Line data Source code
1 : /*-------------------------------------------------------------------------
2 : *
3 : * nbtdedup.c
4 : * Deduplicate or bottom-up delete items in Postgres btrees.
5 : *
6 : * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group
7 : * Portions Copyright (c) 1994, Regents of the University of California
8 : *
9 : *
10 : * IDENTIFICATION
11 : * src/backend/access/nbtree/nbtdedup.c
12 : *
13 : *-------------------------------------------------------------------------
14 : */
15 : #include "postgres.h"
16 :
17 : #include "access/nbtree.h"
18 : #include "access/nbtxlog.h"
19 : #include "access/tableam.h"
20 : #include "access/xloginsert.h"
21 : #include "miscadmin.h"
22 : #include "utils/rel.h"
23 :
24 : static void _bt_bottomupdel_finish_pending(Page page, BTDedupState state,
25 : TM_IndexDeleteOp *delstate);
26 : static bool _bt_do_singleval(Relation rel, Page page, BTDedupState state,
27 : OffsetNumber minoff, IndexTuple newitem);
28 : static void _bt_singleval_fillfactor(Page page, BTDedupState state,
29 : Size newitemsz);
30 : #ifdef USE_ASSERT_CHECKING
31 : static bool _bt_posting_valid(IndexTuple posting);
32 : #endif
33 :
34 : /*
35 : * Perform a deduplication pass.
36 : *
37 : * The general approach taken here is to perform as much deduplication as
38 : * possible to free as much space as possible. Note, however, that "single
39 : * value" strategy is used for !bottomupdedup callers when the page is full of
40 : * tuples of a single value. Deduplication passes that apply the strategy
41 : * will leave behind a few untouched tuples at the end of the page, preparing
42 : * the page for an anticipated page split that uses nbtsplitloc.c's own single
43 : * value strategy. Our high level goal is to delay merging the untouched
44 : * tuples until after the page splits.
45 : *
46 : * When a call to _bt_bottomupdel_pass() just took place (and failed), our
47 : * high level goal is to prevent a page split entirely by buying more time.
48 : * We still hope that a page split can be avoided altogether. That's why
49 : * single value strategy is not even considered for bottomupdedup callers.
50 : *
51 : * The page will have to be split if we cannot successfully free at least
52 : * newitemsz (we also need space for newitem's line pointer, which isn't
53 : * included in caller's newitemsz).
54 : *
55 : * Note: Caller should have already deleted all existing items with their
56 : * LP_DEAD bits set.
57 : */
58 : void
59 14269 : _bt_dedup_pass(Relation rel, Buffer buf, IndexTuple newitem, Size newitemsz,
60 : bool bottomupdedup)
61 : {
62 : OffsetNumber offnum,
63 : minoff,
64 : maxoff;
65 14269 : Page page = BufferGetPage(buf);
66 14269 : BTPageOpaque opaque = BTPageGetOpaque(page);
67 : Page newpage;
68 : BTDedupState state;
69 14269 : Size pagesaving PG_USED_FOR_ASSERTS_ONLY = 0;
70 14269 : bool singlevalstrat = false;
71 14269 : int nkeyatts = IndexRelationGetNumberOfKeyAttributes(rel);
72 :
73 : /* Passed-in newitemsz is MAXALIGNED but does not include line pointer */
74 14269 : newitemsz += sizeof(ItemIdData);
75 :
76 : /*
77 : * Initialize deduplication state.
78 : *
79 : * It would be possible for maxpostingsize (limit on posting list tuple
80 : * size) to be set to one third of the page. However, it seems like a
81 : * good idea to limit the size of posting lists to one sixth of a page.
82 : * That ought to leave us with a good split point when pages full of
83 : * duplicates can be split several times.
84 : */
85 14269 : state = palloc_object(BTDedupStateData);
86 14269 : state->deduplicate = true;
87 14269 : state->nmaxitems = 0;
88 14269 : state->maxpostingsize = Min(BTMaxItemSize / 2, INDEX_SIZE_MASK);
89 : /* Metadata about base tuple of current pending posting list */
90 14269 : state->base = NULL;
91 14269 : state->baseoff = InvalidOffsetNumber;
92 14269 : state->basetupsize = 0;
93 : /* Metadata about current pending posting list TIDs */
94 14269 : state->htids = palloc(state->maxpostingsize);
95 14269 : state->nhtids = 0;
96 14269 : state->nitems = 0;
97 : /* Size of all physical tuples to be replaced by pending posting list */
98 14269 : state->phystupsize = 0;
99 : /* nintervals should be initialized to zero */
100 14269 : state->nintervals = 0;
101 :
102 14269 : minoff = P_FIRSTDATAKEY(opaque);
103 14269 : maxoff = PageGetMaxOffsetNumber(page);
104 :
105 : /*
106 : * Consider applying "single value" strategy, though only if the page
107 : * seems likely to be split in the near future
108 : */
109 14269 : if (!bottomupdedup)
110 12475 : singlevalstrat = _bt_do_singleval(rel, page, state, minoff, newitem);
111 :
112 : /*
113 : * Deduplicate items from page, and write them to newpage.
114 : *
115 : * Copy the original page's LSN into newpage copy. This will become the
116 : * updated version of the page. We need this because XLogInsert will
117 : * examine the LSN and possibly dump it in a page image.
