SQLite

**
*************************************************************************
** This file contains SQLite's grammar for SQL.  Process this file
** using the lemon parser generator to generate C code that runs
** the parser.  Lemon will also generate a header file containing
** numeric codes for all of the tokens.
**
** @(#) $Id: parse.y,v 1.183 2005/11/06 04:06:59 drh Exp $
*/

// All token codes are small integers with #defines that begin with "TK_"
%token_prefix TK_

// The type of the data attached to each token is Token.  This is also the
// default type for non-terminals.

  DATABASE DEFERRED DESC DETACH EACH END EXCLUSIVE EXPLAIN FAIL FOR
  IGNORE IMMEDIATE INITIALLY INSTEAD LIKE_KW MATCH PLAN QUERY KEY
  OF OFFSET PRAGMA RAISE REPLACE RESTRICT ROW STATEMENT
  TEMP TRIGGER VACUUM VIEW
%ifdef SQLITE_OMIT_COMPOUND_SELECT
  EXCEPT INTERSECT UNION
%endif
  REINDEX RENAME CTIME_KW
  .

// Define operator precedence early so that this is the first occurance
// of the operator tokens in the grammer.  Keeping the operators together
// causes them to be assigned integer values that are close together,
// which keeps parser tables smaller.
//

%left STAR SLASH REM.
%left CONCAT.
%right UMINUS UPLUS BITNOT.

// And "ids" is an identifer-or-string.
//
%type ids {Token}
ids(A) ::= ID|STRING(X).   {A = X;}


// The name of a column or table can be any of the following:
//
%type nm {Token}
nm(A) ::= ID(X).         {A = X;}
nm(A) ::= STRING(X).     {A = X;}
nm(A) ::= JOIN_KW(X).    {A = X;}

  if( Z ){
    Z->op = Y;
    Z->pPrior = X;
  }
  A = Z;
}
%type multiselect_op {int}
multiselect_op(A) ::= UNION(OP).             {A = @OP;}
multiselect_op(A) ::= UNION ALL.             {A = TK_ALL;}

multiselect_op(A) ::= EXCEPT|INTERSECT(OP).  {A = @OP;}
%endif // SQLITE_OMIT_COMPOUND_SELECT
oneselect(A) ::= SELECT distinct(D) selcollist(W) from(X) where_opt(Y)
                 groupby_opt(P) having_opt(Q) orderby_opt(Z) limit_opt(L). {
  A = sqlite3SelectNew(W,X,Y,P,Q,Z,D,L.pLimit,L.pOffset);
}

// The "distinct" nonterminal is true (1) if the DISTINCT keyword is

%type fullname {SrcList*}
%destructor fullname {sqlite3SrcListDelete($$);}
fullname(A) ::= nm(X) dbnm(Y).  {A = sqlite3SrcListAppend(0,&X,&Y);}

%type joinop {int}
%type joinop2 {int}
joinop(X) ::= COMMA|JOIN.              { X = JT_INNER; }

joinop(X) ::= JOIN_KW(A) JOIN.         { X = sqlite3JoinType(pParse,&A,0,0); }
joinop(X) ::= JOIN_KW(A) nm(B) JOIN.   { X = sqlite3JoinType(pParse,&A,&B,0); }
joinop(X) ::= JOIN_KW(A) nm(B) nm(C) JOIN.
                                       { X = sqlite3JoinType(pParse,&A,&B,&C); }

%type on_opt {Expr*}
%destructor on_opt {sqlite3ExprDelete($$);}

expr(A) ::= nm(X) DOT nm(Y) DOT nm(Z). {
  Expr *temp1 = sqlite3Expr(TK_ID, 0, 0, &X);
  Expr *temp2 = sqlite3Expr(TK_ID, 0, 0, &Y);
  Expr *temp3 = sqlite3Expr(TK_ID, 0, 0, &Z);
  Expr *temp4 = sqlite3Expr(TK_DOT, temp2, temp3, 0);
  A = sqlite3Expr(TK_DOT, temp1, temp4, 0);
}
term(A) ::= INTEGER|FLOAT|BLOB(X).      {A = sqlite3Expr(@X, 0, 0, &X);}

term(A) ::= STRING(X).       {A = sqlite3Expr(@X, 0, 0, &X);}

expr(A) ::= REGISTER(X).     {A = sqlite3RegisterExpr(pParse, &X);}
expr(A) ::= VARIABLE(X).     {
  Token *pToken = &X;
  Expr *pExpr = A = sqlite3Expr(TK_VARIABLE, 0, 0, pToken);
  sqlite3ExprAssignVarNumber(pParse, pExpr);
}
%ifndef SQLITE_OMIT_CAST

