| 1 | /*------------------------------------------------------------------------- |
| 2 | * |
| 3 | * gistget.c |
| 4 | * fetch tuples from a GiST scan. |
| 5 | * |
| 6 | * |
| 7 | * Portions Copyright (c) 1996-2019, PostgreSQL Global Development Group |
| 8 | * Portions Copyright (c) 1994, Regents of the University of California |
| 9 | * |
| 10 | * IDENTIFICATION |
| 11 | * src/backend/access/gist/gistget.c |
| 12 | * |
| 13 | *------------------------------------------------------------------------- |
| 14 | */ |
| 15 | #include "postgres.h" |
| 16 | |
| 17 | #include "access/genam.h" |
| 18 | #include "access/gist_private.h" |
| 19 | #include "access/relscan.h" |
| 20 | #include "miscadmin.h" |
| 21 | #include "storage/lmgr.h" |
| 22 | #include "storage/predicate.h" |
| 23 | #include "pgstat.h" |
| 24 | #include "lib/pairingheap.h" |
| 25 | #include "utils/float.h" |
| 26 | #include "utils/memutils.h" |
| 27 | #include "utils/rel.h" |
| 28 | |
| 29 | /* |
| 30 | * gistkillitems() -- set LP_DEAD state for items an indexscan caller has |
| 31 | * told us were killed. |
| 32 | * |
| 33 | * We re-read page here, so it's important to check page LSN. If the page |
| 34 | * has been modified since the last read (as determined by LSN), we cannot |
| 35 | * flag any entries because it is possible that the old entry was vacuumed |
| 36 | * away and the TID was re-used by a completely different heap tuple. |
| 37 | */ |
| 38 | static void |
| 39 | gistkillitems(IndexScanDesc scan) |
| 40 | { |
| 41 | GISTScanOpaque so = (GISTScanOpaque) scan->opaque; |
| 42 | Buffer buffer; |
| 43 | Page page; |
| 44 | OffsetNumber offnum; |
| 45 | ItemId iid; |
| 46 | int i; |
| 47 | bool killedsomething = false; |
| 48 | |
| 49 | Assert(so->curBlkno != InvalidBlockNumber); |
| 50 | Assert(!XLogRecPtrIsInvalid(so->curPageLSN)); |
| 51 | Assert(so->killedItems != NULL); |
| 52 | |
| 53 | buffer = ReadBuffer(scan->indexRelation, so->curBlkno); |
| 54 | if (!BufferIsValid(buffer)) |
| 55 | return; |
| 56 | |
| 57 | LockBuffer(buffer, GIST_SHARE); |
| 58 | gistcheckpage(scan->indexRelation, buffer); |
| 59 | page = BufferGetPage(buffer); |
| 60 | |
| 61 | /* |
| 62 | * If page LSN differs it means that the page was modified since the last |
| 63 | * read. killedItems could be not valid so LP_DEAD hints applying is not |
| 64 | * safe. |
| 65 | */ |
| 66 | if (BufferGetLSNAtomic(buffer) != so->curPageLSN) |
| 67 | { |
| 68 | UnlockReleaseBuffer(buffer); |
| 69 | so->numKilled = 0; /* reset counter */ |
| 70 | return; |
| 71 | } |
| 72 | |
| 73 | Assert(GistPageIsLeaf(page)); |
| 74 | |
| 75 | /* |
| 76 | * Mark all killedItems as dead. We need no additional recheck, because, |
| 77 | * if page was modified, pageLSN must have changed. |
| 78 | */ |
| 79 | for (i = 0; i < so->numKilled; i++) |
| 80 | { |
| 81 | offnum = so->killedItems[i]; |
| 82 | iid = PageGetItemId(page, offnum); |
| 83 | ItemIdMarkDead(iid); |
| 84 | killedsomething = true; |
| 85 | } |
| 86 | |
| 87 | if (killedsomething) |
| 88 | { |
| 89 | GistMarkPageHasGarbage(page); |
| 90 | MarkBufferDirtyHint(buffer, true); |
| 91 | } |
| 92 | |
| 93 | UnlockReleaseBuffer(buffer); |
| 94 | |
| 95 | /* |
| 96 | * Always reset the scan state, so we don't look for same items on other |
| 97 | * pages. |
| 98 | */ |
| 99 | so->numKilled = 0; |
| 100 | } |
| 101 | |
| 102 | /* |
| 103 | * gistindex_keytest() -- does this index tuple satisfy the scan key(s)? |
| 104 | * |
