selfuncs.c 14.4 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14
/*-------------------------------------------------------------------------
 *
 * selfuncs.c--
 *    Selectivity functions for system catalogs and builtin types
 *
 *    These routines are registered in the operator catalog in the
 *    "oprrest" and "oprjoin" attributes.
 *
 *    XXX check all the functions--I suspect them to be 1-based.
 *
 * Copyright (c) 1994, Regents of the University of California
 *
 *
 * IDENTIFICATION
15
 *    $Header: /cvsroot/pgsql/src/backend/utils/adt/selfuncs.c,v 1.8 1997/08/21 03:02:04 momjian Exp $
16 17 18 19 20 21
 *
 *-------------------------------------------------------------------------
 */
#include <stdio.h>
#include <string.h>

22 23
#include "postgres.h"

24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43
#include "access/heapam.h"
#include "utils/tqual.h"	/* for NowTimeQual */
#include "fmgr.h"
#include "utils/builtins.h"	/* for textout() prototype 
				  and where the declarations go */
#include "utils/palloc.h"

#include "catalog/catname.h"
#include "utils/syscache.h"
#include "utils/lsyscache.h"	/* for get_oprrest() */
#include "catalog/pg_statistic.h"

/* N is not a valid var/constant or relation id */
#define	NONVALUE(N)	((N) == -1)

/* 
 * generalize the test for functional index selectivity request
 */
#define FunctionalSelectivity(nIndKeys,attNum) (attNum==InvalidAttrNumber)

44
static float32data getattdisbursion(Oid relid, AttrNumber attnum);
45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63
static void gethilokey(Oid relid, AttrNumber attnum, Oid opid,
		       char **high, char **low);


/*
 *	eqsel		- Selectivity of "=" for any data type.
 */
float64
eqsel(Oid opid,
      Oid relid,
      AttrNumber attno,
      char *value,
      int32 flag)
{
    float64		result;
    
    result = (float64) palloc(sizeof(float64data));
    if (NONVALUE(attno) || NONVALUE(relid))
	*result = 0.1;
64
    else
65
	*result = (float64data)getattdisbursion(relid, (int) attno);
66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96
    return(result);
}

/*
 *	neqsel		- Selectivity of "!=" for any data type.
 */
float64
neqsel(Oid opid,
       Oid relid,
       AttrNumber attno,
       char *value,
       int32 flag)
{
    float64		result;
    
    result = eqsel(opid, relid, attno, value, flag);
    *result = 1.0 - *result;
    return(result);
}

/*
 *	intltsel	- Selectivity of "<" for integers.
 *			  Should work for both longs and shorts.
 */
float64
intltsel(Oid opid,
	 Oid relid,
	 AttrNumber attno,
	 int32 value,
	 int32 flag)
{
97
    float64 		result;
98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115
    char		*highchar, *lowchar;
    long		val, high, low, top, bottom;
    
    result = (float64) palloc(sizeof(float64data));
    if (NONVALUE(attno) || NONVALUE(relid))
	*result = 1.0 / 3;
    else {
	/* XXX          val = atol(value);*/
	val = value;
	gethilokey(relid, (int) attno, opid, &highchar, &lowchar);
	if (*highchar == 'n' || *lowchar == 'n') {
	    *result = 1.0/3.0;
	    return (result);
	}
	high = atol(highchar);
	low = atol(lowchar);
	if ((flag & SEL_RIGHT && val < low) ||
	    (!(flag & SEL_RIGHT) && val > high)) {
116
	    float32data nvals;
117
	    nvals = getattdisbursion(relid, (int) attno);
118 119
	    if (nvals == 0)
		*result = 1.0 / 3.0;
120 121 122 123 124
	    else {
		*result = 3.0 * (float64data)nvals;
		if (*result > 1.0)
		    *result = 1;
	    }
125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177
	}else {
	    bottom = high - low;
	    if (bottom == 0)
		++bottom;
	    if (flag & SEL_RIGHT)
		top = val - low;
	    else
		top = high - val;
	    if (top > bottom)
		*result = 1.0;
	    else {
		if (top == 0)
		    ++top;
		*result = ((1.0 * top) / bottom);
	    }
	}
    }
    return(result);
}

/*
 *	intgtsel	- Selectivity of ">" for integers.
 *			  Should work for both longs and shorts.
 */
float64
intgtsel(Oid opid,
	 Oid relid,
	 AttrNumber attno,
	 int32 value,
	 int32 flag)
{
    float64		result;
    int		notflag;
    
    if (flag & 0)
	notflag = flag & ~SEL_RIGHT;
    else
	notflag = flag | SEL_RIGHT;
    result = intltsel(opid, relid, attno, value, (int32) notflag);
    return(result);
}

