Line data Source code
1 : /*-------------------------------------------------------------------------
2 : *
3 : * tableam.h
4 : * POSTGRES table access method definitions.
5 : *
6 : *
7 : * Portions Copyright (c) 1996-2021, 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 "utils/guc.h"
24 : #include "utils/rel.h"
25 : #include "utils/snapshot.h"
26 :
27 :
28 : #define DEFAULT_TABLE_ACCESS_METHOD "heap"
29 :
30 : /* GUCs */
31 : extern char *default_table_access_method;
32 : extern bool synchronize_seqscans;
33 :
34 :
35 : struct BulkInsertStateData;
36 : struct IndexInfo;
37 : struct SampleScanState;
38 : struct TBMIterateResult;
39 : struct VacuumParams;
40 : struct ValidateIndexState;
41 :
42 : /*
43 : * Bitmask values for the flags argument to the scan_begin callback.
44 : */
45 : typedef enum ScanOptions
46 : {
47 : /* one of SO_TYPE_* may be specified */
48 : SO_TYPE_SEQSCAN = 1 << 0,
49 : SO_TYPE_BITMAPSCAN = 1 << 1,
50 : SO_TYPE_SAMPLESCAN = 1 << 2,
51 : SO_TYPE_TIDSCAN = 1 << 3,
52 : SO_TYPE_ANALYZE = 1 << 4,
53 :
54 : /* several of SO_ALLOW_* may be specified */
55 : /* allow or disallow use of access strategy */
56 : SO_ALLOW_STRAT = 1 << 5,
57 : /* report location to syncscan logic? */
58 : SO_ALLOW_SYNC = 1 << 6,
59 : /* verify visibility page-at-a-time? */
60 : SO_ALLOW_PAGEMODE = 1 << 7,
61 :
62 : /* unregister snapshot at scan end? */
63 : SO_TEMP_SNAPSHOT = 1 << 8
64 : } ScanOptions;
65 :
66 : /*
67 : * Result codes for table_{update,delete,lock_tuple}, and for visibility
68 : * routines inside table AMs.
69 : */
70 : typedef enum TM_Result
71 : {
72 : /*
73 : * Signals that the action succeeded (i.e. update/delete performed, lock
74 : * was acquired)
75 : */
76 : TM_Ok,
77 :
78 : /* The affected tuple wasn't visible to the relevant snapshot */
79 : TM_Invisible,
80 :
81 : /* The affected tuple was already modified by the calling backend */
82 : TM_SelfModified,
83 :
84 : /*
85 : * The affected tuple was updated by another transaction. This includes
86 : * the case where tuple was moved to another partition.
87 : */
88 : TM_Updated,
89 :
90 : /* The affected tuple was deleted by another transaction */
91 : TM_Deleted,
92 :
93 : /*
94 : * The affected tuple is currently being modified by another session. This
95 : * will only be returned if table_(update/delete/lock_tuple) are
96 : * instructed not to wait.
97 : */
98 : TM_BeingModified,
99 :
100 : /* lock couldn't be acquired, action skipped. Only used by lock_tuple */
101 : TM_WouldBlock
102 : } TM_Result;
103 :
104 : /*
105 : * When table_tuple_update, table_tuple_delete, or table_tuple_lock fail
106 : * because the target tuple is already outdated, they fill in this struct to
107 : * provide information to the caller about what happened.
108 : *
109 : * ctid is the target's ctid link: it is the same as the target's TID if the
110 : * target was deleted, or the location of the replacement tuple if the target
111 : * was updated.
112 : *
113 : * xmax is the outdating transaction's XID. If the caller wants to visit the
114 : * replacement tuple, it must check that this matches before believing the
115 : * replacement is really a match.
116 : *
117 : * cmax is the outdating command's CID, but only when the failure code is
118 : * TM_SelfModified (i.e., something in the current transaction outdated the
119 : * tuple); otherwise cmax is zero. (We make this restriction because
120 : * HeapTupleHeaderGetCmax doesn't work for tuples outdated in other
121 : * transactions.)
122 : */
123 : typedef struct TM_FailureData
124 : {
125 : ItemPointerData ctid;
126 : TransactionId xmax;
127 : CommandId cmax;
128 : bool traversed;
129 : } TM_FailureData;
130 :
131 : /*
132 : * State used when calling table_index_delete_tuples().
133 : *
134 : * Represents the status of table tuples, referenced by table TID and taken by
135 : * index AM from index tuples. State consists of high level parameters of the
136 : * deletion operation, plus two mutable palloc()'d arrays for information
137 : * about the status of individual table tuples. These are conceptually one
138 : * single array. Using two arrays keeps the TM_IndexDelete struct small,
139 : * which makes sorting the first array (the deltids array) fast.
140 : *
141 : * Some index AM callers perform simple index tuple deletion (by specifying
142 : * bottomup = false), and include only known-dead deltids. These known-dead
143 : * entries are all marked knowndeletable = true directly (typically these are
144 : * TIDs from LP_DEAD-marked index tuples), but that isn't strictly required.
145 : *
146 : * Callers that specify bottomup = true are "bottom-up index deletion"
147 : * callers. The considerations for the tableam are more subtle with these
148 : * callers because they ask the tableam to perform highly speculative work,
149 : * and might only expect the tableam to check a small fraction of all entries.
150 : * Caller is not allowed to specify knowndeletable = true for any entry
151 : * because everything is highly speculative. Bottom-up caller provides
152 : * context and hints to tableam -- see comments below for details on how index
153 : * AMs and tableams should coordinate during bottom-up index deletion.
154 : *
155 : * Simple index deletion callers may ask the tableam to perform speculative
156 : * work, too. This is a little like bottom-up deletion, but not too much.
157 : * The tableam will only perform speculative work when it's practically free
158 : * to do so in passing for simple deletion caller (while always performing
159 : * whatever work is is needed to enable knowndeletable/LP_DEAD index tuples to
160 : * be deleted within index AM). This is the real reason why it's possible for
161 : * simple index deletion caller to specify knowndeletable = false up front
162 : * (this means "check if it's possible for me to delete corresponding index
163 : * tuple when it's cheap to do so in passing"). The index AM should only
164 : * include "extra" entries for index tuples whose TIDs point to a table block
165 : * that tableam is expected to have to visit anyway (in the event of a block
166 : * orientated tableam). The tableam isn't strictly obligated to check these
167 : * "extra" TIDs, but a block-based AM should always manage to do so in
168 : * practice.
169 : *
170 : * The final contents of the deltids/status arrays are interesting to callers
171 : * that ask tableam to perform speculative work (i.e. when _any_ items have
172 : * knowndeletable set to false up front). These index AM callers will
173 : * naturally need to consult final state to determine which index tuples are
174 : * in fact deletable.
175 : *
176 : * The index AM can keep track of which index tuple relates to which deltid by
177 : * setting idxoffnum (and/or relying on each entry being uniquely identifiable
178 : * using tid), which is important when the final contents of the array will
179 : * need to be interpreted -- the array can shrink from initial size after
180 : * tableam processing and/or have entries in a new order (tableam may sort
181 : * deltids array for its own reasons). Bottom-up callers may find that final
182 : * ndeltids is 0 on return from call to tableam, in which case no index tuple
183 : * deletions are possible. Simple deletion callers can rely on any entries
184 : * they know to be deletable appearing in the final array as deletable.
185 : */
186 : typedef struct TM_IndexDelete
187 : {
188 : ItemPointerData tid; /* table TID from index tuple */
189 : int16 id; /* Offset into TM_IndexStatus array */
190 : } TM_IndexDelete;
191 :
192 : typedef struct TM_IndexStatus
193 : {
194 : OffsetNumber idxoffnum; /* Index am page offset number */
195 : bool knowndeletable; /* Currently known to be deletable? */
196 :
197 : /* Bottom-up index deletion specific fields follow */
198 : bool promising; /* Promising (duplicate) index tuple? */
199 : int16 freespace; /* Space freed in index if deleted */
200 : } TM_IndexStatus;
201 :
202 : /*
203 : * Index AM/tableam coordination is central to the design of bottom-up index
204 : * deletion. The index AM provides hints about where to look to the tableam
205 : * by marking some entries as "promising". Index AM does this with duplicate
206 : * index tuples that are strongly suspected to be old versions left behind by
207 : * UPDATEs that did not logically modify indexed values. Index AM may find it
208 : * helpful to only mark entries as promising when they're thought to have been
209 : * affected by such an UPDATE in the recent past.
210 : *
211 : * Bottom-up index deletion casts a wide net at first, usually by including
212 : * all TIDs on a target index page. It is up to the tableam to worry about
213 : * the cost of checking transaction status information. The tableam is in
214 : * control, but needs careful guidance from the index AM. Index AM requests
215 : * that bottomupfreespace target be met, while tableam measures progress
216 : * towards that goal by tallying the per-entry freespace value for known
217 : * deletable entries. (All !bottomup callers can just set these space related
218 : * fields to zero.)
219 : */
220 : typedef struct TM_IndexDeleteOp
221 : {
222 : bool bottomup; /* Bottom-up (not simple) deletion? */
223 : int bottomupfreespace; /* Bottom-up space target */
224 :
225 : /* Mutable per-TID information follows (index AM initializes entries) */
226 : int ndeltids; /* Current # of deltids/status elements */
227 : TM_IndexDelete *deltids;
228 : TM_IndexStatus *status;
229 : } TM_IndexDeleteOp;
230 :
231 : /* "options" flag bits for table_tuple_insert */
232 : /* TABLE_INSERT_SKIP_WAL was 0x0001; RelationNeedsWAL() now governs */
233 : #define TABLE_INSERT_SKIP_FSM 0x0002
234 : #define TABLE_INSERT_FROZEN 0x0004
235 : #define TABLE_INSERT_NO_LOGICAL 0x0008
236 :
237 : /* flag bits for table_tuple_lock */
238 : /* Follow tuples whose update is in progress if lock modes don't conflict */
239 : #define TUPLE_LOCK_FLAG_LOCK_UPDATE_IN_PROGRESS (1 << 0)
240 : /* Follow update chain and lock latest version of tuple */
241 : #define TUPLE_LOCK_FLAG_FIND_LAST_VERSION (1 << 1)
242 :
243 :
244 : /* Typedef for callback function for table_index_build_scan */
245 : typedef void (*IndexBuildCallback) (Relation index,
246 : ItemPointer tid,
247 : Datum *values,
248 : bool *isnull,
249 : bool tupleIsAlive,
250 : void *state);
251 :
252 : /*
253 : * API struct for a table AM. Note this must be allocated in a
254 : * server-lifetime manner, typically as a static const struct, which then gets
255 : * returned by FormData_pg_am.amhandler.