| 105 | * The index tuple might represent either a heap tuple or a lower index page, |
| 106 | * depending on whether the containing page is a leaf page or not. |
| 107 | * |
| 108 | * On success return for a heap tuple, *recheck_p is set to indicate whether |
| 109 | * the quals need to be rechecked. We recheck if any of the consistent() |
| 110 | * functions request it. recheck is not interesting when examining a non-leaf |
| 111 | * entry, since we must visit the lower index page if there's any doubt. |
| 112 | * Similarly, *recheck_distances_p is set to indicate whether the distances |
| 113 | * need to be rechecked, and it is also ignored for non-leaf entries. |
| 114 | * |
| 115 | * If we are doing an ordered scan, so->distances[] is filled with distance |
| 116 | * data from the distance() functions before returning success. |
| 117 | * |
| 118 | * We must decompress the key in the IndexTuple before passing it to the |
| 119 | * sk_funcs (which actually are the opclass Consistent or Distance methods). |
| 120 | * |
| 121 | * Note that this function is always invoked in a short-lived memory context, |
| 122 | * so we don't need to worry about cleaning up allocated memory, either here |
| 123 | * or in the implementation of any Consistent or Distance methods. |
| 124 | */ |
| 125 | static bool |
| 126 | gistindex_keytest(IndexScanDesc scan, |
| 127 | IndexTuple tuple, |
| 128 | Page page, |
| 129 | OffsetNumber offset, |
| 130 | bool *recheck_p, |
| 131 | bool *recheck_distances_p) |
| 132 | { |
| 133 | GISTScanOpaque so = (GISTScanOpaque) scan->opaque; |
| 134 | GISTSTATE *giststate = so->giststate; |
| 135 | ScanKey key = scan->keyData; |
| 136 | int keySize = scan->numberOfKeys; |
| 137 | IndexOrderByDistance *distance_p; |
| 138 | Relation r = scan->indexRelation; |
| 139 | |
| 140 | *recheck_p = false; |
| 141 | *recheck_distances_p = false; |
| 142 | |
| 143 | /* |
| 144 | * If it's a leftover invalid tuple from pre-9.1, treat it as a match with |
| 145 | * minimum possible distances. This means we'll always follow it to the |
| 146 | * referenced page. |
| 147 | */ |
| 148 | if (GistTupleIsInvalid(tuple)) |
| 149 | { |
| 150 | int i; |
| 151 | |
| 152 | if (GistPageIsLeaf(page)) /* shouldn't happen */ |
| 153 | elog(ERROR, "invalid GiST tuple found on leaf page" ); |
| 154 | for (i = 0; i < scan->numberOfOrderBys; i++) |
| 155 | { |
| 156 | so->distances[i].value = -get_float8_infinity(); |
| 157 | so->distances[i].isnull = false; |
| 158 | } |
| 159 | return true; |
| 160 | } |
| 161 | |
| 162 | /* Check whether it matches according to the Consistent functions */ |
| 163 | while (keySize > 0) |
| 164 | { |
| 165 | Datum datum; |
| 166 | bool isNull; |
| 167 | |
| 168 | datum = index_getattr(tuple, |
| 169 | key->sk_attno, |
| 170 | giststate->leafTupdesc, |
| 171 | &isNull); |
| 172 | |
| 173 | if (key->sk_flags & SK_ISNULL) |
| 174 | { |
| 175 | /* |
| 176 | * On non-leaf page we can't conclude that child hasn't NULL |
| 177 | * values because of assumption in GiST: union (VAL, NULL) is VAL. |
| 178 | * But if on non-leaf page key IS NULL, then all children are |
| 179 | * NULL. |
| 180 | */ |
| 181 | if (key->sk_flags & SK_SEARCHNULL) |
| 182 | { |
| 183 | if (GistPageIsLeaf(page) && !isNull) |
| 184 | return false; |
| 185 | } |
| 186 | else |
| 187 | { |
| 188 | Assert(key->sk_flags & SK_SEARCHNOTNULL); |
| 189 | if (isNull) |
| 190 | return false; |
| 191 | } |
| 192 | } |
| 193 | else if (isNull) |
| 194 | { |
| 195 | return false; |
