1 /*-------------------------------------------------------------------------
4 * header file for postgres btree access method implementation.
7 * Portions Copyright (c) 1996-2011, PostgreSQL Global Development Group
8 * Portions Copyright (c) 1994, Regents of the University of California
10 * src/include/access/nbtree.h
12 *-------------------------------------------------------------------------
17 #include "access/genam.h"
18 #include "access/itup.h"
19 #include "access/sdir.h"
20 #include "access/xlog.h"
21 #include "access/xlogutils.h"
22 #include "catalog/pg_index.h"
24 /* There's room for a 16-bit vacuum cycle ID in BTPageOpaqueData */
25 typedef uint16 BTCycleId;
28 * BTPageOpaqueData -- At the end of every page, we store a pointer
29 * to both siblings in the tree. This is used to do forward/backward
30 * index scans. The next-page link is also critical for recovery when
31 * a search has navigated to the wrong page due to concurrent page splits
32 * or deletions; see src/backend/access/nbtree/README for more info.
34 * In addition, we store the page's btree level (counting upwards from
35 * zero at a leaf page) as well as some flag bits indicating the page type
36 * and status. If the page is deleted, we replace the level with the
37 * next-transaction-ID value indicating when it is safe to reclaim the page.
39 * We also store a "vacuum cycle ID". When a page is split while VACUUM is
40 * processing the index, a nonzero value associated with the VACUUM run is
41 * stored into both halves of the split page. (If VACUUM is not running,
42 * both pages receive zero cycleids.) This allows VACUUM to detect whether
43 * a page was split since it started, with a small probability of false match
44 * if the page was last split some exact multiple of MAX_BT_CYCLE_ID VACUUMs
45 * ago. Also, during a split, the BTP_SPLIT_END flag is cleared in the left
46 * (original) page, and set in the right page, but only if the next page
47 * to its right has a different cycleid.
49 * NOTE: the BTP_LEAF flag bit is redundant since level==0 could be tested
53 typedef struct BTPageOpaqueData
55 BlockNumber btpo_prev; /* left sibling, or P_NONE if leftmost */
56 BlockNumber btpo_next; /* right sibling, or P_NONE if rightmost */
59 uint32 level; /* tree level --- zero for leaf pages */
60 TransactionId xact; /* next transaction ID, if deleted */
62 uint16 btpo_flags; /* flag bits, see below */
63 BTCycleId btpo_cycleid; /* vacuum cycle ID of latest split */
66 typedef BTPageOpaqueData *BTPageOpaque;
68 /* Bits defined in btpo_flags */
69 #define BTP_LEAF (1 << 0) /* leaf page, i.e. not internal page */
70 #define BTP_ROOT (1 << 1) /* root page (has no parent) */
71 #define BTP_DELETED (1 << 2) /* page has been deleted from tree */
72 #define BTP_META (1 << 3) /* meta-page */
73 #define BTP_HALF_DEAD (1 << 4) /* empty, but still in tree */
74 #define BTP_SPLIT_END (1 << 5) /* rightmost page of split group */
75 #define BTP_HAS_GARBAGE (1 << 6) /* page has LP_DEAD tuples */
78 * The max allowed value of a cycle ID is a bit less than 64K. This is
79 * for convenience of pg_filedump and similar utilities: we want to use
80 * the last 2 bytes of special space as an index type indicator, and
81 * restricting cycle ID lets btree use that space for vacuum cycle IDs
82 * while still allowing index type to be identified.
84 #define MAX_BT_CYCLE_ID 0xFF7F
88 * The Meta page is always the first page in the btree index.
89 * Its primary purpose is to point to the location of the btree root page.
90 * We also point to the "fast" root, which is the current effective root;
91 * see README for discussion.
94 typedef struct BTMetaPageData
96 uint32 btm_magic; /* should contain BTREE_MAGIC */
97 uint32 btm_version; /* should contain BTREE_VERSION */
98 BlockNumber btm_root; /* current root location */
99 uint32 btm_level; /* tree level of the root page */
100 BlockNumber btm_fastroot; /* current "fast" root location */
101 uint32 btm_fastlevel; /* tree level of the "fast" root page */
104 #define BTPageGetMeta(p) \
105 ((BTMetaPageData *) PageGetContents(p))
107 #define BTREE_METAPAGE 0 /* first page is meta */
108 #define BTREE_MAGIC 0x053162 /* magic number of btree pages */
109 #define BTREE_VERSION 2 /* current version number */
112 * Maximum size of a btree index entry, including its tuple header.
