heapam.c 50.6 KB
Newer Older
1 2
/*-------------------------------------------------------------------------
 *
3
 * heapam.c
4
 *	  heap access method code
5
 *
Bruce Momjian's avatar
Add:  
Bruce Momjian committed
6 7
 * Portions Copyright (c) 1996-2000, PostgreSQL, Inc
 * Portions Copyright (c) 1994, Regents of the University of California
8 9 10
 *
 *
 * IDENTIFICATION
11
 *	  $Header: /cvsroot/pgsql/src/backend/access/heap/heapam.c,v 1.80 2000/07/21 10:31:30 wieck Exp $
12 13 14
 *
 *
 * INTERFACE ROUTINES
15 16 17 18 19 20 21 22 23 24 25
 *		heapgettup		- fetch next heap tuple from a scan
 *		heap_open		- open a heap relation by relationId
 *		heap_openr		- open a heap relation by name
 *		heap_close		- close a heap relation
 *		heap_beginscan	- begin relation scan
 *		heap_rescan		- restart a relation scan
 *		heap_endscan	- end relation scan
 *		heap_getnext	- retrieve next tuple in scan
 *		heap_fetch		- retrive tuple with tid
 *		heap_insert		- insert tuple into a relation
 *		heap_delete		- delete a tuple from a relation
26
 *		heap_update - replace a tuple in a relation with another tuple
27 28 29
 *		heap_markpos	- mark scan position
 *		heap_restrpos	- restore position to marked location
 *
30
 * NOTES
31 32 33
 *	  This file contains the heap_ routines which implement
 *	  the POSTGRES heap access method used for all POSTGRES
 *	  relations.
34 35
 *
 * OLD COMMENTS
36
 *		struct relscan hints:  (struct should be made AM independent?)
37
 *
38 39 40 41
 *		rs_ctid is the tid of the last tuple returned by getnext.
 *		rs_ptid and rs_ntid are the tids of the previous and next tuples
 *		returned by getnext, respectively.	NULL indicates an end of
 *		scan (either direction); NON indicates an unknow value.
42
 *
43 44 45 46 47 48 49 50 51 52 53 54 55 56 57
 *		possible combinations:
 *		rs_p	rs_c	rs_n			interpretation
 *		NULL	NULL	NULL			empty scan
 *		NULL	NULL	NON				at begining of scan
 *		NULL	NULL	t1				at begining of scan (with cached tid)
 *		NON		NULL	NULL			at end of scan
 *		t1		NULL	NULL			at end of scan (with cached tid)
 *		NULL	t1		NULL			just returned only tuple
 *		NULL	t1		NON				just returned first tuple
 *		NULL	t1		t2				returned first tuple (with cached tid)
 *		NON		t1		NULL			just returned last tuple
 *		t2		t1		NULL			returned last tuple (with cached tid)
 *		t1		t2		NON				in the middle of a forward scan
 *		NON		t2		t1				in the middle of a reverse scan
 *		ti		tj		tk				in the middle of a scan (w cached tid)
58
 *
59 60
 *		Here NULL is ...tup == NULL && ...buf == InvalidBuffer,
 *		and NON is ...tup == NULL && ...buf == UnknownBuffer.
61
 *
62 63 64
 *		Currently, the NONTID values are not cached with their actual
 *		values by getnext.	Values may be cached by markpos since it stores
 *		all three tids.
65
 *
66 67
 *		NOTE:  the calls to elog() must stop.  Should decide on an interface
 *		between the general and specific AM calls.
68
 *
69 70 71 72
 *		XXX probably do not need a free tuple routine for heaps.
 *		Huh?  Free tuple is not necessary for tuples returned by scans, but
 *		is necessary for tuples which are returned by
 *		RelationGetTupleByItemPointer. -hirohama
73 74 75 76
 *
 *-------------------------------------------------------------------------
 */

77
#include "postgres.h"
78

79 80
#include "access/heapam.h"
#include "access/hio.h"
Tom Lane's avatar
Tom Lane committed
81
#include "access/tuptoaster.h"
Bruce Momjian's avatar
Bruce Momjian committed
82
#include "access/valid.h"
83
#include "catalog/catalog.h"
Bruce Momjian's avatar
Bruce Momjian committed
84 85 86
#include "miscadmin.h"
#include "utils/inval.h"
#include "utils/relcache.h"
87

Marc G. Fournier's avatar
Marc G. Fournier committed
88

89
/* ----------------------------------------------------------------
90
 *						 heap support routines
91 92 93 94
 * ----------------------------------------------------------------
 */

/* ----------------
95
 *		initscan - scan code common to heap_beginscan and heap_rescan
96 97 98
 * ----------------
 */
static void
99
initscan(HeapScanDesc scan,
100 101 102 103
		 Relation relation,
		 int atend,
		 unsigned nkeys,
		 ScanKey key)
104
{
105 106 107 108 109 110 111 112 113 114
	/* ----------------
	 *	Make sure we have up-to-date idea of number of blocks in relation.
	 *	It is sufficient to do this once at scan start, since any tuples
	 *	added while the scan is in progress will be invisible to my
	 *	transaction anyway...
	 * ----------------
	 */
	relation->rd_nblocks = RelationGetNumberOfBlocks(relation);

	if (relation->rd_nblocks == 0)
115 116 117 118 119
	{
		/* ----------------
		 *	relation is empty
		 * ----------------
		 */
120
		scan->rs_ntup.t_datamcxt = scan->rs_ctup.t_datamcxt =
121
			scan->rs_ptup.t_datamcxt = NULL;
Bruce Momjian's avatar
Bruce Momjian committed
122
		scan->rs_ntup.t_data = scan->rs_ctup.t_data =
123
			scan->rs_ptup.t_data = NULL;
124
		scan->rs_nbuf = scan->rs_cbuf = scan->rs_pbuf = InvalidBuffer;
125 126 127 128 129 130 131
	}
	else if (atend)
	{
		/* ----------------
		 *	reverse scan
		 * ----------------
		 */
132
		scan->rs_ntup.t_datamcxt = scan->rs_ctup.t_datamcxt = NULL;
133
		scan->rs_ntup.t_data = scan->rs_ctup.t_data = NULL;
134
		scan->rs_nbuf = scan->rs_cbuf = InvalidBuffer;
135
		scan->rs_ptup.t_datamcxt = NULL;
136
		scan->rs_ptup.t_data = NULL;
137
		scan->rs_pbuf = UnknownBuffer;
138 139 140 141 142 143 144
	}
	else
	{
		/* ----------------
		 *	forward scan
		 * ----------------
		 */
145
		scan->rs_ctup.t_datamcxt = scan->rs_ptup.t_datamcxt = NULL;
146
		scan->rs_ctup.t_data = scan->rs_ptup.t_data = NULL;
147
		scan->rs_cbuf = scan->rs_pbuf = InvalidBuffer;
148
		scan->rs_ntup.t_datamcxt = NULL;
149
		scan->rs_ntup.t_data = NULL;
150
		scan->rs_nbuf = UnknownBuffer;
151 152 153
	}							/* invalid too */

	/* we don't have a marked position... */
154 155 156 157
	ItemPointerSetInvalid(&(scan->rs_mptid));
	ItemPointerSetInvalid(&(scan->rs_mctid));
	ItemPointerSetInvalid(&(scan->rs_mntid));
	ItemPointerSetInvalid(&(scan->rs_mcd));
158

159
	/* ----------------
160
	 *	copy the scan key, if appropriate
161 162
	 * ----------------
	 */
163
	if (key != NULL)
164
		memmove(scan->rs_key, key, nkeys * sizeof(ScanKeyData));
165 166 167
}

/* ----------------
168
 *		unpinscan - code common to heap_rescan and heap_endscan
169 170 171
 * ----------------
 */
static void
172
unpinscan(HeapScanDesc scan)
173
{
174 175
	if (BufferIsValid(scan->rs_pbuf))
		ReleaseBuffer(scan->rs_pbuf);
176 177

	/* ------------------------------------
178
	 *	Scan will pin buffer once for each non-NULL tuple pointer
179 180 181 182
	 *	(ptup, ctup, ntup), so they have to be unpinned multiple
	 *	times.
	 * ------------------------------------
	 */
183 184
	if (BufferIsValid(scan->rs_cbuf))
		ReleaseBuffer(scan->rs_cbuf);
185

186 187
	if (BufferIsValid(scan->rs_nbuf))
		ReleaseBuffer(scan->rs_nbuf);
188

189 190 191
	/*
	 * we don't bother to clear rs_pbuf etc --- caller must reinitialize
	 * them if scan descriptor is not being deleted.
192
	 */
193 194 195
}

/* ------------------------------------------
196
 *		nextpage
197
 *
198 199 200
 *		figure out the next page to scan after the current page
 *		taking into account of possible adjustment of degrees of
 *		parallelism
201 202 203 204 205
 * ------------------------------------------
 */
static int
nextpage(int page, int dir)
{
206
	return (dir < 0) ? page - 1 : page + 1;
207 208 209
}

/* ----------------
210
 *		heapgettup - fetch next heap tuple
211
 *
212 213
 *		routine used by heap_getnext() which does most of the
 *		real work in scanning tuples.
214 215 216 217 218
 *
 *		The scan routines handle their own buffer lock/unlocking, so
 *		there is no reason to request the buffer number unless
 *		to want to perform some other operation with the result,
 *		like pass it to another function.
219 220
 * ----------------
 */
221
static void
222
heapgettup(Relation relation,
223
		   HeapTuple tuple,
224
		   int dir,
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
225
		   Buffer *buffer,
226
		   Snapshot snapshot,
227 228
		   int nkeys,
		   ScanKey key)
229
{
Bruce Momjian's avatar
Bruce Momjian committed
230 231 232 233 234 235 236 237 238
	ItemId		lpp;
	Page		dp;
	int			page;
	int			pages;
	int			lines;
	OffsetNumber lineoff;
	int			linesleft;
	ItemPointer tid = (tuple->t_data == NULL) ?
	(ItemPointer) NULL : &(tuple->t_self);
239 240
    
	tuple->tableOid = relation->rd_id;
241

242
	/* ----------------
243
	 *	increment access statistics
244 245
	 * ----------------
	 */
246 247 248
	IncrHeapAccessStat(local_heapgettup);
	IncrHeapAccessStat(global_heapgettup);