118 : */
119 14269 : newpage = PageGetTempPageCopySpecial(page);
120 14269 : PageSetLSN(newpage, PageGetLSN(page));
121 :
122 : /* Copy high key, if any */
123 14269 : if (!P_RIGHTMOST(opaque))
124 : {
125 11448 : ItemId hitemid = PageGetItemId(page, P_HIKEY);
126 11448 : Size hitemsz = ItemIdGetLength(hitemid);
127 11448 : IndexTuple hitem = (IndexTuple) PageGetItem(page, hitemid);
128 :
129 11448 : if (PageAddItem(newpage, hitem, hitemsz, P_HIKEY, false, false) == InvalidOffsetNumber)
130 0 : elog(ERROR, "deduplication failed to add highkey");
131 : }
132 :
133 14269 : for (offnum = minoff;
134 3210480 : offnum <= maxoff;
135 3196211 : offnum = OffsetNumberNext(offnum))
136 : {
137 3196211 : ItemId itemid = PageGetItemId(page, offnum);
138 3196211 : IndexTuple itup = (IndexTuple) PageGetItem(page, itemid);
139 :
140 : Assert(!ItemIdIsDead(itemid));
141 :
142 3196211 : if (offnum == minoff)
143 : {
144 : /*
145 : * No previous/base tuple for the data item -- use the data item
146 : * as base tuple of pending posting list
147 : */
148 14269 : _bt_dedup_start_pending(state, itup, offnum);
149 : }
150 6362594 : else if (state->deduplicate &&
151 3615233 : _bt_keep_natts_fast(rel, state->base, itup) > nkeyatts &&
152 434581 : _bt_dedup_save_htid(state, itup))
153 : {
154 : /*
155 : * Tuple is equal to base tuple of pending posting list. Heap
156 : * TID(s) for itup have been saved in state.
157 : */
158 : }
159 : else
160 : {
161 : /*
162 : * Tuple is not equal to pending posting list tuple, or
163 : * _bt_dedup_save_htid() opted to not merge current item into
164 : * pending posting list for some other reason (e.g., adding more
165 : * TIDs would have caused posting list to exceed current
166 : * maxpostingsize).
167 : *
168 : * If state contains pending posting list with more than one item,
169 : * form new posting tuple and add it to our temp page (newpage).
170 : * Else add pending interval's base tuple to the temp page as-is.
171 : */
172 2754041 : pagesaving += _bt_dedup_finish_pending(newpage, state);
173 :
174 2754041 : if (singlevalstrat)
175 : {
176 : /*
177 : * Single value strategy's extra steps.
178 : *
179 : * Lower maxpostingsize for sixth and final large posting list
180 : * tuple at the point where 5 maxpostingsize-capped tuples
181 : * have either been formed or observed.
182 : *
183 : * When a sixth maxpostingsize-capped item is formed/observed,
184 : * stop merging together tuples altogether. The few tuples
185 : * that remain at the end of the page won't be merged together
186 : * at all (at least not until after a future page split takes
187 : * place, when this page's newly allocated right sibling page
188 : * gets its first deduplication pass).
189 : */
190 2857 : if (state->nmaxitems == 5)
191 396 : _bt_singleval_fillfactor(page, state, newitemsz);
192 2461 : else if (state->nmaxitems == 6)
193 : {
194 147 : state->deduplicate = false;
195 147 : singlevalstrat = false; /* won't be back here */
196 : }
197 : }
198 :
199 : /* itup starts new pending posting list */
200 2754041 : _bt_dedup_start_pending(state, itup, offnum);
201 : }
202 : }
203 :
204 : /* Handle the last item */
205 14269 : pagesaving += _bt_dedup_finish_pending(newpage, state);
206 :
207 : /*
208 : * If no items suitable for deduplication were found, newpage must be
209 : * exactly the same as the original page, so just return from function.
210 : *
211 : * We could determine whether or not to proceed on the basis the space
212 : * savings being sufficient to avoid an immediate page split instead. We
213 : * don't do that because there is some small value in nbtsplitloc.c always
214 : * operating against a page that is fully deduplicated (apart from
215 : * newitem). Besides, most of the cost has already been paid.
216 : */
217 14269 : if (state->nintervals == 0)
218 : {
219 : /* cannot leak memory here */
220 2508 : pfree(newpage);
221 2508 : pfree(state->htids);
222 2508 : pfree(state);
223 2508 : return;
224 : }
225 :
226 : /*
227 : * By here, it's clear that deduplication will definitely go ahead.
228 : *
229 : * Clear the BTP_HAS_GARBAGE page flag. The index must be a heapkeyspace
230 : * index, and as such we'll never pay attention to BTP_HAS_GARBAGE anyway.
231 : * But keep things tidy.
232 : */
233 11761 : if (P_HAS_GARBAGE(opaque))
234 : {
235 0 : BTPageOpaque nopaque = BTPageGetOpaque(newpage);
236 :
237 0 : nopaque->btpo_flags &= ~BTP_HAS_GARBAGE;
238 : }
239 :
240 11761 : START_CRIT_SECTION();
241 :
242 11761 : PageRestoreTempPage(newpage, page);
243 11761 : MarkBufferDirty(buf);
244 :
245 : /* XLOG stuff */
246 11761 : if (RelationNeedsWAL(rel))
247 : {
248 : XLogRecPtr recptr;
249 : xl_btree_dedup xlrec_dedup;
250 :
251 11698 : xlrec_dedup.nintervals = state->nintervals;
252 :
253 11698 : XLogBeginInsert();
254 11698 : XLogRegisterBuffer(0, buf, REGBUF_STANDARD);
255 11698 : XLogRegisterData(&xlrec_dedup, SizeOfBtreeDedup);
256 :
257 : /*
258 : * The intervals array is not in the buffer, but pretend that it is.
259 : * When XLogInsert stores the whole buffer, the array need not be
260 : * stored too.
261 : */
262 11698 : XLogRegisterBufData(0, state->intervals,
263 11698 : state->nintervals * sizeof(BTDedupInterval));
264 :
265 11698 : recptr = XLogInsert(RM_BTREE_ID, XLOG_BTREE_DEDUP);
266 :
267 11698 : PageSetLSN(page, recptr);
268 : }
269 :
270 11761 : END_CRIT_SECTION();
271 :
272 : /* Local space accounting should agree with page accounting */
273 : Assert(pagesaving < newitemsz || PageGetExactFreeSpace(page) >= newitemsz);
274 :
275 : /* cannot leak memory here */
276 11761 : pfree(state->htids);
277 11761 : pfree(state);
278 : }
279 :
280 : /*
281 : * Perform bottom-up index deletion pass.
282 : *
283 : * See if duplicate index tuples (plus certain nearby tuples) are eligible to
284 : * be deleted via bottom-up index deletion. The high level goal here is to
285 : * entirely prevent "unnecessary" page splits caused by MVCC version churn
286 : * from UPDATEs (when the UPDATEs don't logically modify any of the columns
287 : * covered by the 'rel' index). This is qualitative, not quantitative -- we
288 : * do not particularly care about once-off opportunities to delete many index
289 : * tuples together.