| 196 | } |
| 197 | else |
| 198 | { |
| 199 | Datum test; |
| 200 | bool recheck; |
| 201 | GISTENTRY de; |
| 202 | |
| 203 | gistdentryinit(giststate, key->sk_attno - 1, &de, |
| 204 | datum, r, page, offset, |
| 205 | false, isNull); |
| 206 | |
| 207 | /* |
| 208 | * Call the Consistent function to evaluate the test. The |
| 209 | * arguments are the index datum (as a GISTENTRY*), the comparison |
| 210 | * datum, the comparison operator's strategy number and subtype |
| 211 | * from pg_amop, and the recheck flag. |
| 212 | * |
| 213 | * (Presently there's no need to pass the subtype since it'll |
| 214 | * always be zero, but might as well pass it for possible future |
| 215 | * use.) |
| 216 | * |
| 217 | * We initialize the recheck flag to true (the safest assumption) |
| 218 | * in case the Consistent function forgets to set it. |
| 219 | */ |
| 220 | recheck = true; |
| 221 | |
| 222 | test = FunctionCall5Coll(&key->sk_func, |
| 223 | key->sk_collation, |
| 224 | PointerGetDatum(&de), |
| 225 | key->sk_argument, |
| 226 | Int16GetDatum(key->sk_strategy), |
| 227 | ObjectIdGetDatum(key->sk_subtype), |
| 228 | PointerGetDatum(&recheck)); |
| 229 | |
| 230 | if (!DatumGetBool(test)) |
| 231 | return false; |
| 232 | *recheck_p |= recheck; |
| 233 | } |
| 234 | |
| 235 | key++; |
| 236 | keySize--; |
| 237 | } |
| 238 | |
| 239 | /* OK, it passes --- now let's compute the distances */ |
| 240 | key = scan->orderByData; |
| 241 | distance_p = so->distances; |
| 242 | keySize = scan->numberOfOrderBys; |
| 243 | while (keySize > 0) |
| 244 | { |
| 245 | Datum datum; |
| 246 | bool isNull; |
| 247 | |
| 248 | datum = index_getattr(tuple, |
| 249 | key->sk_attno, |
| 250 | giststate->leafTupdesc, |
| 251 | &isNull); |
| 252 | |
| 253 | if ((key->sk_flags & SK_ISNULL) || isNull) |
| 254 | { |
| 255 | /* Assume distance computes as null */ |
| 256 | distance_p->value = 0.0; |
| 257 | distance_p->isnull = true; |
| 258 | } |
| 259 | else |
| 260 | { |
| 261 | Datum dist; |
| 262 | bool recheck; |
| 263 | GISTENTRY de; |
| 264 | |
| 265 | gistdentryinit(giststate, key->sk_attno - 1, &de, |
| 266 | datum, r, page, offset, |
| 267 | false, isNull); |
| 268 | |
| 269 | /* |
| 270 | * Call the Distance function to evaluate the distance. The |
| 271 | * arguments are the index datum (as a GISTENTRY*), the comparison |
| 272 | * datum, the ordering operator's strategy number and subtype from |
| 273 | * pg_amop, and the recheck flag. |
| 274 | * |
| 275 | * (Presently there's no need to pass the subtype since it'll |
| 276 | * always be zero, but might as well pass it for possible future |
| 277 | * use.) |
| 278 | * |
| 279 | * If the function sets the recheck flag, the returned distance is |
| 280 | * a lower bound on the true distance and needs to be rechecked. |
| 281 | * We initialize the flag to 'false'. This flag was added in |
| 282 | * version 9.5; distance functions written before that won't know |
| 283 | * about the flag, but are expected to never be lossy. |
| 284 | */ |
| 285 | recheck = false; |
| 286 | dist = FunctionCall5Coll(&key->sk_func, |
| 287 | key->sk_collation, |
| 288 | PointerGetDatum(&de), |
| 289 | key->sk_argument, |
| 290 | Int16GetDatum(key->sk_strategy), |
| 291 | ObjectIdGetDatum(key->sk_subtype), |
| 292 | PointerGetDatum(&recheck)); |
| 293 | *recheck_distances_p |= recheck; |
| 294 | distance_p->value = DatumGetFloat8(dist); |
| 295 | distance_p->isnull = false; |
| 296 | } |
| 297 | |
| 298 | key++; |