249
	/* ----------------
250 251 252 253
	 *	debugging stuff
	 *
	 * check validity of arguments, here and for other functions too
	 * Note: no locking manipulations needed--this is a local function
254 255
	 * ----------------
	 */
256 257 258
#ifdef	HEAPDEBUGALL
	if (ItemPointerIsValid(tid))
	{
259
		elog(DEBUG, "heapgettup(%s, tid=0x%x[%d,%d], dir=%d, ...)",
260 261
			 RelationGetRelationName(relation), tid, tid->ip_blkid,
			 tid->ip_posid, dir);
262
	}
263 264
	else
	{
265
		elog(DEBUG, "heapgettup(%s, tid=0x%x, dir=%d, ...)",
266
			 RelationGetRelationName(relation), tid, dir);
267
	}
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
268
	elog(DEBUG, "heapgettup(..., b=0x%x, nkeys=%d, key=0x%x", buffer, nkeys, key);
269

270
	elog(DEBUG, "heapgettup: relation(%c)=`%s', %p",
271
		 relation->rd_rel->relkind, RelationGetRelationName(relation),
272
		 snapshot);
273
#endif	 /* !defined(HEAPDEBUGALL) */
274 275 276

	if (!ItemPointerIsValid(tid))
		Assert(!PointerIsValid(tid));
277 278

	/* ----------------
279
	 *	return null immediately if relation is empty
280 281
	 * ----------------
	 */
282
	if (!(pages = relation->rd_nblocks))
283
	{
284
		tuple->t_datamcxt = NULL;
285 286 287
		tuple->t_data = NULL;
		return;
	}
288 289 290 291 292 293 294 295 296 297 298 299 300 301

	/* ----------------
	 *	calculate next starting lineoff, given scan direction
	 * ----------------
	 */
	if (!dir)
	{
		/* ----------------
		 * ``no movement'' scan direction
		 * ----------------
		 */
		/* assume it is a valid TID XXX */
		if (ItemPointerIsValid(tid) == false)
		{
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
302
			*buffer = InvalidBuffer;
303
			tuple->t_datamcxt = NULL;
304 305
			tuple->t_data = NULL;
			return;
306
		}
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
307
		*buffer = RelationGetBufferWithBuffer(relation,
Bruce Momjian's avatar
Bruce Momjian committed
308 309
										  ItemPointerGetBlockNumber(tid),
											  *buffer);
310

Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
311
		if (!BufferIsValid(*buffer))
312
			elog(ERROR, "heapgettup: failed ReadBuffer");
313

Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
314 315 316
		LockBuffer(*buffer, BUFFER_LOCK_SHARE);

		dp = (Page) BufferGetPage(*buffer);
317 318 319
		lineoff = ItemPointerGetOffsetNumber(tid);
		lpp = PageGetItemId(dp, lineoff);

320
		tuple->t_datamcxt = NULL;
321 322
		tuple->t_data = (HeapTupleHeader) PageGetItem((Page) dp, lpp);
		tuple->t_len = ItemIdGetLength(lpp);
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
323
		LockBuffer(*buffer, BUFFER_LOCK_UNLOCK);
324
		return;
325

326
	}
327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344
	else if (dir < 0)
	{
		/* ----------------
		 *	reverse scan direction
		 * ----------------
		 */
		if (ItemPointerIsValid(tid) == false)
			tid = NULL;
		if (tid == NULL)
		{
			page = pages - 1;	/* final page */
		}
		else
		{
			page = ItemPointerGetBlockNumber(tid);		/* current page */
		}
		if (page < 0)
		{
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
345
			*buffer = InvalidBuffer;
346 347
			tuple->t_data = NULL;
			return;
348 349
		}

Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
350 351
		*buffer = RelationGetBufferWithBuffer(relation, page, *buffer);
		if (!BufferIsValid(*buffer))
352
			elog(ERROR, "heapgettup: failed ReadBuffer");
353

Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
354 355 356
		LockBuffer(*buffer, BUFFER_LOCK_SHARE);

		dp = (Page) BufferGetPage(*buffer);
357 358 359 360
		lines = PageGetMaxOffsetNumber(dp);
		if (tid == NULL)
		{
			lineoff = lines;	/* final offnum */
361
		}
362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389
		else
		{
			lineoff =			/* previous offnum */
				OffsetNumberPrev(ItemPointerGetOffsetNumber(tid));
		}
		/* page and lineoff now reference the physically previous tid */

	}
	else
	{
		/* ----------------
		 *	forward scan direction
		 * ----------------
		 */
		if (ItemPointerIsValid(tid) == false)
		{
			page = 0;			/* first page */
			lineoff = FirstOffsetNumber;		/* first offnum */
		}
		else
		{
			page = ItemPointerGetBlockNumber(tid);		/* current page */
			lineoff =			/* next offnum */
				OffsetNumberNext(ItemPointerGetOffsetNumber(tid));
		}

		if (page >= pages)
		{
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
390
			*buffer = InvalidBuffer;
391
			tuple->t_datamcxt = NULL;
392 393
			tuple->t_data = NULL;
			return;
394 395 396
		}
		/* page and lineoff now reference the physically next tid */

Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
397 398
		*buffer = RelationGetBufferWithBuffer(relation, page, *buffer);
		if (!BufferIsValid(*buffer))
399
			elog(ERROR, "heapgettup: failed ReadBuffer");
400

Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
401 402 403
		LockBuffer(*buffer, BUFFER_LOCK_SHARE);

		dp = (Page) BufferGetPage(*buffer);
404
		lines = PageGetMaxOffsetNumber(dp);
405
	}
406 407 408

	/* 'dir' is now non-zero */

409
	/* ----------------
410 411
	 *	calculate line pointer and number of remaining items
	 *	to check on this page.
412 413
	 * ----------------
	 */
414 415 416 417 418 419
	lpp = PageGetItemId(dp, lineoff);
	if (dir < 0)
		linesleft = lineoff - 1;
	else
		linesleft = lines - lineoff;

420
	/* ----------------
421 422
	 *	advance the scan until we find a qualifying tuple or
	 *	run out of stuff to scan
423 424
	 * ----------------
	 */
425 426 427 428
	for (;;)
	{
		while (linesleft >= 0)
		{
429
			if (ItemIdIsUsed(lpp))
430
			{
431
				tuple->t_datamcxt = NULL;
432 433 434 435 436 437 438
				tuple->t_data = (HeapTupleHeader) PageGetItem((Page) dp, lpp);
				tuple->t_len = ItemIdGetLength(lpp);
				ItemPointerSet(&(tuple->t_self), page, lineoff);
				/* ----------------
				 *	if current tuple qualifies, return it.
				 * ----------------
				 */
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
439
				HeapTupleSatisfies(tuple, relation, *buffer, (PageHeader) dp,
440 441
								   snapshot, nkeys, key);
				if (tuple->t_data != NULL)
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
442 443
				{
					LockBuffer(*buffer, BUFFER_LOCK_UNLOCK);
444
					return;
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
445
				}
446 447 448 449 450 451 452 453 454 455
			}

			/* ----------------
			 *	otherwise move to the next item on the page
			 * ----------------
			 */
			--linesleft;
			if (dir < 0)
			{
				--lpp;			/* move back in this page's ItemId array */
456
				--lineoff;
457 458 459
			}
			else
			{
Bruce Momjian's avatar
Bruce Momjian committed
460 461
				++lpp;			/* move forward in this page's ItemId
								 * array */
462
				++lineoff;
463 464 465 466 467 468 469 470
			}
		}

		/* ----------------
		 *	if we get here, it means we've exhausted the items on
		 *	this page and it's time to move to the next..
		 * ----------------
		 */
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
471
		LockBuffer(*buffer, BUFFER_LOCK_UNLOCK);
472 473 474 475 476 477 478 479
		page = nextpage(page, dir);

		/* ----------------
		 *	return NULL if we've exhausted all the pages..
		 * ----------------
		 */
		if (page < 0 || page >= pages)
		{
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
480 481 482
			if (BufferIsValid(*buffer))
				ReleaseBuffer(*buffer);
			*buffer = InvalidBuffer;
483
			tuple->t_datamcxt = NULL;
484 485
			tuple->t_data = NULL;
			return;
486 487
		}

Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
488
		*buffer = ReleaseAndReadBuffer(*buffer, relation, page);
489

Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
490
		if (!BufferIsValid(*buffer))
491
			elog(ERROR, "heapgettup: failed ReadBuffer");
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
492 493
		LockBuffer(*buffer, BUFFER_LOCK_SHARE);
		dp = (Page) BufferGetPage(*buffer);
494
		lines = PageGetMaxOffsetNumber((Page) dp);
495 496
		linesleft = lines - 1;
		if (dir < 0)
497 498 499 500
		{
			lineoff = lines;
			lpp = PageGetItemId(dp, lines);
		}
501
		else
502 503
		{
			lineoff = FirstOffsetNumber;
504
			lpp = PageGetItemId(dp, FirstOffsetNumber);
505
		}
506 507 508 509
	}
}


