Line data Source code
1 : /*-------------------------------------------------------------------------
2 : *
3 : * tableam.h
4 : * POSTGRES table access method definitions.
5 : *
6 : *
7 : * Portions Copyright (c) 1996-2024, PostgreSQL Global Development Group
8 : * Portions Copyright (c) 1994, Regents of the University of California
9 : *
10 : * src/include/access/tableam.h
11 : *
12 : * NOTES
13 : * See tableam.sgml for higher level documentation.
14 : *
15 : *-------------------------------------------------------------------------
16 : */
17 : #ifndef TABLEAM_H
18 : #define TABLEAM_H
19 :
20 : #include "access/relscan.h"
21 : #include "access/sdir.h"
22 : #include "access/xact.h"
23 : #include "executor/tuptable.h"
24 : #include "storage/read_stream.h"
25 : #include "utils/rel.h"
26 : #include "utils/snapshot.h"
27 :
28 :
29 : #define DEFAULT_TABLE_ACCESS_METHOD "heap"
30 :
31 : /* GUCs */
32 : extern PGDLLIMPORT char *default_table_access_method;
33 : extern PGDLLIMPORT bool synchronize_seqscans;
34 :
35 :
36 : struct BulkInsertStateData;
37 : struct IndexInfo;
38 : struct SampleScanState;
39 : struct TBMIterateResult;
40 : struct VacuumParams;
41 : struct ValidateIndexState;
42 :
43 : /*
44 : * Bitmask values for the flags argument to the scan_begin callback.
45 : */
46 : typedef enum ScanOptions
47 : {
48 : /* one of SO_TYPE_* may be specified */
49 : SO_TYPE_SEQSCAN = 1 << 0,
50 : SO_TYPE_BITMAPSCAN = 1 << 1,
51 : SO_TYPE_SAMPLESCAN = 1 << 2,
52 : SO_TYPE_TIDSCAN = 1 << 3,
53 : SO_TYPE_TIDRANGESCAN = 1 << 4,
54 : SO_TYPE_ANALYZE = 1 << 5,
55 :
56 : /* several of SO_ALLOW_* may be specified */
57 : /* allow or disallow use of access strategy */
58 : SO_ALLOW_STRAT = 1 << 6,
59 : /* report location to syncscan logic? */
60 : SO_ALLOW_SYNC = 1 << 7,
61 : /* verify visibility page-at-a-time? */
62 : SO_ALLOW_PAGEMODE = 1 << 8,
63 :
64 : /* unregister snapshot at scan end? */
65 : SO_TEMP_SNAPSHOT = 1 << 9,
66 :
67 : /*
68 : * At the discretion of the table AM, bitmap table scans may be able to
69 : * skip fetching a block from the table if none of the table data is
70 : * needed. If table data may be needed, set SO_NEED_TUPLES.
71 : */
72 : SO_NEED_TUPLES = 1 << 10,
73 : } ScanOptions;
74 :
75 : /*
76 : * Result codes for table_{update,delete,lock_tuple}, and for visibility
77 : * routines inside table AMs.
78 : */
79 : typedef enum TM_Result
80 : {
81 : /*
82 : * Signals that the action succeeded (i.e. update/delete performed, lock
83 : * was acquired)
84 : */
85 : TM_Ok,
86 :
87 : /* The affected tuple wasn't visible to the relevant snapshot */
88 : TM_Invisible,
89 :
90 : /* The affected tuple was already modified by the calling backend */
91 : TM_SelfModified,
92 :
93 : /*
94 : * The affected tuple was updated by another transaction. This includes
95 : * the case where tuple was moved to another partition.
96 : */
97 : TM_Updated,
98 :
99 : /* The affected tuple was deleted by another transaction */
100 : TM_Deleted,
101 :
102 : /*
103 : * The affected tuple is currently being modified by another session. This
104 : * will only be returned if table_(update/delete/lock_tuple) are
105 : * instructed not to wait.
106 : */
107 : TM_BeingModified,
108 :
109 : /* lock couldn't be acquired, action skipped. Only used by lock_tuple */
110 : TM_WouldBlock,
111 : } TM_Result;
112 :
113 : /*
114 : * Result codes for table_update(..., update_indexes*..).
115 : * Used to determine which indexes to update.
116 : */
117 : typedef enum TU_UpdateIndexes
118 : {
119 : /* No indexed columns were updated (incl. TID addressing of tuple) */
120 : TU_None,
121 :
122 : /* A non-summarizing indexed column was updated, or the TID has changed */
123 : TU_All,
124 :
125 : /* Only summarized columns were updated, TID is unchanged */
126 : TU_Summarizing,
127 : } TU_UpdateIndexes;
128 :
129 : /*
130 : * When table_tuple_update, table_tuple_delete, or table_tuple_lock fail
131 : * because the target tuple is already outdated, they fill in this struct to
132 : * provide information to the caller about what happened.
133 : *
134 : * ctid is the target's ctid link: it is the same as the target's TID if the
135 : * target was deleted, or the location of the replacement tuple if the target
136 : * was updated.
137 : *
138 : * xmax is the outdating transaction's XID. If the caller wants to visit the
139 : * replacement tuple, it must check that this matches before believing the
140 : * replacement is really a match.
141 : *
142 : * cmax is the outdating command's CID, but only when the failure code is
143 : * TM_SelfModified (i.e., something in the current transaction outdated the
144 : * tuple); otherwise cmax is zero. (We make this restriction because
145 : * HeapTupleHeaderGetCmax doesn't work for tuples outdated in other
146 : * transactions.)
147 : */
148 : typedef struct TM_FailureData
149 : {
150 : ItemPointerData ctid;
151 : TransactionId xmax;
152 : CommandId cmax;
153 : bool traversed;
154 : } TM_FailureData;
155 :
156 : /*
157 : * State used when calling table_index_delete_tuples().
158 : *
159 : * Represents the status of table tuples, referenced by table TID and taken by
160 : * index AM from index tuples. State consists of high level parameters of the
161 : * deletion operation, plus two mutable palloc()'d arrays for information
162 : * about the status of individual table tuples. These are conceptually one
163 : * single array. Using two arrays keeps the TM_IndexDelete struct small,
164 : * which makes sorting the first array (the deltids array) fast.
165 : *
166 : * Some index AM callers perform simple index tuple deletion (by specifying
167 : * bottomup = false), and include only known-dead deltids. These known-dead
168 : * entries are all marked knowndeletable = true directly (typically these are
169 : * TIDs from LP_DEAD-marked index tuples), but that isn't strictly required.
170 : *
171 : * Callers that specify bottomup = true are "bottom-up index deletion"
172 : * callers. The considerations for the tableam are more subtle with these
173 : * callers because they ask the tableam to perform highly speculative work,
174 : * and might only expect the tableam to check a small fraction of all entries.
175 : * Caller is not allowed to specify knowndeletable = true for any entry
176 : * because everything is highly speculative. Bottom-up caller provides
177 : * context and hints to tableam -- see comments below for details on how index
178 : * AMs and tableams should coordinate during bottom-up index deletion.
179 : *
180 : * Simple index deletion callers may ask the tableam to perform speculative
181 : * work, too. This is a little like bottom-up deletion, but not too much.
182 : * The tableam will only perform speculative work when it's practically free
183 : * to do so in passing for simple deletion caller (while always performing
184 : * whatever work is needed to enable knowndeletable/LP_DEAD index tuples to
185 : * be deleted within index AM). This is the real reason why it's possible for
186 : * simple index deletion caller to specify knowndeletable = false up front
187 : * (this means "check if it's possible for me to delete corresponding index
188 : * tuple when it's cheap to do so in passing"). The index AM should only
189 : * include "extra" entries for index tuples whose TIDs point to a table block
190 : * that tableam is expected to have to visit anyway (in the event of a block
191 : * orientated tableam). The tableam isn't strictly obligated to check these
192 : * "extra" TIDs, but a block-based AM should always manage to do so in
193 : * practice.
194 : *
195 : * The final contents of the deltids/status arrays are interesting to callers
196 : * that ask tableam to perform speculative work (i.e. when _any_ items have
197 : * knowndeletable set to false up front). These index AM callers will
198 : * naturally need to consult final state to determine which index tuples are
199 : * in fact deletable.
200 : *
201 : * The index AM can keep track of which index tuple relates to which deltid by
202 : * setting idxoffnum (and/or relying on each entry being uniquely identifiable
203 : * using tid), which is important when the final contents of the array will
204 : * need to be interpreted -- the array can shrink from initial size after
205 : * tableam processing and/or have entries in a new order (tableam may sort
206 : * deltids array for its own reasons). Bottom-up callers may find that final
207 : * ndeltids is 0 on return from call to tableam, in which case no index tuple
208 : * deletions are possible. Simple deletion callers can rely on any entries
209 : * they know to be deletable appearing in the final array as deletable.
210 : */
211 : typedef struct TM_IndexDelete
212 : {
213 : ItemPointerData tid; /* table TID from index tuple */
214 : int16 id; /* Offset into TM_IndexStatus array */
215 : } TM_IndexDelete;
216 :
217 : typedef struct TM_IndexStatus
218 : {
219 : OffsetNumber idxoffnum; /* Index am page offset number */
220 : bool knowndeletable; /* Currently known to be deletable? */
221 :
222 : /* Bottom-up index deletion specific fields follow */
223 : bool promising; /* Promising (duplicate) index tuple? */
224 : int16 freespace; /* Space freed in index if deleted */
225 : } TM_IndexStatus;
226 :
227 : /*
228 : * Index AM/tableam coordination is central to the design of bottom-up index
229 : * deletion. The index AM provides hints about where to look to the tableam
230 : * by marking some entries as "promising". Index AM does this with duplicate
231 : * index tuples that are strongly suspected to be old versions left behind by
232 : * UPDATEs that did not logically modify indexed values. Index AM may find it
233 : * helpful to only mark entries as promising when they're thought to have been
234 : * affected by such an UPDATE in the recent past.
235 : *
236 : * Bottom-up index deletion casts a wide net at first, usually by including
237 : * all TIDs on a target index page. It is up to the tableam to worry about
238 : * the cost of checking transaction status information. The tableam is in
239 : * control, but needs careful guidance from the index AM. Index AM requests
240 : * that bottomupfreespace target be met, while tableam measures progress
241 : * towards that goal by tallying the per-entry freespace value for known
242 : * deletable entries. (All !bottomup callers can just set these space related
243 : * fields to zero.)
244 : */
245 : typedef struct TM_IndexDeleteOp
246 : {
247 : Relation irel; /* Target index relation */
248 : BlockNumber iblknum; /* Index block number (for error reports) */
249 : bool bottomup; /* Bottom-up (not simple) deletion? */
250 : int bottomupfreespace; /* Bottom-up space target */
251 :
252 : /* Mutable per-TID information follows (index AM initializes entries) */
253 : int ndeltids; /* Current # of deltids/status elements */
254 : TM_IndexDelete *deltids;
255 : TM_IndexStatus *status;
256 : } TM_IndexDeleteOp;
257 :
258 : /* "options" flag bits for table_tuple_insert */
259 : /* TABLE_INSERT_SKIP_WAL was 0x0001; RelationNeedsWAL() now governs */
260 : #define TABLE_INSERT_SKIP_FSM 0x0002
261 : #define TABLE_INSERT_FROZEN 0x0004
262 : #define TABLE_INSERT_NO_LOGICAL 0x0008
263 :
264 : /* flag bits for table_tuple_lock */
265 : /* Follow tuples whose update is in progress if lock modes don't conflict */
266 : #define TUPLE_LOCK_FLAG_LOCK_UPDATE_IN_PROGRESS (1 << 0)
267 : /* Follow update chain and lock latest version of tuple */
268 : #define TUPLE_LOCK_FLAG_FIND_LAST_VERSION (1 << 1)
269 :
270 :
271 : /* Typedef for callback function for table_index_build_scan */
272 : typedef void (*IndexBuildCallback) (Relation index,
273 : ItemPointer tid,
274 : Datum *values,
275 : bool *isnull,
276 : bool tupleIsAlive,
277 : void *state);
278 :
279 : /*
280 : * API struct for a table AM. Note this must be allocated in a
281 : * server-lifetime manner, typically as a static const struct, which then gets
282 : * returned by FormData_pg_am.amhandler.