114 * We actually need to be able to fit three items on every page,
115 * so restrict any one item to 1/3 the per-page available space.
117 #define BTMaxItemSize(page) \
118 MAXALIGN_DOWN((PageGetPageSize(page) - \
119 MAXALIGN(SizeOfPageHeaderData + 3*sizeof(ItemIdData)) - \
120 MAXALIGN(sizeof(BTPageOpaqueData))) / 3)
123 * The leaf-page fillfactor defaults to 90% but is user-adjustable.
124 * For pages above the leaf level, we use a fixed 70% fillfactor.
125 * The fillfactor is applied during index build and when splitting
126 * a rightmost page; when splitting non-rightmost pages we try to
127 * divide the data equally.
129 #define BTREE_MIN_FILLFACTOR 10
130 #define BTREE_DEFAULT_FILLFACTOR 90
131 #define BTREE_NONLEAF_FILLFACTOR 70
134 * Test whether two btree entries are "the same".
137 * In addition, we must guarantee that all tuples in the index are unique,
138 * in order to satisfy some assumptions in Lehman and Yao. The way that we
139 * do this is by generating a new OID for every insertion that we do in the
140 * tree. This adds eight bytes to the size of btree index tuples. Note
141 * that we do not use the OID as part of a composite key; the OID only
142 * serves as a unique identifier for a given index tuple (logical position
146 * actually, we must guarantee that all tuples in A LEVEL
147 * are unique, not in ALL INDEX. So, we can use the t_tid
148 * as unique identifier for a given index tuple (logical position
149 * within a level). - vadim 04/09/97
151 #define BTTidSame(i1, i2) \
152 ( (i1).ip_blkid.bi_hi == (i2).ip_blkid.bi_hi && \
153 (i1).ip_blkid.bi_lo == (i2).ip_blkid.bi_lo && \
154 (i1).ip_posid == (i2).ip_posid )
155 #define BTEntrySame(i1, i2) \
156 BTTidSame((i1)->t_tid, (i2)->t_tid)
160 * In general, the btree code tries to localize its knowledge about
161 * page layout to a couple of routines. However, we need a special
162 * value to indicate "no page number" in those places where we expect
163 * page numbers. We can use zero for this because we never need to
164 * make a pointer to the metadata page.
170 * Macros to test whether a page is leftmost or rightmost on its tree level,
171 * as well as other state info kept in the opaque data.
173 #define P_LEFTMOST(opaque) ((opaque)->btpo_prev == P_NONE)
174 #define P_RIGHTMOST(opaque) ((opaque)->btpo_next == P_NONE)
175 #define P_ISLEAF(opaque) ((opaque)->btpo_flags & BTP_LEAF)
176 #define P_ISROOT(opaque) ((opaque)->btpo_flags & BTP_ROOT)
177 #define P_ISDELETED(opaque) ((opaque)->btpo_flags & BTP_DELETED)
178 #define P_ISHALFDEAD(opaque) ((opaque)->btpo_flags & BTP_HALF_DEAD)
179 #define P_IGNORE(opaque) ((opaque)->btpo_flags & (BTP_DELETED|BTP_HALF_DEAD))
180 #define P_HAS_GARBAGE(opaque) ((opaque)->btpo_flags & BTP_HAS_GARBAGE)
183 * Lehman and Yao's algorithm requires a ``high key'' on every non-rightmost
184 * page. The high key is not a data key, but gives info about what range of
185 * keys is supposed to be on this page. The high key on a page is required
186 * to be greater than or equal to any data key that appears on the page.
187 * If we find ourselves trying to insert a key > high key, we know we need
188 * to move right (this should only happen if the page was split since we
189 * examined the parent page).
191 * Our insertion algorithm guarantees that we can use the initial least key
192 * on our right sibling as the high key. Once a page is created, its high
193 * key changes only if the page is split.
195 * On a non-rightmost page, the high key lives in item 1 and data items
196 * start in item 2. Rightmost pages have no high key, so we store data
197 * items beginning in item 1.