/*
 *	eqjoinsel	- Join selectivity of "="
 */
float64
eqjoinsel(Oid opid,
	  Oid relid1,
	  AttrNumber attno1,
	  Oid relid2,
	  AttrNumber attno2)
{
    float64		result;
178
    float32data		num1, num2, max;
179 180 181 182 183 184
    
    result = (float64) palloc(sizeof(float64data));
    if (NONVALUE(attno1) || NONVALUE(relid1) ||
	NONVALUE(attno2) || NONVALUE(relid2))
	*result = 0.1;
    else {
185 186
	num1 = getattdisbursion(relid1, (int) attno1);
	num2 = getattdisbursion(relid2, (int) attno2);
187 188 189 190
	max = (num1 > num2) ? num1 : num2;
	if (max == 0)
	    *result = 1.0;
	else
191
	    *result = (float64data)max;
192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247
    }
    return(result);
}

/*
 *	neqjoinsel	- Join selectivity of "!="
 */
float64
neqjoinsel(Oid opid,
	   Oid relid1,
	   AttrNumber attno1,
	   Oid relid2,
	   AttrNumber attno2)
{
    float64		result;
    
    result = eqjoinsel(opid, relid1, attno1, relid2, attno2);
    *result = 1.0 - *result;
    return(result);
}

/*
 *	intltjoinsel	- Join selectivity of "<"
 */
float64
intltjoinsel(Oid opid,
	     Oid relid1,
	     AttrNumber attno1,
	     Oid relid2,
	     AttrNumber attno2)
{
    float64	result;
    
    result = (float64) palloc(sizeof(float64data));
    *result = 1.0 / 3.0;
    return(result);
}

/*
 *	intgtjoinsel	- Join selectivity of ">"
 */
float64
intgtjoinsel(Oid opid,
	     Oid relid1,
	     AttrNumber attno1,
	     Oid relid2,
	     AttrNumber attno2)
{
    float64	result;
    
    result = (float64) palloc(sizeof(float64data));
    *result = 1.0 / 3.0;
    return(result);
}

/*
248
 *	getattdisbursion	- Retrieves the number of values within an attribute.
249 250
 *
 *	Note:
251
 *		getattdisbursion and gethilokey both currently use keyed
252 253 254 255 256 257 258 259 260
 *		relation scans and amgetattr.  Alternatively,
 *		the relation scan could be non-keyed and the tuple
 *		returned could be cast (struct X *) tuple + tuple->t_hoff.
 *		The first method is good for testing the implementation,
 *		but the second may ultimately be faster?!?  In any case,
 *		using the cast instead of amgetattr would be
 *		more efficient.  However, the cast will not work
 *		for gethilokey which accesses stahikey in struct statistic.
 */
261
static float32data
262
getattdisbursion(Oid relid, AttrNumber attnum)
263 264
{
    HeapTuple	atp;
265 266
    float32data	nvals;
    int32 ntuples;
267 268 269 270 271 272
    
    atp = SearchSysCacheTuple(ATTNUM, 
			      ObjectIdGetDatum(relid), 
			      Int16GetDatum(attnum),
			      0,0);
    if (!HeapTupleIsValid(atp)) {
273
	elog(WARN, "getattdisbursion: no attribute tuple %d %d",
274 275 276
	     relid, attnum);
	return(0);
    }
277
    nvals = ((AttributeTupleForm ) GETSTRUCT(atp))->attdisbursion;
278 279 280 281 282 283 284 285
    if (nvals > 0) return(nvals);
    
    atp = SearchSysCacheTuple(RELOID, ObjectIdGetDatum(relid),
			      0,0,0);
    /* XXX -- use number of tuples as number of distinctive values
       just for now, in case number of distinctive values is
       not cached */
    if (!HeapTupleIsValid(atp)) {
286
	elog(WARN, "getattdisbursion: no relation tuple %d", relid);
287 288
	return(0);
    }
289
    ntuples = ((Form_pg_class) GETSTRUCT(atp))->reltuples;
290
    /* Look above how nvals is used. 	- vadim 04/09/97 */
291 292
    if ( ntuples > 0 )
    	nvals = 1.0 / ntuples;
293
    