283 : *
284 : * In most cases it's not appropriate to call the callbacks directly, use the
285 : * table_* wrapper functions instead.
286 : *
287 : * GetTableAmRoutine() asserts that required callbacks are filled in, remember
288 : * to update when adding a callback.
289 : */
290 : typedef struct TableAmRoutine
291 : {
292 : /* this must be set to T_TableAmRoutine */
293 : NodeTag type;
294 :
295 :
296 : /* ------------------------------------------------------------------------
297 : * Slot related callbacks.
298 : * ------------------------------------------------------------------------
299 : */
300 :
301 : /*
302 : * Return slot implementation suitable for storing a tuple of this AM.
303 : */
304 : const TupleTableSlotOps *(*slot_callbacks) (Relation rel);
305 :
306 :
307 : /* ------------------------------------------------------------------------
308 : * Table scan callbacks.
309 : * ------------------------------------------------------------------------
310 : */
311 :
312 : /*
313 : * Start a scan of `rel`. The callback has to return a TableScanDesc,
314 : * which will typically be embedded in a larger, AM specific, struct.
315 : *
316 : * If nkeys != 0, the results need to be filtered by those scan keys.
317 : *
318 : * pscan, if not NULL, will have already been initialized with
319 : * parallelscan_initialize(), and has to be for the same relation. Will
320 : * only be set coming from table_beginscan_parallel().
321 : *
322 : * `flags` is a bitmask indicating the type of scan (ScanOptions's
323 : * SO_TYPE_*, currently only one may be specified), options controlling
324 : * the scan's behaviour (ScanOptions's SO_ALLOW_*, several may be
325 : * specified, an AM may ignore unsupported ones) and whether the snapshot
326 : * needs to be deallocated at scan_end (ScanOptions's SO_TEMP_SNAPSHOT).
327 : */
328 : TableScanDesc (*scan_begin) (Relation rel,
329 : Snapshot snapshot,
330 : int nkeys, struct ScanKeyData *key,
331 : ParallelTableScanDesc pscan,
332 : uint32 flags);
333 :
334 : /*
335 : * Release resources and deallocate scan. If TableScanDesc.temp_snap,
336 : * TableScanDesc.rs_snapshot needs to be unregistered.
337 : */
338 : void (*scan_end) (TableScanDesc scan);
339 :
340 : /*
341 : * Restart relation scan. If set_params is set to true, allow_{strat,
342 : * sync, pagemode} (see scan_begin) changes should be taken into account.
343 : */
344 : void (*scan_rescan) (TableScanDesc scan, struct ScanKeyData *key,
345 : bool set_params, bool allow_strat,
346 : bool allow_sync, bool allow_pagemode);
347 :
348 : /*
349 : * Return next tuple from `scan`, store in slot.
350 : */
351 : bool (*scan_getnextslot) (TableScanDesc scan,
352 : ScanDirection direction,
353 : TupleTableSlot *slot);
354 :
355 : /*-----------
356 : * Optional functions to provide scanning for ranges of ItemPointers.
357 : * Implementations must either provide both of these functions, or neither
358 : * of them.
359 : *
360 : * Implementations of scan_set_tidrange must themselves handle
361 : * ItemPointers of any value. i.e, they must handle each of the following:
362 : *
363 : * 1) mintid or maxtid is beyond the end of the table; and
364 : * 2) mintid is above maxtid; and
365 : * 3) item offset for mintid or maxtid is beyond the maximum offset
366 : * allowed by the AM.
367 : *
368 : * Implementations can assume that scan_set_tidrange is always called
369 : * before scan_getnextslot_tidrange or after scan_rescan and before any
370 : * further calls to scan_getnextslot_tidrange.
371 : */
372 : void (*scan_set_tidrange) (TableScanDesc scan,
373 : ItemPointer mintid,
374 : ItemPointer maxtid);
375 :
376 : /*
377 : * Return next tuple from `scan` that's in the range of TIDs defined by
378 : * scan_set_tidrange.
379 : */
380 : bool (*scan_getnextslot_tidrange) (TableScanDesc scan,
381 : ScanDirection direction,
382 : TupleTableSlot *slot);
383 :
384 : /* ------------------------------------------------------------------------
385 : * Parallel table scan related functions.
386 : * ------------------------------------------------------------------------
387 : */
388 :
389 : /*
390 : * Estimate the size of shared memory needed for a parallel scan of this
391 : * relation. The snapshot does not need to be accounted for.
392 : */
393 : Size (*parallelscan_estimate) (Relation rel);
394 :
395 : /*
396 : * Initialize ParallelTableScanDesc for a parallel scan of this relation.
397 : * `pscan` will be sized according to parallelscan_estimate() for the same
398 : * relation.
399 : */
400 : Size (*parallelscan_initialize) (Relation rel,
401 : ParallelTableScanDesc pscan);
402 :
403 : /*
404 : * Reinitialize `pscan` for a new scan. `rel` will be the same relation as
405 : * when `pscan` was initialized by parallelscan_initialize.
406 : */
407 : void (*parallelscan_reinitialize) (Relation rel,
408 : ParallelTableScanDesc pscan);
409 :
410 :
411 : /* ------------------------------------------------------------------------
412 : * Index Scan Callbacks
413 : * ------------------------------------------------------------------------
414 : */
415 :
416 : /*
417 : * Prepare to fetch tuples from the relation, as needed when fetching
418 : * tuples for an index scan. The callback has to return an
419 : * IndexFetchTableData, which the AM will typically embed in a larger
420 : * structure with additional information.
421 : *
422 : * Tuples for an index scan can then be fetched via index_fetch_tuple.
423 : */
424 : struct IndexFetchTableData *(*index_fetch_begin) (Relation rel);
425 :
426 : /*
427 : * Reset index fetch. Typically this will release cross index fetch
428 : * resources held in IndexFetchTableData.
429 : */
430 : void (*index_fetch_reset) (struct IndexFetchTableData *data);
431 :
432 : /*
433 : * Release resources and deallocate index fetch.
434 : */
435 : void (*index_fetch_end) (struct IndexFetchTableData *data);
436 :
437 : /*
438 : * Fetch tuple at `tid` into `slot`, after doing a visibility test
439 : * according to `snapshot`. If a tuple was found and passed the visibility
440 : * test, return true, false otherwise.
441 : *
442 : * Note that AMs that do not necessarily update indexes when indexed
443 : * columns do not change, need to return the current/correct version of
444 : * the tuple that is visible to the snapshot, even if the tid points to an
445 : * older version of the tuple.
446 : *
447 : * *call_again is false on the first call to index_fetch_tuple for a tid.
448 : * If there potentially is another tuple matching the tid, *call_again
449 : * needs to be set to true by index_fetch_tuple, signaling to the caller
450 : * that index_fetch_tuple should be called again for the same tid.
451 : *
452 : * *all_dead, if all_dead is not NULL, should be set to true by
453 : * index_fetch_tuple iff it is guaranteed that no backend needs to see
454 : * that tuple. Index AMs can use that to avoid returning that tid in
455 : * future searches.
456 : */
457 : bool (*index_fetch_tuple) (struct IndexFetchTableData *scan,
458 : ItemPointer tid,
459 : Snapshot snapshot,
460 : TupleTableSlot *slot,
461 : bool *call_again, bool *all_dead);
462 :
463 :
464 : /* ------------------------------------------------------------------------
465 : * Callbacks for non-modifying operations on individual tuples
466 : * ------------------------------------------------------------------------
467 : */
468 :
469 : /*
470 : * Fetch tuple at `tid` into `slot`, after doing a visibility test
471 : * according to `snapshot`. If a tuple was found and passed the visibility
472 : * test, returns true, false otherwise.
473 : */
474 : bool (*tuple_fetch_row_version) (Relation rel,
475 : ItemPointer tid,
476 : Snapshot snapshot,
477 : TupleTableSlot *slot);
478 :
479 : /*
480 : * Is tid valid for a scan of this relation.
481 : */
482 : bool (*tuple_tid_valid) (TableScanDesc scan,
483 : ItemPointer tid);
484 :
485 : /*
486 : * Return the latest version of the tuple at `tid`, by updating `tid` to
487 : * point at the newest version.
488 : */
489 : void (*tuple_get_latest_tid) (TableScanDesc scan,
490 : ItemPointer tid);
491 :
492 : /*
493 : * Does the tuple in `slot` satisfy `snapshot`? The slot needs to be of
494 : * the appropriate type for the AM.
495 : */
496 : bool (*tuple_satisfies_snapshot) (Relation rel,
497 : TupleTableSlot *slot,
498 : Snapshot snapshot);
499 :
500 : /* see table_index_delete_tuples() */
501 : TransactionId (*index_delete_tuples) (Relation rel,
502 : TM_IndexDeleteOp *delstate);
503 :
504 :
505 : /* ------------------------------------------------------------------------
506 : * Manipulations of physical tuples.
507 : * ------------------------------------------------------------------------
508 : */
509 :
510 : /* see table_tuple_insert() for reference about parameters */
511 : void (*tuple_insert) (Relation rel, TupleTableSlot *slot,
512 : CommandId cid, int options,
513 : struct BulkInsertStateData *bistate);
514 :
515 : /* see table_tuple_insert_speculative() for reference about parameters */
516 : void (*tuple_insert_speculative) (Relation rel,
517 : TupleTableSlot *slot,
518 : CommandId cid,
519 : int options,
520 : struct BulkInsertStateData *bistate,
521 : uint32 specToken);
522 :
523 : /* see table_tuple_complete_speculative() for reference about parameters */
524 : void (*tuple_complete_speculative) (Relation rel,
525 : TupleTableSlot *slot,
526 : uint32 specToken,
527 : bool succeeded);
528 :
529 : /* see table_multi_insert() for reference about parameters */
530 : void (*multi_insert) (Relation rel, TupleTableSlot **slots, int nslots,
531 : CommandId cid, int options, struct BulkInsertStateData *bistate);
532 :
533 : /* see table_tuple_delete() for reference about parameters */
534 : TM_Result (*tuple_delete) (Relation rel,
535 : ItemPointer tid,
536 : CommandId cid,
537 : Snapshot snapshot,
538 : Snapshot crosscheck,
539 : bool wait,
540 : TM_FailureData *tmfd,
541 : bool changingPart);
542 :
543 : /* see table_tuple_update() for reference about parameters */
544 : TM_Result (*tuple_update) (Relation rel,
545 : ItemPointer otid,
546 : TupleTableSlot *slot,
547 : CommandId cid,
548 : Snapshot snapshot,
549 : Snapshot crosscheck,
550 : bool wait,
551 : TM_FailureData *tmfd,
552 : LockTupleMode *lockmode,
553 : TU_UpdateIndexes *update_indexes);
554 :
555 : /* see table_tuple_lock() for reference about parameters */
556 : TM_Result (*tuple_lock) (Relation rel,
557 : ItemPointer tid,
558 : Snapshot snapshot,
559 : TupleTableSlot *slot,
560 : CommandId cid,
561 : LockTupleMode mode,
562 : LockWaitPolicy wait_policy,
563 : uint8 flags,
564 : TM_FailureData *tmfd);
565 :
566 : /*
567 : * Perform operations necessary to complete insertions made via
568 : * tuple_insert and multi_insert with a BulkInsertState specified. In-tree
569 : * access methods ceased to use this.
