PostgreSQL Source Code  git master
nbtree.h
Go to the documentation of this file.
1 /*-------------------------------------------------------------------------
2  *
3  * nbtree.h
4  * header file for postgres btree access method implementation.
5  *
6  *
7  * Portions Copyright (c) 1996-2023, PostgreSQL Global Development Group
8  * Portions Copyright (c) 1994, Regents of the University of California
9  *
10  * src/include/access/nbtree.h
11  *
12  *-------------------------------------------------------------------------
13  */
14 #ifndef NBTREE_H
15 #define NBTREE_H
16 
17 #include "access/amapi.h"
18 #include "access/itup.h"
19 #include "access/sdir.h"
20 #include "access/tableam.h"
21 #include "access/xlogreader.h"
22 #include "catalog/pg_am_d.h"
23 #include "catalog/pg_index.h"
24 #include "lib/stringinfo.h"
25 #include "storage/bufmgr.h"
26 #include "storage/shm_toc.h"
27 
28 /* There's room for a 16-bit vacuum cycle ID in BTPageOpaqueData */
29 typedef uint16 BTCycleId;
30 
31 /*
32  * BTPageOpaqueData -- At the end of every page, we store a pointer
33  * to both siblings in the tree. This is used to do forward/backward
34  * index scans. The next-page link is also critical for recovery when
35  * a search has navigated to the wrong page due to concurrent page splits
36  * or deletions; see src/backend/access/nbtree/README for more info.
37  *
38  * In addition, we store the page's btree level (counting upwards from
39  * zero at a leaf page) as well as some flag bits indicating the page type
40  * and status. If the page is deleted, a BTDeletedPageData struct is stored
41  * in the page's tuple area, while a standard BTPageOpaqueData struct is
42  * stored in the page special area.
43  *
44  * We also store a "vacuum cycle ID". When a page is split while VACUUM is
45  * processing the index, a nonzero value associated with the VACUUM run is
46  * stored into both halves of the split page. (If VACUUM is not running,
47  * both pages receive zero cycleids.) This allows VACUUM to detect whether
48  * a page was split since it started, with a small probability of false match
49  * if the page was last split some exact multiple of MAX_BT_CYCLE_ID VACUUMs
50  * ago. Also, during a split, the BTP_SPLIT_END flag is cleared in the left
51  * (original) page, and set in the right page, but only if the next page
52  * to its right has a different cycleid.
53  *
54  * NOTE: the BTP_LEAF flag bit is redundant since level==0 could be tested
55  * instead.
56  *
57  * NOTE: the btpo_level field used to be a union type in order to allow
58  * deleted pages to store a 32-bit safexid in the same field. We now store
59  * 64-bit/full safexid values using BTDeletedPageData instead.
60  */
61 
62 typedef struct BTPageOpaqueData
63 {
64  BlockNumber btpo_prev; /* left sibling, or P_NONE if leftmost */
65  BlockNumber btpo_next; /* right sibling, or P_NONE if rightmost */
66  uint32 btpo_level; /* tree level --- zero for leaf pages */
67  uint16 btpo_flags; /* flag bits, see below */
68  BTCycleId btpo_cycleid; /* vacuum cycle ID of latest split */
70 
72 
73 #define BTPageGetOpaque(page) ((BTPageOpaque) PageGetSpecialPointer(page))
74 
75 /* Bits defined in btpo_flags */
76 #define BTP_LEAF (1 << 0) /* leaf page, i.e. not internal page */
77 #define BTP_ROOT (1 << 1) /* root page (has no parent) */
78 #define BTP_DELETED (1 << 2) /* page has been deleted from tree */
79 #define BTP_META (1 << 3) /* meta-page */
80 #define BTP_HALF_DEAD (1 << 4) /* empty, but still in tree */
81 #define BTP_SPLIT_END (1 << 5) /* rightmost page of split group */
82 #define BTP_HAS_GARBAGE (1 << 6) /* page has LP_DEAD tuples (deprecated) */
83 #define BTP_INCOMPLETE_SPLIT (1 << 7) /* right sibling's downlink is missing */
84 #define BTP_HAS_FULLXID (1 << 8) /* contains BTDeletedPageData */
85 
86 /*
87  * The max allowed value of a cycle ID is a bit less than 64K. This is
88  * for convenience of pg_filedump and similar utilities: we want to use
89  * the last 2 bytes of special space as an index type indicator, and
90  * restricting cycle ID lets btree use that space for vacuum cycle IDs
91  * while still allowing index type to be identified.
92  */
93 #define MAX_BT_CYCLE_ID 0xFF7F
94 
95 
96 /*
97  * The Meta page is always the first page in the btree index.
98  * Its primary purpose is to point to the location of the btree root page.
99  * We also point to the "fast" root, which is the current effective root;
100  * see README for discussion.
101  */
102 
103 typedef struct BTMetaPageData
104 {
105  uint32 btm_magic; /* should contain BTREE_MAGIC */
106  uint32 btm_version; /* nbtree version (always <= BTREE_VERSION) */
107  BlockNumber btm_root; /* current root location */
108  uint32 btm_level; /* tree level of the root page */
109  BlockNumber btm_fastroot; /* current "fast" root location */
110  uint32 btm_fastlevel; /* tree level of the "fast" root page */
111  /* remaining fields only valid when btm_version >= BTREE_NOVAC_VERSION */
112 
113  /* number of deleted, non-recyclable pages during last cleanup */
115  /* number of heap tuples during last cleanup (deprecated) */
117 
118  bool btm_allequalimage; /* are all columns "equalimage"? */
120 
121 #define BTPageGetMeta(p) \
122  ((BTMetaPageData *) PageGetContents(p))
123 
124 /*
125  * The current Btree version is 4. That's what you'll get when you create
126  * a new index.
127  *
128  * Btree version 3 was used in PostgreSQL v11. It is mostly the same as
129  * version 4, but heap TIDs were not part of the keyspace. Index tuples
130  * with duplicate keys could be stored in any order. We continue to
131  * support reading and writing Btree versions 2 and 3, so that they don't
132  * need to be immediately re-indexed at pg_upgrade. In order to get the
133  * new heapkeyspace semantics, however, a REINDEX is needed.
134  *
135  * Deduplication is safe to use when the btm_allequalimage field is set to
136  * true. It's safe to read the btm_allequalimage field on version 3, but
137  * only version 4 indexes make use of deduplication. Even version 4
138  * indexes created on PostgreSQL v12 will need a REINDEX to make use of
139  * deduplication, though, since there is no other way to set
140  * btm_allequalimage to true (pg_upgrade hasn't been taught to set the
141  * metapage field).
142  *
143  * Btree version 2 is mostly the same as version 3. There are two new
144  * fields in the metapage that were introduced in version 3. A version 2
145  * metapage will be automatically upgraded to version 3 on the first
146  * insert to it. INCLUDE indexes cannot use version 2.
147  */
148 #define BTREE_METAPAGE 0 /* first page is meta */
149 #define BTREE_MAGIC 0x053162 /* magic number in metapage */
150 #define BTREE_VERSION 4 /* current version number */
151 #define BTREE_MIN_VERSION 2 /* minimum supported version */
152 #define BTREE_NOVAC_VERSION 3 /* version with all meta fields set */
153 
154 /*
155  * Maximum size of a btree index entry, including its tuple header.
156  *
157  * We actually need to be able to fit three items on every page,
158  * so restrict any one item to 1/3 the per-page available space.
159  *
160  * There are rare cases where _bt_truncate() will need to enlarge
161  * a heap index tuple to make space for a tiebreaker heap TID
162  * attribute, which we account for here.
163  */
164 #define BTMaxItemSize(page) \
165  (MAXALIGN_DOWN((PageGetPageSize(page) - \
166  MAXALIGN(SizeOfPageHeaderData + 3*sizeof(ItemIdData)) - \
167  MAXALIGN(sizeof(BTPageOpaqueData))) / 3) - \
168  MAXALIGN(sizeof(ItemPointerData)))
169 #define BTMaxItemSizeNoHeapTid(page) \
170  MAXALIGN_DOWN((PageGetPageSize(page) - \
171  MAXALIGN(SizeOfPageHeaderData + 3*sizeof(ItemIdData)) - \
172  MAXALIGN(sizeof(BTPageOpaqueData))) / 3)
173 
174 /*
175  * MaxTIDsPerBTreePage is an upper bound on the number of heap TIDs tuples
176  * that may be stored on a btree leaf page. It is used to size the
177  * per-page temporary buffers.
178  *
179  * Note: we don't bother considering per-tuple overheads here to keep
180  * things simple (value is based on how many elements a single array of
181  * heap TIDs must have to fill the space between the page header and
182  * special area). The value is slightly higher (i.e. more conservative)
183  * than necessary as a result, which is considered acceptable.
184  */
185 #define MaxTIDsPerBTreePage \
186  (int) ((BLCKSZ - SizeOfPageHeaderData - sizeof(BTPageOpaqueData)) / \
187  sizeof(ItemPointerData))
188 
189 /*
190  * The leaf-page fillfactor defaults to 90% but is user-adjustable.
191  * For pages above the leaf level, we use a fixed 70% fillfactor.
192  * The fillfactor is applied during index build and when splitting
193  * a rightmost page; when splitting non-rightmost pages we try to
194  * divide the data equally. When splitting a page that's entirely
195  * filled with a single value (duplicates), the effective leaf-page
196  * fillfactor is 96%, regardless of whether the page is a rightmost
197  * page.
198  */
199 #define BTREE_MIN_FILLFACTOR 10
200 #define BTREE_DEFAULT_FILLFACTOR 90
201 #define BTREE_NONLEAF_FILLFACTOR 70
202 #define BTREE_SINGLEVAL_FILLFACTOR 96
203 
204 /*
205  * In general, the btree code tries to localize its knowledge about
206  * page layout to a couple of routines. However, we need a special
207  * value to indicate "no page number" in those places where we expect
208  * page numbers. We can use zero for this because we never need to
209  * make a pointer to the metadata page.