200 #define P_HIKEY ((OffsetNumber) 1)
201 #define P_FIRSTKEY ((OffsetNumber) 2)
202 #define P_FIRSTDATAKEY(opaque) (P_RIGHTMOST(opaque) ? P_HIKEY : P_FIRSTKEY)
205 * XLOG records for btree operations
207 * XLOG allows to store some information in high 4 bits of log
208 * record xl_info field
210 #define XLOG_BTREE_INSERT_LEAF 0x00 /* add index tuple without split */
211 #define XLOG_BTREE_INSERT_UPPER 0x10 /* same, on a non-leaf page */
212 #define XLOG_BTREE_INSERT_META 0x20 /* same, plus update metapage */
213 #define XLOG_BTREE_SPLIT_L 0x30 /* add index tuple with split */
214 #define XLOG_BTREE_SPLIT_R 0x40 /* as above, new item on right */
215 #define XLOG_BTREE_SPLIT_L_ROOT 0x50 /* add tuple with split of root */
216 #define XLOG_BTREE_SPLIT_R_ROOT 0x60 /* as above, new item on right */
217 #define XLOG_BTREE_DELETE 0x70 /* delete leaf index tuples for a page */
218 #define XLOG_BTREE_DELETE_PAGE 0x80 /* delete an entire page */
219 #define XLOG_BTREE_DELETE_PAGE_META 0x90 /* same, and update metapage */
220 #define XLOG_BTREE_NEWROOT 0xA0 /* new root page */
221 #define XLOG_BTREE_DELETE_PAGE_HALF 0xB0 /* page deletion that makes
222 * parent half-dead */
223 #define XLOG_BTREE_VACUUM 0xC0 /* delete entries on a page during
225 #define XLOG_BTREE_REUSE_PAGE 0xD0 /* old page is about to be reused from
229 * All that we need to find changed index tuple
231 typedef struct xl_btreetid
234 ItemPointerData tid; /* changed tuple id */
238 * All that we need to regenerate the meta-data page
240 typedef struct xl_btree_metadata
244 BlockNumber fastroot;
249 * This is what we need to know about simple (without split) insert.
251 * This data record is used for INSERT_LEAF, INSERT_UPPER, INSERT_META.
252 * Note that INSERT_META implies it's not a leaf page.
254 typedef struct xl_btree_insert
256 xl_btreetid target; /* inserted tuple id */
257 /* BlockNumber downlink field FOLLOWS IF NOT XLOG_BTREE_INSERT_LEAF */
258 /* xl_btree_metadata FOLLOWS IF XLOG_BTREE_INSERT_META */
259 /* INDEX TUPLE FOLLOWS AT END OF STRUCT */
262 #define SizeOfBtreeInsert (offsetof(xl_btreetid, tid) + SizeOfIptrData)
265 * On insert with split, we save all the items going into the right sibling
266 * so that we can restore it completely from the log record. This way takes
267 * less xlog space than the normal approach, because if we did it standardly,
268 * XLogInsert would almost always think the right page is new and store its
269 * whole page image. The left page, however, is handled in the normal
270 * incremental-update fashion.
272 * Note: the four XLOG_BTREE_SPLIT xl_info codes all use this data record.
273 * The _L and _R variants indicate whether the inserted tuple went into the
274 * left or right split page (and thus, whether newitemoff and the new item
275 * are stored or not). The _ROOT variants indicate that we are splitting
276 * the root page, and thus that a newroot record rather than an insert or
277 * split record should follow. Note that a split record never carries a
278 * metapage update --- we'll do that in the parent-level update.
280 typedef struct xl_btree_split
283 BlockNumber leftsib; /* orig page / new left page */
284 BlockNumber rightsib; /* new right page */
285 BlockNumber rnext; /* next block (orig page's rightlink) */
286 uint32 level; /* tree level of page being split */
287 OffsetNumber firstright; /* first item moved to right page */
290 * If level > 0, BlockIdData downlink follows. (We use BlockIdData rather
291 * than BlockNumber for alignment reasons: SizeOfBtreeSplit is only 16-bit
294 * If level > 0, an IndexTuple representing the HIKEY of the left page
295 * follows. We don't need this on leaf pages, because it's the same as
296 * the leftmost key in the new right page. Also, it's suppressed if
297 * XLogInsert chooses to store the left page's whole page image.