570 : *
571 : * Typically callers of tuple_insert and multi_insert will just pass all
572 : * the flags that apply to them, and each AM has to decide which of them
573 : * make sense for it, and then only take actions in finish_bulk_insert for
574 : * those flags, and ignore others.
575 : *
576 : * Optional callback.
577 : */
578 : void (*finish_bulk_insert) (Relation rel, int options);
579 :
580 :
581 : /* ------------------------------------------------------------------------
582 : * DDL related functionality.
583 : * ------------------------------------------------------------------------
584 : */
585 :
586 : /*
587 : * This callback needs to create new relation storage for `rel`, with
588 : * appropriate durability behaviour for `persistence`.
589 : *
590 : * Note that only the subset of the relcache filled by
591 : * RelationBuildLocalRelation() can be relied upon and that the relation's
592 : * catalog entries will either not yet exist (new relation), or will still
593 : * reference the old relfilelocator.
594 : *
595 : * As output *freezeXid, *minmulti must be set to the values appropriate
596 : * for pg_class.{relfrozenxid, relminmxid}. For AMs that don't need those
597 : * fields to be filled they can be set to InvalidTransactionId and
598 : * InvalidMultiXactId, respectively.
599 : *
600 : * See also table_relation_set_new_filelocator().
601 : */
602 : void (*relation_set_new_filelocator) (Relation rel,
603 : const RelFileLocator *newrlocator,
604 : char persistence,
605 : TransactionId *freezeXid,
606 : MultiXactId *minmulti);
607 :
608 : /*
609 : * This callback needs to remove all contents from `rel`'s current
610 : * relfilelocator. No provisions for transactional behaviour need to be
611 : * made. Often this can be implemented by truncating the underlying
612 : * storage to its minimal size.
613 : *
614 : * See also table_relation_nontransactional_truncate().
615 : */
616 : void (*relation_nontransactional_truncate) (Relation rel);
617 :
618 : /*
619 : * See table_relation_copy_data().
620 : *
621 : * This can typically be implemented by directly copying the underlying
622 : * storage, unless it contains references to the tablespace internally.
623 : */
624 : void (*relation_copy_data) (Relation rel,
625 : const RelFileLocator *newrlocator);
626 :
627 : /* See table_relation_copy_for_cluster() */
628 : void (*relation_copy_for_cluster) (Relation OldTable,
629 : Relation NewTable,
630 : Relation OldIndex,
631 : bool use_sort,
632 : TransactionId OldestXmin,
633 : TransactionId *xid_cutoff,
634 : MultiXactId *multi_cutoff,
635 : double *num_tuples,
636 : double *tups_vacuumed,
637 : double *tups_recently_dead);
638 :
639 : /*
640 : * React to VACUUM command on the relation. The VACUUM can be triggered by
641 : * a user or by autovacuum. The specific actions performed by the AM will
642 : * depend heavily on the individual AM.
643 : *
644 : * On entry a transaction is already established, and the relation is
645 : * locked with a ShareUpdateExclusive lock.
646 : *
647 : * Note that neither VACUUM FULL (and CLUSTER), nor ANALYZE go through
648 : * this routine, even if (for ANALYZE) it is part of the same VACUUM
649 : * command.
650 : *
651 : * There probably, in the future, needs to be a separate callback to
652 : * integrate with autovacuum's scheduling.
653 : */
654 : void (*relation_vacuum) (Relation rel,
655 : struct VacuumParams *params,
656 : BufferAccessStrategy bstrategy);
657 :
658 : /*
659 : * Prepare to analyze block `blockno` of `scan`. The scan has been started
660 : * with table_beginscan_analyze(). See also
661 : * table_scan_analyze_next_block().
662 : *
663 : * The callback may acquire resources like locks that are held until
664 : * table_scan_analyze_next_tuple() returns false. It e.g. can make sense
665 : * to hold a lock until all tuples on a block have been analyzed by
666 : * scan_analyze_next_tuple.
667 : *
668 : * The callback can return false if the block is not suitable for
669 : * sampling, e.g. because it's a metapage that could never contain tuples.
670 : *
671 : * XXX: This obviously is primarily suited for block-based AMs. It's not
672 : * clear what a good interface for non block based AMs would be, so there
673 : * isn't one yet.
674 : */
675 : bool (*scan_analyze_next_block) (TableScanDesc scan,
676 : ReadStream *stream);
677 :
678 : /*
679 : * See table_scan_analyze_next_tuple().
680 : *
681 : * Not every AM might have a meaningful concept of dead rows, in which
682 : * case it's OK to not increment *deadrows - but note that that may
683 : * influence autovacuum scheduling (see comment for relation_vacuum
684 : * callback).
685 : */
686 : bool (*scan_analyze_next_tuple) (TableScanDesc scan,
687 : TransactionId OldestXmin,
688 : double *liverows,
689 : double *deadrows,
690 : TupleTableSlot *slot);
691 :
692 : /* see table_index_build_range_scan for reference about parameters */
693 : double (*index_build_range_scan) (Relation table_rel,
694 : Relation index_rel,
695 : struct IndexInfo *index_info,
696 : bool allow_sync,
697 : bool anyvisible,
698 : bool progress,
699 : BlockNumber start_blockno,
700 : BlockNumber numblocks,
701 : IndexBuildCallback callback,
702 : void *callback_state,
703 : TableScanDesc scan);
704 :
705 : /* see table_index_validate_scan for reference about parameters */
706 : void (*index_validate_scan) (Relation table_rel,
707 : Relation index_rel,
708 : struct IndexInfo *index_info,
709 : Snapshot snapshot,
710 : struct ValidateIndexState *state);
711 :
712 :
713 : /* ------------------------------------------------------------------------
714 : * Miscellaneous functions.
715 : * ------------------------------------------------------------------------
716 : */
717 :
718 : /*
719 : * See table_relation_size().
720 : *
721 : * Note that currently a few callers use the MAIN_FORKNUM size to figure
722 : * out the range of potentially interesting blocks (brin, analyze). It's
723 : * probable that we'll need to revise the interface for those at some
724 : * point.
725 : */
726 : uint64 (*relation_size) (Relation rel, ForkNumber forkNumber);
727 :
728 :
729 : /*
730 : * This callback should return true if the relation requires a TOAST table
731 : * and false if it does not. It may wish to examine the relation's tuple
732 : * descriptor before making a decision, but if it uses some other method
733 : * of storing large values (or if it does not support them) it can simply
734 : * return false.
735 : */
736 : bool (*relation_needs_toast_table) (Relation rel);
737 :
738 : /*
739 : * This callback should return the OID of the table AM that implements
740 : * TOAST tables for this AM. If the relation_needs_toast_table callback
741 : * always returns false, this callback is not required.
742 : */
743 : Oid (*relation_toast_am) (Relation rel);
744 :
745 : /*
746 : * This callback is invoked when detoasting a value stored in a toast
747 : * table implemented by this AM. See table_relation_fetch_toast_slice()
748 : * for more details.
749 : */
750 : void (*relation_fetch_toast_slice) (Relation toastrel, Oid valueid,
751 : int32 attrsize,
752 : int32 sliceoffset,
753 : int32 slicelength,
754 : struct varlena *result);
755 :
756 :
757 : /* ------------------------------------------------------------------------
758 : * Planner related functions.
759 : * ------------------------------------------------------------------------
760 : */
761 :
762 : /*
763 : * See table_relation_estimate_size().
764 : *
765 : * While block oriented, it shouldn't be too hard for an AM that doesn't
766 : * internally use blocks to convert into a usable representation.
767 : *
768 : * This differs from the relation_size callback by returning size
769 : * estimates (both relation size and tuple count) for planning purposes,
770 : * rather than returning a currently correct estimate.
771 : */
772 : void (*relation_estimate_size) (Relation rel, int32 *attr_widths,
773 : BlockNumber *pages, double *tuples,
774 : double *allvisfrac);
775 :
776 :
777 : /* ------------------------------------------------------------------------
778 : * Executor related functions.
779 : * ------------------------------------------------------------------------
780 : */
781 :
782 : /*
783 : * Prepare to fetch / check / return tuples from `tbmres->blockno` as part
784 : * of a bitmap table scan. `scan` was started via table_beginscan_bm().
785 : * Return false if there are no tuples to be found on the page, true
786 : * otherwise.
787 : *
788 : * This will typically read and pin the target block, and do the necessary
789 : * work to allow scan_bitmap_next_tuple() to return tuples (e.g. it might
790 : * make sense to perform tuple visibility checks at this time). For some
791 : * AMs it will make more sense to do all the work referencing `tbmres`
792 : * contents here, for others it might be better to defer more work to
793 : * scan_bitmap_next_tuple.
794 : *
795 : * If `tbmres->blockno` is -1, this is a lossy scan and all visible tuples
796 : * on the page have to be returned, otherwise the tuples at offsets in
797 : * `tbmres->offsets` need to be returned.
798 : *
799 : * XXX: Currently this may only be implemented if the AM uses md.c as its
800 : * storage manager, and uses ItemPointer->ip_blkid in a manner that maps
801 : * blockids directly to the underlying storage. nodeBitmapHeapscan.c
802 : * performs prefetching directly using that interface. This probably
803 : * needs to be rectified at a later point.
804 : *
805 : * XXX: Currently this may only be implemented if the AM uses the
806 : * visibilitymap, as nodeBitmapHeapscan.c unconditionally accesses it to
807 : * perform prefetching. This probably needs to be rectified at a later
808 : * point.
809 : *
810 : * Optional callback, but either both scan_bitmap_next_block and
811 : * scan_bitmap_next_tuple need to exist, or neither.
812 : */
813 : bool (*scan_bitmap_next_block) (TableScanDesc scan,
814 : struct TBMIterateResult *tbmres);
815 :
816 : /*
817 : * Fetch the next tuple of a bitmap table scan into `slot` and return true
818 : * if a visible tuple was found, false otherwise.
819 : *
820 : * For some AMs it will make more sense to do all the work referencing
821 : * `tbmres` contents in scan_bitmap_next_block, for others it might be
822 : * better to defer more work to this callback.
823 : *
824 : * Optional callback, but either both scan_bitmap_next_block and
825 : * scan_bitmap_next_tuple need to exist, or neither.
826 : */
827 : bool (*scan_bitmap_next_tuple) (TableScanDesc scan,
828 : struct TBMIterateResult *tbmres,
829 : TupleTableSlot *slot);
830 :
831 : /*
832 : * Prepare to fetch tuples from the next block in a sample scan. Return
833 : * false if the sample scan is finished, true otherwise. `scan` was
834 : * started via table_beginscan_sampling().
