
rp->rhs = (struct symbol**)&rp[1];
rp->rhsalias = (const char**)&(rp->rhs[psp->nrhs]);
for(i=0; i<psp->nrhs; i++){
rp->rhs[i] = psp->rhs[i];
rp->rhsalias[i] = psp->alias[i];
if( rp->rhsalias[i]!=0 ){ rp->rhs[i]->bContent = 1; }
}
rp->lhs = psp->lhs;
rp->lhsalias = psp->lhsalias;
rp->nrhs = psp->nrhs;
rp->code = 0;
rp->noCode = 1;
rp->precsym = 0;
rp->index = psp->gp->nrule++;
rp->nextlhs = rp->lhs->rule;
rp->lhs->rule = rp;
rp->next = 0;
if( psp->firstrule==0 ){
psp->firstrule = psp->lastrule = rp;
}else{
psp->lastrule->next = rp;
psp->lastrule = rp;
}
psp->prevrule = rp;
}
psp->state = WAITING_FOR_DECL_OR_RULE;
}else if( ISALPHA(x[0]) ){
if( psp->nrhs>=MAXRHS ){
ErrorMsg(psp->filename,psp->tokenlineno,
"Too many symbols on RHS of rule beginning at \"%s\".",
x);
psp->errorcnt++;
psp->state = RESYNC_AFTER_RULE_ERROR;
}else{
psp->rhs[psp->nrhs] = Symbol_new(x);
psp->alias[psp->nrhs] = 0;
psp->nrhs++;
}
}else if( (x[0]=='|' || x[0]=='/') && psp->nrhs>0 ){
struct symbol *msp = psp->rhs[psp->nrhs-1];
if( msp->type!=MULTITERMINAL ){
struct symbol *origsp = msp;
msp = (struct symbol *) calloc(1,sizeof(*msp));
memset(msp, 0, sizeof(*msp));
msp->type = MULTITERMINAL;
msp->nsubsym = 1;
msp->subsym = (struct symbol **) calloc(1,sizeof(struct symbol*));
msp->subsym[0] = origsp;
msp->name = origsp->name;
psp->rhs[psp->nrhs-1] = msp;
}
msp->nsubsym++;
msp->subsym = (struct symbol **) realloc(msp->subsym,
sizeof(struct symbol*)*msp->nsubsym);
msp->subsym[msp->nsubsym-1] = Symbol_new(&x[1]);
if( ISLOWER(x[1]) || ISLOWER(msp->subsym[0]->name[0]) ){
ErrorMsg(psp->filename,psp->tokenlineno,
"Cannot form a compound containing a non-terminal");
psp->errorcnt++;
}
}else if( x[0]=='(' && psp->nrhs>0 ){
psp->state = RHS_ALIAS_1;
}else{
ErrorMsg(psp->filename,psp->tokenlineno,
"Illegal character on RHS of rule: \"%s\".",x);
psp->errorcnt++;
psp->state = RESYNC_AFTER_RULE_ERROR;
}
break;
case RHS_ALIAS_1:
if( ISALPHA(x[0]) ){
psp->alias[psp->nrhs-1] = x;
psp->state = RHS_ALIAS_2;
}else{
ErrorMsg(psp->filename,psp->tokenlineno,
"\"%s\" is not a valid alias for the RHS symbol \"%s\"\n",
x,psp->rhs[psp->nrhs-1]->name);
psp->errorcnt++;
psp->state = RESYNC_AFTER_RULE_ERROR;
}
break;
case RHS_ALIAS_2:
if( x[0]==')' ){
psp->state = IN_RHS;
}else{
ErrorMsg(psp->filename,psp->tokenlineno,
"Missing \")\" following LHS alias name \"%s\".",psp->lhsalias);
psp->errorcnt++;
psp->state = RESYNC_AFTER_RULE_ERROR;
}
break;
case WAITING_FOR_DECL_KEYWORD:
if( ISALPHA(x[0]) ){
psp->declkeyword = x;
psp->declargslot = 0;
psp->decllinenoslot = 0;
psp->insertLineMacro = 1;
psp->state = WAITING_FOR_DECL_ARG;
if( strcmp(x,"name")==0 ){
psp->declargslot = &(psp->gp->name);
psp->insertLineMacro = 0;
}else if( strcmp(x,"include")==0 ){
psp->declargslot = &(psp->gp->include);
}else if( strcmp(x,"code")==0 ){
psp->declargslot = &(psp->gp->extracode);
}else if( strcmp(x,"token_destructor")==0 ){