299 * In the _L variants, next are OffsetNumber newitemoff and the new item.
300 * (In the _R variants, the new item is one of the right page's tuples.)
301 * The new item, but not newitemoff, is suppressed if XLogInsert chooses
302 * to store the left page's whole page image.
304 * Last are the right page's tuples in the form used by _bt_restore_page.
308 #define SizeOfBtreeSplit (offsetof(xl_btree_split, firstright) + sizeof(OffsetNumber))
311 * This is what we need to know about delete of individual leaf index tuples.
312 * The WAL record can represent deletion of any number of index tuples on a
313 * single index page when *not* executed by VACUUM.
315 typedef struct xl_btree_delete
317 RelFileNode node; /* RelFileNode of the index */
319 RelFileNode hnode; /* RelFileNode of the heap the index currently
323 /* TARGET OFFSET NUMBERS FOLLOW AT THE END */
326 #define SizeOfBtreeDelete (offsetof(xl_btree_delete, nitems) + sizeof(int))
329 * This is what we need to know about page reuse within btree.
331 typedef struct xl_btree_reuse_page
335 TransactionId latestRemovedXid;
336 } xl_btree_reuse_page;
338 #define SizeOfBtreeReusePage (sizeof(xl_btree_reuse_page))
341 * This is what we need to know about vacuum of individual leaf index tuples.
342 * The WAL record can represent deletion of any number of index tuples on a
343 * single index page when executed by VACUUM.
345 * The correctness requirement for applying these changes during recovery is
346 * that we must do one of these two things for every block in the index:
347 * * lock the block for cleanup and apply any required changes
348 * * EnsureBlockUnpinned()
349 * The purpose of this is to ensure that no index scans started before we
350 * finish scanning the index are still running by the time we begin to remove
353 * Any changes to any one block are registered on just one WAL record. All
354 * blocks that we need to run EnsureBlockUnpinned() are listed as a block range
355 * starting from the last block vacuumed through until this one. Individual
356 * block numbers aren't given.
358 * Note that the *last* WAL record in any vacuum of an index is allowed to
359 * have a zero length array of offsets. Earlier records must have at least one.
361 typedef struct xl_btree_vacuum
365 BlockNumber lastBlockVacuumed;
367 /* TARGET OFFSET NUMBERS FOLLOW */
370 #define SizeOfBtreeVacuum (offsetof(xl_btree_vacuum, lastBlockVacuumed) + sizeof(BlockNumber))
373 * This is what we need to know about deletion of a btree page. The target
374 * identifies the tuple removed from the parent page (note that we remove
375 * this tuple's downlink and the *following* tuple's key). Note we do not
376 * store any content for the deleted page --- it is just rewritten as empty
377 * during recovery, apart from resetting the btpo.xact.
379 typedef struct xl_btree_delete_page
381 xl_btreetid target; /* deleted tuple id in parent page */
382 BlockNumber deadblk; /* child block being deleted */
383 BlockNumber leftblk; /* child block's left sibling, if any */
384 BlockNumber rightblk; /* child block's right sibling */
385 TransactionId btpo_xact; /* value of btpo.xact for use in recovery */
386 /* xl_btree_metadata FOLLOWS IF XLOG_BTREE_DELETE_PAGE_META */
387 } xl_btree_delete_page;
389 #define SizeOfBtreeDeletePage (offsetof(xl_btree_delete_page, btpo_xact) + sizeof(TransactionId))
392 * New root log record. There are zero tuples if this is to establish an
393 * empty root, or two if it is the result of splitting an old root.
395 * Note that although this implies rewriting the metadata page, we don't need
396 * an xl_btree_metadata record --- the rootblk and level are sufficient.
398 typedef struct xl_btree_newroot
401 BlockNumber rootblk; /* location of new root */
402 uint32 level; /* its tree level */
403 /* 0 or 2 INDEX TUPLES FOLLOW AT END OF STRUCT */
406 #define SizeOfBtreeNewroot (offsetof(xl_btree_newroot, level) + sizeof(uint32))
410 * Operator strategy numbers for B-tree have been moved to access/skey.h,
411 * because many places need to use them in ScanKeyInit() calls.