| 299 | distance_p++; |
| 300 | keySize--; |
| 301 | } |
| 302 | |
| 303 | return true; |
| 304 | } |
| 305 | |
| 306 | /* |
| 307 | * Scan all items on the GiST index page identified by *pageItem, and insert |
| 308 | * them into the queue (or directly to output areas) |
| 309 | * |
| 310 | * scan: index scan we are executing |
| 311 | * pageItem: search queue item identifying an index page to scan |
| 312 | * myDistances: distances array associated with pageItem, or NULL at the root |
| 313 | * tbm: if not NULL, gistgetbitmap's output bitmap |
| 314 | * ntids: if not NULL, gistgetbitmap's output tuple counter |
| 315 | * |
| 316 | * If tbm/ntids aren't NULL, we are doing an amgetbitmap scan, and heap |
| 317 | * tuples should be reported directly into the bitmap. If they are NULL, |
| 318 | * we're doing a plain or ordered indexscan. For a plain indexscan, heap |
| 319 | * tuple TIDs are returned into so->pageData[]. For an ordered indexscan, |
| 320 | * heap tuple TIDs are pushed into individual search queue items. In an |
| 321 | * index-only scan, reconstructed index tuples are returned along with the |
| 322 | * TIDs. |
| 323 | * |
| 324 | * If we detect that the index page has split since we saw its downlink |
| 325 | * in the parent, we push its new right sibling onto the queue so the |
| 326 | * sibling will be processed next. |
| 327 | */ |
| 328 | static void |
| 329 | gistScanPage(IndexScanDesc scan, GISTSearchItem *pageItem, |
| 330 | IndexOrderByDistance *myDistances, TIDBitmap *tbm, int64 *ntids) |
| 331 | { |
| 332 | GISTScanOpaque so = (GISTScanOpaque) scan->opaque; |
| 333 | GISTSTATE *giststate = so->giststate; |
| 334 | Relation r = scan->indexRelation; |
| 335 | Buffer buffer; |
| 336 | Page page; |
| 337 | GISTPageOpaque opaque; |
| 338 | OffsetNumber maxoff; |
| 339 | OffsetNumber i; |
| 340 | MemoryContext oldcxt; |
| 341 | |
| 342 | Assert(!GISTSearchItemIsHeap(*pageItem)); |
| 343 | |
| 344 | buffer = ReadBuffer(scan->indexRelation, pageItem->blkno); |
| 345 | LockBuffer(buffer, GIST_SHARE); |
| 346 | PredicateLockPage(r, BufferGetBlockNumber(buffer), scan->xs_snapshot); |
| 347 | gistcheckpage(scan->indexRelation, buffer); |
| 348 | page = BufferGetPage(buffer); |
| 349 | TestForOldSnapshot(scan->xs_snapshot, r, page); |
| 350 | opaque = GistPageGetOpaque(page); |
| 351 | |
| 352 | /* |
| 353 | * Check if we need to follow the rightlink. We need to follow it if the |
| 354 | * page was concurrently split since we visited the parent (in which case |
| 355 | * parentlsn < nsn), or if the system crashed after a page split but |
| 356 | * before the downlink was inserted into the parent. |
| 357 | */ |
| 358 | if (!XLogRecPtrIsInvalid(pageItem->data.parentlsn) && |
| 359 | (GistFollowRight(page) || |
| 360 | pageItem->data.parentlsn < GistPageGetNSN(page)) && |
| 361 | opaque->rightlink != InvalidBlockNumber /* sanity check */ ) |
| 362 | { |
| 363 | /* There was a page split, follow right link to add pages */ |
| 364 | GISTSearchItem *item; |
| 365 | |
| 366 | /* This can't happen when starting at the root */ |
| 367 | Assert(myDistances != NULL); |
| 368 | |
| 369 | oldcxt = MemoryContextSwitchTo(so->queueCxt); |
| 370 | |
| 371 | /* Create new GISTSearchItem for the right sibling index page */ |
| 372 | item = palloc(SizeOfGISTSearchItem(scan->numberOfOrderBys)); |
| 373 | item->blkno = opaque->rightlink; |
| 374 | item->data.parentlsn = pageItem->data.parentlsn; |
| 375 | |