290 : *
291 : * See nbtree/README for details on the design of nbtree bottom-up deletion.
292 : * See access/tableam.h for a description of how we're expected to cooperate
293 : * with the tableam.
294 : *
295 : * Returns true on success, in which case caller can assume page split will be
296 : * avoided for a reasonable amount of time. Returns false when caller should
297 : * deduplicate the page (if possible at all).
298 : *
299 : * Note: Occasionally we return true despite failing to delete enough items to
300 : * avoid a split. This makes caller skip deduplication and go split the page
301 : * right away. Our return value is always just advisory information.
302 : *
303 : * Note: Caller should have already deleted all existing items with their
304 : * LP_DEAD bits set.
305 : */
306 : bool
307 2037 : _bt_bottomupdel_pass(Relation rel, Buffer buf, Relation heapRel,
308 : Size newitemsz)
309 : {
310 : OffsetNumber offnum,
311 : minoff,
312 : maxoff;
313 2037 : Page page = BufferGetPage(buf);
314 2037 : BTPageOpaque opaque = BTPageGetOpaque(page);
315 : BTDedupState state;
316 : TM_IndexDeleteOp delstate;
317 : bool neverdedup;
318 2037 : int nkeyatts = IndexRelationGetNumberOfKeyAttributes(rel);
319 :
320 : /* Passed-in newitemsz is MAXALIGNED but does not include line pointer */
321 2037 : newitemsz += sizeof(ItemIdData);
322 :
323 : /* Initialize deduplication state */
324 2037 : state = palloc_object(BTDedupStateData);
325 2037 : state->deduplicate = true;
326 2037 : state->nmaxitems = 0;
327 2037 : state->maxpostingsize = BLCKSZ; /* We're not really deduplicating */
328 2037 : state->base = NULL;
329 2037 : state->baseoff = InvalidOffsetNumber;
330 2037 : state->basetupsize = 0;
331 2037 : state->htids = palloc(state->maxpostingsize);
332 2037 : state->nhtids = 0;
333 2037 : state->nitems = 0;
334 2037 : state->phystupsize = 0;
335 2037 : state->nintervals = 0;
336 :
337 : /*
338 : * Initialize tableam state that describes bottom-up index deletion
339 : * operation.
340 : *
341 : * We'll go on to ask the tableam to search for TIDs whose index tuples we
342 : * can safely delete. The tableam will search until our leaf page space
343 : * target is satisfied, or until the cost of continuing with the tableam
344 : * operation seems too high. It focuses its efforts on TIDs associated
345 : * with duplicate index tuples that we mark "promising".
346 : *
347 : * This space target is a little arbitrary. The tableam must be able to
348 : * keep the costs and benefits in balance. We provide the tableam with
349 : * exhaustive information about what might work, without directly
350 : * concerning ourselves with avoiding work during the tableam call. Our
351 : * role in costing the bottom-up deletion process is strictly advisory.
352 : */
353 2037 : delstate.irel = rel;
354 2037 : delstate.iblknum = BufferGetBlockNumber(buf);
355 2037 : delstate.bottomup = true;
356 2037 : delstate.bottomupfreespace = Max(BLCKSZ / 16, newitemsz);
357 2037 : delstate.ndeltids = 0;
358 2037 : delstate.deltids = palloc_array(TM_IndexDelete, MaxTIDsPerBTreePage);
359 2037 : delstate.status = palloc_array(TM_IndexStatus, MaxTIDsPerBTreePage);
360 :
361 2037 : minoff = P_FIRSTDATAKEY(opaque);
362 2037 : maxoff = PageGetMaxOffsetNumber(page);
363 2037 : for (offnum = minoff;
364 588735 : offnum <= maxoff;
365 586698 : offnum = OffsetNumberNext(offnum))
366 : {
367 586698 : ItemId itemid = PageGetItemId(page, offnum);
368 586698 : IndexTuple itup = (IndexTuple) PageGetItem(page, itemid);
369 :
370 : Assert(!ItemIdIsDead(itemid));
371 :
372 586698 : if (offnum == minoff)
373 : {
374 : /* itup starts first pending interval */
375 2037 : _bt_dedup_start_pending(state, itup, offnum);
376 : }
377 660821 : else if (_bt_keep_natts_fast(rel, state->base, itup) > nkeyatts &&
378 76160 : _bt_dedup_save_htid(state, itup))
379 : {
380 : /* Tuple is equal; just added its TIDs to pending interval */
381 : }
382 : else
383 : {
384 : /* Finalize interval -- move its TIDs to delete state */
385 508501 : _bt_bottomupdel_finish_pending(page, state, &delstate);
386 :
387 : /* itup starts new pending interval */
388 508501 : _bt_dedup_start_pending(state, itup, offnum);
389 : }
390 : }
391 : /* Finalize final interval -- move its TIDs to delete state */
392 2037 : _bt_bottomupdel_finish_pending(page, state, &delstate);
393 :
394 : /*
395 : * We don't give up now in the event of having few (or even zero)
396 : * promising tuples for the tableam because it's not up to us as the index
397 : * AM to manage costs (note that the tableam might have heuristics of its
398 : * own that work out what to do). We should at least avoid having our
399 : * caller do a useless deduplication pass after we return in the event of
400 : * zero promising tuples, though.
401 : */
402 2037 : neverdedup = false;
403 2037 : if (state->nintervals == 0)
404 8 : neverdedup = true;
405 :
406 2037 : pfree(state->htids);
407 2037 : pfree(state);
408 :
409 : /* Ask tableam which TIDs are deletable, then physically delete them */
410 2037 : _bt_delitems_delete_check(rel, buf, heapRel, &delstate);
411 :
412 2037 : pfree(delstate.deltids);
413 2037 : pfree(delstate.status);
414 :
415 : /* Report "success" to caller unconditionally to avoid deduplication */
416 2037 : if (neverdedup)
417 8 : return true;
418 :
419 : /* Don't dedup when we won't end up back here any time soon anyway */
420 2029 : return PageGetExactFreeSpace(page) >= Max(BLCKSZ / 24, newitemsz);
421 : }
422 :
423 : /*
424 : * Create a new pending posting list tuple based on caller's base tuple.