psp->declargslot = &psp->gp->tokendest;
}else if( strcmp(x,"default_destructor")==0 ){
psp->declargslot = &psp->gp->vardest;
}else if( strcmp(x,"token_prefix")==0 ){
psp->declargslot = &psp->gp->tokenprefix;
psp->insertLineMacro = 0;
}else if( strcmp(x,"syntax_error")==0 ){
psp->declargslot = &(psp->gp->error);
}else if( strcmp(x,"parse_accept")==0 ){
psp->declargslot = &(psp->gp->accept);
}else if( strcmp(x,"parse_failure")==0 ){
psp->declargslot = &(psp->gp->failure);
}else if( strcmp(x,"stack_overflow")==0 ){
psp->declargslot = &(psp->gp->overflow);
}else if( strcmp(x,"extra_argument")==0 ){
psp->declargslot = &(psp->gp->arg);
psp->insertLineMacro = 0;
}else if( strcmp(x,"extra_context")==0 ){
psp->declargslot = &(psp->gp->ctx);
psp->insertLineMacro = 0;
}else if( strcmp(x,"token_type")==0 ){
psp->declargslot = &(psp->gp->tokentype);
psp->insertLineMacro = 0;
}else if( strcmp(x,"default_type")==0 ){
psp->declargslot = &(psp->gp->vartype);
psp->insertLineMacro = 0;
}else if( strcmp(x,"stack_size")==0 ){
psp->declargslot = &(psp->gp->stacksize);
psp->insertLineMacro = 0;
}else if( strcmp(x,"start_symbol")==0 ){
psp->declargslot = &(psp->gp->start);
psp->insertLineMacro = 0;
}else if( strcmp(x,"left")==0 ){
psp->preccounter++;
psp->declassoc = LEFT;
psp->state = WAITING_FOR_PRECEDENCE_SYMBOL;
}else if( strcmp(x,"right")==0 ){
psp->preccounter++;
psp->declassoc = RIGHT;
psp->state = WAITING_FOR_PRECEDENCE_SYMBOL;
}else if( strcmp(x,"nonassoc")==0 ){
psp->preccounter++;
psp->declassoc = NONE;
psp->state = WAITING_FOR_PRECEDENCE_SYMBOL;
}else if( strcmp(x,"destructor")==0 ){
psp->state = WAITING_FOR_DESTRUCTOR_SYMBOL;
}else if( strcmp(x,"type")==0 ){
psp->state = WAITING_FOR_DATATYPE_SYMBOL;
}else if( strcmp(x,"fallback")==0 ){
psp->fallback = 0;
psp->state = WAITING_FOR_FALLBACK_ID;
}else if( strcmp(x,"token")==0 ){
psp->state = WAITING_FOR_TOKEN_NAME;
}else if( strcmp(x,"wildcard")==0 ){
psp->state = WAITING_FOR_WILDCARD_ID;
}else if( strcmp(x,"token_class")==0 ){
psp->state = WAITING_FOR_CLASS_ID;
}else{
ErrorMsg(psp->filename,psp->tokenlineno,
"Unknown declaration keyword: \"%%%s\".",x);
psp->errorcnt++;
psp->state = RESYNC_AFTER_DECL_ERROR;
}
}else{
ErrorMsg(psp->filename,psp->tokenlineno,
"Illegal declaration keyword: \"%s\".",x);
psp->errorcnt++;
psp->state = RESYNC_AFTER_DECL_ERROR;
}
break;
case WAITING_FOR_DESTRUCTOR_SYMBOL:
if( !ISALPHA(x[0]) ){
ErrorMsg(psp->filename,psp->tokenlineno,
"Symbol name missing after %%destructor keyword");
psp->errorcnt++;
psp->state = RESYNC_AFTER_DECL_ERROR;
}else{
struct symbol *sp = Symbol_new(x);
psp->declargslot = &sp->destructor;
psp->decllinenoslot = &sp->destLineno;
psp->insertLineMacro = 1;
psp->state = WAITING_FOR_DECL_ARG;
}
break;
case WAITING_FOR_DATATYPE_SYMBOL:
if( !ISALPHA(x[0]) ){
ErrorMsg(psp->filename,psp->tokenlineno,
"Symbol name missing after %%type keyword");