835 : *
836 : * Typically this will first determine the target block by calling the
837 : * TsmRoutine's NextSampleBlock() callback if not NULL, or alternatively
838 : * perform a sequential scan over all blocks. The determined block is
839 : * then typically read and pinned.
840 : *
841 : * As the TsmRoutine interface is block based, a block needs to be passed
842 : * to NextSampleBlock(). If that's not appropriate for an AM, it
843 : * internally needs to perform mapping between the internal and a block
844 : * based representation.
845 : *
846 : * Note that it's not acceptable to hold deadlock prone resources such as
847 : * lwlocks until scan_sample_next_tuple() has exhausted the tuples on the
848 : * block - the tuple is likely to be returned to an upper query node, and
849 : * the next call could be off a long while. Holding buffer pins and such
850 : * is obviously OK.
851 : *
852 : * Currently it is required to implement this interface, as there's no
853 : * alternative way (contrary e.g. to bitmap scans) to implement sample
854 : * scans. If infeasible to implement, the AM may raise an error.
855 : */
856 : bool (*scan_sample_next_block) (TableScanDesc scan,
857 : struct SampleScanState *scanstate);
858 :
859 : /*
860 : * This callback, only called after scan_sample_next_block has returned
861 : * true, should determine the next tuple to be returned from the selected
862 : * block using the TsmRoutine's NextSampleTuple() callback.
863 : *
864 : * The callback needs to perform visibility checks, and only return
865 : * visible tuples. That obviously can mean calling NextSampleTuple()
866 : * multiple times.
867 : *
868 : * The TsmRoutine interface assumes that there's a maximum offset on a
869 : * given page, so if that doesn't apply to an AM, it needs to emulate that
870 : * assumption somehow.
871 : */
872 : bool (*scan_sample_next_tuple) (TableScanDesc scan,
873 : struct SampleScanState *scanstate,
874 : TupleTableSlot *slot);
875 :
876 : } TableAmRoutine;
877 :
878 :
879 : /* ----------------------------------------------------------------------------
880 : * Slot functions.
881 : * ----------------------------------------------------------------------------
882 : */
883 :
884 : /*
885 : * Returns slot callbacks suitable for holding tuples of the appropriate type
886 : * for the relation. Works for tables, views, foreign tables and partitioned
887 : * tables.
888 : */
889 : extern const TupleTableSlotOps *table_slot_callbacks(Relation relation);
890 :
891 : /*
892 : * Returns slot using the callbacks returned by table_slot_callbacks(), and
893 : * registers it on *reglist.
894 : */
895 : extern TupleTableSlot *table_slot_create(Relation relation, List **reglist);
896 :
897 :
898 : /* ----------------------------------------------------------------------------
899 : * Table scan functions.
900 : * ----------------------------------------------------------------------------
901 : */
902 :
903 : /*
904 : * Start a scan of `rel`. Returned tuples pass a visibility test of
905 : * `snapshot`, and if nkeys != 0, the results are filtered by those scan keys.
906 : */
907 : static inline TableScanDesc
908 169556 : table_beginscan(Relation rel, Snapshot snapshot,
909 : int nkeys, struct ScanKeyData *key)
910 : {
911 169556 : uint32 flags = SO_TYPE_SEQSCAN |
912 : SO_ALLOW_STRAT | SO_ALLOW_SYNC | SO_ALLOW_PAGEMODE;
913 :
914 169556 : return rel->rd_tableam->scan_begin(rel, snapshot, nkeys, key, NULL, flags);
915 : }
916 :
917 : /*
918 : * Like table_beginscan(), but for scanning catalog. It'll automatically use a
919 : * snapshot appropriate for scanning catalog relations.
920 : */
921 : extern TableScanDesc table_beginscan_catalog(Relation relation, int nkeys,
922 : struct ScanKeyData *key);
923 :
924 : /*
925 : * Like table_beginscan(), but table_beginscan_strat() offers an extended API
926 : * that lets the caller control whether a nondefault buffer access strategy
927 : * can be used, and whether syncscan can be chosen (possibly resulting in the
928 : * scan not starting from block zero). Both of these default to true with
929 : * plain table_beginscan.
930 : */
931 : static inline TableScanDesc
932 317988 : table_beginscan_strat(Relation rel, Snapshot snapshot,
933 : int nkeys, struct ScanKeyData *key,
934 : bool allow_strat, bool allow_sync)
935 : {
936 317988 : uint32 flags = SO_TYPE_SEQSCAN | SO_ALLOW_PAGEMODE;
937 :
938 317988 : if (allow_strat)
939 317988 : flags |= SO_ALLOW_STRAT;
940 317988 : if (allow_sync)
941 45646 : flags |= SO_ALLOW_SYNC;
942 :
943 317988 : return rel->rd_tableam->scan_begin(rel, snapshot, nkeys, key, NULL, flags);
944 : }
945 :
946 : /*
947 : * table_beginscan_bm is an alternative entry point for setting up a
948 : * TableScanDesc for a bitmap heap scan. Although that scan technology is
949 : * really quite unlike a standard seqscan, there is just enough commonality to
950 : * make it worth using the same data structure.
951 : */
952 : static inline TableScanDesc
953 13570 : table_beginscan_bm(Relation rel, Snapshot snapshot,
954 : int nkeys, struct ScanKeyData *key, bool need_tuple)
955 : {
956 13570 : uint32 flags = SO_TYPE_BITMAPSCAN | SO_ALLOW_PAGEMODE;
957 :
958 13570 : if (need_tuple)
959 10706 : flags |= SO_NEED_TUPLES;
960 :
961 13570 : return rel->rd_tableam->scan_begin(rel, snapshot, nkeys, key, NULL, flags);
962 : }
963 :
964 : /*
965 : * table_beginscan_sampling is an alternative entry point for setting up a
966 : * TableScanDesc for a TABLESAMPLE scan. As with bitmap scans, it's worth
967 : * using the same data structure although the behavior is rather different.
968 : * In addition to the options offered by table_beginscan_strat, this call
969 : * also allows control of whether page-mode visibility checking is used.
970 : */
971 : static inline TableScanDesc
972 146 : table_beginscan_sampling(Relation rel, Snapshot snapshot,
973 : int nkeys, struct ScanKeyData *key,
974 : bool allow_strat, bool allow_sync,
975 : bool allow_pagemode)
976 : {
977 146 : uint32 flags = SO_TYPE_SAMPLESCAN;
978 :
979 146 : if (allow_strat)
980 134 : flags |= SO_ALLOW_STRAT;
981 146 : if (allow_sync)
982 66 : flags |= SO_ALLOW_SYNC;
983 146 : if (allow_pagemode)
984 122 : flags |= SO_ALLOW_PAGEMODE;
985 :
986 146 : return rel->rd_tableam->scan_begin(rel, snapshot, nkeys, key, NULL, flags);
987 : }
988 :
989 : /*
990 : * table_beginscan_tid is an alternative entry point for setting up a
991 : * TableScanDesc for a Tid scan. As with bitmap scans, it's worth using
992 : * the same data structure although the behavior is rather different.
993 : */
994 : static inline TableScanDesc
995 682 : table_beginscan_tid(Relation rel, Snapshot snapshot)
996 : {
997 682 : uint32 flags = SO_TYPE_TIDSCAN;
998 :
999 682 : return rel->rd_tableam->scan_begin(rel, snapshot, 0, NULL, NULL, flags);
1000 : }
1001 :
1002 : /*
1003 : * table_beginscan_analyze is an alternative entry point for setting up a
1004 : * TableScanDesc for an ANALYZE scan. As with bitmap scans, it's worth using
1005 : * the same data structure although the behavior is rather different.
1006 : */
1007 : static inline TableScanDesc
1008 13454 : table_beginscan_analyze(Relation rel)
1009 : {
1010 13454 : uint32 flags = SO_TYPE_ANALYZE;
1011 :
1012 13454 : return rel->rd_tableam->scan_begin(rel, NULL, 0, NULL, NULL, flags);
1013 : }
1014 :
1015 : /*
1016 : * End relation scan.
1017 : */
1018 : static inline void
1019 593848 : table_endscan(TableScanDesc scan)
1020 : {
1021 593848 : scan->rs_rd->rd_tableam->scan_end(scan);
1022 593848 : }
1023 :
1024 : /*
1025 : * Restart a relation scan.
1026 : */
1027 : static inline void
1028 1000138 : table_rescan(TableScanDesc scan,
1029 : struct ScanKeyData *key)
1030 : {
1031 1000138 : scan->rs_rd->rd_tableam->scan_rescan(scan, key, false, false, false, false);
1032 1000138 : }
1033 :
1034 : /*
1035 : * Restart a relation scan after changing params.
1036 : *
1037 : * This call allows changing the buffer strategy, syncscan, and pagemode
1038 : * options before starting a fresh scan. Note that although the actual use of
1039 : * syncscan might change (effectively, enabling or disabling reporting), the
1040 : * previously selected startblock will be kept.
1041 : */
1042 : static inline void
1043 30 : table_rescan_set_params(TableScanDesc scan, struct ScanKeyData *key,
1044 : bool allow_strat, bool allow_sync, bool allow_pagemode)
1045 : {
1046 30 : scan->rs_rd->rd_tableam->scan_rescan(scan, key, true,
1047 : allow_strat, allow_sync,
1048 : allow_pagemode);
1049 30 : }
1050 :
1051 : /*
1052 : * Return next tuple from `scan`, store in slot.
1053 : */
1054 : static inline bool
1055 76347060 : table_scan_getnextslot(TableScanDesc sscan, ScanDirection direction, TupleTableSlot *slot)
1056 : {
1057 76347060 : slot->tts_tableOid = RelationGetRelid(sscan->rs_rd);
1058 :
1059 : /* We don't expect actual scans using NoMovementScanDirection */
1060 : Assert(direction == ForwardScanDirection ||
1061 : direction == BackwardScanDirection);
1062 :
1063 : /*
1064 : * We don't expect direct calls to table_scan_getnextslot with valid
1065 : * CheckXidAlive for catalog or regular tables. See detailed comments in
1066 : * xact.c where these variables are declared.
1067 : */
1068 76347060 : if (unlikely(TransactionIdIsValid(CheckXidAlive) && !bsysscan))
1069 0 : elog(ERROR, "unexpected table_scan_getnextslot call during logical decoding");
1070 :
1071 76347060 : return sscan->rs_rd->rd_tableam->scan_getnextslot(sscan, direction, slot);
1072 : }
1073 :
1074 : /* ----------------------------------------------------------------------------
1075 : * TID Range scanning related functions.
1076 : * ----------------------------------------------------------------------------
1077 : */
1078 :
1079 : /*
1080 : * table_beginscan_tidrange is the entry point for setting up a TableScanDesc
1081 : * for a TID range scan.
1082 : */
1083 : static inline TableScanDesc
1084 112 : table_beginscan_tidrange(Relation rel, Snapshot snapshot,
1085 : ItemPointer mintid,
1086 : ItemPointer maxtid)
1087 : {
1088 : TableScanDesc sscan;
1089 112 : uint32 flags = SO_TYPE_TIDRANGESCAN | SO_ALLOW_PAGEMODE;
1090 :
1091 112 : sscan = rel->rd_tableam->scan_begin(rel, snapshot, 0, NULL, NULL, flags);
1092 :
1093 : /* Set the range of TIDs to scan */
1094 112 : sscan->rs_rd->rd_tableam->scan_set_tidrange(sscan, mintid, maxtid);
1095 :
1096 112 : return sscan;
1097 : }
1098 :
1099 : /*
1100 : * table_rescan_tidrange resets the scan position and sets the minimum and
1101 : * maximum TID range to scan for a TableScanDesc created by
1102 : * table_beginscan_tidrange.