425 : *
426 : * Every tuple processed by deduplication either becomes the base tuple for a
427 : * posting list, or gets its heap TID(s) accepted into a pending posting list.
428 : * A tuple that starts out as the base tuple for a posting list will only
429 : * actually be rewritten within _bt_dedup_finish_pending() when it turns out
430 : * that there are duplicates that can be merged into the base tuple.
431 : */
432 : void
433 6095227 : _bt_dedup_start_pending(BTDedupState state, IndexTuple base,
434 : OffsetNumber baseoff)
435 : {
436 : Assert(state->nhtids == 0);
437 : Assert(state->nitems == 0);
438 : Assert(!BTreeTupleIsPivot(base));
439 :
440 : /*
441 : * Copy heap TID(s) from new base tuple for new candidate posting list
442 : * into working state's array
443 : */
444 6095227 : if (!BTreeTupleIsPosting(base))
445 : {
446 5073950 : memcpy(state->htids, &base->t_tid, sizeof(ItemPointerData));
447 5073950 : state->nhtids = 1;
448 5073950 : state->basetupsize = IndexTupleSize(base);
449 : }
450 : else
451 : {
452 : int nposting;
453 :
454 1021277 : nposting = BTreeTupleGetNPosting(base);
455 1021277 : memcpy(state->htids, BTreeTupleGetPosting(base),
456 : sizeof(ItemPointerData) * nposting);
457 1021277 : state->nhtids = nposting;
458 : /* basetupsize should not include existing posting list */
459 1021277 : state->basetupsize = BTreeTupleGetPostingOffset(base);
460 : }
461 :
462 : /*
463 : * Save new base tuple itself -- it'll be needed if we actually create a
464 : * new posting list from new pending posting list.
465 : *
466 : * Must maintain physical size of all existing tuples (including line
467 : * pointer overhead) so that we can calculate space savings on page.
468 : */
469 6095227 : state->nitems = 1;
470 6095227 : state->base = base;
471 6095227 : state->baseoff = baseoff;
472 6095227 : state->phystupsize = MAXALIGN(IndexTupleSize(base)) + sizeof(ItemIdData);
473 : /* Also save baseoff in pending state for interval */
474 6095227 : state->intervals[state->nintervals].baseoff = state->baseoff;
475 6095227 : }
476 :
477 : /*
478 : * Save itup heap TID(s) into pending posting list where possible.
479 : *
480 : * Returns bool indicating if the pending posting list managed by state now
481 : * includes itup's heap TID(s).
482 : */
483 : bool
484 1290326 : _bt_dedup_save_htid(BTDedupState state, IndexTuple itup)
485 : {
486 : int nhtids;
487 : ItemPointer htids;
488 : Size mergedtupsz;
489 :
490 : Assert(!BTreeTupleIsPivot(itup));
491 :
492 1290326 : if (!BTreeTupleIsPosting(itup))
493 : {
494 1283770 : nhtids = 1;
495 1283770 : htids = &itup->t_tid;
496 : }
497 : else
498 : {
499 6556 : nhtids = BTreeTupleGetNPosting(itup);
500 6556 : htids = BTreeTupleGetPosting(itup);
501 : }
502 :
503 : /*
504 : * Don't append (have caller finish pending posting list as-is) if
505 : * appending heap TID(s) from itup would put us over maxpostingsize limit.
506 : *
507 : * This calculation needs to match the code used within _bt_form_posting()
508 : * for new posting list tuples.
509 : */
510 1290326 : mergedtupsz = MAXALIGN(state->basetupsize +
511 : (state->nhtids + nhtids) * sizeof(ItemPointerData));
512 :
513 1290326 : if (mergedtupsz > state->maxpostingsize)
514 : {
515 : /*
516 : * Count this as an oversized item for single value strategy, though
517 : * only when there are 50 TIDs in the final posting list tuple. This
518 : * limit (which is fairly arbitrary) avoids confusion about how many
519 : * 1/6 of a page tuples have been encountered/created by the current
520 : * deduplication pass.
521 : *
522 : * Note: We deliberately don't consider which deduplication pass
523 : * merged together tuples to create this item (could be a previous
524 : * deduplication pass, or current pass). See _bt_do_singleval()
525 : * comments.
526 : */
527 10657 : if (state->nhtids > 50)
528 10178 : state->nmaxitems++;
529 :
530 10657 : return false;
531 : }
532 :
533 : /*
534 : * Save heap TIDs to pending posting list tuple -- itup can be merged into
535 : * pending posting list
536 : */
537 1279669 : state->nitems++;
538 1279669 : memcpy(state->htids + state->nhtids, htids,
539 : sizeof(ItemPointerData) * nhtids);
540 1279669 : state->nhtids += nhtids;
541 1279669 : state->phystupsize += MAXALIGN(IndexTupleSize(itup)) + sizeof(ItemIdData);
542 :
543 1279669 : return true;
544 : }
545 :
546 : /*
547 : * Finalize pending posting list tuple, and add it to the page. Final tuple
548 : * is based on saved base tuple, and saved list of heap TIDs.
549 : *
550 : * Returns space saving from deduplicating to make a new posting list tuple.
551 : * Note that this includes line pointer overhead. This is zero in the case
552 : * where no deduplication was possible.