256 : *
257 : * In most cases it's not appropriate to call the callbacks directly, use the
258 : * table_* wrapper functions instead.
259 : *
260 : * GetTableAmRoutine() asserts that required callbacks are filled in, remember
261 : * to update when adding a callback.
262 : */
263 : typedef struct TableAmRoutine
264 : {
265 : /* this must be set to T_TableAmRoutine */
266 : NodeTag type;
267 :
268 :
269 : /* ------------------------------------------------------------------------
270 : * Slot related callbacks.
271 : * ------------------------------------------------------------------------
272 : */
273 :
274 : /*
275 : * Return slot implementation suitable for storing a tuple of this AM.
276 : */
277 : const TupleTableSlotOps *(*slot_callbacks) (Relation rel);
278 :
279 :
280 : /* ------------------------------------------------------------------------
281 : * Table scan callbacks.
282 : * ------------------------------------------------------------------------
283 : */
284 :
285 : /*
286 : * Start a scan of `rel`. The callback has to return a TableScanDesc,
287 : * which will typically be embedded in a larger, AM specific, struct.
288 : *
289 : * If nkeys != 0, the results need to be filtered by those scan keys.
290 : *
291 : * pscan, if not NULL, will have already been initialized with
292 : * parallelscan_initialize(), and has to be for the same relation. Will
293 : * only be set coming from table_beginscan_parallel().
294 : *
295 : * `flags` is a bitmask indicating the type of scan (ScanOptions's
296 : * SO_TYPE_*, currently only one may be specified), options controlling
297 : * the scan's behaviour (ScanOptions's SO_ALLOW_*, several may be
298 : * specified, an AM may ignore unsupported ones) and whether the snapshot
299 : * needs to be deallocated at scan_end (ScanOptions's SO_TEMP_SNAPSHOT).
300 : */
301 : TableScanDesc (*scan_begin) (Relation rel,
302 : Snapshot snapshot,
303 : int nkeys, struct ScanKeyData *key,
304 : ParallelTableScanDesc pscan,
305 : uint32 flags);
306 :
307 : /*
308 : * Release resources and deallocate scan. If TableScanDesc.temp_snap,
309 : * TableScanDesc.rs_snapshot needs to be unregistered.
310 : */
311 : void (*scan_end) (TableScanDesc scan);
312 :
313 : /*
314 : * Restart relation scan. If set_params is set to true, allow_{strat,
315 : * sync, pagemode} (see scan_begin) changes should be taken into account.
316 : */
317 : void (*scan_rescan) (TableScanDesc scan, struct ScanKeyData *key,
318 : bool set_params, bool allow_strat,
319 : bool allow_sync, bool allow_pagemode);
320 :
321 : /*
322 : * Return next tuple from `scan`, store in slot.
323 : */
324 : bool (*scan_getnextslot) (TableScanDesc scan,
325 : ScanDirection direction,
326 : TupleTableSlot *slot);
327 :
328 :
329 : /* ------------------------------------------------------------------------
330 : * Parallel table scan related functions.
331 : * ------------------------------------------------------------------------
332 : */
333 :
334 : /*
335 : * Estimate the size of shared memory needed for a parallel scan of this
336 : * relation. The snapshot does not need to be accounted for.
337 : */
338 : Size (*parallelscan_estimate) (Relation rel);
339 :
340 : /*
341 : * Initialize ParallelTableScanDesc for a parallel scan of this relation.
342 : * `pscan` will be sized according to parallelscan_estimate() for the same
343 : * relation.
344 : */
345 : Size (*parallelscan_initialize) (Relation rel,
346 : ParallelTableScanDesc pscan);
347 :
348 : /*
349 : * Reinitialize `pscan` for a new scan. `rel` will be the same relation as
350 : * when `pscan` was initialized by parallelscan_initialize.
351 : */
352 : void (*parallelscan_reinitialize) (Relation rel,
353 : ParallelTableScanDesc pscan);
354 :
355 :
356 : /* ------------------------------------------------------------------------
357 : * Index Scan Callbacks
358 : * ------------------------------------------------------------------------
359 : */
360 :
361 : /*
362 : * Prepare to fetch tuples from the relation, as needed when fetching
363 : * tuples for an index scan. The callback has to return an
364 : * IndexFetchTableData, which the AM will typically embed in a larger
365 : * structure with additional information.
366 : *
367 : * Tuples for an index scan can then be fetched via index_fetch_tuple.
368 : */
369 : struct IndexFetchTableData *(*index_fetch_begin) (Relation rel);
370 :
371 : /*
372 : * Reset index fetch. Typically this will release cross index fetch
373 : * resources held in IndexFetchTableData.
374 : */
375 : void (*index_fetch_reset) (struct IndexFetchTableData *data);
376 :
377 : /*
378 : * Release resources and deallocate index fetch.
379 : */
380 : void (*index_fetch_end) (struct IndexFetchTableData *data);
381 :
382 : /*
383 : * Fetch tuple at `tid` into `slot`, after doing a visibility test
384 : * according to `snapshot`. If a tuple was found and passed the visibility
385 : * test, return true, false otherwise.
386 : *
387 : * Note that AMs that do not necessarily update indexes when indexed
388 : * columns do not change, need to return the current/correct version of
389 : * the tuple that is visible to the snapshot, even if the tid points to an
390 : * older version of the tuple.
391 : *
392 : * *call_again is false on the first call to index_fetch_tuple for a tid.
393 : * If there potentially is another tuple matching the tid, *call_again
394 : * needs to be set to true by index_fetch_tuple, signaling to the caller
395 : * that index_fetch_tuple should be called again for the same tid.
396 : *
397 : * *all_dead, if all_dead is not NULL, should be set to true by
398 : * index_fetch_tuple iff it is guaranteed that no backend needs to see
399 : * that tuple. Index AMs can use that to avoid returning that tid in
400 : * future searches.
401 : */
402 : bool (*index_fetch_tuple) (struct IndexFetchTableData *scan,
403 : ItemPointer tid,
404 : Snapshot snapshot,
405 : TupleTableSlot *slot,
406 : bool *call_again, bool *all_dead);
407 :
408 :
409 : /* ------------------------------------------------------------------------
410 : * Callbacks for non-modifying operations on individual tuples
411 : * ------------------------------------------------------------------------
412 : */
413 :
414 : /*
415 : * Fetch tuple at `tid` into `slot`, after doing a visibility test
416 : * according to `snapshot`. If a tuple was found and passed the visibility
417 : * test, returns true, false otherwise.
418 : */
419 : bool (*tuple_fetch_row_version) (Relation rel,
420 : ItemPointer tid,
421 : Snapshot snapshot,
422 : TupleTableSlot *slot);
423 :
424 : /*
425 : * Is tid valid for a scan of this relation.
426 : */
427 : bool (*tuple_tid_valid) (TableScanDesc scan,
428 : ItemPointer tid);
429 :
430 : /*
431 : * Return the latest version of the tuple at `tid`, by updating `tid` to
432 : * point at the newest version.
433 : */
434 : void (*tuple_get_latest_tid) (TableScanDesc scan,
435 : ItemPointer tid);
436 :
437 : /*
438 : * Does the tuple in `slot` satisfy `snapshot`? The slot needs to be of
439 : * the appropriate type for the AM.
440 : */
441 : bool (*tuple_satisfies_snapshot) (Relation rel,
442 : TupleTableSlot *slot,
443 : Snapshot snapshot);
444 :
445 : /* see table_index_delete_tuples() */
446 : TransactionId (*index_delete_tuples) (Relation rel,
447 : TM_IndexDeleteOp *delstate);
448 :
449 :
450 : /* ------------------------------------------------------------------------
451 : * Manipulations of physical tuples.
452 : * ------------------------------------------------------------------------
453 : */
454 :
455 : /* see table_tuple_insert() for reference about parameters */
456 : void (*tuple_insert) (Relation rel, TupleTableSlot *slot,
457 : CommandId cid, int options,
458 : struct BulkInsertStateData *bistate);
459 :
460 : /* see table_tuple_insert_speculative() for reference about parameters */
461 : void (*tuple_insert_speculative) (Relation rel,
462 : TupleTableSlot *slot,
463 : CommandId cid,
464 : int options,
465 : struct BulkInsertStateData *bistate,
466 : uint32 specToken);
467 :
468 : /* see table_tuple_complete_speculative() for reference about parameters */
469 : void (*tuple_complete_speculative) (Relation rel,
470 : TupleTableSlot *slot,
471 : uint32 specToken,
472 : bool succeeded);
473 :
474 : /* see table_multi_insert() for reference about parameters */
475 : void (*multi_insert) (Relation rel, TupleTableSlot **slots, int nslots,
476 : CommandId cid, int options, struct BulkInsertStateData *bistate);
477 :
478 : /* see table_tuple_delete() for reference about parameters */
479 : TM_Result (*tuple_delete) (Relation rel,
480 : ItemPointer tid,
481 : CommandId cid,
482 : Snapshot snapshot,
483 : Snapshot crosscheck,
484 : bool wait,
485 : TM_FailureData *tmfd,
486 : bool changingPart);
487 :
488 : /* see table_tuple_update() for reference about parameters */
489 : TM_Result (*tuple_update) (Relation rel,
490 : ItemPointer otid,
491 : TupleTableSlot *slot,
492 : CommandId cid,
493 : Snapshot snapshot,
494 : Snapshot crosscheck,
495 : bool wait,
496 : TM_FailureData *tmfd,
497 : LockTupleMode *lockmode,
498 : bool *update_indexes);
499 :
500 : /* see table_tuple_lock() for reference about parameters */
501 : TM_Result (*tuple_lock) (Relation rel,
502 : ItemPointer tid,
503 : Snapshot snapshot,
504 : TupleTableSlot *slot,
505 : CommandId cid,
506 : LockTupleMode mode,
507 : LockWaitPolicy wait_policy,
508 : uint8 flags,
509 : TM_FailureData *tmfd);
510 :
511 : /*
512 : * Perform operations necessary to complete insertions made via
513 : * tuple_insert and multi_insert with a BulkInsertState specified. In-tree
514 : * access methods ceased to use this.