210  */
211 
212 #define P_NONE 0
213 
214 /*
215  * Macros to test whether a page is leftmost or rightmost on its tree level,
216  * as well as other state info kept in the opaque data.
217  */
218 #define P_LEFTMOST(opaque) ((opaque)->btpo_prev == P_NONE)
219 #define P_RIGHTMOST(opaque) ((opaque)->btpo_next == P_NONE)
220 #define P_ISLEAF(opaque) (((opaque)->btpo_flags & BTP_LEAF) != 0)
221 #define P_ISROOT(opaque) (((opaque)->btpo_flags & BTP_ROOT) != 0)
222 #define P_ISDELETED(opaque) (((opaque)->btpo_flags & BTP_DELETED) != 0)
223 #define P_ISMETA(opaque) (((opaque)->btpo_flags & BTP_META) != 0)
224 #define P_ISHALFDEAD(opaque) (((opaque)->btpo_flags & BTP_HALF_DEAD) != 0)
225 #define P_IGNORE(opaque) (((opaque)->btpo_flags & (BTP_DELETED|BTP_HALF_DEAD)) != 0)
226 #define P_HAS_GARBAGE(opaque) (((opaque)->btpo_flags & BTP_HAS_GARBAGE) != 0)
227 #define P_INCOMPLETE_SPLIT(opaque) (((opaque)->btpo_flags & BTP_INCOMPLETE_SPLIT) != 0)
228 #define P_HAS_FULLXID(opaque) (((opaque)->btpo_flags & BTP_HAS_FULLXID) != 0)
229 
230 /*
231  * BTDeletedPageData is the page contents of a deleted page
232  */
233 typedef struct BTDeletedPageData
234 {
235  FullTransactionId safexid; /* See BTPageIsRecyclable() */
237 
238 static inline void
240 {
241  BTPageOpaque opaque;
242  PageHeader header;
243  BTDeletedPageData *contents;
244 
245  opaque = BTPageGetOpaque(page);
246  header = ((PageHeader) page);
247 
248  opaque->btpo_flags &= ~BTP_HALF_DEAD;
251  sizeof(BTDeletedPageData);
252  header->pd_upper = header->pd_special;
253 
254  /* Set safexid in deleted page */
255  contents = ((BTDeletedPageData *) PageGetContents(page));
256  contents->safexid = safexid;
257 }
258 
259 static inline FullTransactionId
261 {
262  BTPageOpaque opaque;
263  BTDeletedPageData *contents;
264 
265  /* We only expect to be called with a deleted page */
266  Assert(!PageIsNew(page));
267  opaque = BTPageGetOpaque(page);
268  Assert(P_ISDELETED(opaque));
269 
270  /* pg_upgrade'd deleted page -- must be safe to delete now */
271  if (!P_HAS_FULLXID(opaque))
273 
274  /* Get safexid from deleted page */
275  contents = ((BTDeletedPageData *) PageGetContents(page));
276  return contents->safexid;
277 }
278 
279 /*
280  * Is an existing page recyclable?
281  *
282  * This exists to centralize the policy on which deleted pages are now safe to
283  * re-use. However, _bt_pendingfsm_finalize() duplicates some of the same
284  * logic because it doesn't work directly with pages -- keep the two in sync.
285  *
286  * Note: PageIsNew() pages are always safe to recycle, but we can't deal with
287  * them here (caller is responsible for that case themselves). Caller might
288  * well need special handling for new pages anyway.
289  */
290 static inline bool
292 {
293  BTPageOpaque opaque;
294 
295  Assert(!PageIsNew(page));
296 
297  /* Recycling okay iff page is deleted and safexid is old enough */
298  opaque = BTPageGetOpaque(page);
299  if (P_ISDELETED(opaque))
300  {
301  /*
302  * The page was deleted, but when? If it was just deleted, a scan
303  * might have seen the downlink to it, and will read the page later.
304  * As long as that can happen, we must keep the deleted page around as
305  * a tombstone.
306  *
307  * For that check if the deletion XID could still be visible to
308  * anyone. If not, then no scan that's still in progress could have
309  * seen its downlink, and we can recycle it.
310  *
311  * XXX: If we had the heap relation we could be more aggressive about
312  * recycling deleted pages in non-catalog relations. For now we just
313  * pass NULL. That is at least simple and consistent.
314  */
316  }
317 
318  return false;
319 }
320 
321 /*
322  * BTVacState and BTPendingFSM are private nbtree.c state used during VACUUM.
323  * They are exported for use by page deletion related code in nbtpage.c.
324  */
325 typedef struct BTPendingFSM
326 {
327  BlockNumber target; /* Page deleted by current VACUUM */
328  FullTransactionId safexid; /* Page's BTDeletedPageData.safexid */
330 
331 typedef struct BTVacState
332 {
339 
340  /*
341  * _bt_pendingfsm_finalize() state
342  */
343  int bufsize; /* pendingpages space (in # elements) */
344  int maxbufsize; /* max bufsize that respects work_mem */
345  BTPendingFSM *pendingpages; /* One entry per newly deleted page */
346  int npendingpages; /* current # valid pendingpages */
348 
349 /*
350  * Lehman and Yao's algorithm requires a ``high key'' on every non-rightmost
351  * page. The high key is not a tuple that is used to visit the heap. It is
352  * a pivot tuple (see "Notes on B-Tree tuple format" below for definition).
353  * The high key on a page is required to be greater than or equal to any
354  * other key that appears on the page. If we find ourselves trying to
355  * insert a key that is strictly > high key, we know we need to move right
356  * (this should only happen if the page was split since we examined the
357  * parent page).
358  *
359  * Our insertion algorithm guarantees that we can use the initial least key
360  * on our right sibling as the high key. Once a page is created, its high
361  * key changes only if the page is split.
362  *
363  * On a non-rightmost page, the high key lives in item 1 and data items
364  * start in item 2. Rightmost pages have no high key, so we store data
365  * items beginning in item 1.
366  */
367 
368 #define P_HIKEY ((OffsetNumber) 1)
369 #define P_FIRSTKEY ((OffsetNumber) 2)
370 #define P_FIRSTDATAKEY(opaque) (P_RIGHTMOST(opaque) ? P_HIKEY : P_FIRSTKEY)
371 
372 /*
373  * Notes on B-Tree tuple format, and key and non-key attributes:
374  *
375  * INCLUDE B-Tree indexes have non-key attributes. These are extra
376  * attributes that may be returned by index-only scans, but do not influence
377  * the order of items in the index (formally, non-key attributes are not
378  * considered to be part of the key space). Non-key attributes are only
379  * present in leaf index tuples whose item pointers actually point to heap
380  * tuples (non-pivot tuples). _bt_check_natts() enforces the rules
381  * described here.
382  *
383  * Non-pivot tuple format (plain/non-posting variant):
384  *
385  * t_tid | t_info | key values | INCLUDE columns, if any
386  *
387  * t_tid points to the heap TID, which is a tiebreaker key column as of
388  * BTREE_VERSION 4.
389  *
390  * Non-pivot tuples complement pivot tuples, which only have key columns.
391  * The sole purpose of pivot tuples is to represent how the key space is
392  * separated. In general, any B-Tree index that has more than one level
393  * (i.e. any index that does not just consist of a metapage and a single
394  * leaf root page) must have some number of pivot tuples, since pivot
395  * tuples are used for traversing the tree. Suffix truncation can omit
396  * trailing key columns when a new pivot is formed, which makes minus
397  * infinity their logical value. Since BTREE_VERSION 4 indexes treat heap
398  * TID as a trailing key column that ensures that all index tuples are
399  * physically unique, it is necessary to represent heap TID as a trailing
400  * key column in pivot tuples, though very often this can be truncated
401  * away, just like any other key column. (Actually, the heap TID is
402  * omitted rather than truncated, since its representation is different to
403  * the non-pivot representation.)
404  *
405  * Pivot tuple format:
406  *
407  * t_tid | t_info | key values | [heap TID]
408  *
409  * We store the number of columns present inside pivot tuples by abusing
410  * their t_tid offset field, since pivot tuples never need to store a real
411  * offset (pivot tuples generally store a downlink in t_tid, though). The
412  * offset field only stores the number of columns/attributes when the
413  * INDEX_ALT_TID_MASK bit is set, which doesn't count the trailing heap
414  * TID column sometimes stored in pivot tuples -- that's represented by
415  * the presence of BT_PIVOT_HEAP_TID_ATTR. The INDEX_ALT_TID_MASK bit in
416  * t_info is always set on BTREE_VERSION 4 pivot tuples, since
417  * BTreeTupleIsPivot() must work reliably on heapkeyspace versions.
418  *
419  * In version 2 or version 3 (!heapkeyspace) indexes, INDEX_ALT_TID_MASK
420  * might not be set in pivot tuples. BTreeTupleIsPivot() won't work
421  * reliably as a result. The number of columns stored is implicitly the
422  * same as the number of columns in the index, just like any non-pivot
423  * tuple. (The number of columns stored should not vary, since suffix
424  * truncation of key columns is unsafe within any !heapkeyspace index.)
425  *
426  * The 12 least significant bits from t_tid's offset number are used to
427  * represent the number of key columns within a pivot tuple. This leaves 4
428  * status bits (BT_STATUS_OFFSET_MASK bits), which are shared by all tuples
429  * that have the INDEX_ALT_TID_MASK bit set (set in t_info) to store basic
430  * tuple metadata. BTreeTupleIsPivot() and BTreeTupleIsPosting() use the
431  * BT_STATUS_OFFSET_MASK bits.