| 376 | /* Insert it into the queue using same distances as for this page */ |
| 377 | memcpy(item->distances, myDistances, |
| 378 | sizeof(item->distances[0]) * scan->numberOfOrderBys); |
| 379 | |
| 380 | pairingheap_add(so->queue, &item->phNode); |
| 381 | |
| 382 | MemoryContextSwitchTo(oldcxt); |
| 383 | } |
| 384 | |
| 385 | /* |
| 386 | * Check if the page was deleted after we saw the downlink. There's |
| 387 | * nothing of interest on a deleted page. Note that we must do this |
| 388 | * after checking the NSN for concurrent splits! It's possible that |
| 389 | * the page originally contained some tuples that are visible to us, |
| 390 | * but was split so that all the visible tuples were moved to another |
| 391 | * page, and then this page was deleted. |
| 392 | */ |
| 393 | if (GistPageIsDeleted(page)) |
| 394 | { |
| 395 | UnlockReleaseBuffer(buffer); |
| 396 | return; |
| 397 | } |
| 398 | |
| 399 | so->nPageData = so->curPageData = 0; |
| 400 | scan->xs_hitup = NULL; /* might point into pageDataCxt */ |
| 401 | if (so->pageDataCxt) |
| 402 | MemoryContextReset(so->pageDataCxt); |
| 403 | |
| 404 | /* |
| 405 | * We save the LSN of the page as we read it, so that we know whether it |
| 406 | * safe to apply LP_DEAD hints to the page later. This allows us to drop |
| 407 | * the pin for MVCC scans, which allows vacuum to avoid blocking. |
| 408 | */ |
| 409 | so->curPageLSN = BufferGetLSNAtomic(buffer); |
| 410 | |
| 411 | /* |
| 412 | * check all tuples on page |
| 413 | */ |
| 414 | maxoff = PageGetMaxOffsetNumber(page); |
| 415 | for (i = FirstOffsetNumber; i <= maxoff; i = OffsetNumberNext(i)) |
| 416 | { |
| 417 | ItemId iid = PageGetItemId(page, i); |
| 418 | IndexTuple it; |
| 419 | bool match; |
| 420 | bool recheck; |
| 421 | bool recheck_distances; |
| 422 | |
| 423 | /* |
| 424 | * If the scan specifies not to return killed tuples, then we treat a |
| 425 | * killed tuple as not passing the qual. |
| 426 | */ |
| 427 | if (scan->ignore_killed_tuples && ItemIdIsDead(iid)) |
| 428 | continue; |
| 429 | |
| 430 | it = (IndexTuple) PageGetItem(page, iid); |
| 431 | |
| 432 | /* |
| 433 | * Must call gistindex_keytest in tempCxt, and clean up any leftover |
| 434 | * junk afterward. |
| 435 | */ |
| 436 | oldcxt = MemoryContextSwitchTo(so->giststate->tempCxt); |
| 437 | |
| 438 | match = gistindex_keytest(scan, it, page, i, |
| 439 | &recheck, &recheck_distances); |
| 440 | |
| 441 | MemoryContextSwitchTo(oldcxt); |
| 442 | MemoryContextReset(so->giststate->tempCxt); |
| 443 | |
| 444 | /* Ignore tuple if it doesn't match */ |
| 445 | if (!match) |
| 446 | continue; |
| 447 | |
| 448 | if (tbm && GistPageIsLeaf(page)) |
| 449 | { |
| 450 | /* |
| 451 | * getbitmap scan, so just push heap tuple TIDs into the bitmap |
| 452 | * without worrying about ordering |
| 453 | */ |
| 454 | tbm_add_tuples(tbm, &it->t_tid, 1, recheck); |
| 455 | (*ntids)++; |
| 456 | } |
| 457 | else if (scan->numberOfOrderBys == 0 && GistPageIsLeaf(page)) |
| 458 | { |
| 459 | /* |
| 460 | * Non-ordered scan, so report tuples in so->pageData[] |
| 461 | */ |
| 462 | so->pageData[so->nPageData].heapPtr = it->t_tid; |
| 463 | so->pageData[so->nPageData].recheck = recheck; |
| 464 | so->pageData[so->nPageData].offnum = i; |
| 465 | |
| 466 | /* |
| 467 | * In an index-only scan, also fetch the data from the tuple. The |
| 468 | * reconstructed tuples are stored in pageDataCxt. |
| 469 | */ |
| 470 | if (scan->xs_want_itup) |
| 471 | { |