553 : */
554 : Size
555 3205516 : _bt_dedup_finish_pending(Page newpage, BTDedupState state)
556 : {
557 : OffsetNumber tupoff;
558 : Size tuplesz;
559 : Size spacesaving;
560 :
561 : Assert(state->nitems > 0);
562 : Assert(state->nitems <= state->nhtids);
563 : Assert(state->intervals[state->nintervals].baseoff == state->baseoff);
564 :
565 3205516 : tupoff = OffsetNumberNext(PageGetMaxOffsetNumber(newpage));
566 3205516 : if (state->nitems == 1)
567 : {
568 : /* Use original, unchanged base tuple */
569 2996609 : tuplesz = IndexTupleSize(state->base);
570 : Assert(tuplesz == MAXALIGN(IndexTupleSize(state->base)));
571 : Assert(tuplesz <= BTMaxItemSize);
572 2996609 : if (PageAddItem(newpage, state->base, tuplesz, tupoff, false, false) == InvalidOffsetNumber)
573 0 : elog(ERROR, "deduplication failed to add tuple to page");
574 :
575 2996609 : spacesaving = 0;
576 : }
577 : else
578 : {
579 : IndexTuple final;
580 :
581 : /* Form a tuple with a posting list */
582 208907 : final = _bt_form_posting(state->base, state->htids, state->nhtids);
583 208907 : tuplesz = IndexTupleSize(final);
584 : Assert(tuplesz <= state->maxpostingsize);
585 :
586 : /* Save final number of items for posting list */
587 208907 : state->intervals[state->nintervals].nitems = state->nitems;
588 :
589 : Assert(tuplesz == MAXALIGN(IndexTupleSize(final)));
590 : Assert(tuplesz <= BTMaxItemSize);
591 208907 : if (PageAddItem(newpage, final, tuplesz, tupoff, false, false) == InvalidOffsetNumber)
592 0 : elog(ERROR, "deduplication failed to add tuple to page");
593 :
594 208907 : pfree(final);
595 208907 : spacesaving = state->phystupsize - (tuplesz + sizeof(ItemIdData));
596 : /* Increment nintervals, since we wrote a new posting list tuple */
597 208907 : state->nintervals++;
598 : Assert(spacesaving > 0 && spacesaving < BLCKSZ);
599 : }
600 :
601 : /* Reset state for next pending posting list */
602 3205516 : state->nhtids = 0;
603 3205516 : state->nitems = 0;
604 3205516 : state->phystupsize = 0;
605 :
606 3205516 : return spacesaving;
607 : }
608 :
609 : /*
610 : * Finalize interval during bottom-up index deletion.
611 : *
612 : * During a bottom-up pass we expect that TIDs will be recorded in dedup state
613 : * first, and then get moved over to delstate (in variable-sized batches) by
614 : * calling here. Call here happens when the number of TIDs in a dedup
615 : * interval is known, and interval gets finalized (i.e. when caller sees next
616 : * tuple on the page is not a duplicate, or when caller runs out of tuples to
617 : * process from leaf page).
618 : *
619 : * This is where bottom-up deletion determines and remembers which entries are
620 : * duplicates. This will be important information to the tableam delete
621 : * infrastructure later on. Plain index tuple duplicates are marked
622 : * "promising" here, per tableam contract.
623 : *
624 : * Our approach to marking entries whose TIDs come from posting lists is more
625 : * complicated. Posting lists can only be formed by a deduplication pass (or
626 : * during an index build), so recent version churn affecting the pointed-to
627 : * logical rows is not particularly likely. We may still give a weak signal
628 : * about posting list tuples' entries (by marking just one of its TIDs/entries
629 : * promising), though this is only a possibility in the event of further
630 : * duplicate index tuples in final interval that covers posting list tuple (as
631 : * in the plain tuple case). A weak signal/hint will be useful to the tableam
632 : * when it has no stronger signal to go with for the deletion operation as a
633 : * whole.
634 : *
635 : * The heuristics we use work well in practice because we only need to give
636 : * the tableam the right _general_ idea about where to look. Garbage tends to
637 : * naturally get concentrated in relatively few table blocks with workloads
638 : * that bottom-up deletion targets. The tableam cannot possibly rank all
639 : * available table blocks sensibly based on the hints we provide, but that's
640 : * okay -- only the extremes matter. The tableam just needs to be able to
641 : * predict which few table blocks will have the most tuples that are safe to
642 : * delete for each deletion operation, with low variance across related
643 : * deletion operations.
644 : */
645 : static void
646 510538 : _bt_bottomupdel_finish_pending(Page page, BTDedupState state,
647 : TM_IndexDeleteOp *delstate)
648 : {
649 510538 : bool dupinterval = (state->nitems > 1);
650 :
651 : Assert(state->nitems > 0);
652 : Assert(state->nitems <= state->nhtids);
653 : Assert(state->intervals[state->nintervals].baseoff == state->baseoff);
654 :
655 1097236 : for (int i = 0; i < state->nitems; i++)
656 : {
657 586698 : OffsetNumber offnum = state->baseoff + i;
658 586698 : ItemId itemid = PageGetItemId(page, offnum);
659 586698 : IndexTuple itup = (IndexTuple) PageGetItem(page, itemid);
660 586698 : TM_IndexDelete *ideltid = &delstate->deltids[delstate->ndeltids];
661 586698 : TM_IndexStatus *istatus = &delstate->status[delstate->ndeltids];
662 :
663 586698 : if (!BTreeTupleIsPosting(itup))
664 : {
665 : /* Simple case: A plain non-pivot tuple */
666 476793 : ideltid->tid = itup->t_tid;
667 476793 : ideltid->id = delstate->ndeltids;
668 476793 : istatus->idxoffnum = offnum;
669 476793 : istatus->knowndeletable = false; /* for now */
670 476793 : istatus->promising = dupinterval; /* simple rule */
671 476793 : istatus->freespace = ItemIdGetLength(itemid) + sizeof(ItemIdData);
672 :
673 476793 : delstate->ndeltids++;
674 : }
675 : else
676 : {
677 : /*
678 : * Complicated case: A posting list tuple.
679 : *
680 : * We make the conservative assumption that there can only be at
681 : * most one affected logical row per posting list tuple. There
682 : * will be at most one promising entry in deltids to represent
683 : * this presumed lone logical row. Note that this isn't even
684 : * considered unless the posting list tuple is also in an interval
685 : * of duplicates -- this complicated rule is just a variant of the
686 : * simple rule used to decide if plain index tuples are promising.