}
term(A) ::= CTIME_KW(OP). {
  /* The CURRENT_TIME, CURRENT_DATE, and CURRENT_TIMESTAMP values are
  ** treated as functions that return constants */
  A = sqlite3ExprFunction(0,&OP);
  if( A ) A->op = TK_CONST_FUNC;  
}
expr(A) ::= expr(X) AND(OP) expr(Y).            {A = sqlite3Expr(@OP, X, Y, 0);}
expr(A) ::= expr(X) OR(OP) expr(Y).             {A = sqlite3Expr(@OP, X, Y, 0);}
expr(A) ::= expr(X) LT|GT|GE|LE(OP) expr(Y).    {A = sqlite3Expr(@OP, X, Y, 0);}
expr(A) ::= expr(X) EQ|NE(OP) expr(Y).          {A = sqlite3Expr(@OP, X, Y, 0);}
expr(A) ::= expr(X) BITAND|BITOR|LSHIFT|RSHIFT(OP) expr(Y).
                                                {A = sqlite3Expr(@OP, X, Y, 0);}
expr(A) ::= expr(X) PLUS|MINUS(OP) expr(Y).     {A = sqlite3Expr(@OP, X, Y, 0);}

expr(A) ::= expr(X) STAR|SLASH|REM(OP) expr(Y). {A = sqlite3Expr(@OP, X, Y, 0);}








expr(A) ::= expr(X) CONCAT(OP) expr(Y).         {A = sqlite3Expr(@OP, X, Y, 0);}
%type likeop {struct LikeOp}
likeop(A) ::= LIKE_KW(X).     {A.operator = X; A.not = 0;}
likeop(A) ::= NOT LIKE_KW(X). {A.operator = X; A.not = 1;}
%type escape {Expr*}
escape(X) ::= ESCAPE expr(A). [ESCAPE] {X = A;}
escape(X) ::= .               [ESCAPE] {X = 0;}
expr(A) ::= expr(X) likeop(OP) expr(Y) escape(E).  [LIKE_KW]  {
  ExprList *pList = sqlite3ExprListAppend(0, Y, 0);
  pList = sqlite3ExprListAppend(pList, X, 0);
  if( E ){
    pList = sqlite3ExprListAppend(pList, E, 0);
  }
  A = sqlite3ExprFunction(pList, &OP.operator);
  if( OP.not ) A = sqlite3Expr(TK_NOT, A, 0, 0);
  sqlite3ExprSpan(A, &X->span, &Y->span);
}

expr(A) ::= expr(X) ISNULL|NOTNULL(E). {
  A = sqlite3Expr(@E, X, 0, 0);
  sqlite3ExprSpan(A,&X->span,&E);
}
expr(A) ::= expr(X) IS NULL(E). {
  A = sqlite3Expr(TK_ISNULL, X, 0, 0);
  sqlite3ExprSpan(A,&X->span,&E);
}




expr(A) ::= expr(X) NOT NULL(E). {
  A = sqlite3Expr(TK_NOTNULL, X, 0, 0);
  sqlite3ExprSpan(A,&X->span,&E);
}
expr(A) ::= expr(X) IS NOT NULL(E). {
  A = sqlite3Expr(TK_NOTNULL, X, 0, 0);
  sqlite3ExprSpan(A,&X->span,&E);
}
expr(A) ::= NOT|BITNOT(B) expr(X). {