| 472 | oldcxt = MemoryContextSwitchTo(so->pageDataCxt); |
| 473 | so->pageData[so->nPageData].recontup = |
| 474 | gistFetchTuple(giststate, r, it); |
| 475 | MemoryContextSwitchTo(oldcxt); |
| 476 | } |
| 477 | so->nPageData++; |
| 478 | } |
| 479 | else |
| 480 | { |
| 481 | /* |
| 482 | * Must push item into search queue. We get here for any lower |
| 483 | * index page, and also for heap tuples if doing an ordered |
| 484 | * search. |
| 485 | */ |
| 486 | GISTSearchItem *item; |
| 487 | int nOrderBys = scan->numberOfOrderBys; |
| 488 | |
| 489 | oldcxt = MemoryContextSwitchTo(so->queueCxt); |
| 490 | |
| 491 | /* Create new GISTSearchItem for this item */ |
| 492 | item = palloc(SizeOfGISTSearchItem(scan->numberOfOrderBys)); |
| 493 | |
| 494 | if (GistPageIsLeaf(page)) |
| 495 | { |
| 496 | /* Creating heap-tuple GISTSearchItem */ |
| 497 | item->blkno = InvalidBlockNumber; |
| 498 | item->data.heap.heapPtr = it->t_tid; |
| 499 | item->data.heap.recheck = recheck; |
| 500 | item->data.heap.recheckDistances = recheck_distances; |
| 501 | |
| 502 | /* |
| 503 | * In an index-only scan, also fetch the data from the tuple. |
| 504 | */ |
| 505 | if (scan->xs_want_itup) |
| 506 | item->data.heap.recontup = gistFetchTuple(giststate, r, it); |
| 507 | } |
| 508 | else |
| 509 | { |
| 510 | /* Creating index-page GISTSearchItem */ |
| 511 | item->blkno = ItemPointerGetBlockNumber(&it->t_tid); |
| 512 | |
| 513 | /* |
| 514 | * LSN of current page is lsn of parent page for child. We |
| 515 | * only have a shared lock, so we need to get the LSN |
| 516 | * atomically. |
| 517 | */ |
| 518 | item->data.parentlsn = BufferGetLSNAtomic(buffer); |
| 519 | } |
| 520 | |
| 521 | /* Insert it into the queue using new distance data */ |
| 522 | memcpy(item->distances, so->distances, |
| 523 | sizeof(item->distances[0]) * nOrderBys); |
| 524 | |
| 525 | pairingheap_add(so->queue, &item->phNode); |
| 526 | |
| 527 | MemoryContextSwitchTo(oldcxt); |
| 528 | } |
| 529 | } |
| 530 | |
| 531 | UnlockReleaseBuffer(buffer); |
| 532 | } |
| 533 | |
| 534 | /* |
| 535 | * Extract next item (in order) from search queue |
| 536 | * |
| 537 | * Returns a GISTSearchItem or NULL. Caller must pfree item when done with it. |
| 538 | */ |
| 539 | static GISTSearchItem * |
| 540 | getNextGISTSearchItem(GISTScanOpaque so) |
| 541 | { |
| 542 | GISTSearchItem *item; |
| 543 | |
| 544 | if (!pairingheap_is_empty(so->queue)) |
| 545 | { |
| 546 | item = (GISTSearchItem *) pairingheap_remove_first(so->queue); |
| 547 | } |
| 548 | else |
| 549 | { |
| 550 | /* Done when both heaps are empty */ |
| 551 | item = NULL; |
| 552 | } |
| 553 | |
| 554 | /* Return item; caller is responsible to pfree it */ |
| 555 | return item; |
| 556 | } |
| 557 | |
| 558 | /* |
| 559 | * Fetch next heap tuple in an ordered search |
| 560 | */ |
| 561 | static bool |
| 562 | getNextNearest(IndexScanDesc scan) |
| 563 | { |
| 564 | GISTScanOpaque so = (GISTScanOpaque) scan->opaque; |
| 565 | bool res = false; |
| 566 | |
| 567 | if (scan->xs_hitup) |
| 568 | { |
| 569 | /* free previously returned tuple */ |
| 570 | pfree(scan->xs_hitup); |
| 571 | scan->xs_hitup = NULL; |
| 572 | } |
| 573 | |
| 574 | do |
| 575 | { |
| 576 | GISTSearchItem *item = getNextGISTSearchItem(so); |
| 577 | |
| 578 | if (!item) |
| 579 | break; |
| 580 | |
| 581 | if (GISTSearchItemIsHeap(*item)) |
| 582 | { |
| 583 | /* found a heap item at currently minimal distance */ |
| 584 | scan->xs_heaptid = item->data.heap.heapPtr; |