413 * The strategy numbers are chosen so that we can commute them by
416 #define BTCommuteStrategyNumber(strat) (BTMaxStrategyNumber + 1 - (strat))
419 * When a new operator class is declared, we require that the user
420 * supply us with an amproc procedure (BTORDER_PROC) for determining
421 * whether, for two keys a and b, a < b, a = b, or a > b. This routine
422 * must return < 0, 0, > 0, respectively, in these three cases. (It must
423 * not return INT_MIN, since we may negate the result before using it.)
425 * To facilitate accelerated sorting, an operator class may choose to
426 * offer a second procedure (BTSORTSUPPORT_PROC). For full details, see
427 * src/include/utils/sortsupport.h.
430 #define BTORDER_PROC 1
431 #define BTSORTSUPPORT_PROC 2
434 * We need to be able to tell the difference between read and write
435 * requests for pages, in order to do locking correctly.
438 #define BT_READ BUFFER_LOCK_SHARE
439 #define BT_WRITE BUFFER_LOCK_EXCLUSIVE
442 * BTStackData -- As we descend a tree, we push the (location, downlink)
443 * pairs from internal pages onto a private stack. If we split a
444 * leaf, we use this stack to walk back up the tree and insert data
445 * into parent pages (and possibly to split them, too). Lehman and
446 * Yao's update algorithm guarantees that under no circumstances can
447 * our private stack give us an irredeemably bad picture up the tree.
448 * Again, see the paper for details.
451 typedef struct BTStackData
453 BlockNumber bts_blkno;
454 OffsetNumber bts_offset;
455 IndexTupleData bts_btentry;
456 struct BTStackData *bts_parent;
459 typedef BTStackData *BTStack;
462 * BTScanOpaqueData is the btree-private state needed for an indexscan.
463 * This consists of preprocessed scan keys (see _bt_preprocess_keys() for
464 * details of the preprocessing), information about the current location
465 * of the scan, and information about the marked location, if any. (We use
466 * BTScanPosData to represent the data needed for each of current and marked
467 * locations.) In addition we can remember some known-killed index entries
468 * that must be marked before we can move off the current page.
470 * Index scans work a page at a time: we pin and read-lock the page, identify
471 * all the matching items on the page and save them in BTScanPosData, then
472 * release the read-lock while returning the items to the caller for
473 * processing. This approach minimizes lock/unlock traffic. Note that we
474 * keep the pin on the index page until the caller is done with all the items
475 * (this is needed for VACUUM synchronization, see nbtree/README). When we
476 * are ready to step to the next page, if the caller has told us any of the
477 * items were killed, we re-lock the page to mark them killed, then unlock.
478 * Finally we drop the pin and step to the next page in the appropriate
481 * If we are doing an index-only scan, we save the entire IndexTuple for each
482 * matched item, otherwise only its heap TID and offset. The IndexTuples go
483 * into a separate workspace array; each BTScanPosItem stores its tuple's
484 * offset within that array.
487 typedef struct BTScanPosItem /* what we remember about each match */
489 ItemPointerData heapTid; /* TID of referenced heap item */
490 OffsetNumber indexOffset; /* index item's location within page */
491 LocationIndex tupleOffset; /* IndexTuple's offset in workspace, if any */
494 typedef struct BTScanPosData
496 Buffer buf; /* if valid, the buffer is pinned */
498 BlockNumber nextPage; /* page's right link when we scanned it */
501 * moreLeft and moreRight track whether we think there may be matching
502 * index entries to the left and right of the current page, respectively.
503 * We can clear the appropriate one of these flags when _bt_checkkeys()
504 * returns continuescan = false.
510 * If we are doing an index-only scan, nextTupleOffset is the first free
511 * location in the associated tuple storage workspace.
516 * The items array is always ordered in index order (ie, increasing
517 * indexoffset). When scanning backwards it is convenient to fill the
518 * array back-to-front, so we start at the last slot and fill downwards.
519 * Hence we need both a first-valid-entry and a last-valid-entry counter.
520 * itemIndex is a cursor showing which entry was last returned to caller.