687 : */
688 109905 : int nitem = BTreeTupleGetNPosting(itup);
689 109905 : bool firstpromising = false;
690 109905 : bool lastpromising = false;
691 :
692 : Assert(_bt_posting_valid(itup));
693 :
694 109905 : if (dupinterval)
695 : {
696 : /*
697 : * Complicated rule: either the first or last TID in the
698 : * posting list gets marked promising (if any at all)
699 : */
700 : BlockNumber minblocklist,
701 : midblocklist,
702 : maxblocklist;
703 : ItemPointer mintid,
704 : midtid,
705 : maxtid;
706 :
707 9681 : mintid = BTreeTupleGetHeapTID(itup);
708 9681 : midtid = BTreeTupleGetPostingN(itup, nitem / 2);
709 9681 : maxtid = BTreeTupleGetMaxHeapTID(itup);
710 9681 : minblocklist = ItemPointerGetBlockNumber(mintid);
711 9681 : midblocklist = ItemPointerGetBlockNumber(midtid);
712 9681 : maxblocklist = ItemPointerGetBlockNumber(maxtid);
713 :
714 : /* Only entry with predominant table block can be promising */
715 9681 : firstpromising = (minblocklist == midblocklist);
716 9681 : lastpromising = (!firstpromising &&
717 : midblocklist == maxblocklist);
718 : }
719 :
720 597049 : for (int p = 0; p < nitem; p++)
721 : {
722 487144 : ItemPointer htid = BTreeTupleGetPostingN(itup, p);
723 :
724 487144 : ideltid->tid = *htid;
725 487144 : ideltid->id = delstate->ndeltids;
726 487144 : istatus->idxoffnum = offnum;
727 487144 : istatus->knowndeletable = false; /* for now */
728 487144 : istatus->promising = false;
729 487144 : if ((firstpromising && p == 0) ||
730 64015 : (lastpromising && p == nitem - 1))
731 6907 : istatus->promising = true;
732 487144 : istatus->freespace = sizeof(ItemPointerData); /* at worst */
733 :
734 487144 : ideltid++;
735 487144 : istatus++;
736 487144 : delstate->ndeltids++;
737 : }
738 : }
739 : }
740 :
741 510538 : if (dupinterval)
742 : {
743 50036 : state->intervals[state->nintervals].nitems = state->nitems;
744 50036 : state->nintervals++;
745 : }
746 :
747 : /* Reset state for next interval */
748 510538 : state->nhtids = 0;
749 510538 : state->nitems = 0;
750 510538 : state->phystupsize = 0;
751 510538 : }
752 :
753 : /*
754 : * Determine if page non-pivot tuples (data items) are all duplicates of the
755 : * same value -- if they are, deduplication's "single value" strategy should
756 : * be applied. The general goal of this strategy is to ensure that
757 : * nbtsplitloc.c (which uses its own single value strategy) will find a useful
758 : * split point as further duplicates are inserted, and successive rightmost
759 : * page splits occur among pages that store the same duplicate value. When
760 : * the page finally splits, it should end up BTREE_SINGLEVAL_FILLFACTOR% full,
761 : * just like it would if deduplication were disabled.
762 : *
763 : * We expect that affected workloads will require _several_ single value
764 : * strategy deduplication passes (over a page that only stores duplicates)
765 : * before the page is finally split. The first deduplication pass should only
766 : * find regular non-pivot tuples. Later deduplication passes will find
767 : * existing maxpostingsize-capped posting list tuples, which must be skipped
768 : * over. The penultimate pass is generally the first pass that actually
769 : * reaches _bt_singleval_fillfactor(), and so will deliberately leave behind a
770 : * few untouched non-pivot tuples. The final deduplication pass won't free
771 : * any space -- it will skip over everything without merging anything (it
772 : * retraces the steps of the penultimate pass).
773 : *
774 : * Fortunately, having several passes isn't too expensive. Each pass (after
775 : * the first pass) won't spend many cycles on the large posting list tuples
776 : * left by previous passes. Each pass will find a large contiguous group of
777 : * smaller duplicate tuples to merge together at the end of the page.
778 : */
779 : static bool
780 12475 : _bt_do_singleval(Relation rel, Page page, BTDedupState state,
781 : OffsetNumber minoff, IndexTuple newitem)
782 : {
783 12475 : int nkeyatts = IndexRelationGetNumberOfKeyAttributes(rel);
784 : ItemId itemid;
785 : IndexTuple itup;
786 :
787 12475 : itemid = PageGetItemId(page, minoff);
788 12475 : itup = (IndexTuple) PageGetItem(page, itemid);
789 :
790 12475 : if (_bt_keep_natts_fast(rel, newitem, itup) > nkeyatts)
791 : {
792 1078 : itemid = PageGetItemId(page, PageGetMaxOffsetNumber(page));
793 1078 : itup = (IndexTuple) PageGetItem(page, itemid);
794 :
795 1078 : if (_bt_keep_natts_fast(rel, newitem, itup) > nkeyatts)
796 697 : return true;
797 : }
798 :
799 11778 : return false;
800 : }
801 :
802 : /*
803 : * Lower maxpostingsize when using "single value" strategy, to avoid a sixth
804 : * and final maxpostingsize-capped tuple. The sixth and final posting list
805 : * tuple will end up somewhat smaller than the first five. (Note: The first
806 : * five tuples could actually just be very large duplicate tuples that
807 : * couldn't be merged together at all. Deduplication will simply not modify
808 : * the page when that happens.)
809 : *
810 : * When there are six posting lists on the page (after current deduplication
811 : * pass goes on to create/observe a sixth very large tuple), caller should end
812 : * its deduplication pass. It isn't useful to try to deduplicate items that
813 : * are supposed to end up on the new right sibling page following the
814 : * anticipated page split. A future deduplication pass of future right
815 : * sibling page might take care of it. (This is why the first single value
816 : * strategy deduplication pass for a given leaf page will generally find only
817 : * plain non-pivot tuples -- see _bt_do_singleval() comments.)