| 585 | scan->xs_recheck = item->data.heap.recheck; |
| 586 | |
| 587 | index_store_float8_orderby_distances(scan, so->orderByTypes, |
| 588 | item->distances, |
| 589 | item->data.heap.recheckDistances); |
| 590 | |
| 591 | /* in an index-only scan, also return the reconstructed tuple. */ |
| 592 | if (scan->xs_want_itup) |
| 593 | scan->xs_hitup = item->data.heap.recontup; |
| 594 | res = true; |
| 595 | } |
| 596 | else |
| 597 | { |
| 598 | /* visit an index page, extract its items into queue */ |
| 599 | CHECK_FOR_INTERRUPTS(); |
| 600 | |
| 601 | gistScanPage(scan, item, item->distances, NULL, NULL); |
| 602 | } |
| 603 | |
| 604 | pfree(item); |
| 605 | } while (!res); |
| 606 | |
| 607 | return res; |
| 608 | } |
| 609 | |
| 610 | /* |
| 611 | * gistgettuple() -- Get the next tuple in the scan |
| 612 | */ |
| 613 | bool |
| 614 | gistgettuple(IndexScanDesc scan, ScanDirection dir) |
| 615 | { |
| 616 | GISTScanOpaque so = (GISTScanOpaque) scan->opaque; |
| 617 | |
| 618 | if (dir != ForwardScanDirection) |
| 619 | elog(ERROR, "GiST only supports forward scan direction" ); |
| 620 | |
| 621 | if (!so->qual_ok) |
| 622 | return false; |
| 623 | |
| 624 | if (so->firstCall) |
| 625 | { |
| 626 | /* Begin the scan by processing the root page */ |
| 627 | GISTSearchItem fakeItem; |
| 628 | |
| 629 | pgstat_count_index_scan(scan->indexRelation); |
| 630 | |
| 631 | so->firstCall = false; |
| 632 | so->curPageData = so->nPageData = 0; |
| 633 | scan->xs_hitup = NULL; |
| 634 | if (so->pageDataCxt) |
| 635 | MemoryContextReset(so->pageDataCxt); |
| 636 | |
| 637 | fakeItem.blkno = GIST_ROOT_BLKNO; |
| 638 | memset(&fakeItem.data.parentlsn, 0, sizeof(GistNSN)); |
| 639 | gistScanPage(scan, &fakeItem, NULL, NULL, NULL); |
| 640 | } |
| 641 | |
| 642 | if (scan->numberOfOrderBys > 0) |
| 643 | { |
| 644 | /* Must fetch tuples in strict distance order */ |
| 645 | return getNextNearest(scan); |
| 646 | } |
| 647 | else |
| 648 | { |
| 649 | /* Fetch tuples index-page-at-a-time */ |
| 650 | for (;;) |
| 651 | { |
| 652 | if (so->curPageData < so->nPageData) |
| 653 | { |
| 654 | if (scan->kill_prior_tuple && so->curPageData > 0) |
| 655 | { |
| 656 | |
| 657 | if (so->killedItems == NULL) |
| 658 | { |
| 659 | MemoryContext oldCxt = |
| 660 | MemoryContextSwitchTo(so->giststate->scanCxt); |
| 661 | |
| 662 | so->killedItems = |
| 663 | (OffsetNumber *) palloc(MaxIndexTuplesPerPage |
| 664 | * sizeof(OffsetNumber)); |
| 665 | |
| 666 | MemoryContextSwitchTo(oldCxt); |
| 667 | } |
| 668 | if (so->numKilled < MaxIndexTuplesPerPage) |
| 669 | so->killedItems[so->numKilled++] = |
| 670 | so->pageData[so->curPageData - 1].offnum; |
| 671 | } |
| 672 | /* continuing to return tuples from a leaf page */ |
| 673 | scan->xs_heaptid = so->pageData[so->curPageData].heapPtr; |
| 674 | scan->xs_recheck = so->pageData[so->curPageData].recheck; |
| 675 | |
| 676 | /* in an index-only scan, also return the reconstructed tuple */ |
| 677 | if (scan->xs_want_itup) |
| 678 | scan->xs_hitup = so->pageData[so->curPageData].recontup; |
| 679 | |
| 680 | so->curPageData++; |
| 681 | |
| 682 | return true; |
| 683 | } |
| 684 | |
| 685 | /* |
| 686 | * Check the last returned tuple and add it to killitems if |
| 687 | * necessary |
| 688 | */ |
| 689 | if (scan->kill_prior_tuple |
| 690 | && so->curPageData > 0 |
| 691 | && so->curPageData == so->nPageData) |
| 692 | { |
| 693 | |
| 694 | if (so->killedItems == NULL) |
| 695 | { |
| 696 | MemoryContext oldCxt = |