515 : *
516 : * Typically callers of tuple_insert and multi_insert will just pass all
517 : * the flags that apply to them, and each AM has to decide which of them
518 : * make sense for it, and then only take actions in finish_bulk_insert for
519 : * those flags, and ignore others.
520 : *
521 : * Optional callback.
522 : */
523 : void (*finish_bulk_insert) (Relation rel, int options);
524 :
525 :
526 : /* ------------------------------------------------------------------------
527 : * DDL related functionality.
528 : * ------------------------------------------------------------------------
529 : */
530 :
531 : /*
532 : * This callback needs to create a new relation filenode for `rel`, with
533 : * appropriate durability behaviour for `persistence`.
534 : *
535 : * Note that only the subset of the relcache filled by
536 : * RelationBuildLocalRelation() can be relied upon and that the relation's
537 : * catalog entries will either not yet exist (new relation), or will still
538 : * reference the old relfilenode.
539 : *
540 : * As output *freezeXid, *minmulti must be set to the values appropriate
541 : * for pg_class.{relfrozenxid, relminmxid}. For AMs that don't need those
542 : * fields to be filled they can be set to InvalidTransactionId and
543 : * InvalidMultiXactId, respectively.
544 : *
545 : * See also table_relation_set_new_filenode().
546 : */
547 : void (*relation_set_new_filenode) (Relation rel,
548 : const RelFileNode *newrnode,
549 : char persistence,
550 : TransactionId *freezeXid,
551 : MultiXactId *minmulti);
552 :
553 : /*
554 : * This callback needs to remove all contents from `rel`'s current
555 : * relfilenode. No provisions for transactional behaviour need to be made.
556 : * Often this can be implemented by truncating the underlying storage to
557 : * its minimal size.
558 : *
559 : * See also table_relation_nontransactional_truncate().
560 : */
561 : void (*relation_nontransactional_truncate) (Relation rel);
562 :
563 : /*
564 : * See table_relation_copy_data().
565 : *
566 : * This can typically be implemented by directly copying the underlying
567 : * storage, unless it contains references to the tablespace internally.
568 : */
569 : void (*relation_copy_data) (Relation rel,
570 : const RelFileNode *newrnode);
571 :
572 : /* See table_relation_copy_for_cluster() */
573 : void (*relation_copy_for_cluster) (Relation NewTable,
574 : Relation OldTable,
575 : Relation OldIndex,
576 : bool use_sort,
577 : TransactionId OldestXmin,
578 : TransactionId *xid_cutoff,
579 : MultiXactId *multi_cutoff,
580 : double *num_tuples,
581 : double *tups_vacuumed,
582 : double *tups_recently_dead);
583 :
584 : /*
585 : * React to VACUUM command on the relation. The VACUUM can be triggered by
586 : * a user or by autovacuum. The specific actions performed by the AM will
587 : * depend heavily on the individual AM.
588 : *
589 : * On entry a transaction is already established, and the relation is
590 : * locked with a ShareUpdateExclusive lock.
591 : *
592 : * Note that neither VACUUM FULL (and CLUSTER), nor ANALYZE go through
593 : * this routine, even if (for ANALYZE) it is part of the same VACUUM
594 : * command.
595 : *
596 : * There probably, in the future, needs to be a separate callback to
597 : * integrate with autovacuum's scheduling.
598 : */
599 : void (*relation_vacuum) (Relation onerel,
600 : struct VacuumParams *params,
601 : BufferAccessStrategy bstrategy);
602 :
603 : /*
604 : * Prepare to analyze block `blockno` of `scan`. The scan has been started
605 : * with table_beginscan_analyze(). See also
606 : * table_scan_analyze_next_block().
607 : *
608 : * The callback may acquire resources like locks that are held until
609 : * table_scan_analyze_next_tuple() returns false. It e.g. can make sense
610 : * to hold a lock until all tuples on a block have been analyzed by
611 : * scan_analyze_next_tuple.
612 : *
613 : * The callback can return false if the block is not suitable for
614 : * sampling, e.g. because it's a metapage that could never contain tuples.
615 : *
616 : * XXX: This obviously is primarily suited for block-based AMs. It's not
617 : * clear what a good interface for non block based AMs would be, so there
618 : * isn't one yet.
619 : */
620 : bool (*scan_analyze_next_block) (TableScanDesc scan,
621 : BlockNumber blockno,
622 : BufferAccessStrategy bstrategy);
623 :
624 : /*
625 : * See table_scan_analyze_next_tuple().
626 : *
627 : * Not every AM might have a meaningful concept of dead rows, in which
628 : * case it's OK to not increment *deadrows - but note that that may
629 : * influence autovacuum scheduling (see comment for relation_vacuum
630 : * callback).
631 : */
632 : bool (*scan_analyze_next_tuple) (TableScanDesc scan,
633 : TransactionId OldestXmin,
634 : double *liverows,
635 : double *deadrows,
636 : TupleTableSlot *slot);
637 :
638 : /* see table_index_build_range_scan for reference about parameters */
639 : double (*index_build_range_scan) (Relation table_rel,
640 : Relation index_rel,
641 : struct IndexInfo *index_info,
642 : bool allow_sync,
643 : bool anyvisible,
644 : bool progress,
645 : BlockNumber start_blockno,
646 : BlockNumber numblocks,
647 : IndexBuildCallback callback,
648 : void *callback_state,
649 : TableScanDesc scan);
650 :
651 : /* see table_index_validate_scan for reference about parameters */
652 : void (*index_validate_scan) (Relation table_rel,
653 : Relation index_rel,
654 : struct IndexInfo *index_info,
655 : Snapshot snapshot,
656 : struct ValidateIndexState *state);
657 :
658 :
659 : /* ------------------------------------------------------------------------
660 : * Miscellaneous functions.
661 : * ------------------------------------------------------------------------
662 : */
663 :
664 : /*
665 : * See table_relation_size().
666 : *
667 : * Note that currently a few callers use the MAIN_FORKNUM size to figure
668 : * out the range of potentially interesting blocks (brin, analyze). It's
669 : * probable that we'll need to revise the interface for those at some
670 : * point.
671 : */
672 : uint64 (*relation_size) (Relation rel, ForkNumber forkNumber);
673 :
674 :
675 : /*
676 : * This callback should return true if the relation requires a TOAST table
677 : * and false if it does not. It may wish to examine the relation's tuple
678 : * descriptor before making a decision, but if it uses some other method
679 : * of storing large values (or if it does not support them) it can simply
680 : * return false.
681 : */
682 : bool (*relation_needs_toast_table) (Relation rel);
683 :
684 : /*
685 : * This callback should return the OID of the table AM that implements
686 : * TOAST tables for this AM. If the relation_needs_toast_table callback
687 : * always returns false, this callback is not required.
688 : */
689 : Oid (*relation_toast_am) (Relation rel);
690 :
691 : /*
692 : * This callback is invoked when detoasting a value stored in a toast
693 : * table implemented by this AM. See table_relation_fetch_toast_slice()
694 : * for more details.
695 : */
696 : void (*relation_fetch_toast_slice) (Relation toastrel, Oid valueid,
697 : int32 attrsize,
698 : int32 sliceoffset,
699 : int32 slicelength,
700 : struct varlena *result);
701 :
702 :
703 : /* ------------------------------------------------------------------------
704 : * Planner related functions.
705 : * ------------------------------------------------------------------------
706 : */
707 :
708 : /*
709 : * See table_relation_estimate_size().
710 : *
711 : * While block oriented, it shouldn't be too hard for an AM that doesn't
712 : * internally use blocks to convert into a usable representation.
713 : *
714 : * This differs from the relation_size callback by returning size
715 : * estimates (both relation size and tuple count) for planning purposes,
716 : * rather than returning a currently correct estimate.
717 : */
718 : void (*relation_estimate_size) (Relation rel, int32 *attr_widths,
719 : BlockNumber *pages, double *tuples,
720 : double *allvisfrac);
721 :
722 :
723 : /* ------------------------------------------------------------------------
724 : * Executor related functions.
725 : * ------------------------------------------------------------------------
726 : */
727 :
728 : /*
729 : * Prepare to fetch / check / return tuples from `tbmres->blockno` as part
730 : * of a bitmap table scan. `scan` was started via table_beginscan_bm().
731 : * Return false if there are no tuples to be found on the page, true
732 : * otherwise.
733 : *
734 : * This will typically read and pin the target block, and do the necessary
735 : * work to allow scan_bitmap_next_tuple() to return tuples (e.g. it might
736 : * make sense to perform tuple visibility checks at this time). For some
737 : * AMs it will make more sense to do all the work referencing `tbmres`
738 : * contents here, for others it might be better to defer more work to
739 : * scan_bitmap_next_tuple.
740 : *
741 : * If `tbmres->blockno` is -1, this is a lossy scan and all visible tuples
742 : * on the page have to be returned, otherwise the tuples at offsets in
743 : * `tbmres->offsets` need to be returned.
744 : *
745 : * XXX: Currently this may only be implemented if the AM uses md.c as its
746 : * storage manager, and uses ItemPointer->ip_blkid in a manner that maps
747 : * blockids directly to the underlying storage. nodeBitmapHeapscan.c
748 : * performs prefetching directly using that interface. This probably
749 : * needs to be rectified at a later point.
750 : *
751 : * XXX: Currently this may only be implemented if the AM uses the
752 : * visibilitymap, as nodeBitmapHeapscan.c unconditionally accesses it to
753 : * perform prefetching. This probably needs to be rectified at a later
754 : * point.
755 : *
756 : * Optional callback, but either both scan_bitmap_next_block and
757 : * scan_bitmap_next_tuple need to exist, or neither.
758 : */
759 : bool (*scan_bitmap_next_block) (TableScanDesc scan,
760 : struct TBMIterateResult *tbmres);
761 :
762 : /*
763 : * Fetch the next tuple of a bitmap table scan into `slot` and return true
764 : * if a visible tuple was found, false otherwise.