522 int firstItem; /* first valid index in items[] */
523 int lastItem; /* last valid index in items[] */
524 int itemIndex; /* current index in items[] */
526 BTScanPosItem items[MaxIndexTuplesPerPage]; /* MUST BE LAST */
529 typedef BTScanPosData *BTScanPos;
531 #define BTScanPosIsValid(scanpos) BufferIsValid((scanpos).buf)
533 /* We need one of these for each equality-type SK_SEARCHARRAY scan key */
534 typedef struct BTArrayKeyInfo
536 int scan_key; /* index of associated key in arrayKeyData */
537 int cur_elem; /* index of current element in elem_values */
538 int num_elems; /* number of elems in current array value */
539 Datum *elem_values; /* array of num_elems Datums */
542 typedef struct BTScanOpaqueData
544 /* these fields are set by _bt_preprocess_keys(): */
545 bool qual_ok; /* false if qual can never be satisfied */
546 int numberOfKeys; /* number of preprocessed scan keys */
547 ScanKey keyData; /* array of preprocessed scan keys */
549 /* workspace for SK_SEARCHARRAY support */
550 ScanKey arrayKeyData; /* modified copy of scan->keyData */
551 int numArrayKeys; /* number of equality-type array keys (-1 if
552 * there are any unsatisfiable array keys) */
553 BTArrayKeyInfo *arrayKeys; /* info about each equality-type array key */
554 MemoryContext arrayContext; /* scan-lifespan context for array data */
556 /* info about killed items if any (killedItems is NULL if never used) */
557 int *killedItems; /* currPos.items indexes of killed items */
558 int numKilled; /* number of currently stored items */
561 * If we are doing an index-only scan, these are the tuple storage
562 * workspaces for the currPos and markPos respectively. Each is of
563 * size BLCKSZ, so it can hold as much as a full page's worth of tuples.
565 char *currTuples; /* tuple storage for currPos */
566 char *markTuples; /* tuple storage for markPos */
569 * If the marked position is on the same page as current position, we
570 * don't use markPos, but just keep the marked itemIndex in markItemIndex
571 * (all the rest of currPos is valid for the mark position). Hence, to
572 * determine if there is a mark, first look at markItemIndex, then at
575 int markItemIndex; /* itemIndex, or -1 if not valid */
577 /* keep these last in struct for efficiency */
578 BTScanPosData currPos; /* current position data */
579 BTScanPosData markPos; /* marked position, if any */
582 typedef BTScanOpaqueData *BTScanOpaque;
585 * We use some private sk_flags bits in preprocessed scan keys. We're allowed
586 * to use bits 16-31 (see skey.h). The uppermost bits are copied from the
587 * index's indoption[] array entry for the index attribute.
589 #define SK_BT_REQFWD 0x00010000 /* required to continue forward scan */
590 #define SK_BT_REQBKWD 0x00020000 /* required to continue backward scan */
591 #define SK_BT_INDOPTION_SHIFT 24 /* must clear the above bits */
592 #define SK_BT_DESC (INDOPTION_DESC << SK_BT_INDOPTION_SHIFT)
593 #define SK_BT_NULLS_FIRST (INDOPTION_NULLS_FIRST << SK_BT_INDOPTION_SHIFT)
596 * prototypes for functions in nbtree.c (external entry points for btree)
598 extern Datum btbuild(PG_FUNCTION_ARGS);
599 extern Datum btbuildempty(PG_FUNCTION_ARGS);
600 extern Datum btinsert(PG_FUNCTION_ARGS);
601 extern Datum btbeginscan(PG_FUNCTION_ARGS);
602 extern Datum btgettuple(PG_FUNCTION_ARGS);
603 extern Datum btgetbitmap(PG_FUNCTION_ARGS);
604 extern Datum btrescan(PG_FUNCTION_ARGS);
605 extern Datum btendscan(PG_FUNCTION_ARGS);
606 extern Datum btmarkpos(PG_FUNCTION_ARGS);
607 extern Datum btrestrpos(PG_FUNCTION_ARGS);
608 extern Datum btbulkdelete(PG_FUNCTION_ARGS);
609 extern Datum btvacuumcleanup(PG_FUNCTION_ARGS);
610 extern Datum btoptions(PG_FUNCTION_ARGS);