510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563
#if defined(DISABLE_COMPLEX_MACRO)
/*
 * This is formatted so oddly so that the correspondence to the macro
 * definition in access/heapam.h is maintained.
 */
Datum
fastgetattr(HeapTuple tup, int attnum, TupleDesc tupleDesc,
			bool *isnull)
{
	return (
			(attnum) > 0 ?
			(
			 ((isnull) ? (*(isnull) = false) : (dummyret) NULL),
			 HeapTupleNoNulls(tup) ?
			 (
			  ((tupleDesc)->attrs[(attnum) - 1]->attcacheoff != -1 ||
			   (attnum) == 1) ?
			  (
			   (Datum) fetchatt(&((tupleDesc)->attrs[(attnum) - 1]),
						 (char *) (tup)->t_data + (tup)->t_data->t_hoff +
								(
								 ((attnum) != 1) ?
							(tupleDesc)->attrs[(attnum) - 1]->attcacheoff
								 :
								 0
								 )
								)
			   )
			  :
			  nocachegetattr((tup), (attnum), (tupleDesc), (isnull))
			  )
			 :
			 (
			  att_isnull((attnum) - 1, (tup)->t_data->t_bits) ?
			  (
			   ((isnull) ? (*(isnull) = true) : (dummyret) NULL),
			   (Datum) NULL
			   )
			  :
			  (
			   nocachegetattr((tup), (attnum), (tupleDesc), (isnull))
			   )
			  )
			 )
			:
			(
			 (Datum) NULL
			 )
	);
}
#endif /* defined(DISABLE_COMPLEX_MACRO)*/



564
/* ----------------------------------------------------------------
565
 *					 heap access method interface
566 567 568
 * ----------------------------------------------------------------
 */
/* ----------------
569
 *		heap_open - open a heap relation by relationId
570
 *
571 572 573 574 575
 *		If lockmode is "NoLock", no lock is obtained on the relation,
 *		and the caller must check for a NULL return value indicating
 *		that no such relation exists.
 *		Otherwise, an error is raised if the relation does not exist,
 *		and the specified kind of lock is obtained on the relation.
576 577 578
 * ----------------
 */
Relation
579
heap_open(Oid relationId, LOCKMODE lockmode)
580
{
581
	Relation	r;
582

583 584
	Assert(lockmode >= NoLock && lockmode < MAX_LOCKMODES);

585 586 587 588 589 590 591
	/* ----------------
	 *	increment access statistics
	 * ----------------
	 */
	IncrHeapAccessStat(local_open);
	IncrHeapAccessStat(global_open);

592 593
	/* The relcache does all the real work... */
	r = RelationIdGetRelation(relationId);
594

595
	/* Under no circumstances will we return an index as a relation. */
596
	if (RelationIsValid(r) && r->rd_rel->relkind == RELKIND_INDEX)
597
		elog(ERROR, "%s is an index relation", RelationGetRelationName(r));
598

599 600 601
	if (lockmode == NoLock)
		return r;				/* caller must check RelationIsValid! */

602
	if (!RelationIsValid(r))
603 604 605 606
		elog(ERROR, "Relation %u does not exist", relationId);

	LockRelation(r, lockmode);

607
	return r;
608 609 610
}

/* ----------------
611
 *		heap_openr - open a heap relation by name
612
 *
613 614 615 616 617
 *		If lockmode is "NoLock", no lock is obtained on the relation,
 *		and the caller must check for a NULL return value indicating
 *		that no such relation exists.
 *		Otherwise, an error is raised if the relation does not exist,
 *		and the specified kind of lock is obtained on the relation.
618 619 620
 * ----------------
 */
Relation
621
heap_openr(const char *relationName, LOCKMODE lockmode)
622
{
623
	Relation	r;
624

625 626
	Assert(lockmode >= NoLock && lockmode < MAX_LOCKMODES);

627

628 629 630 631 632 633 634
	/* ----------------
	 *	increment access statistics
	 * ----------------
	 */
	IncrHeapAccessStat(local_openr);
	IncrHeapAccessStat(global_openr);

635
	/* The relcache does all the real work... */
636 637
	r = RelationNameGetRelation(relationName);

638
	/* Under no circumstances will we return an index as a relation. */
639
	if (RelationIsValid(r) && r->rd_rel->relkind == RELKIND_INDEX)
640
		elog(ERROR, "%s is an index relation", RelationGetRelationName(r));
641

642 643 644
	if (lockmode == NoLock)
		return r;				/* caller must check RelationIsValid! */

645
	if (!RelationIsValid(r))
646 647 648 649
		elog(ERROR, "Relation '%s' does not exist", relationName);

	LockRelation(r, lockmode);

650
	return r;
651 652 653
}

/* ----------------
654
 *		heap_close - close a heap relation
655
 *
656 657 658
 *		If lockmode is not "NoLock", we first release the specified lock.
 *		Note that it is often sensible to hold a lock beyond heap_close;
 *		in that case, the lock is released automatically at xact end.
659 660 661
 * ----------------
 */
void
662
heap_close(Relation relation, LOCKMODE lockmode)
663
{
664 665
	Assert(lockmode >= NoLock && lockmode < MAX_LOCKMODES);

666 667 668 669 670 671 672
	/* ----------------
	 *	increment access statistics
	 * ----------------
	 */
	IncrHeapAccessStat(local_close);
	IncrHeapAccessStat(global_close);

673 674 675 676
	if (lockmode != NoLock)
		UnlockRelation(relation, lockmode);

	/* The relcache does the real work... */
677
	RelationClose(relation);
678 679 680 681
}


/* ----------------
682
 *		heap_beginscan	- begin relation scan
683 684 685 686
 * ----------------
 */
HeapScanDesc
heap_beginscan(Relation relation,
687
			   int atend,
688
			   Snapshot snapshot,
689 690
			   unsigned nkeys,
			   ScanKey key)
691
{
692
	HeapScanDesc scan;
693 694 695 696 697 698 699 700 701 702 703

	/* ----------------
	 *	increment access statistics
	 * ----------------
	 */
	IncrHeapAccessStat(local_beginscan);
	IncrHeapAccessStat(global_beginscan);

	/* ----------------
	 *	sanity checks
	 * ----------------
704
	 */
705
	if (!RelationIsValid(relation))
706
		elog(ERROR, "heap_beginscan: !RelationIsValid(relation)");
707 708 709 710 711 712 713

	/* ----------------
	 *	increment relation ref count while scanning relation
	 * ----------------
	 */
	RelationIncrementReferenceCount(relation);

714 715 716 717
	/* ----------------
	 *	Acquire AccessShareLock for the duration of the scan
	 *
	 *	Note: we could get an SI inval message here and consequently have
718
	 *	to rebuild the relcache entry.	The refcount increment above
719 720 721 722 723 724 725 726 727
	 *	ensures that we will rebuild it and not just flush it...
	 * ----------------
	 */
	LockRelation(relation, AccessShareLock);

	/* XXX someday assert SelfTimeQual if relkind == RELKIND_UNCATALOGED */
	if (relation->rd_rel->relkind == RELKIND_UNCATALOGED)
		snapshot = SnapshotSelf;

728 729 730 731
	/* ----------------
	 *	allocate and initialize scan descriptor
	 * ----------------
	 */
732
	scan = (HeapScanDesc) palloc(sizeof(HeapScanDescData));
733

734
	scan->rs_rd = relation;
735 736 737
	scan->rs_atend = atend;
	scan->rs_snapshot = snapshot;
	scan->rs_nkeys = (short) nkeys;
738 739

	if (nkeys)
740

741
		/*
742 743
		 * we do this here instead of in initscan() because heap_rescan
		 * also calls initscan() and we don't want to allocate memory
744 745
		 * again
		 */
746
		scan->rs_key = (ScanKey) palloc(sizeof(ScanKeyData) * nkeys);
747
	else
748
		scan->rs_key = NULL;
749

750
	initscan(scan, relation, atend, nkeys, key);
751

752
	return scan;
753 754 755
}

/* ----------------
756
 *		heap_rescan		- restart a relation scan
757 758 759
 * ----------------
 */
void
760
heap_rescan(HeapScanDesc scan,
761 762
			bool scanFromEnd,
			ScanKey key)
763
{
764 765 766 767 768 769 770 771 772 773 774 775 776
	/* ----------------
	 *	increment access statistics
	 * ----------------
	 */
	IncrHeapAccessStat(local_rescan);
	IncrHeapAccessStat(global_rescan);

	/* Note: set relation level read lock is still set */

	/* ----------------
	 *	unpin scan buffers
	 * ----------------
	 */
777
	unpinscan(scan);
778 779 780 781 782

	/* ----------------
	 *	reinitialize scan descriptor
	 * ----------------
	 */
783
	scan->rs_atend = (bool) scanFromEnd;
784
	initscan(scan, scan->rs_rd, scanFromEnd, scan->rs_nkeys, key);
785 786 787
}

/* ----------------
788
 *		heap_endscan	- end relation scan
789
 *
790 791
 *		See how to integrate with index scans.
 *		Check handling if reldesc caching.
792 793 794
 * ----------------
 */
void
795
heap_endscan(HeapScanDesc scan)
796
{
797 798 799 800 801 802 803 804 805 806 807 808 809
	/* ----------------
	 *	increment access statistics
	 * ----------------
	 */
	IncrHeapAccessStat(local_endscan);
	IncrHeapAccessStat(global_endscan);

	/* Note: no locking manipulations needed */

	/* ----------------
	 *	unpin scan buffers
	 * ----------------
	 */
810
	unpinscan(scan);
811

812 813 814 815 816 817
	/* ----------------
	 *	Release AccessShareLock acquired by heap_beginscan()
	 * ----------------
	 */
	UnlockRelation(scan->rs_rd, AccessShareLock);

818 819 820 821
	/* ----------------
	 *	decrement relation reference count and free scan descriptor storage
	 * ----------------
	 */
822
	RelationDecrementReferenceCount(scan->rs_rd);
823

824 825 826
	if (scan->rs_key)
		pfree(scan->rs_key);

827
	pfree(scan);
828 829 830
}

/* ----------------
831
 *		heap_getnext	- retrieve next tuple in scan
832
 *
833
 *		Fix to work with index relations.
834 835
 *		We don't return the buffer anymore, but you can get it from the
 *		returned HeapTuple.
836 837 838 839 840
 * ----------------
 */