765 : *
766 : * For some AMs it will make more sense to do all the work referencing
767 : * `tbmres` contents in scan_bitmap_next_block, for others it might be
768 : * better to defer more work to this callback.
769 : *
770 : * Optional callback, but either both scan_bitmap_next_block and
771 : * scan_bitmap_next_tuple need to exist, or neither.
772 : */
773 : bool (*scan_bitmap_next_tuple) (TableScanDesc scan,
774 : struct TBMIterateResult *tbmres,
775 : TupleTableSlot *slot);
776 :
777 : /*
778 : * Prepare to fetch tuples from the next block in a sample scan. Return
779 : * false if the sample scan is finished, true otherwise. `scan` was
780 : * started via table_beginscan_sampling().
781 : *
782 : * Typically this will first determine the target block by calling the
783 : * TsmRoutine's NextSampleBlock() callback if not NULL, or alternatively
784 : * perform a sequential scan over all blocks. The determined block is
785 : * then typically read and pinned.
786 : *
787 : * As the TsmRoutine interface is block based, a block needs to be passed
788 : * to NextSampleBlock(). If that's not appropriate for an AM, it
789 : * internally needs to perform mapping between the internal and a block
790 : * based representation.
791 : *
792 : * Note that it's not acceptable to hold deadlock prone resources such as
793 : * lwlocks until scan_sample_next_tuple() has exhausted the tuples on the
794 : * block - the tuple is likely to be returned to an upper query node, and
795 : * the next call could be off a long while. Holding buffer pins and such
796 : * is obviously OK.
797 : *
798 : * Currently it is required to implement this interface, as there's no
799 : * alternative way (contrary e.g. to bitmap scans) to implement sample
800 : * scans. If infeasible to implement, the AM may raise an error.
801 : */
802 : bool (*scan_sample_next_block) (TableScanDesc scan,
803 : struct SampleScanState *scanstate);
804 :
805 : /*
806 : * This callback, only called after scan_sample_next_block has returned
807 : * true, should determine the next tuple to be returned from the selected
808 : * block using the TsmRoutine's NextSampleTuple() callback.
809 : *
810 : * The callback needs to perform visibility checks, and only return
811 : * visible tuples. That obviously can mean calling NextSampleTuple()
812 : * multiple times.
813 : *
814 : * The TsmRoutine interface assumes that there's a maximum offset on a
815 : * given page, so if that doesn't apply to an AM, it needs to emulate that
816 : * assumption somehow.
817 : */
818 : bool (*scan_sample_next_tuple) (TableScanDesc scan,
819 : struct SampleScanState *scanstate,
820 : TupleTableSlot *slot);
821 :
822 : } TableAmRoutine;
823 :
824 :
825 : /* ----------------------------------------------------------------------------
826 : * Slot functions.
827 : * ----------------------------------------------------------------------------
828 : */
829 :
830 : /*
831 : * Returns slot callbacks suitable for holding tuples of the appropriate type
832 : * for the relation. Works for tables, views, foreign tables and partitioned
833 : * tables.
834 : */
835 : extern const TupleTableSlotOps *table_slot_callbacks(Relation rel);
836 :
837 : /*
838 : * Returns slot using the callbacks returned by table_slot_callbacks(), and
839 : * registers it on *reglist.
840 : */
841 : extern TupleTableSlot *table_slot_create(Relation rel, List **reglist);
842 :
843 :
844 : /* ----------------------------------------------------------------------------
845 : * Table scan functions.
846 : * ----------------------------------------------------------------------------
847 : */
848 :
849 : /*
850 : * Start a scan of `rel`. Returned tuples pass a visibility test of
851 : * `snapshot`, and if nkeys != 0, the results are filtered by those scan keys.
852 : */
853 : static inline TableScanDesc
854 126702 : table_beginscan(Relation rel, Snapshot snapshot,
855 : int nkeys, struct ScanKeyData *key)
856 : {
857 126702 : uint32 flags = SO_TYPE_SEQSCAN |
858 : SO_ALLOW_STRAT | SO_ALLOW_SYNC | SO_ALLOW_PAGEMODE;
859 :
860 126702 : return rel->rd_tableam->scan_begin(rel, snapshot, nkeys, key, NULL, flags);
861 : }
862 :
863 : /*
864 : * Like table_beginscan(), but for scanning catalog. It'll automatically use a
865 : * snapshot appropriate for scanning catalog relations.
866 : */
867 : extern TableScanDesc table_beginscan_catalog(Relation rel, int nkeys,
868 : struct ScanKeyData *key);
869 :
870 : /*
871 : * Like table_beginscan(), but table_beginscan_strat() offers an extended API
872 : * that lets the caller control whether a nondefault buffer access strategy
873 : * can be used, and whether syncscan can be chosen (possibly resulting in the
874 : * scan not starting from block zero). Both of these default to true with
875 : * plain table_beginscan.
876 : */
877 : static inline TableScanDesc
878 743542 : table_beginscan_strat(Relation rel, Snapshot snapshot,
879 : int nkeys, struct ScanKeyData *key,
880 : bool allow_strat, bool allow_sync)
881 : {
882 743542 : uint32 flags = SO_TYPE_SEQSCAN | SO_ALLOW_PAGEMODE;
883 :
884 743542 : if (allow_strat)
885 743542 : flags |= SO_ALLOW_STRAT;
886 743542 : if (allow_sync)
887 85154 : flags |= SO_ALLOW_SYNC;
888 :
889 743542 : return rel->rd_tableam->scan_begin(rel, snapshot, nkeys, key, NULL, flags);
890 : }
891 :
892 : /*
893 : * table_beginscan_bm is an alternative entry point for setting up a
894 : * TableScanDesc for a bitmap heap scan. Although that scan technology is
895 : * really quite unlike a standard seqscan, there is just enough commonality to
896 : * make it worth using the same data structure.
897 : */
898 : static inline TableScanDesc
899 12108 : table_beginscan_bm(Relation rel, Snapshot snapshot,
900 : int nkeys, struct ScanKeyData *key)
901 : {
902 12108 : uint32 flags = SO_TYPE_BITMAPSCAN | SO_ALLOW_PAGEMODE;
903 :
904 12108 : return rel->rd_tableam->scan_begin(rel, snapshot, nkeys, key, NULL, flags);
905 : }
906 :
907 : /*
908 : * table_beginscan_sampling is an alternative entry point for setting up a
909 : * TableScanDesc for a TABLESAMPLE scan. As with bitmap scans, it's worth
910 : * using the same data structure although the behavior is rather different.
911 : * In addition to the options offered by table_beginscan_strat, this call
912 : * also allows control of whether page-mode visibility checking is used.
913 : */
914 : static inline TableScanDesc
915 104 : table_beginscan_sampling(Relation rel, Snapshot snapshot,
916 : int nkeys, struct ScanKeyData *key,
917 : bool allow_strat, bool allow_sync,
918 : bool allow_pagemode)
919 : {
920 104 : uint32 flags = SO_TYPE_SAMPLESCAN;
921 :
922 104 : if (allow_strat)
923 96 : flags |= SO_ALLOW_STRAT;
924 104 : if (allow_sync)
925 44 : flags |= SO_ALLOW_SYNC;
926 104 : if (allow_pagemode)
927 88 : flags |= SO_ALLOW_PAGEMODE;
928 :
929 104 : return rel->rd_tableam->scan_begin(rel, snapshot, nkeys, key, NULL, flags);
930 : }
931 :
932 : /*
933 : * table_beginscan_tid is an alternative entry point for setting up a
934 : * TableScanDesc for a Tid scan. As with bitmap scans, it's worth using
935 : * the same data structure although the behavior is rather different.
936 : */
937 : static inline TableScanDesc
938 432 : table_beginscan_tid(Relation rel, Snapshot snapshot)
939 : {
940 432 : uint32 flags = SO_TYPE_TIDSCAN;
941 :
942 432 : return rel->rd_tableam->scan_begin(rel, snapshot, 0, NULL, NULL, flags);
943 : }
944 :
945 : /*
946 : * table_beginscan_analyze is an alternative entry point for setting up a
947 : * TableScanDesc for an ANALYZE scan. As with bitmap scans, it's worth using
948 : * the same data structure although the behavior is rather different.
949 : */
950 : static inline TableScanDesc
951 32136 : table_beginscan_analyze(Relation rel)
952 : {
953 32136 : uint32 flags = SO_TYPE_ANALYZE;
954 :
955 32136 : return rel->rd_tableam->scan_begin(rel, NULL, 0, NULL, NULL, flags);
956 : }
957 :
958 : /*
959 : * End relation scan.
960 : */
961 : static inline void
962 1071388 : table_endscan(TableScanDesc scan)
963 : {
964 1071388 : scan->rs_rd->rd_tableam->scan_end(scan);
965 1071388 : }
966 :
967 : /*
968 : * Restart a relation scan.
969 : */
970 : static inline void
971 934806 : table_rescan(TableScanDesc scan,
972 : struct ScanKeyData *key)
973 : {
974 934806 : scan->rs_rd->rd_tableam->scan_rescan(scan, key, false, false, false, false);
975 934806 : }
976 :
977 : /*
978 : * Restart a relation scan after changing params.
979 : *
980 : * This call allows changing the buffer strategy, syncscan, and pagemode
981 : * options before starting a fresh scan. Note that although the actual use of
982 : * syncscan might change (effectively, enabling or disabling reporting), the
983 : * previously selected startblock will be kept.
984 : */
985 : static inline void
986 26 : table_rescan_set_params(TableScanDesc scan, struct ScanKeyData *key,
987 : bool allow_strat, bool allow_sync, bool allow_pagemode)
988 : {
989 26 : scan->rs_rd->rd_tableam->scan_rescan(scan, key, true,
990 : allow_strat, allow_sync,
991 : allow_pagemode);
992 26 : }
993 :
994 : /*
995 : * Update snapshot used by the scan.
996 : */
997 : extern void table_scan_update_snapshot(TableScanDesc scan, Snapshot snapshot);
998 :
999 : /*
1000 : * Return next tuple from `scan`, store in slot.