432  *
433  * Sometimes non-pivot tuples also use a representation that repurposes
434  * t_tid to store metadata rather than a TID. PostgreSQL v13 introduced a
435  * new non-pivot tuple format to support deduplication: posting list
436  * tuples. Deduplication merges together multiple equal non-pivot tuples
437  * into a logically equivalent, space efficient representation. A posting
438  * list is an array of ItemPointerData elements. Non-pivot tuples are
439  * merged together to form posting list tuples lazily, at the point where
440  * we'd otherwise have to split a leaf page.
441  *
442  * Posting tuple format (alternative non-pivot tuple representation):
443  *
444  * t_tid | t_info | key values | posting list (TID array)
445  *
446  * Posting list tuples are recognized as such by having the
447  * INDEX_ALT_TID_MASK status bit set in t_info and the BT_IS_POSTING status
448  * bit set in t_tid's offset number. These flags redefine the content of
449  * the posting tuple's t_tid to store the location of the posting list
450  * (instead of a block number), as well as the total number of heap TIDs
451  * present in the tuple (instead of a real offset number).
452  *
453  * The 12 least significant bits from t_tid's offset number are used to
454  * represent the number of heap TIDs present in the tuple, leaving 4 status
455  * bits (the BT_STATUS_OFFSET_MASK bits). Like any non-pivot tuple, the
456  * number of columns stored is always implicitly the total number in the
457  * index (in practice there can never be non-key columns stored, since
458  * deduplication is not supported with INCLUDE indexes).
459  */
460 #define INDEX_ALT_TID_MASK INDEX_AM_RESERVED_BIT
461 
462 /* Item pointer offset bit masks */
463 #define BT_OFFSET_MASK 0x0FFF
464 #define BT_STATUS_OFFSET_MASK 0xF000
465 /* BT_STATUS_OFFSET_MASK status bits */
466 #define BT_PIVOT_HEAP_TID_ATTR 0x1000
467 #define BT_IS_POSTING 0x2000
468 
469 /*
470  * Mask allocated for number of keys in index tuple must be able to fit
471  * maximum possible number of index attributes
472  */
474  "BT_OFFSET_MASK can't fit INDEX_MAX_KEYS");
475 
476 /*
477  * Note: BTreeTupleIsPivot() can have false negatives (but not false
478  * positives) when used with !heapkeyspace indexes
479  */
480 static inline bool
482 {
483  if ((itup->t_info & INDEX_ALT_TID_MASK) == 0)
484  return false;
485  /* absence of BT_IS_POSTING in offset number indicates pivot tuple */
487  return false;
488 
489  return true;
490 }
491 
492 static inline bool
494 {
495  if ((itup->t_info & INDEX_ALT_TID_MASK) == 0)
496  return false;
497  /* presence of BT_IS_POSTING in offset number indicates posting tuple */
499  return false;
500 
501  return true;
502 }
503 
504 static inline void
505 BTreeTupleSetPosting(IndexTuple itup, uint16 nhtids, int postingoffset)
506 {
507  Assert(nhtids > 1);
508  Assert((nhtids & BT_STATUS_OFFSET_MASK) == 0);
509  Assert((size_t) postingoffset == MAXALIGN(postingoffset));
510  Assert(postingoffset < INDEX_SIZE_MASK);
511  Assert(!BTreeTupleIsPivot(itup));
512 
513  itup->t_info |= INDEX_ALT_TID_MASK;
514  ItemPointerSetOffsetNumber(&itup->t_tid, (nhtids | BT_IS_POSTING));
515  ItemPointerSetBlockNumber(&itup->t_tid, postingoffset);
516 }
517 
518 static inline uint16
520 {
521  OffsetNumber existing;
522 
523  Assert(BTreeTupleIsPosting(posting));
524 
525  existing = ItemPointerGetOffsetNumberNoCheck(&posting->t_tid);
526  return (existing & BT_OFFSET_MASK);
527 }
528 
529 static inline uint32
531 {
532  Assert(BTreeTupleIsPosting(posting));
533 
534  return ItemPointerGetBlockNumberNoCheck(&posting->t_tid);
535 }
536 
537 static inline ItemPointer
539 {
540  return (ItemPointer) ((char *) posting +
541  BTreeTupleGetPostingOffset(posting));
542 }
543 
544 static inline ItemPointer
546 {
547  return BTreeTupleGetPosting(posting) + n;
548 }
549 
550 /*
551  * Get/set downlink block number in pivot tuple.
552  *
553  * Note: Cannot assert that tuple is a pivot tuple. If we did so then
554  * !heapkeyspace indexes would exhibit false positive assertion failures.
555  */
556 static inline BlockNumber
558 {
559  return ItemPointerGetBlockNumberNoCheck(&pivot->t_tid);
560 }
561 
562 static inline void
564 {
565  ItemPointerSetBlockNumber(&pivot->t_tid, blkno);
566 }
567 
568 /*
569  * Get number of attributes within tuple.
570  *
571  * Note that this does not include an implicit tiebreaker heap TID
572  * attribute, if any. Note also that the number of key attributes must be
573  * explicitly represented in all heapkeyspace pivot tuples.
574  *
575  * Note: This is defined as a macro rather than an inline function to
576  * avoid including rel.h.
577  */
578 #define BTreeTupleGetNAtts(itup, rel) \
579  ( \
580  (BTreeTupleIsPivot(itup)) ? \
581  ( \
582  ItemPointerGetOffsetNumberNoCheck(&(itup)->t_tid) & BT_OFFSET_MASK \
583  ) \
584  : \
585  IndexRelationGetNumberOfAttributes(rel) \
586  )
587 
588 /*
589  * Set number of key attributes in tuple.
590  *
591  * The heap TID tiebreaker attribute bit may also be set here, indicating that
592  * a heap TID value will be stored at the end of the tuple (i.e. using the
593  * special pivot tuple representation).
594  */
595 static inline void
596 BTreeTupleSetNAtts(IndexTuple itup, uint16 nkeyatts, bool heaptid)
597 {
598  Assert(nkeyatts <= INDEX_MAX_KEYS);
599  Assert((nkeyatts & BT_STATUS_OFFSET_MASK) == 0);
600  Assert(!heaptid || nkeyatts > 0);
601  Assert(!BTreeTupleIsPivot(itup) || nkeyatts == 0);
602 
603  itup->t_info |= INDEX_ALT_TID_MASK;
604 
605  if (heaptid)
606  nkeyatts |= BT_PIVOT_HEAP_TID_ATTR;
607 
608  /* BT_IS_POSTING bit is deliberately unset here */
609  ItemPointerSetOffsetNumber(&itup->t_tid, nkeyatts);
610  Assert(BTreeTupleIsPivot(itup));
611 }
612 
613 /*
614  * Get/set leaf page's "top parent" link from its high key. Used during page
615  * deletion.
616  *
617  * Note: Cannot assert that tuple is a pivot tuple. If we did so then
618  * !heapkeyspace indexes would exhibit false positive assertion failures.
619  */
620 static inline BlockNumber
622 {
623  return ItemPointerGetBlockNumberNoCheck(&leafhikey->t_tid);
624 }
625 
626 static inline void
628 {
629  ItemPointerSetBlockNumber(&leafhikey->t_tid, blkno);
630  BTreeTupleSetNAtts(leafhikey, 0, false);
631 }
632 
633 /*
634  * Get tiebreaker heap TID attribute, if any.
635  *
636  * This returns the first/lowest heap TID in the case of a posting list tuple.
637  */
638 static inline ItemPointer
640 {
641  if (BTreeTupleIsPivot(itup))
642  {
643  /* Pivot tuple heap TID representation? */
646  return (ItemPointer) ((char *) itup + IndexTupleSize(itup) -
647  sizeof(ItemPointerData));
648 
649  /* Heap TID attribute was truncated */
650  return NULL;
651  }
652  else if (BTreeTupleIsPosting(itup))
653  return BTreeTupleGetPosting(itup);
654 
655  return &itup->t_tid;
656 }
657 
658 /*
659  * Get maximum heap TID attribute, which could be the only TID in the case of
660  * a non-pivot tuple that does not have a posting list tuple.
661  *
662  * Works with non-pivot tuples only.
663  */
664 static inline ItemPointer
666 {
667  Assert(!BTreeTupleIsPivot(itup));
668 
669  if (BTreeTupleIsPosting(itup))
670  {
671  uint16 nposting = BTreeTupleGetNPosting(itup);
672 
673  return BTreeTupleGetPostingN(itup, nposting - 1);
674  }
675 
676  return &itup->t_tid;
677 }
678 
679 /*
680  * Operator strategy numbers for B-tree have been moved to access/stratnum.h,
681  * because many places need to use them in ScanKeyInit() calls.
682  *
683  * The strategy numbers are chosen so that we can commute them by
684  * subtraction, thus:
685  */
686 #define BTCommuteStrategyNumber(strat) (BTMaxStrategyNumber + 1 - (strat))
687 
688 /*
689  * When a new operator class is declared, we require that the user
690  * supply us with an amproc procedure (BTORDER_PROC) for determining
691  * whether, for two keys a and b, a < b, a = b, or a > b. This routine
692  * must return < 0, 0, > 0, respectively, in these three cases.
693  *
694  * To facilitate accelerated sorting, an operator class may choose to
695  * offer a second procedure (BTSORTSUPPORT_PROC). For full details, see
696  * src/include/utils/sortsupport.h.
697  *
698  * To support window frames defined by "RANGE offset PRECEDING/FOLLOWING",
699  * an operator class may choose to offer a third amproc procedure
700  * (BTINRANGE_PROC), independently of whether it offers sortsupport.
701  * For full details, see doc/src/sgml/btree.sgml.
702  *
703  * To facilitate B-Tree deduplication, an operator class may choose to
704  * offer a forth amproc procedure (BTEQUALIMAGE_PROC). For full details,
705  * see doc/src/sgml/btree.sgml.