613 * prototypes for functions in nbtinsert.c
615 extern bool _bt_doinsert(Relation rel, IndexTuple itup,
616 IndexUniqueCheck checkUnique, Relation heapRel);
617 extern Buffer _bt_getstackbuf(Relation rel, BTStack stack, int access);
618 extern void _bt_insert_parent(Relation rel, Buffer buf, Buffer rbuf,
619 BTStack stack, bool is_root, bool is_only);
622 * prototypes for functions in nbtpage.c
624 extern void _bt_initmetapage(Page page, BlockNumber rootbknum, uint32 level);
625 extern Buffer _bt_getroot(Relation rel, int access);
626 extern Buffer _bt_gettrueroot(Relation rel);
627 extern void _bt_checkpage(Relation rel, Buffer buf);
628 extern Buffer _bt_getbuf(Relation rel, BlockNumber blkno, int access);
629 extern Buffer _bt_relandgetbuf(Relation rel, Buffer obuf,
630 BlockNumber blkno, int access);
631 extern void _bt_relbuf(Relation rel, Buffer buf);
632 extern void _bt_pageinit(Page page, Size size);
633 extern bool _bt_page_recyclable(Page page);
634 extern void _bt_delitems_delete(Relation rel, Buffer buf,
635 OffsetNumber *itemnos, int nitems, Relation heapRel);
636 extern void _bt_delitems_vacuum(Relation rel, Buffer buf,
637 OffsetNumber *itemnos, int nitems, BlockNumber lastBlockVacuumed);
638 extern int _bt_pagedel(Relation rel, Buffer buf, BTStack stack);
641 * prototypes for functions in nbtsearch.c
643 extern BTStack _bt_search(Relation rel,
644 int keysz, ScanKey scankey, bool nextkey,
645 Buffer *bufP, int access);
646 extern Buffer _bt_moveright(Relation rel, Buffer buf, int keysz,
647 ScanKey scankey, bool nextkey, int access);
648 extern OffsetNumber _bt_binsrch(Relation rel, Buffer buf, int keysz,
649 ScanKey scankey, bool nextkey);
650 extern int32 _bt_compare(Relation rel, int keysz, ScanKey scankey,
651 Page page, OffsetNumber offnum);
652 extern bool _bt_first(IndexScanDesc scan, ScanDirection dir);
653 extern bool _bt_next(IndexScanDesc scan, ScanDirection dir);
654 extern Buffer _bt_get_endpoint(Relation rel, uint32 level, bool rightmost);
657 * prototypes for functions in nbtutils.c
659 extern ScanKey _bt_mkscankey(Relation rel, IndexTuple itup);
660 extern ScanKey _bt_mkscankey_nodata(Relation rel);
661 extern void _bt_freeskey(ScanKey skey);
662 extern void _bt_freestack(BTStack stack);
663 extern void _bt_preprocess_array_keys(IndexScanDesc scan);
664 extern void _bt_start_array_keys(IndexScanDesc scan, ScanDirection dir);
665 extern bool _bt_advance_array_keys(IndexScanDesc scan, ScanDirection dir);
666 extern void _bt_preprocess_keys(IndexScanDesc scan);
667 extern IndexTuple _bt_checkkeys(IndexScanDesc scan,
668 Page page, OffsetNumber offnum,
669 ScanDirection dir, bool *continuescan);
670 extern void _bt_killitems(IndexScanDesc scan, bool haveLock);
671 extern BTCycleId _bt_vacuum_cycleid(Relation rel);
672 extern BTCycleId _bt_start_vacuum(Relation rel);
673 extern void _bt_end_vacuum(Relation rel);
674 extern void _bt_end_vacuum_callback(int code, Datum arg);
675 extern Size BTreeShmemSize(void);
676 extern void BTreeShmemInit(void);
679 * prototypes for functions in nbtsort.c
681 typedef struct BTSpool BTSpool; /* opaque type known only within nbtsort.c */
683 extern BTSpool *_bt_spoolinit(Relation index, bool isunique, bool isdead);
684 extern void _bt_spooldestroy(BTSpool *btspool);
685 extern void _bt_spool(IndexTuple itup, BTSpool *btspool);
686 extern void _bt_leafbuild(BTSpool *btspool, BTSpool *spool2);
689 * prototypes for functions in nbtxlog.c
691 extern void btree_redo(XLogRecPtr lsn, XLogRecord *record);
692 extern void btree_desc(StringInfo buf, uint8 xl_info, char *rec);
693 extern void btree_xlog_startup(void);
694 extern void btree_xlog_cleanup(void);
695 extern bool btree_safe_restartpoint(void);
697 #endif /* NBTREE_H */