1001 : */
1002 : static inline bool
1003 64299576 : table_scan_getnextslot(TableScanDesc sscan, ScanDirection direction, TupleTableSlot *slot)
1004 : {
1005 64299576 : slot->tts_tableOid = RelationGetRelid(sscan->rs_rd);
1006 :
1007 : /*
1008 : * We don't expect direct calls to table_scan_getnextslot with valid
1009 : * CheckXidAlive for catalog or regular tables. See detailed comments in
1010 : * xact.c where these variables are declared.
1011 : */
1012 64299576 : if (unlikely(TransactionIdIsValid(CheckXidAlive) && !bsysscan))
1013 0 : elog(ERROR, "unexpected table_scan_getnextslot call during logical decoding");
1014 :
1015 64299576 : return sscan->rs_rd->rd_tableam->scan_getnextslot(sscan, direction, slot);
1016 : }
1017 :
1018 :
1019 : /* ----------------------------------------------------------------------------
1020 : * Parallel table scan related functions.
1021 : * ----------------------------------------------------------------------------
1022 : */
1023 :
1024 : /*
1025 : * Estimate the size of shared memory needed for a parallel scan of this
1026 : * relation.
1027 : */
1028 : extern Size table_parallelscan_estimate(Relation rel, Snapshot snapshot);
1029 :
1030 : /*
1031 : * Initialize ParallelTableScanDesc for a parallel scan of this
1032 : * relation. `pscan` needs to be sized according to parallelscan_estimate()
1033 : * for the same relation. Call this just once in the leader process; then,
1034 : * individual workers attach via table_beginscan_parallel.
1035 : */
1036 : extern void table_parallelscan_initialize(Relation rel,
1037 : ParallelTableScanDesc pscan,
1038 : Snapshot snapshot);
1039 :
1040 : /*
1041 : * Begin a parallel scan. `pscan` needs to have been initialized with
1042 : * table_parallelscan_initialize(), for the same relation. The initialization
1043 : * does not need to have happened in this backend.
1044 : *
1045 : * Caller must hold a suitable lock on the relation.
1046 : */
1047 : extern TableScanDesc table_beginscan_parallel(Relation rel,
1048 : ParallelTableScanDesc pscan);
1049 :
1050 : /*
1051 : * Restart a parallel scan. Call this in the leader process. Caller is
1052 : * responsible for making sure that all workers have finished the scan
1053 : * beforehand.
1054 : */
1055 : static inline void
1056 152 : table_parallelscan_reinitialize(Relation rel, ParallelTableScanDesc pscan)
1057 : {
1058 152 : rel->rd_tableam->parallelscan_reinitialize(rel, pscan);
1059 152 : }
1060 :
1061 :
1062 : /* ----------------------------------------------------------------------------
1063 : * Index scan related functions.
1064 : * ----------------------------------------------------------------------------
1065 : */
1066 :
1067 : /*
1068 : * Prepare to fetch tuples from the relation, as needed when fetching tuples
1069 : * for an index scan.
1070 : *
1071 : * Tuples for an index scan can then be fetched via table_index_fetch_tuple().
1072 : */
1073 : static inline IndexFetchTableData *
1074 20075922 : table_index_fetch_begin(Relation rel)
1075 : {
1076 20075922 : return rel->rd_tableam->index_fetch_begin(rel);
1077 : }
1078 :
1079 : /*
1080 : * Reset index fetch. Typically this will release cross index fetch resources
1081 : * held in IndexFetchTableData.
1082 : */
1083 : static inline void
1084 17353772 : table_index_fetch_reset(struct IndexFetchTableData *scan)
1085 : {
1086 17353772 : scan->rel->rd_tableam->index_fetch_reset(scan);
1087 17353772 : }
1088 :
1089 : /*
1090 : * Release resources and deallocate index fetch.
1091 : */
1092 : static inline void
1093 20075184 : table_index_fetch_end(struct IndexFetchTableData *scan)
1094 : {
1095 20075184 : scan->rel->rd_tableam->index_fetch_end(scan);
1096 20075184 : }
1097 :
1098 : /*
1099 : * Fetches, as part of an index scan, tuple at `tid` into `slot`, after doing
1100 : * a visibility test according to `snapshot`. If a tuple was found and passed
1101 : * the visibility test, returns true, false otherwise. Note that *tid may be
1102 : * modified when we return true (see later remarks on multiple row versions
1103 : * reachable via a single index entry).
1104 : *
1105 : * *call_again needs to be false on the first call to table_index_fetch_tuple() for
1106 : * a tid. If there potentially is another tuple matching the tid, *call_again
1107 : * will be set to true, signaling that table_index_fetch_tuple() should be called
1108 : * again for the same tid.
1109 : *
1110 : * *all_dead, if all_dead is not NULL, will be set to true by
1111 : * table_index_fetch_tuple() iff it is guaranteed that no backend needs to see
1112 : * that tuple. Index AMs can use that to avoid returning that tid in future
1113 : * searches.
1114 : *
1115 : * The difference between this function and table_tuple_fetch_row_version()
1116 : * is that this function returns the currently visible version of a row if
1117 : * the AM supports storing multiple row versions reachable via a single index
1118 : * entry (like heap's HOT). Whereas table_tuple_fetch_row_version() only
1119 : * evaluates the tuple exactly at `tid`. Outside of index entry ->table tuple
1120 : * lookups, table_tuple_fetch_row_version() is what's usually needed.
1121 : */
1122 : static inline bool
1123 26098706 : table_index_fetch_tuple(struct IndexFetchTableData *scan,
1124 : ItemPointer tid,
1125 : Snapshot snapshot,
1126 : TupleTableSlot *slot,
1127 : bool *call_again, bool *all_dead)
1128 : {
1129 : /*
1130 : * We don't expect direct calls to table_index_fetch_tuple with valid
1131 : * CheckXidAlive for catalog or regular tables. See detailed comments in
1132 : * xact.c where these variables are declared.
1133 : */
1134 26098706 : if (unlikely(TransactionIdIsValid(CheckXidAlive) && !bsysscan))
1135 0 : elog(ERROR, "unexpected table_index_fetch_tuple call during logical decoding");
1136 :
1137 26098706 : return scan->rel->rd_tableam->index_fetch_tuple(scan, tid, snapshot,
1138 : slot, call_again,
1139 : all_dead);
1140 : }
1141 :
1142 : /*
1143 : * This is a convenience wrapper around table_index_fetch_tuple() which
1144 : * returns whether there are table tuple items corresponding to an index
1145 : * entry. This likely is only useful to verify if there's a conflict in a
1146 : * unique index.
1147 : */
1148 : extern bool table_index_fetch_tuple_check(Relation rel,
1149 : ItemPointer tid,
1150 : Snapshot snapshot,
1151 : bool *all_dead);
1152 :
1153 :
1154 : /* ------------------------------------------------------------------------
1155 : * Functions for non-modifying operations on individual tuples
1156 : * ------------------------------------------------------------------------
1157 : */
1158 :
1159 :
1160 : /*
1161 : * Fetch tuple at `tid` into `slot`, after doing a visibility test according to
1162 : * `snapshot`. If a tuple was found and passed the visibility test, returns
1163 : * true, false otherwise.
1164 : *
1165 : * See table_index_fetch_tuple's comment about what the difference between
1166 : * these functions is. It is correct to use this function outside of index
1167 : * entry->table tuple lookups.
1168 : */
1169 : static inline bool
1170 14722 : table_tuple_fetch_row_version(Relation rel,
1171 : ItemPointer tid,
1172 : Snapshot snapshot,
1173 : TupleTableSlot *slot)
1174 : {
1175 : /*
1176 : * We don't expect direct calls to table_tuple_fetch_row_version with
1177 : * valid CheckXidAlive for catalog or regular tables. See detailed
1178 : * comments in xact.c where these variables are declared.
1179 : */
1180 14722 : if (unlikely(TransactionIdIsValid(CheckXidAlive) && !bsysscan))
1181 0 : elog(ERROR, "unexpected table_tuple_fetch_row_version call during logical decoding");
1182 :
1183 14722 : return rel->rd_tableam->tuple_fetch_row_version(rel, tid, snapshot, slot);
1184 : }
1185 :
1186 : /*
1187 : * Verify that `tid` is a potentially valid tuple identifier. That doesn't
1188 : * mean that the pointed to row needs to exist or be visible, but that
1189 : * attempting to fetch the row (e.g. with table_tuple_get_latest_tid() or
1190 : * table_tuple_fetch_row_version()) should not error out if called with that
1191 : * tid.
1192 : *
1193 : * `scan` needs to have been started via table_beginscan().
1194 : */
1195 : static inline bool
1196 186 : table_tuple_tid_valid(TableScanDesc scan, ItemPointer tid)
1197 : {
1198 186 : return scan->rs_rd->rd_tableam->tuple_tid_valid(scan, tid);
1199 : }
1200 :
1201 : /*
1202 : * Return the latest version of the tuple at `tid`, by updating `tid` to
1203 : * point at the newest version.
1204 : */
1205 : extern void table_tuple_get_latest_tid(TableScanDesc scan, ItemPointer tid);
1206 :
1207 : /*
1208 : * Return true iff tuple in slot satisfies the snapshot.
1209 : *
1210 : * This assumes the slot's tuple is valid, and of the appropriate type for the
1211 : * AM.
1212 : *
1213 : * Some AMs might modify the data underlying the tuple as a side-effect. If so
1214 : * they ought to mark the relevant buffer dirty.
1215 : */
1216 : static inline bool
1217 109958 : table_tuple_satisfies_snapshot(Relation rel, TupleTableSlot *slot,
1218 : Snapshot snapshot)
1219 : {
1220 109958 : return rel->rd_tableam->tuple_satisfies_snapshot(rel, slot, snapshot);
1221 : }
1222 :
1223 : /*
1224 : * Determine which index tuples are safe to delete based on their table TID.
1225 : *
1226 : * Determines which entries from index AM caller's TM_IndexDeleteOp state
1227 : * point to vacuumable table tuples. Entries that are found by tableam to be
1228 : * vacuumable are naturally safe for index AM to delete, and so get directly
1229 : * marked as deletable. See comments above TM_IndexDelete and comments above
1230 : * TM_IndexDeleteOp for full details.