1103 : */
1104 : static inline void
1105 66 : table_rescan_tidrange(TableScanDesc sscan, ItemPointer mintid,
1106 : ItemPointer maxtid)
1107 : {
1108 : /* Ensure table_beginscan_tidrange() was used. */
1109 : Assert((sscan->rs_flags & SO_TYPE_TIDRANGESCAN) != 0);
1110 :
1111 66 : sscan->rs_rd->rd_tableam->scan_rescan(sscan, NULL, false, false, false, false);
1112 66 : sscan->rs_rd->rd_tableam->scan_set_tidrange(sscan, mintid, maxtid);
1113 66 : }
1114 :
1115 : /*
1116 : * Fetch the next tuple from `sscan` for a TID range scan created by
1117 : * table_beginscan_tidrange(). Stores the tuple in `slot` and returns true,
1118 : * or returns false if no more tuples exist in the range.
1119 : */
1120 : static inline bool
1121 5940 : table_scan_getnextslot_tidrange(TableScanDesc sscan, ScanDirection direction,
1122 : TupleTableSlot *slot)
1123 : {
1124 : /* Ensure table_beginscan_tidrange() was used. */
1125 : Assert((sscan->rs_flags & SO_TYPE_TIDRANGESCAN) != 0);
1126 :
1127 : /* We don't expect actual scans using NoMovementScanDirection */
1128 : Assert(direction == ForwardScanDirection ||
1129 : direction == BackwardScanDirection);
1130 :
1131 5940 : return sscan->rs_rd->rd_tableam->scan_getnextslot_tidrange(sscan,
1132 : direction,
1133 : slot);
1134 : }
1135 :
1136 :
1137 : /* ----------------------------------------------------------------------------
1138 : * Parallel table scan related functions.
1139 : * ----------------------------------------------------------------------------
1140 : */
1141 :
1142 : /*
1143 : * Estimate the size of shared memory needed for a parallel scan of this
1144 : * relation.
1145 : */
1146 : extern Size table_parallelscan_estimate(Relation rel, Snapshot snapshot);
1147 :
1148 : /*
1149 : * Initialize ParallelTableScanDesc for a parallel scan of this
1150 : * relation. `pscan` needs to be sized according to parallelscan_estimate()
1151 : * for the same relation. Call this just once in the leader process; then,
1152 : * individual workers attach via table_beginscan_parallel.
1153 : */
1154 : extern void table_parallelscan_initialize(Relation rel,
1155 : ParallelTableScanDesc pscan,
1156 : Snapshot snapshot);
1157 :
1158 : /*
1159 : * Begin a parallel scan. `pscan` needs to have been initialized with
1160 : * table_parallelscan_initialize(), for the same relation. The initialization
1161 : * does not need to have happened in this backend.
1162 : *
1163 : * Caller must hold a suitable lock on the relation.
1164 : */
1165 : extern TableScanDesc table_beginscan_parallel(Relation relation,
1166 : ParallelTableScanDesc pscan);
1167 :
1168 : /*
1169 : * Restart a parallel scan. Call this in the leader process. Caller is
1170 : * responsible for making sure that all workers have finished the scan
1171 : * beforehand.
1172 : */
1173 : static inline void
1174 228 : table_parallelscan_reinitialize(Relation rel, ParallelTableScanDesc pscan)
1175 : {
1176 228 : rel->rd_tableam->parallelscan_reinitialize(rel, pscan);
1177 228 : }
1178 :
1179 :
1180 : /* ----------------------------------------------------------------------------
1181 : * Index scan related functions.
1182 : * ----------------------------------------------------------------------------
1183 : */
1184 :
1185 : /*
1186 : * Prepare to fetch tuples from the relation, as needed when fetching tuples
1187 : * for an index scan.
1188 : *
1189 : * Tuples for an index scan can then be fetched via table_index_fetch_tuple().
1190 : */
1191 : static inline IndexFetchTableData *
1192 22055210 : table_index_fetch_begin(Relation rel)
1193 : {
1194 22055210 : return rel->rd_tableam->index_fetch_begin(rel);
1195 : }
1196 :
1197 : /*
1198 : * Reset index fetch. Typically this will release cross index fetch resources
1199 : * held in IndexFetchTableData.
1200 : */
1201 : static inline void
1202 16426592 : table_index_fetch_reset(struct IndexFetchTableData *scan)
1203 : {
1204 16426592 : scan->rel->rd_tableam->index_fetch_reset(scan);
1205 16426592 : }
1206 :
1207 : /*
1208 : * Release resources and deallocate index fetch.
1209 : */
1210 : static inline void
1211 22053856 : table_index_fetch_end(struct IndexFetchTableData *scan)
1212 : {
1213 22053856 : scan->rel->rd_tableam->index_fetch_end(scan);
1214 22053856 : }
1215 :
1216 : /*
1217 : * Fetches, as part of an index scan, tuple at `tid` into `slot`, after doing
1218 : * a visibility test according to `snapshot`. If a tuple was found and passed
1219 : * the visibility test, returns true, false otherwise. Note that *tid may be
1220 : * modified when we return true (see later remarks on multiple row versions
1221 : * reachable via a single index entry).
1222 : *
1223 : * *call_again needs to be false on the first call to table_index_fetch_tuple() for
1224 : * a tid. If there potentially is another tuple matching the tid, *call_again
1225 : * will be set to true, signaling that table_index_fetch_tuple() should be called
1226 : * again for the same tid.
1227 : *
1228 : * *all_dead, if all_dead is not NULL, will be set to true by
1229 : * table_index_fetch_tuple() iff it is guaranteed that no backend needs to see
1230 : * that tuple. Index AMs can use that to avoid returning that tid in future
1231 : * searches.
1232 : *
1233 : * The difference between this function and table_tuple_fetch_row_version()
1234 : * is that this function returns the currently visible version of a row if
1235 : * the AM supports storing multiple row versions reachable via a single index
1236 : * entry (like heap's HOT). Whereas table_tuple_fetch_row_version() only
1237 : * evaluates the tuple exactly at `tid`. Outside of index entry ->table tuple
1238 : * lookups, table_tuple_fetch_row_version() is what's usually needed.
1239 : */
1240 : static inline bool
1241 30916284 : table_index_fetch_tuple(struct IndexFetchTableData *scan,
1242 : ItemPointer tid,
1243 : Snapshot snapshot,
1244 : TupleTableSlot *slot,
1245 : bool *call_again, bool *all_dead)
1246 : {
1247 : /*
1248 : * We don't expect direct calls to table_index_fetch_tuple with valid
1249 : * CheckXidAlive for catalog or regular tables. See detailed comments in
1250 : * xact.c where these variables are declared.
1251 : */
1252 30916284 : if (unlikely(TransactionIdIsValid(CheckXidAlive) && !bsysscan))
1253 0 : elog(ERROR, "unexpected table_index_fetch_tuple call during logical decoding");
1254 :
1255 30916284 : return scan->rel->rd_tableam->index_fetch_tuple(scan, tid, snapshot,
1256 : slot, call_again,
1257 : all_dead);
1258 : }
1259 :
1260 : /*
1261 : * This is a convenience wrapper around table_index_fetch_tuple() which
1262 : * returns whether there are table tuple items corresponding to an index
1263 : * entry. This likely is only useful to verify if there's a conflict in a
1264 : * unique index.
1265 : */
1266 : extern bool table_index_fetch_tuple_check(Relation rel,
1267 : ItemPointer tid,
1268 : Snapshot snapshot,
1269 : bool *all_dead);
1270 :
1271 :
1272 : /* ------------------------------------------------------------------------
1273 : * Functions for non-modifying operations on individual tuples
1274 : * ------------------------------------------------------------------------
1275 : */
1276 :
1277 :
1278 : /*
1279 : * Fetch tuple at `tid` into `slot`, after doing a visibility test according to
1280 : * `snapshot`. If a tuple was found and passed the visibility test, returns
1281 : * true, false otherwise.
1282 : *
1283 : * See table_index_fetch_tuple's comment about what the difference between
1284 : * these functions is. It is correct to use this function outside of index
1285 : * entry->table tuple lookups.
1286 : */
1287 : static inline bool
1288 340842 : table_tuple_fetch_row_version(Relation rel,
1289 : ItemPointer tid,
1290 : Snapshot snapshot,
1291 : TupleTableSlot *slot)
1292 : {
1293 : /*
1294 : * We don't expect direct calls to table_tuple_fetch_row_version with
1295 : * valid CheckXidAlive for catalog or regular tables. See detailed
1296 : * comments in xact.c where these variables are declared.
1297 : */
1298 340842 : if (unlikely(TransactionIdIsValid(CheckXidAlive) && !bsysscan))
1299 0 : elog(ERROR, "unexpected table_tuple_fetch_row_version call during logical decoding");
1300 :
1301 340842 : return rel->rd_tableam->tuple_fetch_row_version(rel, tid, snapshot, slot);
1302 : }
1303 :
1304 : /*
1305 : * Verify that `tid` is a potentially valid tuple identifier. That doesn't
1306 : * mean that the pointed to row needs to exist or be visible, but that
1307 : * attempting to fetch the row (e.g. with table_tuple_get_latest_tid() or
1308 : * table_tuple_fetch_row_version()) should not error out if called with that
1309 : * tid.
1310 : *
1311 : * `scan` needs to have been started via table_beginscan().
1312 : */
1313 : static inline bool
1314 346 : table_tuple_tid_valid(TableScanDesc scan, ItemPointer tid)
1315 : {
1316 346 : return scan->rs_rd->rd_tableam->tuple_tid_valid(scan, tid);
1317 : }
1318 :
1319 : /*
1320 : * Return the latest version of the tuple at `tid`, by updating `tid` to
1321 : * point at the newest version.
1322 : */
1323 : extern void table_tuple_get_latest_tid(TableScanDesc scan, ItemPointer tid);
1324 :
1325 : /*
1326 : * Return true iff tuple in slot satisfies the snapshot.
1327 : *
1328 : * This assumes the slot's tuple is valid, and of the appropriate type for the
1329 : * AM.
1330 : *
1331 : * Some AMs might modify the data underlying the tuple as a side-effect. If so
1332 : * they ought to mark the relevant buffer dirty.
1333 : */
1334 : static inline bool
1335 188926 : table_tuple_satisfies_snapshot(Relation rel, TupleTableSlot *slot,
1336 : Snapshot snapshot)
1337 : {
1338 188926 : return rel->rd_tableam->tuple_satisfies_snapshot(rel, slot, snapshot);
1339 : }
1340 :
1341 : /*
1342 : * Determine which index tuples are safe to delete based on their table TID.
1343 : *
1344 : * Determines which entries from index AM caller's TM_IndexDeleteOp state
1345 : * point to vacuumable table tuples. Entries that are found by tableam to be
1346 : * vacuumable are naturally safe for index AM to delete, and so get directly
1347 : * marked as deletable. See comments above TM_IndexDelete and comments above
1348 : * TM_IndexDeleteOp for full details.