#ifdef HEAPDEBUGALL
#define HEAPDEBUG_1 \
841
elog(DEBUG, "heap_getnext([%s,nkeys=%d],backw=%d) called", \
842
	 RelationGetRelationName(scan->rs_rd), scan->rs_nkeys, backw)
843

844
#define HEAPDEBUG_2 \
845 846
	 elog(DEBUG, "heap_getnext called with backw (no tracing yet)")

847
#define HEAPDEBUG_3 \
848 849
	 elog(DEBUG, "heap_getnext returns NULL at end")

850
#define HEAPDEBUG_4 \
851 852
	 elog(DEBUG, "heap_getnext valid buffer UNPIN'd")

853
#define HEAPDEBUG_5 \
854 855
	 elog(DEBUG, "heap_getnext next tuple was cached")

856
#define HEAPDEBUG_6 \
857 858
	 elog(DEBUG, "heap_getnext returning EOS")

859
#define HEAPDEBUG_7 \
860
	 elog(DEBUG, "heap_getnext returning tuple");
861 862 863 864 865 866 867 868
#else
#define HEAPDEBUG_1
#define HEAPDEBUG_2
#define HEAPDEBUG_3
#define HEAPDEBUG_4
#define HEAPDEBUG_5
#define HEAPDEBUG_6
#define HEAPDEBUG_7
869
#endif	 /* !defined(HEAPDEBUGALL) */
870 871


872
HeapTuple
873
heap_getnext(HeapScanDesc scandesc, int backw)
874
{
875
	HeapScanDesc scan = scandesc;
876

877
	/* ----------------
878
	 *	increment access statistics
879 880
	 * ----------------
	 */
881 882 883 884 885 886 887 888
	IncrHeapAccessStat(local_getnext);
	IncrHeapAccessStat(global_getnext);

	/* Note: no locking manipulations needed */

	/* ----------------
	 *	argument checks
	 * ----------------
889
	 */
890
	if (scan == NULL)
891
		elog(ERROR, "heap_getnext: NULL relscan");
892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907

	/* ----------------
	 *	initialize return buffer to InvalidBuffer
	 * ----------------
	 */

	HEAPDEBUG_1;				/* heap_getnext( info ) */

	if (backw)
	{
		/* ----------------
		 *	handle reverse scan
		 * ----------------
		 */
		HEAPDEBUG_2;			/* heap_getnext called with backw */

908
		if (scan->rs_ptup.t_data == scan->rs_ctup.t_data &&
909
			BufferIsInvalid(scan->rs_pbuf))
910
			return NULL;
911 912 913 914 915

		/*
		 * Copy the "current" tuple/buffer to "next". Pin/unpin the
		 * buffers accordingly
		 */
916
		if (scan->rs_nbuf != scan->rs_cbuf)
917
		{
918 919 920 921
			if (BufferIsValid(scan->rs_nbuf))
				ReleaseBuffer(scan->rs_nbuf);
			if (BufferIsValid(scan->rs_cbuf))
				IncrBufferRefCount(scan->rs_cbuf);
922
		}
923 924
		scan->rs_ntup = scan->rs_ctup;
		scan->rs_nbuf = scan->rs_cbuf;
925

926
		if (scan->rs_ptup.t_data != NULL)
927
		{
928
			if (scan->rs_cbuf != scan->rs_pbuf)
929
			{
930 931 932 933
				if (BufferIsValid(scan->rs_cbuf))
					ReleaseBuffer(scan->rs_cbuf);
				if (BufferIsValid(scan->rs_pbuf))
					IncrBufferRefCount(scan->rs_pbuf);
934
			}
935 936
			scan->rs_ctup = scan->rs_ptup;
			scan->rs_cbuf = scan->rs_pbuf;
937 938 939
		}
		else
		{						/* NONTUP */
Bruce Momjian's avatar
Bruce Momjian committed
940

941
			/*
942
			 * Don't release scan->rs_cbuf at this point, because
943 944 945 946 947 948 949
			 * heapgettup doesn't increase PrivateRefCount if it is
			 * already set. On a backward scan, both rs_ctup and rs_ntup
			 * usually point to the same buffer page, so
			 * PrivateRefCount[rs_cbuf] should be 2 (or more, if for
			 * instance ctup is stored in a TupleTableSlot).  - 01/09/94
			 */

950 951 952 953 954 955 956
			heapgettup(scan->rs_rd,
					   &(scan->rs_ctup),
					   -1,
					   &(scan->rs_cbuf),
					   scan->rs_snapshot,
					   scan->rs_nkeys,
					   scan->rs_key);
957 958
		}

959
		if (scan->rs_ctup.t_data == NULL && !BufferIsValid(scan->rs_cbuf))
960
		{
961 962
			if (BufferIsValid(scan->rs_pbuf))
				ReleaseBuffer(scan->rs_pbuf);
963
			scan->rs_ptup.t_datamcxt = NULL;
964
			scan->rs_ptup.t_data = NULL;
965
			scan->rs_pbuf = InvalidBuffer;
966
			return NULL;
967 968
		}

969 970
		if (BufferIsValid(scan->rs_pbuf))
			ReleaseBuffer(scan->rs_pbuf);
971
		scan->rs_ptup.t_datamcxt = NULL;
972
		scan->rs_ptup.t_data = NULL;
973
		scan->rs_pbuf = UnknownBuffer;
974 975 976 977 978 979 980 981

	}
	else
	{
		/* ----------------
		 *	handle forward scan
		 * ----------------
		 */
982
		if (scan->rs_ctup.t_data == scan->rs_ntup.t_data &&
983
			BufferIsInvalid(scan->rs_nbuf))
984 985
		{
			HEAPDEBUG_3;		/* heap_getnext returns NULL at end */
986
			return NULL;
987 988 989 990 991 992
		}

		/*
		 * Copy the "current" tuple/buffer to "previous". Pin/unpin the
		 * buffers accordingly
		 */
993
		if (scan->rs_pbuf != scan->rs_cbuf)
994
		{
995 996 997 998
			if (BufferIsValid(scan->rs_pbuf))
				ReleaseBuffer(scan->rs_pbuf);
			if (BufferIsValid(scan->rs_cbuf))
				IncrBufferRefCount(scan->rs_cbuf);
999
		}
1000 1001
		scan->rs_ptup = scan->rs_ctup;
		scan->rs_pbuf = scan->rs_cbuf;
1002

1003
		if (scan->rs_ntup.t_data != NULL)
1004
		{
1005
			if (scan->rs_cbuf != scan->rs_nbuf)
1006
			{
1007 1008 1009 1010
				if (BufferIsValid(scan->rs_cbuf))
					ReleaseBuffer(scan->rs_cbuf);
				if (BufferIsValid(scan->rs_nbuf))
					IncrBufferRefCount(scan->rs_nbuf);
1011
			}
1012 1013
			scan->rs_ctup = scan->rs_ntup;
			scan->rs_cbuf = scan->rs_nbuf;
1014 1015 1016 1017
			HEAPDEBUG_5;		/* heap_getnext next tuple was cached */
		}
		else
		{						/* NONTUP */
Bruce Momjian's avatar
Bruce Momjian committed
1018

1019
			/*
1020
			 * Don't release scan->rs_cbuf at this point, because
1021 1022 1023 1024 1025 1026 1027
			 * heapgettup doesn't increase PrivateRefCount if it is
			 * already set. On a forward scan, both rs_ctup and rs_ptup
			 * usually point to the same buffer page, so
			 * PrivateRefCount[rs_cbuf] should be 2 (or more, if for
			 * instance ctup is stored in a TupleTableSlot).  - 01/09/93
			 */

1028 1029 1030 1031 1032 1033 1034
			heapgettup(scan->rs_rd,
					   &(scan->rs_ctup),
					   1,
					   &scan->rs_cbuf,
					   scan->rs_snapshot,
					   scan->rs_nkeys,
					   scan->rs_key);
1035 1036
		}

1037
		if (scan->rs_ctup.t_data == NULL && !BufferIsValid(scan->rs_cbuf))
1038
		{
1039 1040
			if (BufferIsValid(scan->rs_nbuf))
				ReleaseBuffer(scan->rs_nbuf);
1041
			scan->rs_ntup.t_datamcxt = NULL;
1042
			scan->rs_ntup.t_data = NULL;
1043
			scan->rs_nbuf = InvalidBuffer;
1044
			HEAPDEBUG_6;		/* heap_getnext returning EOS */
1045
			return NULL;
1046 1047
		}

1048 1049
		if (BufferIsValid(scan->rs_nbuf))
			ReleaseBuffer(scan->rs_nbuf);
1050
		scan->rs_ntup.t_datamcxt = NULL;
1051
		scan->rs_ntup.t_data = NULL;
1052
		scan->rs_nbuf = UnknownBuffer;
1053 1054
	}

1055
	/* ----------------
1056 1057
	 *	if we get here it means we have a new current scan tuple, so
	 *	point to the proper return buffer and return the tuple.
1058 1059
	 * ----------------
	 */
1060 1061 1062

	HEAPDEBUG_7;				/* heap_getnext returning tuple */

1063
	return ((scan->rs_ctup.t_data == NULL) ? NULL : &(scan->rs_ctup));
1064 1065 1066
}

/* ----------------
1067
 *		heap_fetch		- retrive tuple with tid
1068
 *
1069
 *		Currently ignores LP_IVALID during processing!
1070 1071 1072 1073
 *
 *		Because this is not part of a scan, there is no way to
 *		automatically lock/unlock the shared buffers.
 *		For this reason, we require that the user retrieve the buffer
1074
 *		value, and they are required to BufferRelease() it when they
1075 1076
 *		are done.  If they want to make a copy of it before releasing it,
 *		they can call heap_copytyple().
1077 1078
 * ----------------
 */
1079
void
1080
heap_fetch(Relation relation,
1081
		   Snapshot snapshot,
1082
		   HeapTuple tuple,
1083
		   Buffer *userbuf)
1084
{
Bruce Momjian's avatar
Bruce Momjian committed
1085 1086 1087 1088 1089
	ItemId		lp;
	Buffer		buffer;
	PageHeader	dp;
	ItemPointer tid = &(tuple->t_self);
	OffsetNumber offnum;
1090