1231 : *
1232 : * Returns a latestRemovedXid transaction ID that caller generally places in
1233 : * its index deletion WAL record. This might be used during subsequent REDO
1234 : * of the WAL record when in Hot Standby mode -- a recovery conflict for the
1235 : * index deletion operation might be required on the standby.
1236 : */
1237 : static inline TransactionId
1238 7864 : table_index_delete_tuples(Relation rel, TM_IndexDeleteOp *delstate)
1239 : {
1240 7864 : return rel->rd_tableam->index_delete_tuples(rel, delstate);
1241 : }
1242 :
1243 :
1244 : /* ----------------------------------------------------------------------------
1245 : * Functions for manipulations of physical tuples.
1246 : * ----------------------------------------------------------------------------
1247 : */
1248 :
1249 : /*
1250 : * Insert a tuple from a slot into table AM routine.
1251 : *
1252 : * The options bitmask allows the caller to specify options that may change the
1253 : * behaviour of the AM. The AM will ignore options that it does not support.
1254 : *
1255 : * If the TABLE_INSERT_SKIP_FSM option is specified, AMs are free to not reuse
1256 : * free space in the relation. This can save some cycles when we know the
1257 : * relation is new and doesn't contain useful amounts of free space.
1258 : * TABLE_INSERT_SKIP_FSM is commonly passed directly to
1259 : * RelationGetBufferForTuple. See that method for more information.
1260 : *
1261 : * TABLE_INSERT_FROZEN should only be specified for inserts into
1262 : * relfilenodes created during the current subtransaction and when
1263 : * there are no prior snapshots or pre-existing portals open.
1264 : * This causes rows to be frozen, which is an MVCC violation and
1265 : * requires explicit options chosen by user.
1266 : *
1267 : * TABLE_INSERT_NO_LOGICAL force-disables the emitting of logical decoding
1268 : * information for the tuple. This should solely be used during table rewrites
1269 : * where RelationIsLogicallyLogged(relation) is not yet accurate for the new
1270 : * relation.
1271 : *
1272 : * Note that most of these options will be applied when inserting into the
1273 : * heap's TOAST table, too, if the tuple requires any out-of-line data.
1274 : *
1275 : * The BulkInsertState object (if any; bistate can be NULL for default
1276 : * behavior) is also just passed through to RelationGetBufferForTuple. If
1277 : * `bistate` is provided, table_finish_bulk_insert() needs to be called.
1278 : *
1279 : * On return the slot's tts_tid and tts_tableOid are updated to reflect the
1280 : * insertion. But note that any toasting of fields within the slot is NOT
1281 : * reflected in the slots contents.
1282 : */
1283 : static inline void
1284 13056960 : table_tuple_insert(Relation rel, TupleTableSlot *slot, CommandId cid,
1285 : int options, struct BulkInsertStateData *bistate)
1286 : {
1287 13056960 : rel->rd_tableam->tuple_insert(rel, slot, cid, options,
1288 : bistate);
1289 13056938 : }
1290 :
1291 : /*
1292 : * Perform a "speculative insertion". These can be backed out afterwards
1293 : * without aborting the whole transaction. Other sessions can wait for the
1294 : * speculative insertion to be confirmed, turning it into a regular tuple, or
1295 : * aborted, as if it never existed. Speculatively inserted tuples behave as
1296 : * "value locks" of short duration, used to implement INSERT .. ON CONFLICT.
1297 : *
1298 : * A transaction having performed a speculative insertion has to either abort,
1299 : * or finish the speculative insertion with
1300 : * table_tuple_complete_speculative(succeeded = ...).
1301 : */
1302 : static inline void
1303 3898 : table_tuple_insert_speculative(Relation rel, TupleTableSlot *slot,
1304 : CommandId cid, int options,
1305 : struct BulkInsertStateData *bistate,
1306 : uint32 specToken)
1307 : {
1308 3898 : rel->rd_tableam->tuple_insert_speculative(rel, slot, cid, options,
1309 : bistate, specToken);
1310 3898 : }
1311 :
1312 : /*
1313 : * Complete "speculative insertion" started in the same transaction. If
1314 : * succeeded is true, the tuple is fully inserted, if false, it's removed.
1315 : */
1316 : static inline void
1317 3894 : table_tuple_complete_speculative(Relation rel, TupleTableSlot *slot,
1318 : uint32 specToken, bool succeeded)
1319 : {
1320 3894 : rel->rd_tableam->tuple_complete_speculative(rel, slot, specToken,
1321 : succeeded);
1322 3894 : }
1323 :
1324 : /*
1325 : * Insert multiple tuples into a table.
1326 : *
1327 : * This is like table_tuple_insert(), but inserts multiple tuples in one
1328 : * operation. That's often faster than calling table_tuple_insert() in a loop,
1329 : * because e.g. the AM can reduce WAL logging and page locking overhead.
1330 : *
1331 : * Except for taking `nslots` tuples as input, and an array of TupleTableSlots
1332 : * in `slots`, the parameters for table_multi_insert() are the same as for
1333 : * table_tuple_insert().
1334 : *
1335 : * Note: this leaks memory into the current memory context. You can create a
1336 : * temporary context before calling this, if that's a problem.
1337 : */
1338 : static inline void
1339 1922 : table_multi_insert(Relation rel, TupleTableSlot **slots, int nslots,
1340 : CommandId cid, int options, struct BulkInsertStateData *bistate)
1341 : {
1342 1922 : rel->rd_tableam->multi_insert(rel, slots, nslots,
1343 : cid, options, bistate);
1344 1922 : }
1345 :
1346 : /*
1347 : * Delete a tuple.
1348 : *
1349 : * NB: do not call this directly unless prepared to deal with
1350 : * concurrent-update conditions. Use simple_table_tuple_delete instead.
1351 : *
1352 : * Input parameters:
1353 : * relation - table to be modified (caller must hold suitable lock)
1354 : * tid - TID of tuple to be deleted
1355 : * cid - delete command ID (used for visibility test, and stored into
1356 : * cmax if successful)
1357 : * crosscheck - if not InvalidSnapshot, also check tuple against this
1358 : * wait - true if should wait for any conflicting update to commit/abort
1359 : * Output parameters:
1360 : * tmfd - filled in failure cases (see below)
1361 : * changingPart - true iff the tuple is being moved to another partition
1362 : * table due to an update of the partition key. Otherwise, false.
1363 : *
1364 : * Normal, successful return value is TM_Ok, which means we did actually
1365 : * delete it. Failure return codes are TM_SelfModified, TM_Updated, and
1366 : * TM_BeingModified (the last only possible if wait == false).
1367 : *
1368 : * In the failure cases, the routine fills *tmfd with the tuple's t_ctid,
1369 : * t_xmax, and, if possible, and, if possible, t_cmax. See comments for
1370 : * struct TM_FailureData for additional info.
1371 : */
1372 : static inline TM_Result
1373 993956 : table_tuple_delete(Relation rel, ItemPointer tid, CommandId cid,
1374 : Snapshot snapshot, Snapshot crosscheck, bool wait,
1375 : TM_FailureData *tmfd, bool changingPart)
1376 : {
1377 993956 : return rel->rd_tableam->tuple_delete(rel, tid, cid,
1378 : snapshot, crosscheck,
1379 : wait, tmfd, changingPart);
1380 : }
1381 :
1382 : /*
1383 : * Update a tuple.
1384 : *
1385 : * NB: do not call this directly unless you are prepared to deal with
1386 : * concurrent-update conditions. Use simple_table_tuple_update instead.
1387 : *
1388 : * Input parameters:
1389 : * relation - table to be modified (caller must hold suitable lock)
1390 : * otid - TID of old tuple to be replaced
1391 : * slot - newly constructed tuple data to store
1392 : * cid - update command ID (used for visibility test, and stored into
1393 : * cmax/cmin if successful)
1394 : * crosscheck - if not InvalidSnapshot, also check old tuple against this
1395 : * wait - true if should wait for any conflicting update to commit/abort
1396 : * Output parameters:
1397 : * tmfd - filled in failure cases (see below)
1398 : * lockmode - filled with lock mode acquired on tuple
1399 : * update_indexes - in success cases this is set to true if new index entries
1400 : * are required for this tuple
1401 : *
1402 : * Normal, successful return value is TM_Ok, which means we did actually
1403 : * update it. Failure return codes are TM_SelfModified, TM_Updated, and
1404 : * TM_BeingModified (the last only possible if wait == false).
1405 : *
1406 : * On success, the slot's tts_tid and tts_tableOid are updated to match the new
1407 : * stored tuple; in particular, slot->tts_tid is set to the TID where the
1408 : * new tuple was inserted, and its HEAP_ONLY_TUPLE flag is set iff a HOT
1409 : * update was done. However, any TOAST changes in the new tuple's
1410 : * data are not reflected into *newtup.
1411 : *
1412 : * In the failure cases, the routine fills *tmfd with the tuple's t_ctid,
1413 : * t_xmax, and, if possible, t_cmax. See comments for struct TM_FailureData
1414 : * for additional info.
1415 : */
1416 : static inline TM_Result
1417 217878 : table_tuple_update(Relation rel, ItemPointer otid, TupleTableSlot *slot,
1418 : CommandId cid, Snapshot snapshot, Snapshot crosscheck,
1419 : bool wait, TM_FailureData *tmfd, LockTupleMode *lockmode,
1420 : bool *update_indexes)
1421 : {
1422 217878 : return rel->rd_tableam->tuple_update(rel, otid, slot,
1423 : cid, snapshot, crosscheck,
1424 : wait, tmfd,
1425 : lockmode, update_indexes);
1426 : }
1427 :
1428 : /*
1429 : * Lock a tuple in the specified mode.
1430 : *
1431 : * Input parameters:
1432 : * relation: relation containing tuple (caller must hold suitable lock)
1433 : * tid: TID of tuple to lock
1434 : * snapshot: snapshot to use for visibility determinations
1435 : * cid: current command ID (used for visibility test, and stored into
1436 : * tuple's cmax if lock is successful)
1437 : * mode: lock mode desired
1438 : * wait_policy: what to do if tuple lock is not available
1439 : * flags:
1440 : * If TUPLE_LOCK_FLAG_LOCK_UPDATE_IN_PROGRESS, follow the update chain to
1441 : * also lock descendant tuples if lock modes don't conflict.