1349 : *
1350 : * Returns a snapshotConflictHorizon transaction ID that caller places in
1351 : * its index deletion WAL record. This might be used during subsequent REDO
1352 : * of the WAL record when in Hot Standby mode -- a recovery conflict for the
1353 : * index deletion operation might be required on the standby.
1354 : */
1355 : static inline TransactionId
1356 9900 : table_index_delete_tuples(Relation rel, TM_IndexDeleteOp *delstate)
1357 : {
1358 9900 : return rel->rd_tableam->index_delete_tuples(rel, delstate);
1359 : }
1360 :
1361 :
1362 : /* ----------------------------------------------------------------------------
1363 : * Functions for manipulations of physical tuples.
1364 : * ----------------------------------------------------------------------------
1365 : */
1366 :
1367 : /*
1368 : * Insert a tuple from a slot into table AM routine.
1369 : *
1370 : * The options bitmask allows the caller to specify options that may change the
1371 : * behaviour of the AM. The AM will ignore options that it does not support.
1372 : *
1373 : * If the TABLE_INSERT_SKIP_FSM option is specified, AMs are free to not reuse
1374 : * free space in the relation. This can save some cycles when we know the
1375 : * relation is new and doesn't contain useful amounts of free space.
1376 : * TABLE_INSERT_SKIP_FSM is commonly passed directly to
1377 : * RelationGetBufferForTuple. See that method for more information.
1378 : *
1379 : * TABLE_INSERT_FROZEN should only be specified for inserts into
1380 : * relation storage created during the current subtransaction and when
1381 : * there are no prior snapshots or pre-existing portals open.
1382 : * This causes rows to be frozen, which is an MVCC violation and
1383 : * requires explicit options chosen by user.
1384 : *
1385 : * TABLE_INSERT_NO_LOGICAL force-disables the emitting of logical decoding
1386 : * information for the tuple. This should solely be used during table rewrites
1387 : * where RelationIsLogicallyLogged(relation) is not yet accurate for the new
1388 : * relation.
1389 : *
1390 : * Note that most of these options will be applied when inserting into the
1391 : * heap's TOAST table, too, if the tuple requires any out-of-line data.
1392 : *
1393 : * The BulkInsertState object (if any; bistate can be NULL for default
1394 : * behavior) is also just passed through to RelationGetBufferForTuple. If
1395 : * `bistate` is provided, table_finish_bulk_insert() needs to be called.
1396 : *
1397 : * On return the slot's tts_tid and tts_tableOid are updated to reflect the
1398 : * insertion. But note that any toasting of fields within the slot is NOT
1399 : * reflected in the slots contents.
1400 : */
1401 : static inline void
1402 13648472 : table_tuple_insert(Relation rel, TupleTableSlot *slot, CommandId cid,
1403 : int options, struct BulkInsertStateData *bistate)
1404 : {
1405 13648472 : rel->rd_tableam->tuple_insert(rel, slot, cid, options,
1406 : bistate);
1407 13648448 : }
1408 :
1409 : /*
1410 : * Perform a "speculative insertion". These can be backed out afterwards
1411 : * without aborting the whole transaction. Other sessions can wait for the
1412 : * speculative insertion to be confirmed, turning it into a regular tuple, or
1413 : * aborted, as if it never existed. Speculatively inserted tuples behave as
1414 : * "value locks" of short duration, used to implement INSERT .. ON CONFLICT.
1415 : *
1416 : * A transaction having performed a speculative insertion has to either abort,
1417 : * or finish the speculative insertion with
1418 : * table_tuple_complete_speculative(succeeded = ...).
1419 : */
1420 : static inline void
1421 4026 : table_tuple_insert_speculative(Relation rel, TupleTableSlot *slot,
1422 : CommandId cid, int options,
1423 : struct BulkInsertStateData *bistate,
1424 : uint32 specToken)
1425 : {
1426 4026 : rel->rd_tableam->tuple_insert_speculative(rel, slot, cid, options,
1427 : bistate, specToken);
1428 4026 : }
1429 :
1430 : /*
1431 : * Complete "speculative insertion" started in the same transaction. If
1432 : * succeeded is true, the tuple is fully inserted, if false, it's removed.
1433 : */
1434 : static inline void
1435 4020 : table_tuple_complete_speculative(Relation rel, TupleTableSlot *slot,
1436 : uint32 specToken, bool succeeded)
1437 : {
1438 4020 : rel->rd_tableam->tuple_complete_speculative(rel, slot, specToken,
1439 : succeeded);
1440 4020 : }
1441 :
1442 : /*
1443 : * Insert multiple tuples into a table.
1444 : *
1445 : * This is like table_tuple_insert(), but inserts multiple tuples in one
1446 : * operation. That's often faster than calling table_tuple_insert() in a loop,
1447 : * because e.g. the AM can reduce WAL logging and page locking overhead.
1448 : *
1449 : * Except for taking `nslots` tuples as input, and an array of TupleTableSlots
1450 : * in `slots`, the parameters for table_multi_insert() are the same as for
1451 : * table_tuple_insert().
1452 : *
1453 : * Note: this leaks memory into the current memory context. You can create a
1454 : * temporary context before calling this, if that's a problem.
1455 : */
1456 : static inline void
1457 2322 : table_multi_insert(Relation rel, TupleTableSlot **slots, int nslots,
1458 : CommandId cid, int options, struct BulkInsertStateData *bistate)
1459 : {
1460 2322 : rel->rd_tableam->multi_insert(rel, slots, nslots,
1461 : cid, options, bistate);
1462 2322 : }
1463 :
1464 : /*
1465 : * Delete a tuple.
1466 : *
1467 : * NB: do not call this directly unless prepared to deal with
1468 : * concurrent-update conditions. Use simple_table_tuple_delete instead.
1469 : *
1470 : * Input parameters:
1471 : * relation - table to be modified (caller must hold suitable lock)
1472 : * tid - TID of tuple to be deleted
1473 : * cid - delete command ID (used for visibility test, and stored into
1474 : * cmax if successful)
1475 : * crosscheck - if not InvalidSnapshot, also check tuple against this
1476 : * wait - true if should wait for any conflicting update to commit/abort
1477 : * Output parameters:
1478 : * tmfd - filled in failure cases (see below)
1479 : * changingPart - true iff the tuple is being moved to another partition
1480 : * table due to an update of the partition key. Otherwise, false.
1481 : *
1482 : * Normal, successful return value is TM_Ok, which means we did actually
1483 : * delete it. Failure return codes are TM_SelfModified, TM_Updated, and
1484 : * TM_BeingModified (the last only possible if wait == false).
1485 : *
1486 : * In the failure cases, the routine fills *tmfd with the tuple's t_ctid,
1487 : * t_xmax, and, if possible, t_cmax. See comments for struct
1488 : * TM_FailureData for additional info.
1489 : */
1490 : static inline TM_Result
1491 1606746 : table_tuple_delete(Relation rel, ItemPointer tid, CommandId cid,
1492 : Snapshot snapshot, Snapshot crosscheck, bool wait,
1493 : TM_FailureData *tmfd, bool changingPart)
1494 : {
1495 1606746 : return rel->rd_tableam->tuple_delete(rel, tid, cid,
1496 : snapshot, crosscheck,
1497 : wait, tmfd, changingPart);
1498 : }
1499 :
1500 : /*
1501 : * Update a tuple.
1502 : *
1503 : * NB: do not call this directly unless you are prepared to deal with
1504 : * concurrent-update conditions. Use simple_table_tuple_update instead.
1505 : *
1506 : * Input parameters:
1507 : * relation - table to be modified (caller must hold suitable lock)
1508 : * otid - TID of old tuple to be replaced
1509 : * slot - newly constructed tuple data to store
1510 : * cid - update command ID (used for visibility test, and stored into
1511 : * cmax/cmin if successful)
1512 : * crosscheck - if not InvalidSnapshot, also check old tuple against this
1513 : * wait - true if should wait for any conflicting update to commit/abort
1514 : * Output parameters:
1515 : * tmfd - filled in failure cases (see below)
1516 : * lockmode - filled with lock mode acquired on tuple
1517 : * update_indexes - in success cases this is set to true if new index entries
1518 : * are required for this tuple
1519 : *
1520 : * Normal, successful return value is TM_Ok, which means we did actually
1521 : * update it. Failure return codes are TM_SelfModified, TM_Updated, and
1522 : * TM_BeingModified (the last only possible if wait == false).
1523 : *
1524 : * On success, the slot's tts_tid and tts_tableOid are updated to match the new
1525 : * stored tuple; in particular, slot->tts_tid is set to the TID where the
1526 : * new tuple was inserted, and its HEAP_ONLY_TUPLE flag is set iff a HOT
1527 : * update was done. However, any TOAST changes in the new tuple's
1528 : * data are not reflected into *newtup.
1529 : *
1530 : * In the failure cases, the routine fills *tmfd with the tuple's t_ctid,
1531 : * t_xmax, and, if possible, t_cmax. See comments for struct TM_FailureData
1532 : * for additional info.
1533 : */
1534 : static inline TM_Result
1535 376254 : table_tuple_update(Relation rel, ItemPointer otid, TupleTableSlot *slot,
1536 : CommandId cid, Snapshot snapshot, Snapshot crosscheck,
1537 : bool wait, TM_FailureData *tmfd, LockTupleMode *lockmode,
1538 : TU_UpdateIndexes *update_indexes)
1539 : {
1540 376254 : return rel->rd_tableam->tuple_update(rel, otid, slot,
1541 : cid, snapshot, crosscheck,
1542 : wait, tmfd,
1543 : lockmode, update_indexes);
1544 : }
1545 :
1546 : /*
1547 : * Lock a tuple in the specified mode.
1548 : *
1549 : * Input parameters:
1550 : * relation: relation containing tuple (caller must hold suitable lock)
1551 : * tid: TID of tuple to lock
1552 : * snapshot: snapshot to use for visibility determinations
1553 : * cid: current command ID (used for visibility test, and stored into
1554 : * tuple's cmax if lock is successful)
1555 : * mode: lock mode desired
1556 : * wait_policy: what to do if tuple lock is not available
1557 : * flags:
1558 : * If TUPLE_LOCK_FLAG_LOCK_UPDATE_IN_PROGRESS, follow the update chain to
1559 : * also lock descendant tuples if lock modes don't conflict.
1560 : * If TUPLE_LOCK_FLAG_FIND_LAST_VERSION, follow the update chain and lock
1561 : * latest version.
1562 : *
1563 : * Output parameters:
1564 : * *slot: contains the target tuple
1565 : * *tmfd: filled in failure cases (see below)
1566 : *
1567 : * Function result may be:
1568 : * TM_Ok: lock was successfully acquired
1569 : * TM_Invisible: lock failed because tuple was never visible to us
1570 : * TM_SelfModified: lock failed because tuple updated by self
1571 : * TM_Updated: lock failed because tuple updated by other xact
1572 : * TM_Deleted: lock failed because tuple deleted by other xact
1573 : * TM_WouldBlock: lock couldn't be acquired and wait_policy is skip
1574 : *
1575 : * In the failure cases other than TM_Invisible and TM_Deleted, the routine
1576 : * fills *tmfd with the tuple's t_ctid, t_xmax, and, if possible, t_cmax. See
1577 : * comments for struct TM_FailureData for additional info.