706  */
707 
708 #define BTORDER_PROC 1
709 #define BTSORTSUPPORT_PROC 2
710 #define BTINRANGE_PROC 3
711 #define BTEQUALIMAGE_PROC 4
712 #define BTOPTIONS_PROC 5
713 #define BTNProcs 5
714 
715 /*
716  * We need to be able to tell the difference between read and write
717  * requests for pages, in order to do locking correctly.
718  */
719 
720 #define BT_READ BUFFER_LOCK_SHARE
721 #define BT_WRITE BUFFER_LOCK_EXCLUSIVE
722 
723 /*
724  * BTStackData -- As we descend a tree, we push the location of pivot
725  * tuples whose downlink we are about to follow onto a private stack. If
726  * we split a leaf, we use this stack to walk back up the tree and insert
727  * data into its parent page at the correct location. We also have to
728  * recursively insert into the grandparent page if and when the parent page
729  * splits. Our private stack can become stale due to concurrent page
730  * splits and page deletions, but it should never give us an irredeemably
731  * bad picture.
732  */
733 typedef struct BTStackData
734 {
739 
741 
742 /*
743  * BTScanInsertData is the btree-private state needed to find an initial
744  * position for an indexscan, or to insert new tuples -- an "insertion
745  * scankey" (not to be confused with a search scankey). It's used to descend
746  * a B-Tree using _bt_search.
747  *
748  * heapkeyspace indicates if we expect all keys in the index to be physically
749  * unique because heap TID is used as a tiebreaker attribute, and if index may
750  * have truncated key attributes in pivot tuples. This is actually a property
751  * of the index relation itself (not an indexscan). heapkeyspace indexes are
752  * indexes whose version is >= version 4. It's convenient to keep this close
753  * by, rather than accessing the metapage repeatedly.
754  *
755  * allequalimage is set to indicate that deduplication is safe for the index.
756  * This is also a property of the index relation rather than an indexscan.
757  *
758  * anynullkeys indicates if any of the keys had NULL value when scankey was
759  * built from index tuple (note that already-truncated tuple key attributes
760  * set NULL as a placeholder key value, which also affects value of
761  * anynullkeys). This is a convenience for unique index non-pivot tuple
762  * insertion, which usually temporarily unsets scantid, but shouldn't iff
763  * anynullkeys is true. Value generally matches non-pivot tuple's HasNulls
764  * bit, but may not when inserting into an INCLUDE index (tuple header value
765  * is affected by the NULL-ness of both key and non-key attributes).
766  *
767  * When nextkey is false (the usual case), _bt_search and _bt_binsrch will
768  * locate the first item >= scankey. When nextkey is true, they will locate
769  * the first item > scan key.
770  *
771  * pivotsearch is set to true by callers that want to re-find a leaf page
772  * using a scankey built from a leaf page's high key. Most callers set this
773  * to false.
774  *
775  * scantid is the heap TID that is used as a final tiebreaker attribute. It
776  * is set to NULL when index scan doesn't need to find a position for a
777  * specific physical tuple. Must be set when inserting new tuples into
778  * heapkeyspace indexes, since every tuple in the tree unambiguously belongs
779  * in one exact position (it's never set with !heapkeyspace indexes, though).
780  * Despite the representational difference, nbtree search code considers
781  * scantid to be just another insertion scankey attribute.
782  *
783  * scankeys is an array of scan key entries for attributes that are compared
784  * before scantid (user-visible attributes). keysz is the size of the array.
785  * During insertion, there must be a scan key for every attribute, but when
786  * starting a regular index scan some can be omitted. The array is used as a
787  * flexible array member, though it's sized in a way that makes it possible to
788  * use stack allocations. See nbtree/README for full details.
789  */
790 typedef struct BTScanInsertData
791 {
795  bool nextkey;
797  ItemPointer scantid; /* tiebreaker for scankeys */
798  int keysz; /* Size of scankeys array */
799  ScanKeyData scankeys[INDEX_MAX_KEYS]; /* Must appear last */
801 
803 
804 /*
805  * BTInsertStateData is a working area used during insertion.
806  *
807  * This is filled in after descending the tree to the first leaf page the new
808  * tuple might belong on. Tracks the current position while performing
809  * uniqueness check, before we have determined which exact page to insert
810  * to.
811  *
812  * (This should be private to nbtinsert.c, but it's also used by
813  * _bt_binsrch_insert)
814  */
815 typedef struct BTInsertStateData
816 {
817  IndexTuple itup; /* Item we're inserting */
818  Size itemsz; /* Size of itup -- should be MAXALIGN()'d */
819  BTScanInsert itup_key; /* Insertion scankey */
820 
821  /* Buffer containing leaf page we're likely to insert itup on */
823 
824  /*
825  * Cache of bounds within the current buffer. Only used for insertions
826  * where _bt_check_unique is called. See _bt_binsrch_insert and
827  * _bt_findinsertloc for details.
828  */
832 
833  /*
834  * if _bt_binsrch_insert found the location inside existing posting list,
835  * save the position inside the list. -1 sentinel value indicates overlap
836  * with an existing posting list tuple that has its LP_DEAD bit set.
837  */
840 
842 
843 /*
844  * State used to representing an individual pending tuple during
845  * deduplication.
846  */
847 typedef struct BTDedupInterval
848 {
852 
853 /*
854  * BTDedupStateData is a working area used during deduplication.
855  *
856  * The status info fields track the state of a whole-page deduplication pass.
857  * State about the current pending posting list is also tracked.
858  *
859  * A pending posting list is comprised of a contiguous group of equal items
860  * from the page, starting from page offset number 'baseoff'. This is the
861  * offset number of the "base" tuple for new posting list. 'nitems' is the
862  * current total number of existing items from the page that will be merged to
863  * make a new posting list tuple, including the base tuple item. (Existing
864  * items may themselves be posting list tuples, or regular non-pivot tuples.)
865  *
866  * The total size of the existing tuples to be freed when pending posting list
867  * is processed gets tracked by 'phystupsize'. This information allows
868  * deduplication to calculate the space saving for each new posting list
869  * tuple, and for the entire pass over the page as a whole.
870  */
871 typedef struct BTDedupStateData
872 {
873  /* Deduplication status info for entire pass over page */
874  bool deduplicate; /* Still deduplicating page? */
875  int nmaxitems; /* Number of max-sized tuples so far */
876  Size maxpostingsize; /* Limit on size of final tuple */
877 
878  /* Metadata about base tuple of current pending posting list */
879  IndexTuple base; /* Use to form new posting list */
880  OffsetNumber baseoff; /* page offset of base */
881  Size basetupsize; /* base size without original posting list */
882 
883  /* Other metadata about pending posting list */
884  ItemPointer htids; /* Heap TIDs in pending posting list */
885  int nhtids; /* Number of heap TIDs in htids array */
886  int nitems; /* Number of existing tuples/line pointers */
887  Size phystupsize; /* Includes line pointer overhead */
888 
889  /*
890  * Array of tuples to go on new version of the page. Contains one entry
891  * for each group of consecutive items. Note that existing tuples that
892  * will not become posting list tuples do not appear in the array (they
893  * are implicitly unchanged by deduplication pass).
894  */
895  int nintervals; /* current number of intervals in array */
898 
900 
901 /*
902  * BTVacuumPostingData is state that represents how to VACUUM (or delete) a
903  * posting list tuple when some (though not all) of its TIDs are to be
904  * deleted.
905  *
906  * Convention is that itup field is the original posting list tuple on input,
907  * and palloc()'d final tuple used to overwrite existing tuple on output.
908  */
909 typedef struct BTVacuumPostingData
910 {
911  /* Tuple that will be/was updated */
914 
915  /* State needed to describe final itup in WAL */
919 
921 
922 /*
923  * BTScanOpaqueData is the btree-private state needed for an indexscan.
924  * This consists of preprocessed scan keys (see _bt_preprocess_keys() for
925  * details of the preprocessing), information about the current location
926  * of the scan, and information about the marked location, if any. (We use
927  * BTScanPosData to represent the data needed for each of current and marked
928  * locations.) In addition we can remember some known-killed index entries
929  * that must be marked before we can move off the current page.
930  *
931  * Index scans work a page at a time: we pin and read-lock the page, identify
932  * all the matching items on the page and save them in BTScanPosData, then
933  * release the read-lock while returning the items to the caller for
934  * processing. This approach minimizes lock/unlock traffic. Note that we
935  * keep the pin on the index page until the caller is done with all the items
936  * (this is needed for VACUUM synchronization, see nbtree/README). When we
937  * are ready to step to the next page, if the caller has told us any of the
938  * items were killed, we re-lock the page to mark them killed, then unlock.
939  * Finally we drop the pin and step to the next page in the appropriate
940  * direction.
941  *
942  * If we are doing an index-only scan, we save the entire IndexTuple for each
943  * matched item, otherwise only its heap TID and offset. The IndexTuples go
944  * into a separate workspace array; each BTScanPosItem stores its tuple's
945  * offset within that array. Posting list tuples store a "base" tuple once,
946  * allowing the same key to be returned for each TID in the posting list
947  * tuple.
948  */
949 
950 typedef struct BTScanPosItem /* what we remember about each match */
951 {
952  ItemPointerData heapTid; /* TID of referenced heap item */
953  OffsetNumber indexOffset; /* index item's location within page */
954  LocationIndex tupleOffset; /* IndexTuple's offset in workspace, if any */
956 
957 typedef struct BTScanPosData
958 {
959  Buffer buf; /* if valid, the buffer is pinned */
960 
961  XLogRecPtr lsn; /* pos in the WAL stream when page was read */
962  BlockNumber currPage; /* page referenced by items array */
963  BlockNumber nextPage; /* page's right link when we scanned it */
964 
965  /*
966  * moreLeft and moreRight track whether we think there may be matching
967  * index entries to the left and right of the current page, respectively.