818 : */
819 : static void
820 396 : _bt_singleval_fillfactor(Page page, BTDedupState state, Size newitemsz)
821 : {
822 : Size leftfree;
823 : int reduction;
824 :
825 : /* This calculation needs to match nbtsplitloc.c */
826 396 : leftfree = PageGetPageSize(page) - SizeOfPageHeaderData -
827 : MAXALIGN(sizeof(BTPageOpaqueData));
828 : /* Subtract size of new high key (includes pivot heap TID space) */
829 396 : leftfree -= newitemsz + MAXALIGN(sizeof(ItemPointerData));
830 :
831 : /*
832 : * Reduce maxpostingsize by an amount equal to target free space on left
833 : * half of page
834 : */
835 396 : reduction = leftfree * ((100 - BTREE_SINGLEVAL_FILLFACTOR) / 100.0);
836 396 : if (state->maxpostingsize > reduction)
837 396 : state->maxpostingsize -= reduction;
838 : else
839 0 : state->maxpostingsize = 0;
840 396 : }
841 :
842 : /*
843 : * Build a posting list tuple based on caller's "base" index tuple and list of
844 : * heap TIDs. When nhtids == 1, builds a standard non-pivot tuple without a
845 : * posting list. (Posting list tuples can never have a single heap TID, partly
846 : * because that ensures that deduplication always reduces final MAXALIGN()'d
847 : * size of entire tuple.)
848 : *
849 : * Convention is that posting list starts at a MAXALIGN()'d offset (rather
850 : * than a SHORTALIGN()'d offset), in line with the approach taken when
851 : * appending a heap TID to new pivot tuple/high key during suffix truncation.
852 : * This sometimes wastes a little space that was only needed as alignment
853 : * padding in the original tuple. Following this convention simplifies the
854 : * space accounting used when deduplicating a page (the same convention
855 : * simplifies the accounting for choosing a point to split a page at).
856 : *
857 : * Note: Caller's "htids" array must be unique and already in ascending TID
858 : * order. Any existing heap TIDs from "base" won't automatically appear in
859 : * returned posting list tuple (they must be included in htids array.)
860 : */
861 : IndexTuple
862 255942 : _bt_form_posting(IndexTuple base, const ItemPointerData *htids, int nhtids)
863 : {
864 : uint32 keysize,
865 : newsize;
866 : IndexTuple itup;
867 :
868 255942 : if (BTreeTupleIsPosting(base))
869 66915 : keysize = BTreeTupleGetPostingOffset(base);
870 : else
871 189027 : keysize = IndexTupleSize(base);
872 :
873 : Assert(!BTreeTupleIsPivot(base));
874 : Assert(nhtids > 0 && nhtids <= PG_UINT16_MAX);
875 : Assert(keysize == MAXALIGN(keysize));
876 :
877 : /* Determine final size of new tuple */
878 255942 : if (nhtids > 1)
879 229095 : newsize = MAXALIGN(keysize +
880 : nhtids * sizeof(ItemPointerData));
881 : else
882 26847 : newsize = keysize;
883 :
884 : Assert(newsize <= INDEX_SIZE_MASK);
885 : Assert(newsize == MAXALIGN(newsize));
886 :
887 : /* Allocate memory using palloc0() (matches index_form_tuple()) */
888 255942 : itup = palloc0(newsize);
889 255942 : memcpy(itup, base, keysize);
890 255942 : itup->t_info &= ~INDEX_SIZE_MASK;
891 255942 : itup->t_info |= newsize;
892 255942 : if (nhtids > 1)
893 : {
894 : /* Form posting list tuple */
895 229095 : BTreeTupleSetPosting(itup, nhtids, keysize);
896 229095 : memcpy(BTreeTupleGetPosting(itup), htids,
897 : sizeof(ItemPointerData) * nhtids);
898 : Assert(_bt_posting_valid(itup));
899 : }
900 : else
901 : {
902 : /* Form standard non-pivot tuple */
903 26847 : itup->t_info &= ~INDEX_ALT_TID_MASK;
904 26847 : ItemPointerCopy(htids, &itup->t_tid);
905 : Assert(ItemPointerIsValid(&itup->t_tid));
906 : }
907 :
908 255942 : return itup;
909 : }
910 :
911 : /*
912 : * Generate a replacement tuple by "updating" a posting list tuple so that it
913 : * no longer has TIDs that need to be deleted.
914 : *
915 : * Used by both VACUUM and index deletion. Caller's vacposting argument
916 : * points to the existing posting list tuple to be updated.
917 : *
918 : * On return, caller's vacposting argument will point to final "updated"
919 : * tuple, which will be palloc()'d in caller's memory context.
920 : */
921 : void
922 36058 : _bt_update_posting(BTVacuumPosting vacposting)
923 : {
924 36058 : IndexTuple origtuple = vacposting->itup;
925 : uint32 keysize,
926 : newsize;
927 : IndexTuple itup;
928 : int nhtids;
929 : int ui,
930 : d;
931 : ItemPointer htids;
932 :
933 36058 : nhtids = BTreeTupleGetNPosting(origtuple) - vacposting->ndeletedtids;
934 :
935 : Assert(_bt_posting_valid(origtuple));
936 : Assert(nhtids > 0 && nhtids < BTreeTupleGetNPosting(origtuple));
937 :
938 : /*
939 : * Determine final size of new tuple.
940 : *
941 : * This calculation needs to match the code used within _bt_form_posting()
942 : * for new posting list tuples. We avoid calling _bt_form_posting() here
943 : * to save ourselves a second memory allocation for a htids workspace.
944 : */
945 36058 : keysize = BTreeTupleGetPostingOffset(origtuple);
946 36058 : if (nhtids > 1)
947 7092 : newsize = MAXALIGN(keysize +
948 : nhtids * sizeof(ItemPointerData));
949 : else
950 28966 : newsize = keysize;
951 :
952 : Assert(newsize <= INDEX_SIZE_MASK);
953 : Assert(newsize == MAXALIGN(newsize));
954 :
955 : /* Allocate memory using palloc0() (matches index_form_tuple()) */
956 36058 : itup = palloc0(newsize);
957 36058 : memcpy(itup, origtuple, keysize);
958 36058 : itup->t_info &= ~INDEX_SIZE_MASK;
959 36058 : itup->t_info |= newsize;
960 :
961 36058 : if (nhtids > 1)
962 : {
963 : /* Form posting list tuple */
964 7092 : BTreeTupleSetPosting(itup, nhtids, keysize);
965 7092 : htids = BTreeTupleGetPosting(itup);
966 : }
967 : else
968 : {
969 : /* Form standard non-pivot tuple */
970 28966 : itup->t_info &= ~INDEX_ALT_TID_MASK;
971 28966 : htids = &itup->t_tid;
972 : }
973 :
974 36058 : ui = 0;
975 36058 : d = 0;
976 194650 : for (int i = 0; i < BTreeTupleGetNPosting(origtuple); i++)
977 : {
978 158592 : if (d < vacposting->ndeletedtids && vacposting->deletetids[d] == i)
979 : {
980 77000 : d++;
981 77000 : continue;
982 : }
983 81592 : htids[ui++] = *BTreeTupleGetPostingN(origtuple, i);
984 : }
985 : Assert(ui == nhtids);
986 : Assert(d == vacposting->ndeletedtids);
987 : Assert(nhtids == 1 || _bt_posting_valid(itup));
988 : Assert(nhtids > 1 || ItemPointerIsValid(&itup->t_tid));
989 :
990 : /* vacposting arg's itup will now point to updated version */
991 36058 : vacposting->itup = itup;
992 36058 : }
993 :
994 : /*
995 : * Prepare for a posting list split by swapping heap TID in newitem with heap
996 : * TID from original posting list (the 'oposting' heap TID located at offset
997 : * 'postingoff'). Modifies newitem, so caller should pass their own private
998 : * copy that can safely be modified.