1091
	tuple->tableOid = relation->rd_id;
1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107
	/* ----------------
	 *	increment access statistics
	 * ----------------
	 */
	IncrHeapAccessStat(local_fetch);
	IncrHeapAccessStat(global_fetch);

	/* ----------------
	 *	get the buffer from the relation descriptor
	 *	Note that this does a buffer pin.
	 * ----------------
	 */

	buffer = ReadBuffer(relation, ItemPointerGetBlockNumber(tid));

	if (!BufferIsValid(buffer))
1108
		elog(ERROR, "heap_fetch: %s relation: ReadBuffer(%lx) failed",
1109
			 RelationGetRelationName(relation), (long) tid);
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1110 1111

	LockBuffer(buffer, BUFFER_LOCK_SHARE);
1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125

	/* ----------------
	 *	get the item line pointer corresponding to the requested tid
	 * ----------------
	 */
	dp = (PageHeader) BufferGetPage(buffer);
	offnum = ItemPointerGetOffsetNumber(tid);
	lp = PageGetItemId(dp, offnum);

	/* ----------------
	 *	more sanity checks
	 * ----------------
	 */

1126 1127
	if (!ItemIdIsUsed(lp))
	{
1128
		LockBuffer(buffer, BUFFER_LOCK_UNLOCK);
1129 1130
		ReleaseBuffer(buffer);
		*userbuf = InvalidBuffer;
1131
		tuple->t_datamcxt = NULL;
1132 1133 1134
		tuple->t_data = NULL;
		return;
	}
1135

1136
	tuple->t_datamcxt = NULL;
1137 1138 1139
	tuple->t_data = (HeapTupleHeader) PageGetItem((Page) dp, lp);
	tuple->t_len = ItemIdGetLength(lp);

1140 1141 1142 1143 1144
	/* ----------------
	 *	check time qualification of tid
	 * ----------------
	 */

1145 1146
	HeapTupleSatisfies(tuple, relation, buffer, dp,
					   snapshot, 0, (ScanKey) NULL);
1147

Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1148 1149
	LockBuffer(buffer, BUFFER_LOCK_UNLOCK);

1150
	if (tuple->t_data == NULL)
1151
	{
1152
		/* Tuple failed time check, so we can release now. */
1153
		ReleaseBuffer(buffer);
1154 1155 1156 1157
		*userbuf = InvalidBuffer;
	}
	else
	{
1158 1159 1160 1161

		/*
		 * All checks passed, so return the tuple as valid. Caller is now
		 * responsible for releasing the buffer.
1162 1163
		 */
		*userbuf = buffer;
1164 1165 1166
	}
}

1167 1168 1169 1170 1171 1172 1173
/* ----------------
 *	heap_get_latest_tid -  get the latest tid of a specified tuple
 *
 * ----------------
 */
ItemPointer
heap_get_latest_tid(Relation relation,
1174 1175
					Snapshot snapshot,
					ItemPointer tid)
1176
{
1177
	ItemId		lp = NULL;
1178 1179
	Buffer		buffer;
	PageHeader	dp;
1180 1181 1182 1183 1184 1185
	OffsetNumber offnum;
	HeapTupleData tp;
	HeapTupleHeader t_data;
	ItemPointerData ctid;
	bool		invalidBlock,
				linkend;
1186

1187
	tp.tableOid = relation->rd_id;
1188 1189 1190 1191 1192 1193 1194 1195 1196 1197
	/* ----------------
	 *	get the buffer from the relation descriptor
	 *	Note that this does a buffer pin.
	 * ----------------
	 */

	buffer = ReadBuffer(relation, ItemPointerGetBlockNumber(tid));

	if (!BufferIsValid(buffer))
		elog(ERROR, "heap_get_latest_tid: %s relation: ReadBuffer(%lx) failed",
1198
			 RelationGetRelationName(relation), (long) tid);
1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215

	LockBuffer(buffer, BUFFER_LOCK_SHARE);

	/* ----------------
	 *	get the item line pointer corresponding to the requested tid
	 * ----------------
	 */
	dp = (PageHeader) BufferGetPage(buffer);
	offnum = ItemPointerGetOffsetNumber(tid);
	invalidBlock = true;
	if (!PageIsNew(dp))
	{
		lp = PageGetItemId(dp, offnum);
		if (ItemIdIsUsed(lp))
			invalidBlock = false;
	}
	if (invalidBlock)
1216
	{
1217 1218 1219
		LockBuffer(buffer, BUFFER_LOCK_UNLOCK);
		ReleaseBuffer(buffer);
		return NULL;
1220
	}
1221 1222 1223 1224 1225 1226

	/* ----------------
	 *	more sanity checks
	 * ----------------
	 */

1227
	tp.t_datamcxt = NULL;
1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241
	t_data = tp.t_data = (HeapTupleHeader) PageGetItem((Page) dp, lp);
	tp.t_len = ItemIdGetLength(lp);
	tp.t_self = *tid;
	ctid = tp.t_data->t_ctid;

	/* ----------------
	 *	check time qualification of tid
	 * ----------------
	 */

	HeapTupleSatisfies(&tp, relation, buffer, dp,
					   snapshot, 0, (ScanKey) NULL);

	linkend = true;
1242
	if ((t_data->t_infomask & HEAP_XMAX_COMMITTED) &&
1243 1244 1245 1246 1247 1248 1249 1250 1251 1252
		!ItemPointerEquals(tid, &ctid))
		linkend = false;

	LockBuffer(buffer, BUFFER_LOCK_UNLOCK);
	ReleaseBuffer(buffer);

	if (tp.t_data == NULL)
	{
		if (linkend)
			return NULL;
1253
		return heap_get_latest_tid(relation, snapshot, &ctid);
1254 1255 1256 1257 1258
	}

	return tid;
}

1259
/* ----------------
1260
 *		heap_insert		- insert tuple
1261
 *
1262 1263
 *		The assignment of t_min (and thus the others) should be
 *		removed eventually.
1264
 *
1265 1266 1267 1268
 *		Currently places the tuple onto the last page.	If there is no room,
 *		it is placed on new pages.	(Heap relations)
 *		Note that concurrent inserts during a scan will probably have
 *		unexpected results, though this will be fixed eventually.
1269
 *
1270
 *		Fix to work with indexes.
1271 1272 1273 1274 1275
 * ----------------
 */
Oid
heap_insert(Relation relation, HeapTuple tup)
{
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1276
	Buffer buffer;
1277 1278 1279 1280
#ifndef TOAST_INDICES
	HeapTupleHeader plaintdata = NULL;
	int32			plaintlen = 0;
#endif
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1281 1282

	/* increment access statistics */
1283
	tup->tableOid = relation->rd_id;
1284 1285
	IncrHeapAccessStat(local_insert);
	IncrHeapAccessStat(global_insert);
1286

1287 1288 1289 1290 1291 1292 1293 1294 1295 1296
	/* ----------------
	 *	If the object id of this tuple has already been assigned, trust
	 *	the caller.  There are a couple of ways this can happen.  At initial
	 *	db creation, the backend program sets oids for tuples.	When we
	 *	define an index, we set the oid.  Finally, in the future, we may
	 *	allow users to set their own object ids in order to support a
	 *	persistent object store (objects need to contain pointers to one
	 *	another).
	 * ----------------
	 */
1297 1298
	if (!OidIsValid(tup->t_data->t_oid))
		tup->t_data->t_oid = newoid();
1299
	else
1300
		CheckMaxObjectId(tup->t_data->t_oid);
1301

1302 1303 1304 1305 1306
	TransactionIdStore(GetCurrentTransactionId(), &(tup->t_data->t_xmin));
	tup->t_data->t_cmin = GetCurrentCommandId();
	StoreInvalidTransactionId(&(tup->t_data->t_xmax));
	tup->t_data->t_infomask &= ~(HEAP_XACT_MASK);
	tup->t_data->t_infomask |= HEAP_XMAX_INVALID;
1307

Jan Wieck's avatar
TOAST  
Jan Wieck committed
1308 1309 1310 1311 1312 1313 1314 1315
#ifdef TUPLE_TOASTER_ACTIVE
	/* ----------
	 * If the new tuple is too big for storage or contains already
	 * toasted attributes from some other relation, invoke the toaster.
	 * ----------
	 */
    if (HeapTupleHasExtended(tup) ||
				(MAXALIGN(tup->t_len) > (MaxTupleSize / 4)))
1316
#ifdef TOAST_INDICES
Jan Wieck's avatar
TOAST  
Jan Wieck committed
1317
		heap_tuple_toast_attrs(relation, tup, NULL);
1318 1319 1320
#else
		heap_tuple_toast_attrs(relation, tup, NULL, &plaintdata, &plaintlen);
#endif
Jan Wieck's avatar
TOAST  
Jan Wieck committed
1321 1322
#endif

Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1323 1324 1325 1326 1327
	/* Find buffer for this tuple */
	buffer = RelationGetBufferForTuple(relation, tup->t_len, InvalidBuffer);

	/* NO ELOG(ERROR) from here till changes are logged */
	RelationPutHeapTuple(relation, buffer, tup);
1328

Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1329 1330 1331 1332 1333 1334
#ifdef XLOG
	/* XLOG stuff */
	{
		xl_heap_insert	xlrec;
		xlrec.itid.dbId = relation->rd_lockInfo.lockRelId.dbId;
		xlrec.itid.relId = relation->rd_lockInfo.lockRelId.relId;
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1335 1336
		xlrec.itid.cid = GetCurrentCommandId();
		xlrec.itid.tid = tup->t_self;
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1337 1338 1339 1340 1341 1342
		xlrec.t_natts = tup->t_data->t_natts;
		xlrec.t_oid = tup->t_data->t_oid;
		xlrec.t_hoff = tup->t_data->t_hoff;
		xlrec.mask = tup->t_data->t_infomask;
		