968  * We can clear the appropriate one of these flags when _bt_checkkeys()
969  * returns continuescan = false.
970  */
971  bool moreLeft;
972  bool moreRight;
973 
974  /*
975  * If we are doing an index-only scan, nextTupleOffset is the first free
976  * location in the associated tuple storage workspace.
977  */
979 
980  /*
981  * The items array is always ordered in index order (ie, increasing
982  * indexoffset). When scanning backwards it is convenient to fill the
983  * array back-to-front, so we start at the last slot and fill downwards.
984  * Hence we need both a first-valid-entry and a last-valid-entry counter.
985  * itemIndex is a cursor showing which entry was last returned to caller.
986  */
987  int firstItem; /* first valid index in items[] */
988  int lastItem; /* last valid index in items[] */
989  int itemIndex; /* current index in items[] */
990 
993 
995 
996 #define BTScanPosIsPinned(scanpos) \
997 ( \
998  AssertMacro(BlockNumberIsValid((scanpos).currPage) || \
999  !BufferIsValid((scanpos).buf)), \
1000  BufferIsValid((scanpos).buf) \
1001 )
1002 #define BTScanPosUnpin(scanpos) \
1003  do { \
1004  ReleaseBuffer((scanpos).buf); \
1005  (scanpos).buf = InvalidBuffer; \
1006  } while (0)
1007 #define BTScanPosUnpinIfPinned(scanpos) \
1008  do { \
1009  if (BTScanPosIsPinned(scanpos)) \
1010  BTScanPosUnpin(scanpos); \
1011  } while (0)
1012 
1013 #define BTScanPosIsValid(scanpos) \
1014 ( \
1015  AssertMacro(BlockNumberIsValid((scanpos).currPage) || \
1016  !BufferIsValid((scanpos).buf)), \
1017  BlockNumberIsValid((scanpos).currPage) \
1018 )
1019 #define BTScanPosInvalidate(scanpos) \
1020  do { \
1021  (scanpos).currPage = InvalidBlockNumber; \
1022  (scanpos).nextPage = InvalidBlockNumber; \
1023  (scanpos).buf = InvalidBuffer; \
1024  (scanpos).lsn = InvalidXLogRecPtr; \
1025  (scanpos).nextTupleOffset = 0; \
1026  } while (0)
1027 
1028 /* We need one of these for each equality-type SK_SEARCHARRAY scan key */
1029 typedef struct BTArrayKeyInfo
1030 {
1031  int scan_key; /* index of associated key in arrayKeyData */
1032  int cur_elem; /* index of current element in elem_values */
1033  int mark_elem; /* index of marked element in elem_values */
1034  int num_elems; /* number of elems in current array value */
1035  Datum *elem_values; /* array of num_elems Datums */
1037 
1038 typedef struct BTScanOpaqueData
1039 {
1040  /* these fields are set by _bt_preprocess_keys(): */
1041  bool qual_ok; /* false if qual can never be satisfied */
1042  int numberOfKeys; /* number of preprocessed scan keys */
1043  ScanKey keyData; /* array of preprocessed scan keys */
1044 
1045  /* workspace for SK_SEARCHARRAY support */
1046  ScanKey arrayKeyData; /* modified copy of scan->keyData */
1047  int numArrayKeys; /* number of equality-type array keys (-1 if
1048  * there are any unsatisfiable array keys) */
1049  int arrayKeyCount; /* count indicating number of array scan keys
1050  * processed */
1051  BTArrayKeyInfo *arrayKeys; /* info about each equality-type array key */
1052  MemoryContext arrayContext; /* scan-lifespan context for array data */
1053 
1054  /* info about killed items if any (killedItems is NULL if never used) */
1055  int *killedItems; /* currPos.items indexes of killed items */
1056  int numKilled; /* number of currently stored items */
1057 
1058  /*
1059  * If we are doing an index-only scan, these are the tuple storage
1060  * workspaces for the currPos and markPos respectively. Each is of size
1061  * BLCKSZ, so it can hold as much as a full page's worth of tuples.
1062  */
1063  char *currTuples; /* tuple storage for currPos */
1064  char *markTuples; /* tuple storage for markPos */
1065 
1066  /*
1067  * If the marked position is on the same page as current position, we
1068  * don't use markPos, but just keep the marked itemIndex in markItemIndex
1069  * (all the rest of currPos is valid for the mark position). Hence, to
1070  * determine if there is a mark, first look at markItemIndex, then at
1071  * markPos.
1072  */
1073  int markItemIndex; /* itemIndex, or -1 if not valid */
1074 
1075  /* keep these last in struct for efficiency */
1076  BTScanPosData currPos; /* current position data */
1077  BTScanPosData markPos; /* marked position, if any */
1079 
1081 
1082 /*
1083  * We use some private sk_flags bits in preprocessed scan keys. We're allowed
1084  * to use bits 16-31 (see skey.h). The uppermost bits are copied from the
1085  * index's indoption[] array entry for the index attribute.
1086  */
1087 #define SK_BT_REQFWD 0x00010000 /* required to continue forward scan */
1088 #define SK_BT_REQBKWD 0x00020000 /* required to continue backward scan */
1089 #define SK_BT_INDOPTION_SHIFT 24 /* must clear the above bits */
1090 #define SK_BT_DESC (INDOPTION_DESC << SK_BT_INDOPTION_SHIFT)
1091 #define SK_BT_NULLS_FIRST (INDOPTION_NULLS_FIRST << SK_BT_INDOPTION_SHIFT)
1092 
1093 typedef struct BTOptions
1094 {
1095  int32 varlena_header_; /* varlena header (do not touch directly!) */
1096  int fillfactor; /* page fill factor in percent (0..100) */
1098  bool deduplicate_items; /* Try to deduplicate items? */
1100 
1101 #define BTGetFillFactor(relation) \
1102  (AssertMacro(relation->rd_rel->relkind == RELKIND_INDEX && \
1103  relation->rd_rel->relam == BTREE_AM_OID), \
1104  (relation)->rd_options ? \
1105  ((BTOptions *) (relation)->rd_options)->fillfactor : \
1106  BTREE_DEFAULT_FILLFACTOR)
1107 #define BTGetTargetPageFreeSpace(relation) \
1108  (BLCKSZ * (100 - BTGetFillFactor(relation)) / 100)
1109 #define BTGetDeduplicateItems(relation) \
1110  (AssertMacro(relation->rd_rel->relkind == RELKIND_INDEX && \
1111  relation->rd_rel->relam == BTREE_AM_OID), \
1112  ((relation)->rd_options ? \
1113  ((BTOptions *) (relation)->rd_options)->deduplicate_items : true))
1114 
1115 /*
1116  * Constant definition for progress reporting. Phase numbers must match
1117  * btbuildphasename.
1118  */
1119 /* PROGRESS_CREATEIDX_SUBPHASE_INITIALIZE is 1 (see progress.h) */
1120 #define PROGRESS_BTREE_PHASE_INDEXBUILD_TABLESCAN 2
1121 #define PROGRESS_BTREE_PHASE_PERFORMSORT_1 3
1122 #define PROGRESS_BTREE_PHASE_PERFORMSORT_2 4
1123 #define PROGRESS_BTREE_PHASE_LEAF_LOAD 5
1124 
1125 /*
1126  * external entry points for btree, in nbtree.c
1127  */
1128 extern void btbuildempty(Relation index);
1129 extern bool btinsert(Relation rel, Datum *values, bool *isnull,
1130  ItemPointer ht_ctid, Relation heapRel,
1131  IndexUniqueCheck checkUnique,
1132  bool indexUnchanged,
1133  struct IndexInfo *indexInfo);
1134 extern IndexScanDesc btbeginscan(Relation rel, int nkeys, int norderbys);
1135 extern Size btestimateparallelscan(void);
1136 extern void btinitparallelscan(void *target);
1137 extern bool btgettuple(IndexScanDesc scan, ScanDirection dir);
1138 extern int64 btgetbitmap(IndexScanDesc scan, TIDBitmap *tbm);
1139 extern void btrescan(IndexScanDesc scan, ScanKey scankey, int nscankeys,
1140  ScanKey orderbys, int norderbys);
1141 extern void btparallelrescan(IndexScanDesc scan);
1142 extern void btendscan(IndexScanDesc scan);
1143 extern void btmarkpos(IndexScanDesc scan);
1144 extern void btrestrpos(IndexScanDesc scan);
1146  IndexBulkDeleteResult *stats,
1148  void *callback_state);
1150  IndexBulkDeleteResult *stats);
1151 extern bool btcanreturn(Relation index, int attno);
1152 
1153 /*
1154  * prototypes for internal functions in nbtree.c
1155  */
1156 extern bool _bt_parallel_seize(IndexScanDesc scan, BlockNumber *pageno);
1157 extern void _bt_parallel_release(IndexScanDesc scan, BlockNumber scan_page);
1158 extern void _bt_parallel_done(IndexScanDesc scan);
1160 
1161 /*
1162  * prototypes for functions in nbtdedup.c
1163  */
1164 extern void _bt_dedup_pass(Relation rel, Buffer buf, Relation heapRel,
1165  IndexTuple newitem, Size newitemsz,
1166  bool bottomupdedup);