294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384
    return(nvals);
}

/*
 *	gethilokey	- Returns a pointer to strings containing
 *			  the high and low keys within an attribute.
 *
 *	Currently returns "0", and "0" in high and low if the statistic
 *	catalog does not contain the proper tuple.  Eventually, the
 *	statistic demon should have the tuple maintained, and it should
 *	elog() if the tuple is missing.
 *
 *	XXX Question: is this worth sticking in the catalog caches,
 *	    or will this get invalidated too often?
 */
static void
gethilokey(Oid relid,
	   AttrNumber attnum,
	   Oid opid,
	   char **high,
	   char **low)
{
    register Relation	rdesc;
    register HeapScanDesc	sdesc;
    static ScanKeyData	key[3] = {
	{ 0, Anum_pg_statistic_starelid, F_OIDEQ },
	{ 0, Anum_pg_statistic_staattnum, F_INT2EQ },
	{ 0, Anum_pg_statistic_staop, F_OIDEQ }
    };
    bool		isnull;
    HeapTuple		tuple;
    
    rdesc = heap_openr(StatisticRelationName);

    key[0].sk_argument = ObjectIdGetDatum(relid);
    key[1].sk_argument = Int16GetDatum((int16) attnum);
    key[2].sk_argument = ObjectIdGetDatum(opid);
    sdesc = heap_beginscan(rdesc, 0, NowTimeQual, 3, key);
    tuple = heap_getnext(sdesc, 0, (Buffer *) NULL);
    if (!HeapTupleIsValid(tuple)) {
	*high = "n";
	*low = "n";
	/* XXX 		elog(WARN, "gethilokey: statistic tuple not found");*/
	return;
    }
    *high = textout((struct varlena *)
		    heap_getattr(tuple,
				 InvalidBuffer,
				 Anum_pg_statistic_stahikey,
				 RelationGetTupleDescriptor(rdesc),
				 &isnull));
    if (isnull)
	elog(DEBUG, "gethilokey: high key is null");
    *low = textout((struct varlena *)
		   heap_getattr(tuple,
				InvalidBuffer,
				Anum_pg_statistic_stalokey,
				RelationGetTupleDescriptor(rdesc),
				&isnull));
    if (isnull)
	elog(DEBUG, "gethilokey: low key is null");
    heap_endscan(sdesc);
    heap_close(rdesc);
}

float64
btreesel(Oid operatorObjectId,
	 Oid indrelid,
	 AttrNumber attributeNumber,
	 char *constValue,
	 int32 constFlag,
	 int32 nIndexKeys,
	 Oid indexrelid)
{
    float64 result;
    float64data resultData;
    
    if (FunctionalSelectivity(nIndexKeys, attributeNumber)) {
	/*
	 * Need to call the functions selectivity
	 * function here.  For now simply assume it's
	 * 1/3 since functions don't currently
	 * have selectivity functions
	 */
	resultData = 1.0 / 3.0;
	result = &resultData;
    }
    else {
	result = (float64)fmgr(get_oprrest (operatorObjectId),
			       (char*)operatorObjectId,
			       (char*)indrelid,
385
			       (char*)(int)attributeNumber,
386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426
			       (char*)constValue,
			       (char*)constFlag,
			       NULL);
    }
    
    if (!PointerIsValid(result))
	elog(WARN, "Btree Selectivity: bad pointer");
    if (*result < 0.0 || *result > 1.0)
	elog(WARN, "Btree Selectivity: bad value %lf", *result);
    
    return(result);
}

float64
btreenpage(Oid operatorObjectId,
	   Oid indrelid,
	   AttrNumber attributeNumber,
	   char *constValue,
	   int32 constFlag,
	   int32 nIndexKeys,
	   Oid indexrelid)
{
    float64 temp, result;
    float64data tempData;
    HeapTuple atp;
    int npage;
    
    if (FunctionalSelectivity(nIndexKeys, attributeNumber)) {
	/*
	 * Need to call the functions selectivity
	 * function here.  For now simply assume it's
	 * 1/3 since functions don't currently
	 * have selectivity functions
	 */
	tempData = 1.0 / 3.0;
	temp = &tempData;
    }
    else {
	temp = (float64)fmgr(get_oprrest (operatorObjectId),
			     (char*)operatorObjectId,
			     (char*)indrelid,
427
			     (char*)(int)attributeNumber,
428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488
			     (char*)constValue,
			     (char*)constFlag,
			     NULL);
    }
    atp = SearchSysCacheTuple(RELOID, 
			      ObjectIdGetDatum(indexrelid),
			      0,0,0);
    if (!HeapTupleIsValid(atp)) {
	elog(WARN, "btreenpage: no index tuple %d", indexrelid);
	return(0);
    }
    
    npage = ((Form_pg_class) GETSTRUCT(atp))->relpages;
    result = (float64)palloc(sizeof(float64data));
    *result = *temp * npage;
    return(result);
}

float64
hashsel(Oid operatorObjectId,
	Oid indrelid,
	AttrNumber attributeNumber,
	char *constValue,
	int32 constFlag,
	int32 nIndexKeys,
	Oid indexrelid)
{
    
