lemon.c

sqlite-3.4.1,嵌入式数据库.是一个功能强大的开源数据库,给学习和研发以及小型公司的发展带来了全所未有的好处.

}

/* Allocate a new acttab structure */
acttab *acttab_alloc(void){
  acttab *p = malloc( sizeof(*p) );
  if( p==0 ){
    fprintf(stderr,"Unable to allocate memory for a new acttab.");
    exit(1);
  }
  memset(p, 0, sizeof(*p));
  return p;
}

/* Add a new action to the current transaction set
*/
void acttab_action(acttab *p, int lookahead, int action){
  if( p->nLookahead>=p->nLookaheadAlloc ){
    p->nLookaheadAlloc += 25;
    p->aLookahead = 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.
*/
int acttab_insert(acttab *p){
  int i, j, k, n;
  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->mxLookahead + 1;
  if( p->nAction + n >= p->nActionAlloc ){
    int oldAlloc = p->nActionAlloc;
    p->nActionAlloc = p->nAction + n + p->nActionAlloc + 20;
    p->aAction = 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 where we can
  ** insert the current transaction set.  Fall out of the loop when that
  ** offset is found.  In the worst case, we fall out of the loop when
  ** i reaches p->nAction, which means we append the new transaction set.
  **
  ** i is the index in p->aAction[] where p->mnLookahead is inserted.
  */
  for(i=0; i<p->nAction+p->mnLookahead; 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 */
      }
    }else if( p->aAction[i].lookahead==p->mnLookahead ){
      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;
      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;  /* Same as a prior transaction set */
      }
    }
  }
  /* Insert transaction set at index i. */
  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;
  }
  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;
}

/********************** 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(xp)
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(lemp)
struct lemon *lemp;
{
  int i, j;
  struct rule *rp;
  int progress;

  for(i=0; i<lemp->nsymbol; i++){
    lemp->symbols[i]->lambda = B_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];
         if( sp->type!=TERMINAL || sp->lambda==B_FALSE ) break;
      }
      if( i==rp->nrhs ){
        rp->lhs->lambda = B_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==B_FALSE ) break;
	}else{
          progress += SetUnion(s1->firstset,s2->firstset);
          if( s2->lambda==B_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(lemp)
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->rule->lhs->name);
      lemp->errorcnt++;
      sp = lemp->rule->lhs;
    }
  }else{
    sp = lemp->rule->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;
    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(lemp)
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(a,b)
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(lemp,stp)
struct lemon *lemp;
struct state *stp;     /* The state from which successors are computed */
{
  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 *new;  /* */
  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 */
      new = Configlist_addbasis(bcfp->rp,bcfp->dot+1);
      Plink_add(&new->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(lemp)
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;
    }
  }