  A = sqlite3Expr(@B, X, 0, 0);
  sqlite3ExprSpan(A,&B,&X->span);
}
expr(A) ::= MINUS(B) expr(X). [UMINUS] {
  A = sqlite3Expr(TK_UMINUS, X, 0, 0);
  sqlite3ExprSpan(A,&B,&X->span);
}

  sqlite3Pragma(pParse,&X,&Z,&Y,1);
}
cmd ::= PRAGMA nm(X) dbnm(Z) LP nm(Y) RP. {sqlite3Pragma(pParse,&X,&Z,&Y,0);}
cmd ::= PRAGMA nm(X) dbnm(Z).             {sqlite3Pragma(pParse,&X,&Z,0,0);}
%endif // SQLITE_OMIT_PRAGMA
plus_num(A) ::= plus_opt number(X).   {A = X;}
minus_num(A) ::= MINUS number(X).     {A = X;}
number(A) ::= INTEGER|FLOAT(X).       {A = X;}

plus_opt ::= PLUS.
plus_opt ::= .

//////////////////////////// The CREATE TRIGGER command /////////////////////

%ifndef SQLITE_OMIT_TRIGGER

trigger_time(A) ::= BEFORE.      { A = TK_BEFORE; }
trigger_time(A) ::= AFTER.       { A = TK_AFTER;  }
trigger_time(A) ::= INSTEAD OF.  { A = TK_INSTEAD;}
trigger_time(A) ::= .            { A = TK_BEFORE; }

%type trigger_event {struct TrigEvent}
%destructor trigger_event {sqlite3IdListDelete($$.b);}
trigger_event(A) ::= DELETE|INSERT(OP).       {A.a = @OP; A.b = 0;}

trigger_event(A) ::= UPDATE(OP).              {A.a = @OP; A.b = 0;}
trigger_event(A) ::= UPDATE OF inscollist(X). {A.a = TK_UPDATE; A.b = X;}

%type foreach_clause {int}
foreach_clause(A) ::= .                   { A = TK_ROW; }
foreach_clause(A) ::= FOR EACH ROW.       { A = TK_ROW; }
foreach_clause(A) ::= FOR EACH STATEMENT. { A = TK_STATEMENT; }

/* Symbols (terminals and nonterminals) of the grammar are stored
** in the following: */
struct symbol {
  char *name;              /* Name of the symbol */
  int index;               /* Index number for this symbol */
  enum {
    TERMINAL,
    NONTERMINAL,
    MULTITERMINAL
  } 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 {
    LEFT,
    RIGHT,
    NONE,
    UNK
  } assoc;                 /* Associativity if predecence is defined */
  char *firstset;          /* First-set for all rules of this symbol */
  Boolean lambda;          /* True if NT and can generate an empty string */
  char *destructor;        /* Code which executes whenever this symbol is
                           ** popped from the stack during error processing */
  int destructorln;        /* Line number of destructor code */
  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 */
  /* 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 */
  char *lhsalias;          /* Alias for the LHS (NULL if none) */

*/
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;
	}

  /* 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++;
      }
    }

    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 */

    /* 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)

      }
      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==B_FALSE ) break;
	  }
	}
        if( i==rp->nrhs ) Plink_add(&cfp->fplp,newcfp);

	  }
          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 or 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 = malloc(sizeof(*msp));
          memset(msp, 0, sizeof(*msp));
          msp->type = MULTITERMINAL;
          msp->nsubsym = 1;
          msp->subsym = malloc(sizeof(struct symbol*));
          msp->subsym[0] = origsp;
          msp->name = origsp->name;
          psp->rhs[psp->nrhs-1] = msp;
        }
        msp->nsubsym++;
        msp->subsym = 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;

      }
    }else if( isalnum(c) ){          /* Identifiers */
      while( (c= *cp)!=0 && (isalnum(c) || c=='_') ) cp++;
      nextcp = cp;
    }else if( c==':' && cp[1]==':' && cp[2]=='=' ){ /* The operator "::=" */
      cp += 3;
      nextcp = cp;
    }else if( (c=='/' || c=='|') && isalpha(cp[1]) ){
      cp += 2;
      while( (c = *cp)!=0 && (isalnum(c) || c=='_') ) cp++;
      nextcp = cp;
    }else{                          /* All other (one character) operators */
      cp++;
      nextcp = cp;
    }
    c = *cp;
    *cp = 0;                        /* Null terminate the token */
    parseonetoken(&ps);             /* Parse the token */