| 697 | MemoryContextSwitchTo(so->giststate->scanCxt); |
| 698 | |
| 699 | so->killedItems = |
| 700 | (OffsetNumber *) palloc(MaxIndexTuplesPerPage |
| 701 | * sizeof(OffsetNumber)); |
| 702 | |
| 703 | MemoryContextSwitchTo(oldCxt); |
| 704 | } |
| 705 | if (so->numKilled < MaxIndexTuplesPerPage) |
| 706 | so->killedItems[so->numKilled++] = |
| 707 | so->pageData[so->curPageData - 1].offnum; |
| 708 | } |
| 709 | /* find and process the next index page */ |
| 710 | do |
| 711 | { |
| 712 | GISTSearchItem *item; |
| 713 | |
| 714 | if ((so->curBlkno != InvalidBlockNumber) && (so->numKilled > 0)) |
| 715 | gistkillitems(scan); |
| 716 | |
| 717 | item = getNextGISTSearchItem(so); |
| 718 | |
| 719 | if (!item) |
| 720 | return false; |
| 721 | |
| 722 | CHECK_FOR_INTERRUPTS(); |
| 723 | |
| 724 | /* save current item BlockNumber for next gistkillitems() call */ |
| 725 | so->curBlkno = item->blkno; |
| 726 | |
| 727 | /* |
| 728 | * While scanning a leaf page, ItemPointers of matching heap |
| 729 | * tuples are stored in so->pageData. If there are any on |
| 730 | * this page, we fall out of the inner "do" and loop around to |
| 731 | * return them. |
| 732 | */ |
| 733 | gistScanPage(scan, item, item->distances, NULL, NULL); |
| 734 | |
| 735 | pfree(item); |
| 736 | } while (so->nPageData == 0); |
| 737 | } |
| 738 | } |
| 739 | } |
| 740 | |
| 741 | /* |
| 742 | * gistgetbitmap() -- Get a bitmap of all heap tuple locations |
| 743 | */ |
| 744 | int64 |
| 745 | gistgetbitmap(IndexScanDesc scan, TIDBitmap *tbm) |
| 746 | { |
| 747 | GISTScanOpaque so = (GISTScanOpaque) scan->opaque; |
| 748 | int64 ntids = 0; |
| 749 | GISTSearchItem fakeItem; |
| 750 | |
| 751 | if (!so->qual_ok) |
| 752 | return 0; |
| 753 | |
| 754 | pgstat_count_index_scan(scan->indexRelation); |
| 755 | |
| 756 | /* Begin the scan by processing the root page */ |
| 757 | so->curPageData = so->nPageData = 0; |
| 758 | scan->xs_hitup = NULL; |
| 759 | if (so->pageDataCxt) |
| 760 | MemoryContextReset(so->pageDataCxt); |
| 761 | |
| 762 | fakeItem.blkno = GIST_ROOT_BLKNO; |
| 763 | memset(&fakeItem.data.parentlsn, 0, sizeof(GistNSN)); |
| 764 | gistScanPage(scan, &fakeItem, NULL, tbm, &ntids); |
| 765 | |
| 766 | /* |
| 767 | * While scanning a leaf page, ItemPointers of matching heap tuples will |
| 768 | * be stored directly into tbm, so we don't need to deal with them here. |
| 769 | */ |
| 770 | for (;;) |
| 771 | { |
| 772 | GISTSearchItem *item = getNextGISTSearchItem(so); |
| 773 | |
| 774 | if (!item) |
| 775 | break; |
| 776 | |
| 777 | CHECK_FOR_INTERRUPTS(); |
| 778 | |
| 779 | gistScanPage(scan, item, item->distances, tbm, &ntids); |
| 780 | |
| 781 | pfree(item); |
| 782 | } |
| 783 | |
| 784 | return ntids; |
| 785 | } |
| 786 | |
| 787 | /* |
| 788 | * Can we do index-only scans on the given index column? |
| 789 | * |
| 790 | * Opclasses that implement a fetch function support index-only scans. |
| 791 | * Opclasses without compression functions also support index-only scans. |
| 792 | * Included attributes always can be fetched for index-only scans. |
| 793 | */ |
| 794 | bool |
| 795 | gistcanreturn(Relation index, int attno) |
| 796 | { |
| 797 | if (attno > IndexRelationGetNumberOfKeyAttributes(index) || |
| 798 | OidIsValid(index_getprocid(index, attno, GIST_FETCH_PROC)) || |
| 799 | !OidIsValid(index_getprocid(index, attno, GIST_COMPRESS_PROC))) |
| 800 | return true; |
| 801 | else |
| 802 | return false; |
| 803 | } |
| 804 | |