999 : *
1000 : * Returns new posting list tuple, which is palloc()'d in caller's context.
1001 : * This is guaranteed to be the same size as 'oposting'. Modified newitem is
1002 : * what caller actually inserts. (This happens inside the same critical
1003 : * section that performs an in-place update of old posting list using new
1004 : * posting list returned here.)
1005 : *
1006 : * While the keys from newitem and oposting must be opclass equal, and must
1007 : * generate identical output when run through the underlying type's output
1008 : * function, it doesn't follow that their representations match exactly.
1009 : * Caller must avoid assuming that there can't be representational differences
1010 : * that make datums from oposting bigger or smaller than the corresponding
1011 : * datums from newitem. For example, differences in TOAST input state might
1012 : * break a faulty assumption about tuple size (the executor is entitled to
1013 : * apply TOAST compression based on its own criteria). It also seems possible
1014 : * that further representational variation will be introduced in the future,
1015 : * in order to support nbtree features like page-level prefix compression.
1016 : *
1017 : * See nbtree/README for details on the design of posting list splits.
1018 : */
1019 : IndexTuple
1020 17891 : _bt_swap_posting(IndexTuple newitem, IndexTuple oposting, int postingoff)
1021 : {
1022 : int nhtids;
1023 : char *replacepos;
1024 : char *replaceposright;
1025 : Size nmovebytes;
1026 : IndexTuple nposting;
1027 :
1028 17891 : nhtids = BTreeTupleGetNPosting(oposting);
1029 : Assert(_bt_posting_valid(oposting));
1030 :
1031 : /*
1032 : * The postingoff argument originated as a _bt_binsrch_posting() return
1033 : * value. It will be 0 in the event of corruption that makes a leaf page
1034 : * contain a non-pivot tuple that's somehow identical to newitem (no two
1035 : * non-pivot tuples should ever have the same TID). This has been known
1036 : * to happen in the field from time to time.
1037 : *
1038 : * Perform a basic sanity check to catch this case now.
1039 : */
1040 17891 : if (!(postingoff > 0 && postingoff < nhtids))
1041 0 : elog(ERROR, "posting list tuple with %d items cannot be split at offset %d",
1042 : nhtids, postingoff);
1043 :
1044 : /*
1045 : * Move item pointers in posting list to make a gap for the new item's
1046 : * heap TID. We shift TIDs one place to the right, losing original
1047 : * rightmost TID. (nmovebytes must not include TIDs to the left of
1048 : * postingoff, nor the existing rightmost/max TID that gets overwritten.)
1049 : */
1050 17891 : nposting = CopyIndexTuple(oposting);
1051 17891 : replacepos = (char *) BTreeTupleGetPostingN(nposting, postingoff);
1052 17891 : replaceposright = (char *) BTreeTupleGetPostingN(nposting, postingoff + 1);
1053 17891 : nmovebytes = (nhtids - postingoff - 1) * sizeof(ItemPointerData);
1054 17891 : memmove(replaceposright, replacepos, nmovebytes);
1055 :
1056 : /* Fill the gap at postingoff with TID of new item (original new TID) */
1057 : Assert(!BTreeTupleIsPivot(newitem) && !BTreeTupleIsPosting(newitem));
1058 17891 : ItemPointerCopy(&newitem->t_tid, (ItemPointer) replacepos);
1059 :
1060 : /* Now copy oposting's rightmost/max TID into new item (final new TID) */
1061 17891 : ItemPointerCopy(BTreeTupleGetMaxHeapTID(oposting), &newitem->t_tid);
1062 :
1063 : Assert(ItemPointerCompare(BTreeTupleGetMaxHeapTID(nposting),
1064 : BTreeTupleGetHeapTID(newitem)) < 0);
1065 : Assert(_bt_posting_valid(nposting));
1066 :
1067 17891 : return nposting;
1068 : }
1069 :
1070 : /*
1071 : * Verify posting list invariants for "posting", which must be a posting list
1072 : * tuple. Used within assertions.
1073 : */
1074 : #ifdef USE_ASSERT_CHECKING
1075 : static bool
1076 : _bt_posting_valid(IndexTuple posting)
1077 : {
1078 : ItemPointerData last;
1079 : ItemPointer htid;
1080 :
1081 : if (!BTreeTupleIsPosting(posting) || BTreeTupleGetNPosting(posting) < 2)
1082 : return false;
1083 :
1084 : /* Remember first heap TID for loop */
1085 : ItemPointerCopy(BTreeTupleGetHeapTID(posting), &last);
1086 : if (!ItemPointerIsValid(&last))
1087 : return false;
1088 :
1089 : /* Iterate, starting from second TID */
1090 : for (int i = 1; i < BTreeTupleGetNPosting(posting); i++)
1091 : {
1092 : htid = BTreeTupleGetPostingN(posting, i);
1093 :
1094 : if (!ItemPointerIsValid(htid))
1095 : return false;
1096 : if (ItemPointerCompare(htid, &last) <= 0)
1097 : return false;
1098 : ItemPointerCopy(htid, &last);
1099 : }
1100 :
1101 : return true;
1102 : }
1103 : #endif
|