      assert( sp->index==j );
      printf(" %3d %-*.*s",j,maxlen,maxlen,sp->name);
    }
    printf("\n");
  }
  for(rp=lemp->rule; rp; rp=rp->next){
    printf("%s",rp->lhs->name);
    /*    if( rp->lhsalias ) printf("(%s)",rp->lhsalias); */
    printf(" ::=");
    for(i=0; i<rp->nrhs; i++){
      sp = rp->rhs[i];
      printf(" %s", sp->name);
      if( sp->type==MULTITERMINAL ){
        for(j=1; j<sp->nsubsym; j++){
          printf("|%s", sp->subsym[j]->name);
        }
      }
      /* if( rp->rhsalias[i] ) printf("(%s)",rp->rhsalias[i]); */
    }
    printf(".");
    if( rp->precsym ) printf(" [%s]",rp->precsym->name);
    /* if( rp->code ) printf("\n    %s",rp->code); */
    printf("\n");
  }
}

void ConfigPrint(fp,cfp)
FILE *fp;
struct config *cfp;
{
  struct rule *rp;
  struct symbol *sp;
  int i, j;
  rp = cfp->rp;
  fprintf(fp,"%s ::=",rp->lhs->name);
  for(i=0; i<=rp->nrhs; i++){
    if( i==cfp->dot ) fprintf(fp," *");
    if( i==rp->nrhs ) break;
    sp = rp->rhs[i];
    fprintf(fp," %s", sp->name);
    if( sp->type==MULTITERMINAL ){
      for(j=1; j<sp->nsubsym; j++){
        fprintf(fp,"|%s",sp->subsym[j]->name);
      }
    }
  }
}

/* #define TEST */
#if 0
/* Print a set */
PRIVATE void SetPrint(out,set,lemp)
FILE *out;
char *set;
struct lemon *lemp;
{
  int i;

        sprintf(buf,"(%d)",cfp->rp->index);
        fprintf(fp,"    %5s ",buf);
      }else{
        fprintf(fp,"          ");
      }
      ConfigPrint(fp,cfp);
      fprintf(fp,"\n");
#if 0
      SetPrint(fp,cfp->fws,lemp);
      PlinkPrint(fp,cfp->fplp,"To  ");
      PlinkPrint(fp,cfp->bplp,"From");
#endif
      if( lemp->basisflag ) cfp=cfp->bp;
      else                  cfp=cfp->next;
    }

        for(i=0; i<rp->nrhs; i++){
          if( rp->rhsalias[i] && strcmp(cp,rp->rhsalias[i])==0 ){
            if( cp!=rp->code && cp[-1]=='@' ){
              /* If the argument is of the form @X then substituted
              ** the token number of X, not the value of X */
              append_str("yymsp[%d].major",-1,i-rp->nrhs+1,0);
            }else{
              struct symbol *sp = rp->rhs[i];
              int dtnum;
              if( sp->type==MULTITERMINAL ){
                dtnum = sp->subsym[0]->dtnum;
              }else{
                dtnum = sp->dtnum;
              }
              append_str("yymsp[%d].minor.yy%d",0,i-rp->nrhs+1, dtnum);

            }
            cp = xp;
            used[i] = 1;
            break;
          }
        }
      }

  /* Generate a table containing a text string that describes every
  ** rule in the rule set of the grammer.  This information is used
  ** when tracing REDUCE actions.
  */
  for(i=0, rp=lemp->rule; rp; rp=rp->next, i++){
    assert( rp->index==i );
    fprintf(out," /* %3d */ \"%s ::=", i, rp->lhs->name);
    for(j=0; j<rp->nrhs; j++){
      struct symbol *sp = rp->rhs[j];
      fprintf(out," %s", sp->name);
      if( sp->type==MULTITERMINAL ){
        int k;
        for(k=1; k<sp->nsubsym; k++){
          fprintf(out,"|%s",sp->subsym[k]->name);
        }
      }
    }
    fprintf(out,"\",\n"); lineno++;
  }
  tplt_xfer(lemp->name,in,out,&lineno);

  /* Generate code which executes every time a symbol is popped from
  ** the stack while processing errors or while destroying the parser. 
  ** (In other words, generate the %destructor actions)