1578 : */
1579 : static inline TM_Result
1580 169030 : table_tuple_lock(Relation rel, ItemPointer tid, Snapshot snapshot,
1581 : TupleTableSlot *slot, CommandId cid, LockTupleMode mode,
1582 : LockWaitPolicy wait_policy, uint8 flags,
1583 : TM_FailureData *tmfd)
1584 : {
1585 169030 : return rel->rd_tableam->tuple_lock(rel, tid, snapshot, slot,
1586 : cid, mode, wait_policy,
1587 : flags, tmfd);
1588 : }
1589 :
1590 : /*
1591 : * Perform operations necessary to complete insertions made via
1592 : * tuple_insert and multi_insert with a BulkInsertState specified.
1593 : */
1594 : static inline void
1595 3724 : table_finish_bulk_insert(Relation rel, int options)
1596 : {
1597 : /* optional callback */
1598 3724 : if (rel->rd_tableam && rel->rd_tableam->finish_bulk_insert)
1599 0 : rel->rd_tableam->finish_bulk_insert(rel, options);
1600 3724 : }
1601 :
1602 :
1603 : /* ------------------------------------------------------------------------
1604 : * DDL related functionality.
1605 : * ------------------------------------------------------------------------
1606 : */
1607 :
1608 : /*
1609 : * Create storage for `rel` in `newrlocator`, with persistence set to
1610 : * `persistence`.
1611 : *
1612 : * This is used both during relation creation and various DDL operations to
1613 : * create new rel storage that can be filled from scratch. When creating
1614 : * new storage for an existing relfilelocator, this should be called before the
1615 : * relcache entry has been updated.
1616 : *
1617 : * *freezeXid, *minmulti are set to the xid / multixact horizon for the table
1618 : * that pg_class.{relfrozenxid, relminmxid} have to be set to.
1619 : */
1620 : static inline void
1621 56290 : table_relation_set_new_filelocator(Relation rel,
1622 : const RelFileLocator *newrlocator,
1623 : char persistence,
1624 : TransactionId *freezeXid,
1625 : MultiXactId *minmulti)
1626 : {
1627 56290 : rel->rd_tableam->relation_set_new_filelocator(rel, newrlocator,
1628 : persistence, freezeXid,
1629 : minmulti);
1630 56290 : }
1631 :
1632 : /*
1633 : * Remove all table contents from `rel`, in a non-transactional manner.
1634 : * Non-transactional meaning that there's no need to support rollbacks. This
1635 : * commonly only is used to perform truncations for relation storage created in
1636 : * the current transaction.
1637 : */
1638 : static inline void
1639 576 : table_relation_nontransactional_truncate(Relation rel)
1640 : {
1641 576 : rel->rd_tableam->relation_nontransactional_truncate(rel);
1642 576 : }
1643 :
1644 : /*
1645 : * Copy data from `rel` into the new relfilelocator `newrlocator`. The new
1646 : * relfilelocator may not have storage associated before this function is
1647 : * called. This is only supposed to be used for low level operations like
1648 : * changing a relation's tablespace.
1649 : */
1650 : static inline void
1651 98 : table_relation_copy_data(Relation rel, const RelFileLocator *newrlocator)
1652 : {
1653 98 : rel->rd_tableam->relation_copy_data(rel, newrlocator);
1654 98 : }
1655 :
1656 : /*
1657 : * Copy data from `OldTable` into `NewTable`, as part of a CLUSTER or VACUUM
1658 : * FULL.
1659 : *
1660 : * Additional Input parameters:
1661 : * - use_sort - if true, the table contents are sorted appropriate for
1662 : * `OldIndex`; if false and OldIndex is not InvalidOid, the data is copied
1663 : * in that index's order; if false and OldIndex is InvalidOid, no sorting is
1664 : * performed
1665 : * - OldIndex - see use_sort
1666 : * - OldestXmin - computed by vacuum_get_cutoffs(), even when
1667 : * not needed for the relation's AM
1668 : * - *xid_cutoff - ditto
1669 : * - *multi_cutoff - ditto
1670 : *
1671 : * Output parameters:
1672 : * - *xid_cutoff - rel's new relfrozenxid value, may be invalid
1673 : * - *multi_cutoff - rel's new relminmxid value, may be invalid
1674 : * - *tups_vacuumed - stats, for logging, if appropriate for AM
1675 : * - *tups_recently_dead - stats, for logging, if appropriate for AM
1676 : */
1677 : static inline void
1678 526 : table_relation_copy_for_cluster(Relation OldTable, Relation NewTable,
1679 : Relation OldIndex,
1680 : bool use_sort,
1681 : TransactionId OldestXmin,
1682 : TransactionId *xid_cutoff,
1683 : MultiXactId *multi_cutoff,
1684 : double *num_tuples,
1685 : double *tups_vacuumed,
1686 : double *tups_recently_dead)
1687 : {
1688 526 : OldTable->rd_tableam->relation_copy_for_cluster(OldTable, NewTable, OldIndex,
1689 : use_sort, OldestXmin,
1690 : xid_cutoff, multi_cutoff,
1691 : num_tuples, tups_vacuumed,
1692 : tups_recently_dead);
1693 526 : }
1694 :
1695 : /*
1696 : * Perform VACUUM on the relation. The VACUUM can be triggered by a user or by
1697 : * autovacuum. The specific actions performed by the AM will depend heavily on
1698 : * the individual AM.
1699 : *
1700 : * On entry a transaction needs to already been established, and the
1701 : * table is locked with a ShareUpdateExclusive lock.
1702 : *
1703 : * Note that neither VACUUM FULL (and CLUSTER), nor ANALYZE go through this
1704 : * routine, even if (for ANALYZE) it is part of the same VACUUM command.
1705 : */
1706 : static inline void
1707 19308 : table_relation_vacuum(Relation rel, struct VacuumParams *params,
1708 : BufferAccessStrategy bstrategy)
1709 : {
1710 19308 : rel->rd_tableam->relation_vacuum(rel, params, bstrategy);
1711 19308 : }
1712 :
1713 : /*
1714 : * Prepare to analyze the next block in the read stream. The scan needs to
1715 : * have been started with table_beginscan_analyze(). Note that this routine
1716 : * might acquire resources like locks that are held until
1717 : * table_scan_analyze_next_tuple() returns false.
1718 : *
1719 : * Returns false if block is unsuitable for sampling, true otherwise.
1720 : */
1721 : static inline bool
1722 115902 : table_scan_analyze_next_block(TableScanDesc scan, ReadStream *stream)
1723 : {
1724 115902 : return scan->rs_rd->rd_tableam->scan_analyze_next_block(scan, stream);
1725 : }
1726 :
1727 : /*
1728 : * Iterate over tuples in the block selected with
1729 : * table_scan_analyze_next_block() (which needs to have returned true, and
1730 : * this routine may not have returned false for the same block before). If a
1731 : * tuple that's suitable for sampling is found, true is returned and a tuple
1732 : * is stored in `slot`.
1733 : *
1734 : * *liverows and *deadrows are incremented according to the encountered
1735 : * tuples.
1736 : */
1737 : static inline bool
1738 8805702 : table_scan_analyze_next_tuple(TableScanDesc scan, TransactionId OldestXmin,
1739 : double *liverows, double *deadrows,
1740 : TupleTableSlot *slot)
1741 : {
1742 8805702 : return scan->rs_rd->rd_tableam->scan_analyze_next_tuple(scan, OldestXmin,
1743 : liverows, deadrows,
1744 : slot);
1745 : }
1746 :
1747 : /*
1748 : * table_index_build_scan - scan the table to find tuples to be indexed
1749 : *
1750 : * This is called back from an access-method-specific index build procedure
1751 : * after the AM has done whatever setup it needs. The parent table relation
1752 : * is scanned to find tuples that should be entered into the index. Each
1753 : * such tuple is passed to the AM's callback routine, which does the right
1754 : * things to add it to the new index. After we return, the AM's index
1755 : * build procedure does whatever cleanup it needs.
1756 : *
1757 : * The total count of live tuples is returned. This is for updating pg_class
1758 : * statistics. (It's annoying not to be able to do that here, but we want to
1759 : * merge that update with others; see index_update_stats.) Note that the
1760 : * index AM itself must keep track of the number of index tuples; we don't do
1761 : * so here because the AM might reject some of the tuples for its own reasons,
1762 : * such as being unable to store NULLs.
1763 : *
1764 : * If 'progress', the PROGRESS_SCAN_BLOCKS_TOTAL counter is updated when
1765 : * starting the scan, and PROGRESS_SCAN_BLOCKS_DONE is updated as we go along.
1766 : *
1767 : * A side effect is to set indexInfo->ii_BrokenHotChain to true if we detect
1768 : * any potentially broken HOT chains. Currently, we set this if there are any
1769 : * RECENTLY_DEAD or DELETE_IN_PROGRESS entries in a HOT chain, without trying
1770 : * very hard to detect whether they're really incompatible with the chain tip.
1771 : * This only really makes sense for heap AM, it might need to be generalized
1772 : * for other AMs later.
1773 : */
1774 : static inline double
1775 46414 : table_index_build_scan(Relation table_rel,
1776 : Relation index_rel,
1777 : struct IndexInfo *index_info,
1778 : bool allow_sync,
1779 : bool progress,
1780 : IndexBuildCallback callback,
1781 : void *callback_state,
1782 : TableScanDesc scan)
1783 : {
1784 46414 : return table_rel->rd_tableam->index_build_range_scan(table_rel,
1785 : index_rel,
1786 : index_info,
1787 : allow_sync,
1788 : false,
1789 : progress,
1790 : 0,
1791 : InvalidBlockNumber,
1792 : callback,
1793 : callback_state,
1794 : scan);
1795 : }
1796 :
1797 : /*
1798 : * As table_index_build_scan(), except that instead of scanning the complete
1799 : * table, only the given number of blocks are scanned. Scan to end-of-rel can
1800 : * be signaled by passing InvalidBlockNumber as numblocks. Note that
1801 : * restricting the range to scan cannot be done when requesting syncscan.
1802 : *
1803 : * When "anyvisible" mode is requested, all tuples visible to any transaction
1804 : * are indexed and counted as live, including those inserted or deleted by
1805 : * transactions that are still in progress.
1806 : */
1807 : static inline double
1808 2934 : table_index_build_range_scan(Relation table_rel,
1809 : Relation index_rel,
1810 : struct IndexInfo *index_info,
1811 : bool allow_sync,
1812 : bool anyvisible,
1813 : bool progress,
1814 : BlockNumber start_blockno,
1815 : BlockNumber numblocks,
1816 : IndexBuildCallback callback,
1817 : void *callback_state,
1818 : TableScanDesc scan)
1819 : {
1820 2934 : return table_rel->rd_tableam->index_build_range_scan(table_rel,
1821 : index_rel,
1822 : index_info,
1823 : allow_sync,
1824 : anyvisible,
1825 : progress,
1826 : start_blockno,
1827 : numblocks,
1828 : callback,
1829 : callback_state,
1830 : scan);
1831 : }
1832 :
1833 : /*
1834 : * table_index_validate_scan - second table scan for concurrent index build
1835 : *
1836 : * See validate_index() for an explanation.
1837 : */
1838 : static inline void
1839 594 : table_index_validate_scan(Relation table_rel,
1840 : Relation index_rel,
1841 : struct IndexInfo *index_info,
1842 : Snapshot snapshot,
1843 : struct ValidateIndexState *state)
1844 : {
1845 594 : table_rel->rd_tableam->index_validate_scan(table_rel,
1846 : index_rel,
1847 : index_info,
1848 : snapshot,
1849 : state);
1850 594 : }
1851 :
1852 :
1853 : /* ----------------------------------------------------------------------------
1854 : * Miscellaneous functionality
1855 : * ----------------------------------------------------------------------------
1856 : */
1857 :
1858 : /*
1859 : * Return the current size of `rel` in bytes. If `forkNumber` is
1860 : * InvalidForkNumber, return the relation's overall size, otherwise the size
1861 : * for the indicated fork.