    float64 result;
    float64data resultData;
    HeapTuple atp;
    int ntuples;
    
    if (FunctionalSelectivity(nIndexKeys, attributeNumber)) {
	/*
	 * Need to call the functions selectivity
	 * function here.  For now simply use 1/Number of Tuples
	 * since functions don't currently
	 * have selectivity functions
	 */
	
	atp = SearchSysCacheTuple(RELOID, ObjectIdGetDatum(indexrelid),
				  0,0,0);
	if (!HeapTupleIsValid(atp)) {
	    elog(WARN, "hashsel: no index tuple %d", indexrelid);
	    return(0);
	}
	ntuples = ((Form_pg_class) GETSTRUCT(atp))->reltuples;
	if (ntuples > 0) {
	    resultData =  1.0 / (float64data) ntuples;
	}
	else {
	    resultData = (float64data) (1.0 / 100.0);
	}
	result = &resultData;
	
    }
    else {
	result = (float64)fmgr(get_oprrest (operatorObjectId),
			       (char*)operatorObjectId,
			       (char*)indrelid,
489
			       (char*)(int)attributeNumber,
490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549
			       (char*)constValue,
			       (char*)constFlag,
			       NULL);
    }
    
    if (!PointerIsValid(result))
	elog(WARN, "Hash Table Selectivity: bad pointer");
    if (*result < 0.0 || *result > 1.0)
	elog(WARN, "Hash Table Selectivity: bad value %lf", *result);
    
    return(result);
    
    
}

float64
hashnpage(Oid operatorObjectId,
	  Oid indrelid,
	  AttrNumber attributeNumber,
	  char *constValue,
	  int32 constFlag,
	  int32 nIndexKeys,
	  Oid indexrelid)
{
    float64 temp, result;
    float64data tempData;
    HeapTuple atp;
    int npage;
    int ntuples;
    
    atp = SearchSysCacheTuple(RELOID, ObjectIdGetDatum(indexrelid),
			      0,0,0);
    if (!HeapTupleIsValid(atp)) {
	elog(WARN, "hashsel: no index tuple %d", indexrelid);
	return(0);
    }
    
    
    if (FunctionalSelectivity(nIndexKeys, attributeNumber)) {
	/*
	 * Need to call the functions selectivity
	 * function here.  For now, use 1/Number of Tuples
	 * since functions don't currently
	 * have selectivity functions
	 */
	
	ntuples = ((Form_pg_class) GETSTRUCT(atp))->reltuples;
	if (ntuples > 0) {
	    tempData =  1.0 / (float64data) ntuples;
	}
	else {
	    tempData = (float64data) (1.0 / 100.0);
	}
	temp = &tempData;
	
    }
    else {
	temp = (float64)fmgr(get_oprrest (operatorObjectId),
			     (char*)operatorObjectId,
			     (char*)indrelid,
550
			     (char*)(int)attributeNumber,
551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587
			     (char*)constValue,
			     (char*)constFlag,
			     NULL);
    }
    
    npage = ((Form_pg_class) GETSTRUCT(atp))->relpages;
    result = (float64)palloc(sizeof(float64data));
    *result = *temp * npage;
    return(result);
}


float64
rtsel(Oid operatorObjectId,
      Oid indrelid,
      AttrNumber attributeNumber,
      char *constValue,
      int32 constFlag,
      int32 nIndexKeys,
      Oid indexrelid)
{
    return (btreesel(operatorObjectId, indrelid, attributeNumber,
		     constValue, constFlag, nIndexKeys, indexrelid));
}

float64
rtnpage(Oid operatorObjectId,
	Oid indrelid,
	AttrNumber attributeNumber,
	char *constValue,
	int32 constFlag,
	int32 nIndexKeys,
	Oid indexrelid)
{
    return (btreenpage(operatorObjectId, indrelid, attributeNumber,
		       constValue, constFlag, nIndexKeys, indexrelid));
}
588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613

float64
gistsel(Oid operatorObjectId,
      Oid indrelid,
      AttrNumber attributeNumber,
      char *constValue,
      int32 constFlag,
      int32 nIndexKeys,
      Oid indexrelid)
{
    return (btreesel(operatorObjectId, indrelid, attributeNumber,
		     constValue, constFlag, nIndexKeys, indexrelid));
}

float64
gistnpage(Oid operatorObjectId,
	  Oid indrelid,
	  AttrNumber attributeNumber,
	  char *constValue,
	  int32 constFlag,
	  int32 nIndexKeys,
	  Oid indexrelid)
{
    return (btreenpage(operatorObjectId, indrelid, attributeNumber,
		       constValue, constFlag, nIndexKeys, indexrelid));
}