1167 extern bool _bt_bottomupdel_pass(Relation rel, Buffer buf, Relation heapRel,
1168  Size newitemsz);
1170  OffsetNumber baseoff);
1174  int nhtids);
1175 extern void _bt_update_posting(BTVacuumPosting vacposting);
1176 extern IndexTuple _bt_swap_posting(IndexTuple newitem, IndexTuple oposting,
1177  int postingoff);
1178 
1179 /*
1180  * prototypes for functions in nbtinsert.c
1181  */
1182 extern bool _bt_doinsert(Relation rel, IndexTuple itup,
1183  IndexUniqueCheck checkUnique, bool indexUnchanged,
1184  Relation heapRel);
1185 extern void _bt_finish_split(Relation rel, Buffer lbuf, BTStack stack);
1186 extern Buffer _bt_getstackbuf(Relation rel, BTStack stack, BlockNumber child);
1187 
1188 /*
1189  * prototypes for functions in nbtsplitloc.c
1190  */
1191 extern OffsetNumber _bt_findsplitloc(Relation rel, Page origpage,
1192  OffsetNumber newitemoff, Size newitemsz, IndexTuple newitem,
1193  bool *newitemonleft);
1194 
1195 /*
1196  * prototypes for functions in nbtpage.c
1197  */
1198 extern void _bt_initmetapage(Page page, BlockNumber rootbknum, uint32 level,
1199  bool allequalimage);
1200 extern bool _bt_vacuum_needs_cleanup(Relation rel);
1201 extern void _bt_set_cleanup_info(Relation rel, BlockNumber num_delpages);
1202 extern void _bt_upgrademetapage(Page page);
1203 extern Buffer _bt_getroot(Relation rel, int access);
1204 extern Buffer _bt_gettrueroot(Relation rel);
1205 extern int _bt_getrootheight(Relation rel);
1206 extern void _bt_metaversion(Relation rel, bool *heapkeyspace,
1207  bool *allequalimage);
1208 extern void _bt_checkpage(Relation rel, Buffer buf);
1209 extern Buffer _bt_getbuf(Relation rel, BlockNumber blkno, int access);
1210 extern Buffer _bt_relandgetbuf(Relation rel, Buffer obuf,
1211  BlockNumber blkno, int access);
1212 extern void _bt_relbuf(Relation rel, Buffer buf);
1213 extern void _bt_lockbuf(Relation rel, Buffer buf, int access);
1214 extern void _bt_unlockbuf(Relation rel, Buffer buf);
1215 extern bool _bt_conditionallockbuf(Relation rel, Buffer buf);
1216 extern void _bt_upgradelockbufcleanup(Relation rel, Buffer buf);
1217 extern void _bt_pageinit(Page page, Size size);
1218 extern void _bt_delitems_vacuum(Relation rel, Buffer buf,
1219  OffsetNumber *deletable, int ndeletable,
1220  BTVacuumPosting *updatable, int nupdatable);
1221 extern void _bt_delitems_delete_check(Relation rel, Buffer buf,
1222  Relation heapRel,
1223  TM_IndexDeleteOp *delstate);
1224 extern void _bt_pagedel(Relation rel, Buffer leafbuf, BTVacState *vstate);
1225 extern void _bt_pendingfsm_init(Relation rel, BTVacState *vstate,
1226  bool cleanuponly);
1227 extern void _bt_pendingfsm_finalize(Relation rel, BTVacState *vstate);
1228 
1229 /*
1230  * prototypes for functions in nbtsearch.c
1231  */
1232 extern BTStack _bt_search(Relation rel, BTScanInsert key, Buffer *bufP,
1233  int access, Snapshot snapshot);
1235  bool forupdate, BTStack stack, int access, Snapshot snapshot);
1236 extern OffsetNumber _bt_binsrch_insert(Relation rel, BTInsertState insertstate);
1237 extern int32 _bt_compare(Relation rel, BTScanInsert key, Page page, OffsetNumber offnum);
1238 extern bool _bt_first(IndexScanDesc scan, ScanDirection dir);
1239 extern bool _bt_next(IndexScanDesc scan, ScanDirection dir);
1240 extern Buffer _bt_get_endpoint(Relation rel, uint32 level, bool rightmost,
1241  Snapshot snapshot);
1242 
1243 /*
1244  * prototypes for functions in nbtutils.c
1245  */
1246 extern BTScanInsert _bt_mkscankey(Relation rel, IndexTuple itup);
1247 extern void _bt_freestack(BTStack stack);
1248 extern void _bt_preprocess_array_keys(IndexScanDesc scan);
1249 extern void _bt_start_array_keys(IndexScanDesc scan, ScanDirection dir);
1250 extern bool _bt_advance_array_keys(IndexScanDesc scan, ScanDirection dir);
1251 extern void _bt_mark_array_keys(IndexScanDesc scan);
1252 extern void _bt_restore_array_keys(IndexScanDesc scan);
1253 extern void _bt_preprocess_keys(IndexScanDesc scan);
1254 extern bool _bt_checkkeys(IndexScanDesc scan, IndexTuple tuple,
1255  int tupnatts, ScanDirection dir, bool *continuescan);
1256 extern void _bt_killitems(IndexScanDesc scan);
1258 extern BTCycleId _bt_start_vacuum(Relation rel);
1259 extern void _bt_end_vacuum(Relation rel);
1260 extern void _bt_end_vacuum_callback(int code, Datum arg);
1261 extern Size BTreeShmemSize(void);
1262 extern void BTreeShmemInit(void);
1263 extern bytea *btoptions(Datum reloptions, bool validate);
1264 extern bool btproperty(Oid index_oid, int attno,
1265  IndexAMProperty prop, const char *propname,
1266  bool *res, bool *isnull);
1267 extern char *btbuildphasename(int64 phasenum);
1268 extern IndexTuple _bt_truncate(Relation rel, IndexTuple lastleft,
1269  IndexTuple firstright, BTScanInsert itup_key);
1270 extern int _bt_keep_natts_fast(Relation rel, IndexTuple lastleft,
1271  IndexTuple firstright);
1272 extern bool _bt_check_natts(Relation rel, bool heapkeyspace, Page page,
1273  OffsetNumber offnum);
1274 extern void _bt_check_third_page(Relation rel, Relation heap,
1275  bool needheaptidspace, Page page, IndexTuple newtup);
1276 extern bool _bt_allequalimage(Relation rel, bool debugmessage);
1277 
1278 /*
1279  * prototypes for functions in nbtvalidate.c
1280  */
1281 extern bool btvalidate(Oid opclassoid);
1282 extern void btadjustmembers(Oid opfamilyoid,
1283  Oid opclassoid,
1284  List *operators,
1285  List *functions);
1286 
1287 /*
1288  * prototypes for functions in nbtsort.c
1289  */
1291  struct IndexInfo *indexInfo);
1292 extern void _bt_parallel_build_main(dsm_segment *seg, shm_toc *toc);
1293 
1294 #endif /* NBTREE_H */
IndexAMProperty
Definition: amapi.h:35
uint32 BlockNumber
Definition: block.h:31
static Datum values[MAXATTR]
Definition: bootstrap.c:156
int Buffer
Definition: buf.h:23
PageHeaderData * PageHeader
Definition: bufpage.h:170
static char * PageGetContents(Page page)
Definition: bufpage.h:254
Pointer Page
Definition: bufpage.h:78
#define SizeOfPageHeaderData
Definition: bufpage.h:213
static bool PageIsNew(Page page)
Definition: bufpage.h:230
uint16 LocationIndex
Definition: bufpage.h:87
unsigned short uint16
Definition: c.h:489
unsigned int uint32
Definition: c.h:490
#define MAXALIGN(LEN)
Definition: c.h:795
signed int int32
Definition: c.h:478
double float8
Definition: c.h:614
#define FLEXIBLE_ARRAY_MEMBER
Definition: c.h:382
size_t Size
Definition: c.h:589
bool(* IndexBulkDeleteCallback)(ItemPointer itemptr, void *state)
Definition: genam.h:86
IndexUniqueCheck
Definition: genam.h:115
static void ItemPointerSetOffsetNumber(ItemPointerData *pointer, OffsetNumber offsetNumber)
Definition: itemptr.h:158
static void ItemPointerSetBlockNumber(ItemPointerData *pointer, BlockNumber blockNumber)
Definition: itemptr.h:147
static OffsetNumber ItemPointerGetOffsetNumberNoCheck(const ItemPointerData *pointer)
Definition: itemptr.h:114
static BlockNumber ItemPointerGetBlockNumberNoCheck(const ItemPointerData *pointer)
Definition: itemptr.h:93
struct ItemPointerData ItemPointerData
#define IndexTupleSize(itup)
Definition: itup.h:70
#define MaxIndexTuplesPerPage
Definition: itup.h:165
#define INDEX_SIZE_MASK
Definition: itup.h:65
Assert(fmt[strlen(fmt) - 1] !='\n')
#define P_HAS_FULLXID(opaque)
Definition: nbtree.h:228
bool btinsert(Relation rel, Datum *values, bool *isnull, ItemPointer ht_ctid, Relation heapRel, IndexUniqueCheck checkUnique, bool indexUnchanged, struct IndexInfo *indexInfo)
Definition: nbtree.c:188
Buffer _bt_relandgetbuf(Relation rel, Buffer obuf, BlockNumber blkno, int access)
Definition: nbtpage.c:1015
void _bt_check_third_page(Relation rel, Relation heap, bool needheaptidspace, Page page, IndexTuple newtup)
Definition: nbtutils.c:2631
bool btcanreturn(Relation index, int attno)
Definition: nbtree.c:1419
void _bt_parallel_release(IndexScanDesc scan, BlockNumber scan_page)
Definition: nbtree.c:694
BTPageOpaqueData * BTPageOpaque