1862 : *
1863 : * Note that the overall size might not be the equivalent of the sum of sizes
1864 : * for the individual forks for some AMs, e.g. because the AMs storage does
1865 : * not neatly map onto the builtin types of forks.
1866 : */
1867 : static inline uint64
1868 2123352 : table_relation_size(Relation rel, ForkNumber forkNumber)
1869 : {
1870 2123352 : return rel->rd_tableam->relation_size(rel, forkNumber);
1871 : }
1872 :
1873 : /*
1874 : * table_relation_needs_toast_table - does this relation need a toast table?
1875 : */
1876 : static inline bool
1877 38844 : table_relation_needs_toast_table(Relation rel)
1878 : {
1879 38844 : return rel->rd_tableam->relation_needs_toast_table(rel);
1880 : }
1881 :
1882 : /*
1883 : * Return the OID of the AM that should be used to implement the TOAST table
1884 : * for this relation.
1885 : */
1886 : static inline Oid
1887 15510 : table_relation_toast_am(Relation rel)
1888 : {
1889 15510 : return rel->rd_tableam->relation_toast_am(rel);
1890 : }
1891 :
1892 : /*
1893 : * Fetch all or part of a TOAST value from a TOAST table.
1894 : *
1895 : * If this AM is never used to implement a TOAST table, then this callback
1896 : * is not needed. But, if toasted values are ever stored in a table of this
1897 : * type, then you will need this callback.
1898 : *
1899 : * toastrel is the relation in which the toasted value is stored.
1900 : *
1901 : * valueid identifies which toast value is to be fetched. For the heap,
1902 : * this corresponds to the values stored in the chunk_id column.
1903 : *
1904 : * attrsize is the total size of the toast value to be fetched.
1905 : *
1906 : * sliceoffset is the offset within the toast value of the first byte that
1907 : * should be fetched.
1908 : *
1909 : * slicelength is the number of bytes from the toast value that should be
1910 : * fetched.
1911 : *
1912 : * result is caller-allocated space into which the fetched bytes should be
1913 : * stored.
1914 : */
1915 : static inline void
1916 18520 : table_relation_fetch_toast_slice(Relation toastrel, Oid valueid,
1917 : int32 attrsize, int32 sliceoffset,
1918 : int32 slicelength, struct varlena *result)
1919 : {
1920 18520 : toastrel->rd_tableam->relation_fetch_toast_slice(toastrel, valueid,
1921 : attrsize,
1922 : sliceoffset, slicelength,
1923 : result);
1924 18520 : }
1925 :
1926 :
1927 : /* ----------------------------------------------------------------------------
1928 : * Planner related functionality
1929 : * ----------------------------------------------------------------------------
1930 : */
1931 :
1932 : /*
1933 : * Estimate the current size of the relation, as an AM specific workhorse for
1934 : * estimate_rel_size(). Look there for an explanation of the parameters.
1935 : */
1936 : static inline void
1937 374832 : table_relation_estimate_size(Relation rel, int32 *attr_widths,
1938 : BlockNumber *pages, double *tuples,
1939 : double *allvisfrac)
1940 : {
1941 374832 : rel->rd_tableam->relation_estimate_size(rel, attr_widths, pages, tuples,
1942 : allvisfrac);
1943 374832 : }
1944 :
1945 :
1946 : /* ----------------------------------------------------------------------------
1947 : * Executor related functionality
1948 : * ----------------------------------------------------------------------------
1949 : */
1950 :
1951 : /*
1952 : * Prepare to fetch / check / return tuples from `tbmres->blockno` as part of
1953 : * a bitmap table scan. `scan` needs to have been started via
1954 : * table_beginscan_bm(). Returns false if there are no tuples to be found on
1955 : * the page, true otherwise.
1956 : *
1957 : * Note, this is an optionally implemented function, therefore should only be
1958 : * used after verifying the presence (at plan time or such).
1959 : */
1960 : static inline bool
1961 387434 : table_scan_bitmap_next_block(TableScanDesc scan,
1962 : struct TBMIterateResult *tbmres)
1963 : {
1964 : /*
1965 : * We don't expect direct calls to table_scan_bitmap_next_block with valid
1966 : * CheckXidAlive for catalog or regular tables. See detailed comments in
1967 : * xact.c where these variables are declared.
1968 : */
1969 387434 : if (unlikely(TransactionIdIsValid(CheckXidAlive) && !bsysscan))
1970 0 : elog(ERROR, "unexpected table_scan_bitmap_next_block call during logical decoding");
1971 :
1972 387434 : return scan->rs_rd->rd_tableam->scan_bitmap_next_block(scan,
1973 : tbmres);
1974 : }
1975 :
1976 : /*
1977 : * Fetch the next tuple of a bitmap table scan into `slot` and return true if
1978 : * a visible tuple was found, false otherwise.
1979 : * table_scan_bitmap_next_block() needs to previously have selected a
1980 : * block (i.e. returned true), and no previous
1981 : * table_scan_bitmap_next_tuple() for the same block may have
1982 : * returned false.
1983 : */
1984 : static inline bool
1985 6717148 : table_scan_bitmap_next_tuple(TableScanDesc scan,
1986 : struct TBMIterateResult *tbmres,
1987 : TupleTableSlot *slot)
1988 : {
1989 : /*
1990 : * We don't expect direct calls to table_scan_bitmap_next_tuple with valid
1991 : * CheckXidAlive for catalog or regular tables. See detailed comments in
1992 : * xact.c where these variables are declared.
1993 : */
1994 6717148 : if (unlikely(TransactionIdIsValid(CheckXidAlive) && !bsysscan))
1995 0 : elog(ERROR, "unexpected table_scan_bitmap_next_tuple call during logical decoding");
1996 :
1997 6717148 : return scan->rs_rd->rd_tableam->scan_bitmap_next_tuple(scan,
1998 : tbmres,
1999 : slot);
2000 : }
2001 :
2002 : /*
2003 : * Prepare to fetch tuples from the next block in a sample scan. Returns false
2004 : * if the sample scan is finished, true otherwise. `scan` needs to have been
2005 : * started via table_beginscan_sampling().
2006 : *
2007 : * This will call the TsmRoutine's NextSampleBlock() callback if necessary
2008 : * (i.e. NextSampleBlock is not NULL), or perform a sequential scan over the
2009 : * underlying relation.
2010 : */
2011 : static inline bool
2012 12910 : table_scan_sample_next_block(TableScanDesc scan,
2013 : struct SampleScanState *scanstate)
2014 : {
2015 : /*
2016 : * We don't expect direct calls to table_scan_sample_next_block with valid
2017 : * CheckXidAlive for catalog or regular tables. See detailed comments in
2018 : * xact.c where these variables are declared.
2019 : */
2020 12910 : if (unlikely(TransactionIdIsValid(CheckXidAlive) && !bsysscan))
2021 0 : elog(ERROR, "unexpected table_scan_sample_next_block call during logical decoding");
2022 12910 : return scan->rs_rd->rd_tableam->scan_sample_next_block(scan, scanstate);
2023 : }
2024 :
2025 : /*
2026 : * Fetch the next sample tuple into `slot` and return true if a visible tuple
2027 : * was found, false otherwise. table_scan_sample_next_block() needs to
2028 : * previously have selected a block (i.e. returned true), and no previous
2029 : * table_scan_sample_next_tuple() for the same block may have returned false.
2030 : *
2031 : * This will call the TsmRoutine's NextSampleTuple() callback.
2032 : */
2033 : static inline bool
2034 253894 : table_scan_sample_next_tuple(TableScanDesc scan,
2035 : struct SampleScanState *scanstate,
2036 : TupleTableSlot *slot)
2037 : {
2038 : /*
2039 : * We don't expect direct calls to table_scan_sample_next_tuple with valid
2040 : * CheckXidAlive for catalog or regular tables. See detailed comments in
2041 : * xact.c where these variables are declared.
2042 : */
2043 253894 : if (unlikely(TransactionIdIsValid(CheckXidAlive) && !bsysscan))
2044 0 : elog(ERROR, "unexpected table_scan_sample_next_tuple call during logical decoding");
2045 253894 : return scan->rs_rd->rd_tableam->scan_sample_next_tuple(scan, scanstate,
2046 : slot);
2047 : }
2048 :
2049 :
2050 : /* ----------------------------------------------------------------------------
2051 : * Functions to make modifications a bit simpler.
2052 : * ----------------------------------------------------------------------------
2053 : */
2054 :
2055 : extern void simple_table_tuple_insert(Relation rel, TupleTableSlot *slot);
2056 : extern void simple_table_tuple_delete(Relation rel, ItemPointer tid,
2057 : Snapshot snapshot);
2058 : extern void simple_table_tuple_update(Relation rel, ItemPointer otid,
2059 : TupleTableSlot *slot, Snapshot snapshot,
2060 : TU_UpdateIndexes *update_indexes);
2061 :
2062 :
2063 : /* ----------------------------------------------------------------------------
2064 : * Helper functions to implement parallel scans for block oriented AMs.
2065 : * ----------------------------------------------------------------------------
2066 : */
2067 :
2068 : extern Size table_block_parallelscan_estimate(Relation rel);
2069 : extern Size table_block_parallelscan_initialize(Relation rel,
2070 : ParallelTableScanDesc pscan);
2071 : extern void table_block_parallelscan_reinitialize(Relation rel,
2072 : ParallelTableScanDesc pscan);
2073 : extern BlockNumber table_block_parallelscan_nextpage(Relation rel,
2074 : ParallelBlockTableScanWorker pbscanwork,
2075 : ParallelBlockTableScanDesc pbscan);
2076 : extern void table_block_parallelscan_startblock_init(Relation rel,
2077 : ParallelBlockTableScanWorker pbscanwork,
2078 : ParallelBlockTableScanDesc pbscan);
2079 :
2080 :
2081 : /* ----------------------------------------------------------------------------
2082 : * Helper functions to implement relation sizing for block oriented AMs.
2083 : * ----------------------------------------------------------------------------
2084 : */
2085 :
2086 : extern uint64 table_block_relation_size(Relation rel, ForkNumber forkNumber);
2087 : extern void table_block_relation_estimate_size(Relation rel,
2088 : int32 *attr_widths,
2089 : BlockNumber *pages,
2090 : double *tuples,
2091 : double *allvisfrac,
2092 : Size overhead_bytes_per_tuple,
2093 : Size usable_bytes_per_page);
2094 :
2095 : /* ----------------------------------------------------------------------------
2096 : * Functions in tableamapi.c
2097 : * ----------------------------------------------------------------------------
2098 : */
2099 :
2100 : extern const TableAmRoutine *GetTableAmRoutine(Oid amhandler);
2101 :
2102 : /* ----------------------------------------------------------------------------
2103 : * Functions in heapam_handler.c
2104 : * ----------------------------------------------------------------------------
2105 : */
2106 :
2107 : extern const TableAmRoutine *GetHeapamTableAmRoutine(void);
2108 :
2109 : #endif /* TABLEAM_H */
|