		XLogRecPtr recptr = XLogInsert(RM_HEAP_ID, XLOG_HEAP_INSERT,
1343 1344 1345
			(char*) xlrec, SizeOfHeapInsert, 
			(char*) tup->t_data + offsetof(HeapTupleHeaderData, t_bits), 
			tup->t_len - offsetof(HeapTupleHeaderData, t_bits));
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1346

Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1347 1348
		((PageHeader) BufferGetPage(buffer))->pd_lsn = recptr;
		((PageHeader) BufferGetPage(buffer))->pd_sui = ThisStartUpID;
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1349 1350 1351
	}
#endif

Jan Wieck's avatar
TOAST  
Jan Wieck committed
1352
	LockBuffer(buffer, BUFFER_LOCK_UNLOCK);
1353
	WriteBuffer(buffer);
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1354

1355
	if (IsSystemRelationName(RelationGetRelationName(relation)))
1356
		RelationMark4RollbackHeapTuple(relation, tup);
1357

1358 1359 1360 1361 1362 1363 1364 1365 1366 1367
#ifndef TOAST_INDICES
    if (plaintdata != NULL && tup->t_data != plaintdata)
	{
		if (tup->t_datamcxt != NULL && (char *) (tup->t_data) !=
					((char *) tup + HEAPTUPLESIZE))
			pfree(tup->t_data);
	    tup->t_data = plaintdata;
		tup->t_len  = plaintlen;
	}
#endif
1368
	return tup->t_data->t_oid;
1369 1370
}

Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1371 1372
/*
 *	heap_delete		- delete a tuple
1373
 */
1374
int
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1375
heap_delete(Relation relation, ItemPointer tid, ItemPointer ctid)
1376
{
Bruce Momjian's avatar
Bruce Momjian committed
1377 1378 1379 1380 1381
	ItemId		lp;
	HeapTupleData tp;
	PageHeader	dp;
	Buffer		buffer;
	int			result;
1382

1383
	tp.tableOid = relation->rd_id;
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1384
	/* increment access statistics */
1385 1386 1387 1388 1389
	IncrHeapAccessStat(local_delete);
	IncrHeapAccessStat(global_delete);

	Assert(ItemPointerIsValid(tid));

Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1390
	buffer = ReadBuffer(relation, ItemPointerGetBlockNumber(tid));
1391

Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1392
	if (!BufferIsValid(buffer))
1393
		elog(ERROR, "heap_delete: failed ReadBuffer");
1394

Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1395
	LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
1396

Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1397 1398
	dp = (PageHeader) BufferGetPage(buffer);
	lp = PageGetItemId(dp, ItemPointerGetOffsetNumber(tid));
1399
	tp.t_datamcxt = NULL;
1400 1401 1402
	tp.t_data = (HeapTupleHeader) PageGetItem((Page) dp, lp);
	tp.t_len = ItemIdGetLength(lp);
	tp.t_self = *tid;
Bruce Momjian's avatar
Bruce Momjian committed
1403

Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1404 1405
l1:
	result = HeapTupleSatisfiesUpdate(&tp);
Bruce Momjian's avatar
Bruce Momjian committed
1406

Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1407
	if (result == HeapTupleInvisible)
1408
	{
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1409 1410 1411
		LockBuffer(buffer, BUFFER_LOCK_UNLOCK);
		ReleaseBuffer(buffer);
		elog(ERROR, "heap_delete: (am)invalid tid");
1412
	}
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1413
	else if (result == HeapTupleBeingUpdated)
1414
	{
Bruce Momjian's avatar
Bruce Momjian committed
1415
		TransactionId xwait = tp.t_data->t_xmax;
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1416

Bruce Momjian's avatar
Bruce Momjian committed
1417
		/* sleep until concurrent transaction ends */
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1418 1419 1420 1421 1422 1423
		LockBuffer(buffer, BUFFER_LOCK_UNLOCK);
		XactLockTableWait(xwait);

		LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
		if (TransactionIdDidAbort(xwait))
			goto l1;
1424 1425 1426 1427 1428

		/*
		 * xwait is committed but if xwait had just marked the tuple for
		 * update then some other xaction could update this tuple before
		 * we got to this point.
1429 1430 1431
		 */
		if (tp.t_data->t_xmax != xwait)
			goto l1;
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450
		if (!(tp.t_data->t_infomask & HEAP_XMAX_COMMITTED))
		{
			tp.t_data->t_infomask |= HEAP_XMAX_COMMITTED;
			SetBufferCommitInfoNeedsSave(buffer);
		}
		/* if tuple was marked for update but not updated... */
		if (tp.t_data->t_infomask & HEAP_MARKED_FOR_UPDATE)
			result = HeapTupleMayBeUpdated;
		else
			result = HeapTupleUpdated;
	}
	if (result != HeapTupleMayBeUpdated)
	{
		Assert(result == HeapTupleSelfUpdated || result == HeapTupleUpdated);
		if (ctid != NULL)
			*ctid = tp.t_data->t_ctid;
		LockBuffer(buffer, BUFFER_LOCK_UNLOCK);
		ReleaseBuffer(buffer);
		return result;
1451 1452
	}

Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1453 1454 1455 1456 1457 1458
#ifdef XLOG
	/* XLOG stuff */
	{
		xl_heap_delete	xlrec;
		xlrec.dtid.dbId = relation->rd_lockInfo.lockRelId.dbId;
		xlrec.dtid.relId = relation->rd_lockInfo.lockRelId.relId;
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1459
		xlrec.dtid.cid = GetCurrentCommandId();
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1460 1461
		xlrec.dtid.tid = tp.t_self;
		XLogRecPtr recptr = XLogInsert(RM_HEAP_ID, XLOG_HEAP_DELETE,
1462
			(char*) xlrec, SizeOfHeapDelete, NULL, 0);
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1463 1464

		dp->pd_lsn = recptr;
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1465
		dp->pd_sui = ThisStartUpID;
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1466 1467 1468
	}
#endif

Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1469
	/* store transaction information of xact deleting the tuple */
1470 1471
	TransactionIdStore(GetCurrentTransactionId(), &(tp.t_data->t_xmax));
	tp.t_data->t_cmax = GetCurrentCommandId();
Bruce Momjian's avatar
Bruce Momjian committed
1472 1473
	tp.t_data->t_infomask &= ~(HEAP_XMAX_COMMITTED |
							 HEAP_XMAX_INVALID | HEAP_MARKED_FOR_UPDATE);
1474

Jan Wieck's avatar
TOAST  
Jan Wieck committed
1475 1476 1477 1478 1479 1480 1481
#ifdef TUPLE_TOASTER_ACTIVE
	/* ----------
	 * If the relation has toastable attributes, we need to delete
	 * no longer needed items there too.
	 * ----------
	 */
	if (HeapTupleHasExtended(&tp))
1482
#ifdef TOAST_INDICES
Jan Wieck's avatar
TOAST  
Jan Wieck committed
1483
		heap_tuple_toast_attrs(relation, NULL, &(tp));
1484 1485 1486
#else
		heap_tuple_toast_attrs(relation, NULL, &(tp), NULL, NULL);
#endif
Jan Wieck's avatar
TOAST  
Jan Wieck committed
1487 1488
#endif

Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1489 1490 1491
	LockBuffer(buffer, BUFFER_LOCK_UNLOCK);

	/* invalidate caches */
1492
	RelationInvalidateHeapTuple(relation, &tp);
1493

Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1494
	WriteBuffer(buffer);
1495

Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1496
	return HeapTupleMayBeUpdated;
1497 1498
}

Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1499
/*
1500
 *	heap_update - replace a tuple
1501 1502
 */
int
1503
heap_update(Relation relation, ItemPointer otid, HeapTuple newtup,
1504
			ItemPointer ctid)
1505
{
Bruce Momjian's avatar
Bruce Momjian committed
1506 1507 1508
	ItemId		lp;
	HeapTupleData oldtup;
	PageHeader	dp;
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1509
	Buffer		buffer, newbuf;
Bruce Momjian's avatar
Bruce Momjian committed
1510
	int			result;
1511 1512 1513 1514
#ifndef TOAST_INDICES
	HeapTupleHeader plaintdata = NULL;
	int32			plaintlen  = 0;
#endif
1515

1516
	newtup->tableOid = relation->rd_id;
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1517
	/* increment access statistics */
1518 1519 1520 1521 1522 1523 1524
	IncrHeapAccessStat(local_replace);
	IncrHeapAccessStat(global_replace);

	Assert(ItemPointerIsValid(otid));

	buffer = ReadBuffer(relation, ItemPointerGetBlockNumber(otid));
	if (!BufferIsValid(buffer))
1525
		elog(ERROR, "amreplace: failed ReadBuffer");
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1526
	LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
1527

Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1528
	dp = (PageHeader) BufferGetPage(buffer);
1529 1530
	lp = PageGetItemId(dp, ItemPointerGetOffsetNumber(otid));

1531
	oldtup.t_datamcxt = NULL;
1532 1533 1534
	oldtup.t_data = (HeapTupleHeader) PageGetItem(dp, lp);
	oldtup.t_len = ItemIdGetLength(lp);
	oldtup.t_self = *otid;
1535

Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1536 1537
l2:
	result = HeapTupleSatisfiesUpdate(&oldtup);
Bruce Momjian's avatar
Bruce Momjian committed
1538

Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1539
	if (result == HeapTupleInvisible)
1540
	{
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1541
		LockBuffer(buffer, BUFFER_LOCK_UNLOCK);
1542
		ReleaseBuffer(buffer);
1543
		elog(ERROR, "heap_update: (am)invalid tid");
1544
	}
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1545
	else if (result == HeapTupleBeingUpdated)
1546
	{
Bruce Momjian's avatar
Bruce Momjian committed
1547
		TransactionId xwait = oldtup.t_data->t_xmax;
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1548 1549 1550 1551 1552 1553 1554 1555