Definition: nbtree.h:71
#define BT_PIVOT_HEAP_TID_ATTR
Definition: nbtree.h:466
static uint16 BTreeTupleGetNPosting(IndexTuple posting)
Definition: nbtree.h:519
void _bt_upgrademetapage(Page page)
Definition: nbtpage.c:109
void _bt_relbuf(Relation rel, Buffer buf)
Definition: nbtpage.c:1035
IndexTuple _bt_swap_posting(IndexTuple newitem, IndexTuple oposting, int postingoff)
Definition: nbtdedup.c:1022
static bool BTreeTupleIsPivot(IndexTuple itup)
Definition: nbtree.h:481
Buffer _bt_gettrueroot(Relation rel)
Definition: nbtpage.c:577
int _bt_getrootheight(Relation rel)
Definition: nbtpage.c:672
BTStackData * BTStack
Definition: nbtree.h:740
void _bt_end_vacuum(Relation rel)
Definition: nbtutils.c:2033
bool btvalidate(Oid opclassoid)
Definition: nbtvalidate.c:41
void _bt_pageinit(Page page, Size size)
Definition: nbtpage.c:1141
bool _bt_first(IndexScanDesc scan, ScanDirection dir)
Definition: nbtsearch.c:860
OffsetNumber _bt_findsplitloc(Relation rel, Page origpage, OffsetNumber newitemoff, Size newitemsz, IndexTuple newitem, bool *newitemonleft)
Definition: nbtsplitloc.c:129
static FullTransactionId BTPageGetDeleteXid(Page page)
Definition: nbtree.h:260
#define BTP_HAS_FULLXID
Definition: nbtree.h:84
void _bt_update_posting(BTVacuumPosting vacposting)
Definition: nbtdedup.c:924
bool _bt_bottomupdel_pass(Relation rel, Buffer buf, Relation heapRel, Size newitemsz)
Definition: nbtdedup.c:307
char * btbuildphasename(int64 phasenum)
Definition: nbtutils.c:2157
#define BTP_HALF_DEAD
Definition: nbtree.h:80
static bool BTPageIsRecyclable(Page page)
Definition: nbtree.h:291
bool _bt_dedup_save_htid(BTDedupState state, IndexTuple itup)
Definition: nbtdedup.c:484
void _bt_end_vacuum_callback(int code, Datum arg)
Definition: nbtutils.c:2061
void _bt_pagedel(Relation rel, Buffer leafbuf, BTVacState *vstate)
Definition: nbtpage.c:1816
struct BTPageOpaqueData BTPageOpaqueData
IndexScanDesc btbeginscan(Relation rel, int nkeys, int norderbys)
Definition: nbtree.c:344
void _bt_delitems_vacuum(Relation rel, Buffer buf, OffsetNumber *deletable, int ndeletable, BTVacuumPosting *updatable, int nupdatable)
Definition: nbtpage.c:1166
static void BTreeTupleSetTopParent(IndexTuple leafhikey, BlockNumber blkno)
Definition: nbtree.h:627
Buffer _bt_getroot(Relation rel, int access)
Definition: nbtpage.c:343
void _bt_freestack(BTStack stack)
Definition: nbtutils.c:182
static void BTreeTupleSetPosting(IndexTuple itup, uint16 nhtids, int postingoffset)
Definition: nbtree.h:505
void BTreeShmemInit(void)
Definition: nbtutils.c:2083
void _bt_parallel_done(IndexScanDesc scan)
Definition: nbtree.c:717
IndexBulkDeleteResult * btbulkdelete(IndexVacuumInfo *info, IndexBulkDeleteResult *stats, IndexBulkDeleteCallback callback, void *callback_state)
Definition: nbtree.c:786
void _bt_dedup_start_pending(BTDedupState state, IndexTuple base, OffsetNumber baseoff)
Definition: nbtdedup.c:433
#define BTPageGetOpaque(page)
Definition: nbtree.h:73
#define P_ISDELETED(opaque)
Definition: nbtree.h:222
bytea * btoptions(Datum reloptions, bool validate)
Definition: nbtutils.c:2111
static ItemPointer BTreeTupleGetPosting(IndexTuple posting)
Definition: nbtree.h:538
void _bt_checkpage(Relation rel, Buffer buf)
Definition: nbtpage.c:794
bool _bt_advance_array_keys(IndexScanDesc scan, ScanDirection dir)
Definition: nbtutils.c:551
BTCycleId _bt_vacuum_cycleid(Relation rel)
Definition: nbtutils.c:1942
void _bt_metaversion(Relation rel, bool *heapkeyspace, bool *allequalimage)
Definition: nbtpage.c:736
static BlockNumber BTreeTupleGetTopParent(IndexTuple leafhikey)
Definition: nbtree.h:621
struct BTArrayKeyInfo BTArrayKeyInfo
void _bt_parallel_advance_array_keys(IndexScanDesc scan)
Definition: nbtree.c:758
void btadjustmembers(Oid opfamilyoid, Oid opclassoid, List *operators, List *functions)
Definition: nbtvalidate.c:293
BTScanInsert _bt_mkscankey(Relation rel, IndexTuple itup)
Definition: nbtutils.c:90
struct BTPendingFSM BTPendingFSM
void _bt_killitems(IndexScanDesc scan)
Definition: nbtutils.c:1725
#define BT_OFFSET_MASK
Definition: nbtree.h:463
IndexBulkDeleteResult * btvacuumcleanup(IndexVacuumInfo *info, IndexBulkDeleteResult *stats)
Definition: nbtree.c:816
IndexTuple _bt_form_posting(IndexTuple base, ItemPointer htids, int nhtids)
Definition: nbtdedup.c:864
bool _bt_doinsert(Relation rel, IndexTuple itup, IndexUniqueCheck checkUnique, bool indexUnchanged, Relation heapRel)
Definition: nbtinsert.c:100
#define MaxTIDsPerBTreePage
Definition: nbtree.h:185
void _bt_finish_split(Relation rel, Buffer lbuf, BTStack stack)
Definition: nbtinsert.c:2230
static void BTreeTupleSetDownLink(IndexTuple pivot, BlockNumber blkno)
Definition: nbtree.h:563
bool _bt_checkkeys(IndexScanDesc scan, IndexTuple tuple, int tupnatts, ScanDirection dir, bool *continuescan)
Definition: nbtutils.c:1362
bool btgettuple(IndexScanDesc scan, ScanDirection dir)
Definition: nbtree.c:212
void btparallelrescan(IndexScanDesc scan)
Definition: nbtree.c:595
struct BTVacState BTVacState
void _bt_preprocess_array_keys(IndexScanDesc scan)
Definition: nbtutils.c:210
bool _bt_check_natts(Relation rel, bool heapkeyspace, Page page, OffsetNumber offnum)
Definition: nbtutils.c:2471
IndexTuple _bt_truncate(Relation rel, IndexTuple lastleft, IndexTuple firstright, BTScanInsert itup_key)
Definition: nbtutils.c:2205
void _bt_set_cleanup_info(Relation rel, BlockNumber num_delpages)
Definition: nbtpage.c:234
#define BT_STATUS_OFFSET_MASK
Definition: nbtree.h:464
uint16 BTCycleId
Definition: nbtree.h:29
static uint32 BTreeTupleGetPostingOffset(IndexTuple posting)
Definition: nbtree.h:530
struct BTScanInsertData BTScanInsertData
void btbuildempty(Relation index)
Definition: nbtree.c:152
bool _bt_parallel_seize(IndexScanDesc scan, BlockNumber *pageno)
Definition: nbtree.c:636
struct BTMetaPageData BTMetaPageData
bool _bt_conditionallockbuf(Relation rel, Buffer buf)
Definition: nbtpage.c:1105
Buffer _bt_getbuf(Relation rel, BlockNumber blkno, int access)
Definition: nbtpage.c:871
BTScanInsertData * BTScanInsert
Definition: nbtree.h:802
Size btestimateparallelscan(void)
Definition: nbtree.c:571
int _bt_keep_natts_fast(Relation rel, IndexTuple lastleft, IndexTuple firstright)
Definition: nbtutils.c:2424
static ItemPointer BTreeTupleGetPostingN(IndexTuple posting, int n)
Definition: nbtree.h:545
IndexBuildResult * btbuild(Relation heap, Relation index, struct IndexInfo *indexInfo)
Definition: nbtsort.c:301
void btinitparallelscan(void *target)
Definition: nbtree.c:580
void _bt_parallel_build_main(dsm_segment *seg, shm_toc *toc)
Definition: nbtsort.c:1798
void _bt_unlockbuf(Relation rel, Buffer buf)
Definition: nbtpage.c:1082
static BlockNumber BTreeTupleGetDownLink(IndexTuple pivot)
Definition: nbtree.h:557
void _bt_restore_array_keys(IndexScanDesc scan)
Definition: nbtutils.c:629
#define INDEX_ALT_TID_MASK
Definition: nbtree.h:460
void _bt_upgradelockbufcleanup(Relation rel, Buffer buf)
Definition: nbtpage.c:1121
void _bt_initmetapage(Page page, BlockNumber rootbknum, uint32 level, bool allequalimage)
Definition: nbtpage.c:69
Buffer _bt_getstackbuf(Relation rel, BTStack stack, BlockNumber child)
Definition: nbtinsert.c:2307
BTVacuumPostingData * BTVacuumPosting
Definition: nbtree.h:920
void _bt_delitems_delete_check(Relation rel, Buffer buf, Relation heapRel, TM_IndexDeleteOp *delstate)
Definition: nbtpage.c:1529
struct BTDeletedPageData BTDeletedPageData
bool _bt_vacuum_needs_cleanup(Relation rel)
Definition: nbtpage.c:181
OffsetNumber _bt_binsrch_insert(Relation rel, BTInsertState insertstate)
Definition: nbtsearch.c:442
bool _bt_next(IndexScanDesc scan, ScanDirection dir)
Definition: nbtsearch.c:1466
struct BTDedupInterval BTDedupInterval
int32 _bt_compare(Relation rel, BTScanInsert key, Page page, OffsetNumber offnum)
Definition: nbtsearch.c:656
struct BTScanPosItem BTScanPosItem
static ItemPointer BTreeTupleGetMaxHeapTID(IndexTuple itup)