1442 : * If TUPLE_LOCK_FLAG_FIND_LAST_VERSION, follow the update chain and lock
1443 : * latest version.
1444 : *
1445 : * Output parameters:
1446 : * *slot: contains the target tuple
1447 : * *tmfd: filled in failure cases (see below)
1448 : *
1449 : * Function result may be:
1450 : * TM_Ok: lock was successfully acquired
1451 : * TM_Invisible: lock failed because tuple was never visible to us
1452 : * TM_SelfModified: lock failed because tuple updated by self
1453 : * TM_Updated: lock failed because tuple updated by other xact
1454 : * TM_Deleted: lock failed because tuple deleted by other xact
1455 : * TM_WouldBlock: lock couldn't be acquired and wait_policy is skip
1456 : *
1457 : * In the failure cases other than TM_Invisible and TM_Deleted, the routine
1458 : * fills *tmfd with the tuple's t_ctid, t_xmax, and, if possible, t_cmax. See
1459 : * comments for struct TM_FailureData for additional info.
1460 : */
1461 : static inline TM_Result
1462 69338 : table_tuple_lock(Relation rel, ItemPointer tid, Snapshot snapshot,
1463 : TupleTableSlot *slot, CommandId cid, LockTupleMode mode,
1464 : LockWaitPolicy wait_policy, uint8 flags,
1465 : TM_FailureData *tmfd)
1466 : {
1467 69338 : return rel->rd_tableam->tuple_lock(rel, tid, snapshot, slot,
1468 : cid, mode, wait_policy,
1469 : flags, tmfd);
1470 : }
1471 :
1472 : /*
1473 : * Perform operations necessary to complete insertions made via
1474 : * tuple_insert and multi_insert with a BulkInsertState specified.
1475 : */
1476 : static inline void
1477 2348 : table_finish_bulk_insert(Relation rel, int options)
1478 : {
1479 : /* optional callback */
1480 2348 : if (rel->rd_tableam && rel->rd_tableam->finish_bulk_insert)
1481 0 : rel->rd_tableam->finish_bulk_insert(rel, options);
1482 2348 : }
1483 :
1484 :
1485 : /* ------------------------------------------------------------------------
1486 : * DDL related functionality.
1487 : * ------------------------------------------------------------------------
1488 : */
1489 :
1490 : /*
1491 : * Create storage for `rel` in `newrnode`, with persistence set to
1492 : * `persistence`.
1493 : *
1494 : * This is used both during relation creation and various DDL operations to
1495 : * create a new relfilenode that can be filled from scratch. When creating
1496 : * new storage for an existing relfilenode, this should be called before the
1497 : * relcache entry has been updated.
1498 : *
1499 : * *freezeXid, *minmulti are set to the xid / multixact horizon for the table
1500 : * that pg_class.{relfrozenxid, relminmxid} have to be set to.
1501 : */
1502 : static inline void
1503 70710 : table_relation_set_new_filenode(Relation rel,
1504 : const RelFileNode *newrnode,
1505 : char persistence,
1506 : TransactionId *freezeXid,
1507 : MultiXactId *minmulti)
1508 : {
1509 70710 : rel->rd_tableam->relation_set_new_filenode(rel, newrnode, persistence,
1510 : freezeXid, minmulti);
1511 70710 : }
1512 :
1513 : /*
1514 : * Remove all table contents from `rel`, in a non-transactional manner.
1515 : * Non-transactional meaning that there's no need to support rollbacks. This
1516 : * commonly only is used to perform truncations for relfilenodes created in the
1517 : * current transaction.
1518 : */
1519 : static inline void
1520 404 : table_relation_nontransactional_truncate(Relation rel)
1521 : {
1522 404 : rel->rd_tableam->relation_nontransactional_truncate(rel);
1523 404 : }
1524 :
1525 : /*
1526 : * Copy data from `rel` into the new relfilenode `newrnode`. The new
1527 : * relfilenode may not have storage associated before this function is
1528 : * called. This is only supposed to be used for low level operations like
1529 : * changing a relation's tablespace.
1530 : */
1531 : static inline void
1532 22 : table_relation_copy_data(Relation rel, const RelFileNode *newrnode)
1533 : {
1534 22 : rel->rd_tableam->relation_copy_data(rel, newrnode);
1535 22 : }
1536 :
1537 : /*
1538 : * Copy data from `OldTable` into `NewTable`, as part of a CLUSTER or VACUUM
1539 : * FULL.
1540 : *
1541 : * Additional Input parameters:
1542 : * - use_sort - if true, the table contents are sorted appropriate for
1543 : * `OldIndex`; if false and OldIndex is not InvalidOid, the data is copied
1544 : * in that index's order; if false and OldIndex is InvalidOid, no sorting is
1545 : * performed
1546 : * - OldIndex - see use_sort
1547 : * - OldestXmin - computed by vacuum_set_xid_limits(), even when
1548 : * not needed for the relation's AM
1549 : * - *xid_cutoff - ditto
1550 : * - *multi_cutoff - ditto
1551 : *
1552 : * Output parameters:
1553 : * - *xid_cutoff - rel's new relfrozenxid value, may be invalid
1554 : * - *multi_cutoff - rel's new relminmxid value, may be invalid
1555 : * - *tups_vacuumed - stats, for logging, if appropriate for AM
1556 : * - *tups_recently_dead - stats, for logging, if appropriate for AM
1557 : */
1558 : static inline void
1559 328 : table_relation_copy_for_cluster(Relation OldTable, Relation NewTable,
1560 : Relation OldIndex,
1561 : bool use_sort,
1562 : TransactionId OldestXmin,
1563 : TransactionId *xid_cutoff,
1564 : MultiXactId *multi_cutoff,
1565 : double *num_tuples,
1566 : double *tups_vacuumed,
1567 : double *tups_recently_dead)
1568 : {
1569 328 : OldTable->rd_tableam->relation_copy_for_cluster(OldTable, NewTable, OldIndex,
1570 : use_sort, OldestXmin,
1571 : xid_cutoff, multi_cutoff,
1572 : num_tuples, tups_vacuumed,
1573 : tups_recently_dead);
1574 328 : }
1575 :
1576 : /*
1577 : * Perform VACUUM on the relation. The VACUUM can be triggered by a user or by
1578 : * autovacuum. The specific actions performed by the AM will depend heavily on
1579 : * the individual AM.
1580 : *
1581 : * On entry a transaction needs to already been established, and the
1582 : * table is locked with a ShareUpdateExclusive lock.
1583 : *
1584 : * Note that neither VACUUM FULL (and CLUSTER), nor ANALYZE go through this
1585 : * routine, even if (for ANALYZE) it is part of the same VACUUM command.
1586 : */
1587 : static inline void
1588 49264 : table_relation_vacuum(Relation rel, struct VacuumParams *params,
1589 : BufferAccessStrategy bstrategy)
1590 : {
1591 49264 : rel->rd_tableam->relation_vacuum(rel, params, bstrategy);
1592 49264 : }
1593 :
1594 : /*
1595 : * Prepare to analyze block `blockno` of `scan`. The scan needs to have been
1596 : * started with table_beginscan_analyze(). Note that this routine might
1597 : * acquire resources like locks that are held until
1598 : * table_scan_analyze_next_tuple() returns false.
1599 : *
1600 : * Returns false if block is unsuitable for sampling, true otherwise.
1601 : */
1602 : static inline bool
1603 199734 : table_scan_analyze_next_block(TableScanDesc scan, BlockNumber blockno,
1604 : BufferAccessStrategy bstrategy)
1605 : {
1606 199734 : return scan->rs_rd->rd_tableam->scan_analyze_next_block(scan, blockno,
1607 : bstrategy);
1608 : }
1609 :
1610 : /*
1611 : * Iterate over tuples in the block selected with
1612 : * table_scan_analyze_next_block() (which needs to have returned true, and
1613 : * this routine may not have returned false for the same block before). If a
1614 : * tuple that's suitable for sampling is found, true is returned and a tuple
1615 : * is stored in `slot`.
1616 : *
1617 : * *liverows and *deadrows are incremented according to the encountered
1618 : * tuples.
1619 : */
1620 : static inline bool
1621 15251722 : table_scan_analyze_next_tuple(TableScanDesc scan, TransactionId OldestXmin,
1622 : double *liverows, double *deadrows,
1623 : TupleTableSlot *slot)
1624 : {
1625 15251722 : return scan->rs_rd->rd_tableam->scan_analyze_next_tuple(scan, OldestXmin,
1626 : liverows, deadrows,
1627 : slot);
1628 : }
1629 :
1630 : /*
1631 : * table_index_build_scan - scan the table to find tuples to be indexed
1632 : *
1633 : * This is called back from an access-method-specific index build procedure
1634 : * after the AM has done whatever setup it needs. The parent table relation
1635 : * is scanned to find tuples that should be entered into the index. Each
1636 : * such tuple is passed to the AM's callback routine, which does the right
1637 : * things to add it to the new index. After we return, the AM's index
1638 : * build procedure does whatever cleanup it needs.
1639 : *
1640 : * The total count of live tuples is returned. This is for updating pg_class
1641 : * statistics. (It's annoying not to be able to do that here, but we want to
1642 : * merge that update with others; see index_update_stats.) Note that the
1643 : * index AM itself must keep track of the number of index tuples; we don't do
1644 : * so here because the AM might reject some of the tuples for its own reasons,
1645 : * such as being unable to store NULLs.
1646 : *
1647 : * If 'progress', the PROGRESS_SCAN_BLOCKS_TOTAL counter is updated when
1648 : * starting the scan, and PROGRESS_SCAN_BLOCKS_DONE is updated as we go along.
1649 : *
1650 : * A side effect is to set indexInfo->ii_BrokenHotChain to true if we detect
1651 : * any potentially broken HOT chains. Currently, we set this if there are any
1652 : * RECENTLY_DEAD or DELETE_IN_PROGRESS entries in a HOT chain, without trying
1653 : * very hard to detect whether they're really incompatible with the chain tip.
1654 : * This only really makes sense for heap AM, it might need to be generalized
1655 : * for other AMs later.