		/* sleep untill concurrent transaction ends */
		LockBuffer(buffer, BUFFER_LOCK_UNLOCK);
		XactLockTableWait(xwait);

		LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
		if (TransactionIdDidAbort(xwait))
			goto l2;
1556 1557 1558 1559 1560

		/*
		 * xwait is committed but if xwait had just marked the tuple for
		 * update then some other xaction could update this tuple before
		 * we got to this point.
1561 1562 1563
		 */
		if (oldtup.t_data->t_xmax != xwait)
			goto l2;
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580
		if (!(oldtup.t_data->t_infomask & HEAP_XMAX_COMMITTED))
		{
			oldtup.t_data->t_infomask |= HEAP_XMAX_COMMITTED;
			SetBufferCommitInfoNeedsSave(buffer);
		}
		/* if tuple was marked for update but not updated... */
		if (oldtup.t_data->t_infomask & HEAP_MARKED_FOR_UPDATE)
			result = HeapTupleMayBeUpdated;
		else
			result = HeapTupleUpdated;
	}
	if (result != HeapTupleMayBeUpdated)
	{
		Assert(result == HeapTupleSelfUpdated || result == HeapTupleUpdated);
		if (ctid != NULL)
			*ctid = oldtup.t_data->t_ctid;
		LockBuffer(buffer, BUFFER_LOCK_UNLOCK);
1581
		ReleaseBuffer(buffer);
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1582
		return result;
1583 1584 1585
	}

	/* XXX order problems if not atomic assignment ??? */
1586 1587 1588 1589 1590
	newtup->t_data->t_oid = oldtup.t_data->t_oid;
	TransactionIdStore(GetCurrentTransactionId(), &(newtup->t_data->t_xmin));
	newtup->t_data->t_cmin = GetCurrentCommandId();
	StoreInvalidTransactionId(&(newtup->t_data->t_xmax));
	newtup->t_data->t_infomask &= ~(HEAP_XACT_MASK);
1591
	newtup->t_data->t_infomask |= (HEAP_XMAX_INVALID | HEAP_UPDATED);
1592

1593 1594 1595 1596 1597 1598 1599 1600 1601 1602
#ifdef TUPLE_TOASTER_ACTIVE
	/* ----------
	 * If this relation is enabled for toasting, let the toaster
	 * delete not any longer needed entries and create new ones to
	 * make the new tuple fit again.
	 * ----------
	 */
	if (HeapTupleHasExtended(&oldtup) || 
			HeapTupleHasExtended(newtup) ||
			(MAXALIGN(newtup->t_len) > (MaxTupleSize / 4)))
1603
#ifdef TOAST_INDICES
1604
		heap_tuple_toast_attrs(relation, newtup, &oldtup);
1605 1606 1607
#else
		heap_tuple_toast_attrs(relation, newtup, &oldtup, &plaintdata, &plaintlen);
#endif
1608 1609
#endif

Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622
	/* Find buffer for new tuple */

	if ((unsigned) MAXALIGN(newtup->t_len) <= PageGetFreeSpace((Page) dp))
		newbuf = buffer;
	else
		newbuf = RelationGetBufferForTuple(relation, newtup->t_len, buffer);

	/* NO ELOG(ERROR) from here till changes are logged */

	/* insert new tuple */
	RelationPutHeapTuple(relation, newbuf, newtup);

	/* logically delete old tuple */
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1623 1624
	TransactionIdStore(GetCurrentTransactionId(), &(oldtup.t_data->t_xmax));
	oldtup.t_data->t_cmax = GetCurrentCommandId();
Bruce Momjian's avatar
Bruce Momjian committed
1625 1626
	oldtup.t_data->t_infomask &= ~(HEAP_XMAX_COMMITTED |
							 HEAP_XMAX_INVALID | HEAP_MARKED_FOR_UPDATE);
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1627

Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1628 1629 1630 1631 1632
	/* record address of new tuple in t_ctid of old one */
	oldtup.t_data->t_ctid = newtup->t_self;

#ifdef XLOG
	/* XLOG stuff */
1633
	{
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1634 1635 1636 1637 1638 1639 1640 1641 1642 1643
		xl_heap_update	xlrec;
		xlrec.dtid.dbId = relation->rd_lockInfo.lockRelId.dbId;
		xlrec.dtid.relId = relation->rd_lockInfo.lockRelId.relId;
		xlrec.dtid.cid = GetCurrentCommandId();
		xlrec.itid.tid = newtup->t_self;
		xlrec.t_natts = newtup->t_data->t_natts;
		xlrec.t_hoff = newtup->t_data->t_hoff;
		xlrec.mask = newtup->t_data->t_infomask;
		
		XLogRecPtr recptr = XLogInsert(RM_HEAP_ID, XLOG_HEAP_UPDATE,
1644 1645 1646
			(char*) xlrec, SizeOfHeapUpdate, 
			(char*) newtup->t_data + offsetof(HeapTupleHeaderData, t_bits), 
			newtup->t_len - offsetof(HeapTupleHeaderData, t_bits));
Bruce Momjian's avatar
Bruce Momjian committed
1647

Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1648 1649 1650 1651 1652 1653 1654
		if (newbuf != buffer)
		{
			((PageHeader) BufferGetPage(newbuf))->pd_lsn = recptr;
			((PageHeader) BufferGetPage(newbuf))->pd_sui = ThisStartUpID;
		}
		((PageHeader) BufferGetPage(buffer))->pd_lsn = recptr;
		((PageHeader) BufferGetPage(buffer))->pd_sui = ThisStartUpID;
1655
	}
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1656
#endif
1657

Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1658 1659 1660 1661 1662
	if (newbuf != buffer)
	{
		LockBuffer(newbuf, BUFFER_LOCK_UNLOCK);
		WriteBuffer(newbuf);
	}
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1663
	LockBuffer(buffer, BUFFER_LOCK_UNLOCK);
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1664
	WriteBuffer(buffer);
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1665 1666

	/* invalidate caches */
1667
	RelationInvalidateHeapTuple(relation, &oldtup);
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1668
	RelationMark4RollbackHeapTuple(relation, newtup);
1669

1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680
#ifndef TOAST_INDICES
    if (plaintdata != NULL && newtup->t_data != plaintdata)
	{
		if (newtup->t_datamcxt != NULL && (char *) (newtup->t_data) !=
					((char *) newtup + HEAPTUPLESIZE))
			pfree(newtup->t_data);
	    newtup->t_data = plaintdata;
		newtup->t_len  = plaintlen;
	}
#endif

Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1681 1682 1683 1684 1685 1686 1687 1688 1689
	return HeapTupleMayBeUpdated;
}

/*
 *	heap_mark4update		- mark a tuple for update
 */
int
heap_mark4update(Relation relation, HeapTuple tuple, Buffer *buffer)
{
Bruce Momjian's avatar
Bruce Momjian committed
1690 1691 1692 1693
	ItemPointer tid = &(tuple->t_self);
	ItemId		lp;
	PageHeader	dp;
	int			result;
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1694

1695
	tuple->tableOid = relation->rd_id;
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708
	/* increment access statistics */
	IncrHeapAccessStat(local_mark4update);
	IncrHeapAccessStat(global_mark4update);

	*buffer = ReadBuffer(relation, ItemPointerGetBlockNumber(tid));

	if (!BufferIsValid(*buffer))
		elog(ERROR, "heap_mark4update: failed ReadBuffer");

	LockBuffer(*buffer, BUFFER_LOCK_EXCLUSIVE);

	dp = (PageHeader) BufferGetPage(*buffer);
	lp = PageGetItemId(dp, ItemPointerGetOffsetNumber(tid));
1709
	tuple->t_datamcxt = NULL;
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1710 1711
	tuple->t_data = (HeapTupleHeader) PageGetItem((Page) dp, lp);
	tuple->t_len = ItemIdGetLength(lp);
Bruce Momjian's avatar
Bruce Momjian committed
1712

Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1713 1714
l3:
	result = HeapTupleSatisfiesUpdate(tuple);
Bruce Momjian's avatar
Bruce Momjian committed
1715

Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1716 1717 1718 1719 1720 1721 1722 1723
	if (result == HeapTupleInvisible)
	{
		LockBuffer(*buffer, BUFFER_LOCK_UNLOCK);
		ReleaseBuffer(*buffer);
		elog(ERROR, "heap_mark4update: (am)invalid tid");
	}
	else if (result == HeapTupleBeingUpdated)
	{
Bruce Momjian's avatar
Bruce Momjian committed
1724
		TransactionId xwait = tuple->t_data->t_xmax;
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1725 1726 1727 1728 1729 1730 1731 1732

		/* sleep untill concurrent transaction ends */
		LockBuffer(*buffer, BUFFER_LOCK_UNLOCK);
		XactLockTableWait(xwait);

		LockBuffer(*buffer, BUFFER_LOCK_EXCLUSIVE);
		if (TransactionIdDidAbort(xwait))
			goto l3;
1733 1734 1735 1736 1737

		/*
		 * xwait is committed but if xwait had just marked the tuple for
		 * update then some other xaction could update this tuple before
		 * we got to this point.
1738 1739 1740
		 */
		if (tuple->t_data->t_xmax != xwait)
			goto l3;
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754
		if (!(tuple->t_data->t_infomask & HEAP_XMAX_COMMITTED))
		{
			tuple->t_data->t_infomask |= HEAP_XMAX_COMMITTED;
			SetBufferCommitInfoNeedsSave(*buffer);
		}
		/* if tuple was marked for update but not updated... */
		if (tuple->t_data->t_infomask & HEAP_MARKED_FOR_UPDATE)
			result = HeapTupleMayBeUpdated;
		else
			result = HeapTupleUpdated;
	}
	if (result != HeapTupleMayBeUpdated)
	{
		Assert(result == HeapTupleSelfUpdated || result == HeapTupleUpdated);
1755
		tuple->t_self = tuple->t_data->t_ctid;
Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1756 1757 1758 1759
		LockBuffer(*buffer, BUFFER_LOCK_UNLOCK);
		return result;
	}

Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1760 1761 1762 1763 1764 1765 1766 1767
#ifdef XLOG
	/*
	 * XLOG stuff: no logging is required as long as we have no
	 * savepoints. For savepoints private log could be used...
	 */
	((PageHeader) BufferGetPage(*buffer))->pd_sui = ThisStartUpID;
#endif

Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1768 1769 1770 1771 1772 1773 1774 1775 1776
	/* store transaction information of xact marking the tuple */
	TransactionIdStore(GetCurrentTransactionId(), &(tuple->t_data->t_xmax));
	tuple->t_data->t_cmax = GetCurrentCommandId();
	tuple->t_data->t_infomask &= ~(HEAP_XMAX_COMMITTED | HEAP_XMAX_INVALID);
	tuple->t_data->t_infomask |= HEAP_MARKED_FOR_UPDATE;

	LockBuffer(*buffer, BUFFER_LOCK_UNLOCK);

	WriteNoReleaseBuffer(*buffer);
1777

Vadim B. Mikheev's avatar
Vadim B. Mikheev committed
1778
	return HeapTupleMayBeUpdated;
1779 1780 1781
}

/* ----------------
1782
 *		heap_markpos	- mark scan position
1783
 *
1784 1785 1786 1787 1788 1789
 *		Note:
 *				Should only one mark be maintained per scan at one time.
 *		Check if this can be done generally--say calls to get the
 *		next/previous tuple and NEVER pass struct scandesc to the
 *		user AM's.  Now, the mark is sent to the executor for safekeeping.
 *		Probably can store this info into a GENERAL scan structure.
1790
 *
1791 1792 1793
 *		May be best to change this call to store the marked position
 *		(up to 2?) in the scan structure itself.
 *		Fix to use the proper caching structure.
1794 1795 1796
 * ----------------
 */
void
1797
heap_markpos(HeapScanDesc scan)
1798
{
1799 1800 1801 1802 1803 1804 1805 1806 1807 1808

	/* ----------------
	 *	increment access statistics
	 * ----------------
	 */
	IncrHeapAccessStat(local_markpos);
	IncrHeapAccessStat(global_markpos);

	/* Note: no locking manipulations needed */

1809
	if (scan->rs_ptup.t_data == NULL &&
1810
		BufferIsUnknown(scan->rs_pbuf))
1811
	{							/* == NONTUP */
1812 1813 1814 1815 1816 1817 1818 1819
		scan->rs_ptup = scan->rs_ctup;
		heapgettup(scan->rs_rd,
				   &(scan->rs_ptup),
				   -1,
				   &scan->rs_pbuf,
				   scan->rs_snapshot,
				   scan->rs_nkeys,
				   scan->rs_key);
1820 1821

	}
1822
	else if (scan->rs_ntup.t_data == NULL &&
1823
			 BufferIsUnknown(scan->rs_nbuf))
1824
	{							/* == NONTUP */
1825 1826 1827 1828 1829 1830 1831 1832
		scan->rs_ntup = scan->rs_ctup;
		heapgettup(scan->rs_rd,
				   &(scan->rs_ntup),
				   1,
				   &scan->rs_nbuf,
				   scan->rs_snapshot,
				   scan->rs_nkeys,
				   scan->rs_key);
1833 1834 1835 1836 1837 1838
	}

	/* ----------------
	 * Should not unpin the buffer pages.  They may still be in use.
	 * ----------------
	 */
1839 1840
	if (scan->rs_ptup.t_data != NULL)
		scan->rs_mptid = scan->rs_ptup.t_self;
1841
	else
1842
		ItemPointerSetInvalid(&scan->rs_mptid);
1843 1844
	if (scan->rs_ctup.t_data != NULL)
		scan->rs_mctid = scan->rs_ctup.t_self;
1845
	else
1846
		ItemPointerSetInvalid(&scan->rs_mctid);
1847 1848
	if (scan->rs_ntup.t_data != NULL)
		scan->rs_mntid = scan->rs_ntup.t_self;
1849
	else
1850
		ItemPointerSetInvalid(&scan->rs_mntid);
1851 1852 1853
}

/* ----------------
1854
 *		heap_restrpos	- restore position to marked location
1855
 *
1856 1857 1858 1859 1860
 *		Note:  there are bad side effects here.  If we were past the end
 *		of a relation when heapmarkpos is called, then if the relation is
 *		extended via insert, then the next call to heaprestrpos will set
 *		cause the added tuples to be visible when the scan continues.
 *		Problems also arise if the TID's are rearranged!!!
1861
 *
1862 1863 1864
 *		Now pins buffer once for each valid tuple pointer (rs_ptup,
 *		rs_ctup, rs_ntup) referencing it.
 *		 - 01/13/94
1865 1866
 *
 * XXX	might be better to do direct access instead of
1867
 *		using the generality of heapgettup().
1868 1869 1870 1871 1872 1873 1874
 *
 * XXX It is very possible that when a scan is restored, that a tuple
 * XXX which previously qualified may fail for time range purposes, unless
 * XXX some form of locking exists (ie., portals currently can act funny.
 * ----------------
 */
void
1875
heap_restrpos(HeapScanDesc scan)
1876
{
1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887
	/* ----------------
	 *	increment access statistics
	 * ----------------
	 */
	IncrHeapAccessStat(local_restrpos);
	IncrHeapAccessStat(global_restrpos);

	/* XXX no amrestrpos checking that ammarkpos called */

	/* Note: no locking manipulations needed */

1888
	unpinscan(scan);
1889 1890

	/* force heapgettup to pin buffer for each loaded tuple */
1891 1892 1893
	scan->rs_pbuf = InvalidBuffer;
	scan->rs_cbuf = InvalidBuffer;
	scan->rs_nbuf = InvalidBuffer;
1894

1895
	if (!ItemPointerIsValid(&scan->rs_mptid))
1896 1897
	{
		scan->rs_ptup.t_datamcxt = NULL;
1898
		scan->rs_ptup.t_data = NULL;
1899
	}
1900 1901
	else
	{
1902
		scan->rs_ptup.t_self = scan->rs_mptid;
1903
		scan->rs_ptup.t_datamcxt = NULL;
1904 1905 1906 1907 1908 1909 1910 1911
		scan->rs_ptup.t_data = (HeapTupleHeader) 0x1;	/* for heapgettup */
		heapgettup(scan->rs_rd,
				   &(scan->rs_ptup),
				   0,
				   &(scan->rs_pbuf),
				   false,
				   0,
				   (ScanKey) NULL);
1912 1913
	}

1914
	if (!ItemPointerIsValid(&scan->rs_mctid))
1915 1916
	{
		scan->rs_ctup.t_datamcxt = NULL;
1917
		scan->rs_ctup.t_data = NULL;
1918
	}
1919 1920
	else
	{
1921
		scan->rs_ctup.t_self = scan->rs_mctid;
1922
		scan->rs_ctup.t_datamcxt = NULL;
1923 1924 1925 1926 1927 1928 1929 1930
		scan->rs_ctup.t_data = (HeapTupleHeader) 0x1;	/* for heapgettup */
		heapgettup(scan->rs_rd,
				   &(scan->rs_ctup),
				   0,
				   &(scan->rs_cbuf),
				   false,
				   0,
				   (ScanKey) NULL);
1931 1932
	}

1933
	if (!ItemPointerIsValid(&scan->rs_mntid))
1934 1935
	{
		scan->rs_ntup.t_datamcxt = NULL;
1936
		scan->rs_ntup.t_data = NULL;
1937
	}
1938 1939
	else
	{
1940
		scan->rs_ntup.t_datamcxt = NULL;
1941 1942 1943 1944 1945 1946 1947 1948 1949
		scan->rs_ntup.t_self = scan->rs_mntid;
		scan->rs_ntup.t_data = (HeapTupleHeader) 0x1;	/* for heapgettup */
		heapgettup(scan->rs_rd,
				   &(scan->rs_ntup),
				   0,
				   &scan->rs_nbuf,
				   false,
				   0,
				   (ScanKey) NULL);
1950
	}
1951
}
1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991

#ifdef XLOG
void heap_redo(XLogRecPtr lsn, XLogRecord *record)
{
	uint8	info = record->xl_info & ~XLR_INFO_MASK;
	
	if (info == XLOG_HEAP_INSERT)
		heap_xlog_insert(true, lsn, record);
	else if (info == XLOG_HEAP_DELETE)
		heap_xlog_delete(true, lsn, record);
	else if (info == XLOG_HEAP_UPDATE)
		heap_xlog_update(true, lsn, record);
	else if (info == XLOG_HEAP_MOVE)
		heap_xlog_move(true, lsn, record);
	else
		elog(STOP, "heap_redo: unknown op code %u", info);
}

void heap_undo(XLogRecPtr lsn, XLogRecord *record)
{
	uint8	info = record->xl_info & ~XLR_INFO_MASK;
	
	if (info == XLOG_HEAP_INSERT)
		heap_xlog_insert(false, lsn, record);
	else if (info == XLOG_HEAP_DELETE)
		heap_xlog_delete(false, lsn, record);
	else if (info == XLOG_HEAP_UPDATE)
		heap_xlog_update(false, lsn, record);
	else if (info == XLOG_HEAP_MOVE)
		heap_xlog_move(false, lsn, record);
	else
		elog(STOP, "heap_undo: unknown op code %u", info);
}

void heap_xlog_insert(bool redo, XLogRecPtr lsn, XLogRecord *record)
{
	xl_heap_insert  xlrec = XLogRecGetData(record);
}

#endif	/* XLOG */