Definition: nbtree.h:665
void _bt_dedup_pass(Relation rel, Buffer buf, Relation heapRel, IndexTuple newitem, Size newitemsz, bool bottomupdedup)
Definition: nbtdedup.c:58
static bool BTreeTupleIsPosting(IndexTuple itup)
Definition: nbtree.h:493
Size BTreeShmemSize(void)
Definition: nbtutils.c:2070
Buffer _bt_get_endpoint(Relation rel, uint32 level, bool rightmost, Snapshot snapshot)
Definition: nbtsearch.c:2307
void _bt_mark_array_keys(IndexScanDesc scan)
Definition: nbtutils.c:610
static void BTPageSetDeleted(Page page, FullTransactionId safexid)
Definition: nbtree.h:239
int64 btgetbitmap(IndexScanDesc scan, TIDBitmap *tbm)
Definition: nbtree.c:286
void btmarkpos(IndexScanDesc scan)
Definition: nbtree.c:485
BTDedupStateData * BTDedupState
Definition: nbtree.h:899
void btendscan(IndexScanDesc scan)
Definition: nbtree.c:449
void _bt_pendingfsm_finalize(Relation rel, BTVacState *vstate)
Definition: nbtpage.c:2956
#define BT_IS_POSTING
Definition: nbtree.h:467
void _bt_lockbuf(Relation rel, Buffer buf, int access)
Definition: nbtpage.c:1051
struct BTOptions BTOptions
void btrescan(IndexScanDesc scan, ScanKey scankey, int nscankeys, ScanKey orderbys, int norderbys)
Definition: nbtree.c:390
bool btproperty(Oid index_oid, int attno, IndexAMProperty prop, const char *propname, bool *res, bool *isnull)
Definition: nbtutils.c:2134
struct BTScanPosData BTScanPosData
static ItemPointer BTreeTupleGetHeapTID(IndexTuple itup)
Definition: nbtree.h:639
void _bt_pendingfsm_init(Relation rel, BTVacState *vstate, bool cleanuponly)
Definition: nbtpage.c:2915
Size _bt_dedup_finish_pending(Page newpage, BTDedupState state)
Definition: nbtdedup.c:555
struct BTScanOpaqueData BTScanOpaqueData
BTStack _bt_search(Relation rel, BTScanInsert key, Buffer *bufP, int access, Snapshot snapshot)
Definition: nbtsearch.c:96
BTInsertStateData * BTInsertState
Definition: nbtree.h:841
static void BTreeTupleSetNAtts(IndexTuple itup, uint16 nkeyatts, bool heaptid)
Definition: nbtree.h:596
void btrestrpos(IndexScanDesc scan)
Definition: nbtree.c:515
bool _bt_allequalimage(Relation rel, bool debugmessage)
Definition: nbtutils.c:2689
struct BTDedupStateData BTDedupStateData
StaticAssertDecl(BT_OFFSET_MASK >=INDEX_MAX_KEYS, "BT_OFFSET_MASK can't fit INDEX_MAX_KEYS")
#define BTP_DELETED
Definition: nbtree.h:78
void _bt_start_array_keys(IndexScanDesc scan, ScanDirection dir)
Definition: nbtutils.c:525
void _bt_preprocess_keys(IndexScanDesc scan)
Definition: nbtutils.c:749
struct BTStackData BTStackData
BTCycleId _bt_start_vacuum(Relation rel)
Definition: nbtutils.c:1976
BTScanPosData * BTScanPos
Definition: nbtree.h:994
struct BTInsertStateData BTInsertStateData
struct BTVacuumPostingData BTVacuumPostingData
Buffer _bt_moveright(Relation rel, BTScanInsert key, Buffer buf, bool forupdate, BTStack stack, int access, Snapshot snapshot)
Definition: nbtsearch.c:236
BTScanOpaqueData * BTScanOpaque
Definition: nbtree.h:1080
uint16 OffsetNumber
Definition: off.h:24
void * arg
#define INDEX_MAX_KEYS
static char * buf
Definition: pg_test_fsync.c:67
uintptr_t Datum
Definition: postgres.h:64
unsigned int Oid
Definition: postgres_ext.h:31
short access
Definition: preproc-type.c:36
bool GlobalVisCheckRemovableFullXid(Relation rel, FullTransactionId fxid)
Definition: procarray.c:4300
static const struct fns functions
Definition: regcomp.c:357
ScanDirection
Definition: sdir.h:25
Datum * elem_values
Definition: nbtree.h:1035
uint16 nitems
Definition: nbtree.h:850
OffsetNumber baseoff
Definition: nbtree.h:849
Size maxpostingsize
Definition: nbtree.h:876
ItemPointer htids
Definition: nbtree.h:884
bool deduplicate
Definition: nbtree.h:874
OffsetNumber baseoff
Definition: nbtree.h:880
Size basetupsize
Definition: nbtree.h:881
BTDedupInterval intervals[MaxIndexTuplesPerPage]
Definition: nbtree.h:896
IndexTuple base
Definition: nbtree.h:879
Size phystupsize
Definition: nbtree.h:887
FullTransactionId safexid
Definition: nbtree.h:235
OffsetNumber stricthigh
Definition: nbtree.h:831
bool bounds_valid
Definition: nbtree.h:829
OffsetNumber low
Definition: nbtree.h:830
IndexTuple itup
Definition: nbtree.h:817
BTScanInsert itup_key
Definition: nbtree.h:819
uint32 btm_last_cleanup_num_delpages
Definition: nbtree.h:114
uint32 btm_level
Definition: nbtree.h:108
float8 btm_last_cleanup_num_heap_tuples
Definition: nbtree.h:116
BlockNumber btm_fastroot
Definition: nbtree.h:109
uint32 btm_version
Definition: nbtree.h:106
uint32 btm_magic
Definition: nbtree.h:105
BlockNumber btm_root
Definition: nbtree.h:107
bool btm_allequalimage
Definition: nbtree.h:118
uint32 btm_fastlevel
Definition: nbtree.h:110
int fillfactor
Definition: nbtree.h:1096
bool deduplicate_items
Definition: nbtree.h:1098
float8 vacuum_cleanup_index_scale_factor
Definition: nbtree.h:1097
int32 varlena_header_
Definition: nbtree.h:1095
BlockNumber btpo_next
Definition: nbtree.h:65
BlockNumber btpo_prev
Definition: nbtree.h:64
uint16 btpo_flags
Definition: nbtree.h:67
uint32 btpo_level
Definition: nbtree.h:66
BTCycleId btpo_cycleid
Definition: nbtree.h:68
FullTransactionId safexid
Definition: nbtree.h:328
BlockNumber target
Definition: nbtree.h:327
bool pivotsearch
Definition: nbtree.h:796
ItemPointer scantid
Definition: nbtree.h:797
bool allequalimage
Definition: nbtree.h:793
bool heapkeyspace
Definition: nbtree.h:792
bool anynullkeys
Definition: nbtree.h:794
ScanKeyData scankeys[INDEX_MAX_KEYS]
Definition: nbtree.h:799
BTArrayKeyInfo * arrayKeys
Definition: nbtree.h:1051
char * markTuples
Definition: nbtree.h:1064
BTScanPosData currPos
Definition: nbtree.h:1076
int * killedItems
Definition: nbtree.h:1055
char * currTuples
Definition: nbtree.h:1063
ScanKey arrayKeyData
Definition: nbtree.h:1046
BTScanPosData markPos
Definition: nbtree.h:1077
ScanKey keyData
Definition: nbtree.h:1043
MemoryContext arrayContext
Definition: nbtree.h:1052
bool moreRight
Definition: nbtree.h:972
Buffer buf
Definition: nbtree.h:959
BlockNumber currPage
Definition: nbtree.h:962
int firstItem
Definition: nbtree.h:987
int nextTupleOffset
Definition: nbtree.h:978
BlockNumber nextPage
Definition: nbtree.h:963
bool moreLeft
Definition: nbtree.h:971
int lastItem
Definition: nbtree.h:988
BTScanPosItem items[MaxTIDsPerBTreePage]
Definition: nbtree.h:991
int itemIndex
Definition: nbtree.h:989
XLogRecPtr lsn
Definition: nbtree.h:961
ItemPointerData heapTid
Definition: nbtree.h:952
LocationIndex tupleOffset
Definition: nbtree.h:954
OffsetNumber indexOffset
Definition: nbtree.h:953
BlockNumber bts_blkno
Definition: nbtree.h:735
struct BTStackData * bts_parent
Definition: nbtree.h:737
OffsetNumber bts_offset
Definition: nbtree.h:736
IndexBulkDeleteResult * stats
Definition: nbtree.h:334
BTCycleId cycleid
Definition: nbtree.h:337
BTPendingFSM * pendingpages
Definition: nbtree.h:345
int npendingpages
Definition: nbtree.h:346
IndexBulkDeleteCallback callback
Definition: nbtree.h:335
MemoryContext pagedelcontext
Definition: nbtree.h:338
IndexVacuumInfo * info
Definition: nbtree.h:333
int bufsize
Definition: nbtree.h:343
int maxbufsize
Definition: nbtree.h:344
void * callback_state
Definition: nbtree.h:336
uint16 deletetids[FLEXIBLE_ARRAY_MEMBER]
Definition: nbtree.h:917
uint16 ndeletedtids
Definition: nbtree.h:916
IndexTuple itup
Definition: nbtree.h:912
OffsetNumber updatedoffset
Definition: nbtree.h:913
ItemPointerData t_tid
Definition: itup.h:37
unsigned short t_info
Definition: itup.h:49
Definition: pg_list.h:54
LocationIndex pd_special
Definition: bufpage.h:164
LocationIndex pd_upper
Definition: bufpage.h:163
LocationIndex pd_lower
Definition: bufpage.h:162
Definition: type.h:95
Definition: regguts.h:318
Definition: c.h:671
static void callback(struct sockaddr *addr, struct sockaddr *mask, void *unused)
Definition: test_ifaddrs.c:46
#define FirstNormalFullTransactionId
Definition: transam.h:57
uint64 XLogRecPtr
Definition: xlogdefs.h:21