psp->errorcnt++;
psp->state = RESYNC_AFTER_DECL_ERROR;
}else{
struct symbol *sp = Symbol_find(x);
if((sp) && (sp->datatype)){
ErrorMsg(psp->filename,psp->tokenlineno,
"Symbol %%type \"%s\" already defined", x);
psp->errorcnt++;
psp->state = RESYNC_AFTER_DECL_ERROR;
}else{
if (!sp){
sp = Symbol_new(x);
}
psp->declargslot = &sp->datatype;
psp->insertLineMacro = 0;
psp->state = WAITING_FOR_DECL_ARG;
}
}
break;
case WAITING_FOR_PRECEDENCE_SYMBOL:
if( x[0]=='.' ){
psp->state = WAITING_FOR_DECL_OR_RULE;
}else if( ISUPPER(x[0]) ){
struct symbol *sp;
sp = Symbol_new(x);
if( sp->prec>=0 ){
ErrorMsg(psp->filename,psp->tokenlineno,
"Symbol \"%s\" has already be given a precedence.",x);
psp->errorcnt++;
}else{
sp->prec = psp->preccounter;
sp->assoc = psp->declassoc;
}
}else{
ErrorMsg(psp->filename,psp->tokenlineno,
"Can't assign a precedence to \"%s\".",x);
psp->errorcnt++;
}
break;
case WAITING_FOR_DECL_ARG:
if( x[0]=='{' || x[0]=='\"' || ISALNUM(x[0]) ){
const char *zOld, *zNew;
char *zBuf, *z;
int nOld, n, nLine = 0, nNew, nBack;
int addLineMacro;
char zLine[50];
zNew = x;
if( zNew[0]=='"' || zNew[0]=='{' ) zNew++;
nNew = lemonStrlen(zNew);
if( *psp->declargslot ){
zOld = *psp->declargslot;
}else{
zOld = "";
}
nOld = lemonStrlen(zOld);
n = nOld + nNew + 20;
addLineMacro = !psp->gp->nolinenosflag && psp->insertLineMacro &&
(psp->decllinenoslot==0 || psp->decllinenoslot[0]!=0);
if( addLineMacro ){
for(z=psp->filename, nBack=0; *z; z++){
if( *z=='\\' ) nBack++;
}
lemon_sprintf(zLine, "#line %d ", psp->tokenlineno);
nLine = lemonStrlen(zLine);
n += nLine + lemonStrlen(psp->filename) + nBack;
}
*psp->declargslot = (char *) realloc(*psp->declargslot, n);
zBuf = *psp->declargslot + nOld;
if( addLineMacro ){
if( nOld && zBuf[-1]!='\n' ){
*(zBuf++) = '\n';
}
memcpy(zBuf, zLine, nLine);
zBuf += nLine;
*(zBuf++) = '"';
for(z=psp->filename; *z; z++){
if( *z=='\\' ){
*(zBuf++) = '\\';
}
*(zBuf++) = *z;
}
*(zBuf++) = '"';
*(zBuf++) = '\n';
}
if( psp->decllinenoslot && psp->decllinenoslot[0]==0 ){
psp->decllinenoslot[0] = psp->tokenlineno;
}
memcpy(zBuf, zNew, nNew);
zBuf += nNew;
*zBuf = 0;
psp->state = WAITING_FOR_DECL_OR_RULE;
}else{
ErrorMsg(psp->filename,psp->tokenlineno,
"Illegal argument to %%%s: %s",psp->declkeyword,x);
psp->errorcnt++;
psp->state = RESYNC_AFTER_DECL_ERROR;
}
break;
case WAITING_FOR_FALLBACK_ID:
if( x[0]=='.' ){
psp->state = WAITING_FOR_DECL_OR_RULE;
}else if( !ISUPPER(x[0]) ){
ErrorMsg(psp->filename, psp->tokenlineno,
"%%fallback argument \"%s\" should be a token", x);
psp->errorcnt++;
}else{
struct symbol *sp = Symbol_new(x);
if( psp->fallback==0 ){
psp->fallback = sp;
}else if( sp->fallback ){
ErrorMsg(psp->filename, psp->tokenlineno,
"More than one fallback assigned to token %s", x);
psp->errorcnt++;
}else{
sp->fallback = psp->fallback;
psp->gp->has_fallback = 1;
}
}
break;
case WAITING_FOR_TOKEN_NAME:
/* Tokens do not have to be declared before use. But they can be
** in order to control their assigned integer number. The number for
** each token is assigned when it is first seen. So by including
**
** %token ONE TWO THREE
**
** early in the grammar file, that assigns small consecutive values
** to each of the tokens ONE TWO and THREE.
*/
if( x[0]=='.' ){
psp->state = WAITING_FOR_DECL_OR_RULE;
}else if( !ISUPPER(x[0]) ){
ErrorMsg(psp->filename, psp->tokenlineno,
"%%token argument \"%s\" should be a token", x);
psp->errorcnt++;
}else{
(void)Symbol_new(x);
}
break;
case WAITING_FOR_WILDCARD_ID:
if( x[0]=='.' ){
psp->state = WAITING_FOR_DECL_OR_RULE;
}else if( !ISUPPER(x[0]) ){
ErrorMsg(psp->filename, psp->tokenlineno,
"%%wildcard argument \"%s\" should be a token", x);
psp->errorcnt++;
}else{
struct symbol *sp = Symbol_new(x);
if( psp->gp->wildcard==0 ){
psp->gp->wildcard = sp;
}else{
ErrorMsg(psp->filename, psp->tokenlineno,
"Extra wildcard to token: %s", x);
psp->errorcnt++;
}
}
break;
case WAITING_FOR_CLASS_ID:
if( !ISLOWER(x[0]) ){
ErrorMsg(psp->filename, psp->tokenlineno,
"%%token_class must be followed by an identifier: %s", x);
psp->errorcnt++;
psp->state = RESYNC_AFTER_DECL_ERROR;
}else if( Symbol_find(x) ){
ErrorMsg(psp->filename, psp->tokenlineno,
"Symbol \"%s\" already used", x);
psp->errorcnt++;
psp->state = RESYNC_AFTER_DECL_ERROR;
}else{
psp->tkclass = Symbol_new(x);
psp->tkclass->type = MULTITERMINAL;
psp->state = WAITING_FOR_CLASS_TOKEN;
}
break;
case WAITING_FOR_CLASS_TOKEN:
if( x[0]=='.' ){
psp->state = WAITING_FOR_DECL_OR_RULE;
}else if( ISUPPER(x[0]) || ((x[0]=='|' || x[0]=='/') && ISUPPER(x[1])) ){
struct symbol *msp = psp->tkclass;
msp->nsubsym++;
msp->subsym = (struct symbol **) realloc(msp->subsym,
sizeof(struct symbol*)*msp->nsubsym);
if( !ISUPPER(x[0]) ) x++;
msp->subsym[msp->nsubsym-1] = Symbol_new(x);
}else{
ErrorMsg(psp->filename, psp->tokenlineno,
"%%token_class argument \"%s\" should be a token", x);
psp->errorcnt++;
psp->state = RESYNC_AFTER_DECL_ERROR;
}
break;
case RESYNC_AFTER_RULE_ERROR:
/* if( x[0]=='.' ) psp->state = WAITING_FOR_DECL_OR_RULE;
** break; */
case RESYNC_AFTER_DECL_ERROR:
if( x[0]=='.' ) psp->state = WAITING_FOR_DECL_OR_RULE;
if( x[0]=='%' ) psp->state = WAITING_FOR_DECL_KEYWORD;
break;
}
}
/* Run the preprocessor over the input file text. The global variables
** azDefine[0] through azDefine[nDefine-1] contains the names of all defined
** macros. This routine looks for "%ifdef" and "%ifndef" and "%endif" and
** comments them out. Text in between is also commented out as appropriate.
*/
static void preprocess_input(char *z){
int i, j, k, n;
int exclude = 0;
int start = 0;
int lineno