1656 : */
1657 : static inline double
1658 85474 : table_index_build_scan(Relation table_rel,
1659 : Relation index_rel,
1660 : struct IndexInfo *index_info,
1661 : bool allow_sync,
1662 : bool progress,
1663 : IndexBuildCallback callback,
1664 : void *callback_state,
1665 : TableScanDesc scan)
1666 : {
1667 85474 : return table_rel->rd_tableam->index_build_range_scan(table_rel,
1668 : index_rel,
1669 : index_info,
1670 : allow_sync,
1671 : false,
1672 : progress,
1673 : 0,
1674 : InvalidBlockNumber,
1675 : callback,
1676 : callback_state,
1677 : scan);
1678 : }
1679 :
1680 : /*
1681 : * As table_index_build_scan(), except that instead of scanning the complete
1682 : * table, only the given number of blocks are scanned. Scan to end-of-rel can
1683 : * be signaled by passing InvalidBlockNumber as numblocks. Note that
1684 : * restricting the range to scan cannot be done when requesting syncscan.
1685 : *
1686 : * When "anyvisible" mode is requested, all tuples visible to any transaction
1687 : * are indexed and counted as live, including those inserted or deleted by
1688 : * transactions that are still in progress.
1689 : */
1690 : static inline double
1691 42 : table_index_build_range_scan(Relation table_rel,
1692 : Relation index_rel,
1693 : struct IndexInfo *index_info,
1694 : bool allow_sync,
1695 : bool anyvisible,
1696 : bool progress,
1697 : BlockNumber start_blockno,
1698 : BlockNumber numblocks,
1699 : IndexBuildCallback callback,
1700 : void *callback_state,
1701 : TableScanDesc scan)
1702 : {
1703 42 : return table_rel->rd_tableam->index_build_range_scan(table_rel,
1704 : index_rel,
1705 : index_info,
1706 : allow_sync,
1707 : anyvisible,
1708 : progress,
1709 : start_blockno,
1710 : numblocks,
1711 : callback,
1712 : callback_state,
1713 : scan);
1714 : }
1715 :
1716 : /*
1717 : * table_index_validate_scan - second table scan for concurrent index build
1718 : *
1719 : * See validate_index() for an explanation.
1720 : */
1721 : static inline void
1722 190 : table_index_validate_scan(Relation table_rel,
1723 : Relation index_rel,
1724 : struct IndexInfo *index_info,
1725 : Snapshot snapshot,
1726 : struct ValidateIndexState *state)
1727 : {
1728 190 : table_rel->rd_tableam->index_validate_scan(table_rel,
1729 : index_rel,
1730 : index_info,
1731 : snapshot,
1732 : state);
1733 190 : }
1734 :
1735 :
1736 : /* ----------------------------------------------------------------------------
1737 : * Miscellaneous functionality
1738 : * ----------------------------------------------------------------------------
1739 : */
1740 :
1741 : /*
1742 : * Return the current size of `rel` in bytes. If `forkNumber` is
1743 : * InvalidForkNumber, return the relation's overall size, otherwise the size
1744 : * for the indicated fork.
1745 : *
1746 : * Note that the overall size might not be the equivalent of the sum of sizes
1747 : * for the individual forks for some AMs, e.g. because the AMs storage does
1748 : * not neatly map onto the builtin types of forks.
1749 : */
1750 : static inline uint64
1751 2561540 : table_relation_size(Relation rel, ForkNumber forkNumber)
1752 : {
1753 2561540 : return rel->rd_tableam->relation_size(rel, forkNumber);
1754 : }
1755 :
1756 : /*
1757 : * table_relation_needs_toast_table - does this relation need a toast table?
1758 : */
1759 : static inline bool
1760 36904 : table_relation_needs_toast_table(Relation rel)
1761 : {
1762 36904 : return rel->rd_tableam->relation_needs_toast_table(rel);
1763 : }
1764 :
1765 : /*
1766 : * Return the OID of the AM that should be used to implement the TOAST table
1767 : * for this relation.
1768 : */
1769 : static inline Oid
1770 22546 : table_relation_toast_am(Relation rel)
1771 : {
1772 22546 : return rel->rd_tableam->relation_toast_am(rel);
1773 : }
1774 :
1775 : /*
1776 : * Fetch all or part of a TOAST value from a TOAST table.
1777 : *
1778 : * If this AM is never used to implement a TOAST table, then this callback
1779 : * is not needed. But, if toasted values are ever stored in a table of this
1780 : * type, then you will need this callback.
1781 : *
1782 : * toastrel is the relation in which the toasted value is stored.
1783 : *
1784 : * valueid identifes which toast value is to be fetched. For the heap,
1785 : * this corresponds to the values stored in the chunk_id column.
1786 : *
1787 : * attrsize is the total size of the toast value to be fetched.
1788 : *
1789 : * sliceoffset is the offset within the toast value of the first byte that
1790 : * should be fetched.
1791 : *
1792 : * slicelength is the number of bytes from the toast value that should be
1793 : * fetched.
1794 : *
1795 : * result is caller-allocated space into which the fetched bytes should be
1796 : * stored.
1797 : */
1798 : static inline void
1799 46722 : table_relation_fetch_toast_slice(Relation toastrel, Oid valueid,
1800 : int32 attrsize, int32 sliceoffset,
1801 : int32 slicelength, struct varlena *result)
1802 : {
1803 46722 : toastrel->rd_tableam->relation_fetch_toast_slice(toastrel, valueid,
1804 : attrsize,
1805 : sliceoffset, slicelength,
1806 : result);
1807 46722 : }
1808 :
1809 :
1810 : /* ----------------------------------------------------------------------------
1811 : * Planner related functionality
1812 : * ----------------------------------------------------------------------------
1813 : */
1814 :
1815 : /*
1816 : * Estimate the current size of the relation, as an AM specific workhorse for
1817 : * estimate_rel_size(). Look there for an explanation of the parameters.
1818 : */
1819 : static inline void
1820 271570 : table_relation_estimate_size(Relation rel, int32 *attr_widths,
1821 : BlockNumber *pages, double *tuples,
1822 : double *allvisfrac)
1823 : {
1824 271570 : rel->rd_tableam->relation_estimate_size(rel, attr_widths, pages, tuples,
1825 : allvisfrac);
1826 271570 : }
1827 :
1828 :
1829 : /* ----------------------------------------------------------------------------
1830 : * Executor related functionality
1831 : * ----------------------------------------------------------------------------
1832 : */
1833 :
1834 : /*
1835 : * Prepare to fetch / check / return tuples from `tbmres->blockno` as part of
1836 : * a bitmap table scan. `scan` needs to have been started via
1837 : * table_beginscan_bm(). Returns false if there are no tuples to be found on
1838 : * the page, true otherwise.
1839 : *
1840 : * Note, this is an optionally implemented function, therefore should only be
1841 : * used after verifying the presence (at plan time or such).
1842 : */
1843 : static inline bool
1844 139080 : table_scan_bitmap_next_block(TableScanDesc scan,
1845 : struct TBMIterateResult *tbmres)
1846 : {
1847 : /*
1848 : * We don't expect direct calls to table_scan_bitmap_next_block with valid
1849 : * CheckXidAlive for catalog or regular tables. See detailed comments in
1850 : * xact.c where these variables are declared.
1851 : */
1852 139080 : if (unlikely(TransactionIdIsValid(CheckXidAlive) && !bsysscan))
1853 0 : elog(ERROR, "unexpected table_scan_bitmap_next_block call during logical decoding");
1854 :
1855 139080 : return scan->rs_rd->rd_tableam->scan_bitmap_next_block(scan,
1856 : tbmres);
1857 : }
1858 :
1859 : /*
1860 : * Fetch the next tuple of a bitmap table scan into `slot` and return true if
1861 : * a visible tuple was found, false otherwise.
1862 : * table_scan_bitmap_next_block() needs to previously have selected a
1863 : * block (i.e. returned true), and no previous
1864 : * table_scan_bitmap_next_tuple() for the same block may have
1865 : * returned false.
1866 : */
1867 : static inline bool
1868 2525398 : table_scan_bitmap_next_tuple(TableScanDesc scan,
1869 : struct TBMIterateResult *tbmres,
1870 : TupleTableSlot *slot)
1871 : {
1872 : /*
1873 : * We don't expect direct calls to table_scan_bitmap_next_tuple with valid
1874 : * CheckXidAlive for catalog or regular tables. See detailed comments in
1875 : * xact.c where these variables are declared.
1876 : */
1877 2525398 : if (unlikely(TransactionIdIsValid(CheckXidAlive) && !bsysscan))
1878 0 : elog(ERROR, "unexpected table_scan_bitmap_next_tuple call during logical decoding");
1879 :
1880 2525398 : return scan->rs_rd->rd_tableam->scan_bitmap_next_tuple(scan,
1881 : tbmres,
1882 : slot);
1883 : }
1884 :
1885 : /*
1886 : * Prepare to fetch tuples from the next block in a sample scan. Returns false
1887 : * if the sample scan is finished, true otherwise. `scan` needs to have been
1888 : * started via table_beginscan_sampling().
1889 : *
1890 : * This will call the TsmRoutine's NextSampleBlock() callback if necessary
1891 : * (i.e. NextSampleBlock is not NULL), or perform a sequential scan over the
1892 : * underlying relation.
1893 : */
1894 : static inline bool
1895 8654 : table_scan_sample_next_block(TableScanDesc scan,
1896 : struct SampleScanState *scanstate)
1897 : {
1898 : /*
1899 : * We don't expect direct calls to table_scan_sample_next_block with valid
1900 : * CheckXidAlive for catalog or regular tables. See detailed comments in
1901 : * xact.c where these variables are declared.
1902 : */
1903 8654 : if (unlikely(TransactionIdIsValid(CheckXidAlive) && !bsysscan))
1904 0 : elog(ERROR, "unexpected table_scan_sample_next_block call during logical decoding");
1905 8654 : return scan->rs_rd->rd_tableam->scan_sample_next_block(scan, scanstate);
1906 : }
1907 :
1908 : /*
1909 : * Fetch the next sample tuple into `slot` and return true if a visible tuple
1910 : * was found, false otherwise. table_scan_sample_next_block() needs to
1911 : * previously have selected a block (i.e. returned true), and no previous
1912 : * table_scan_sample_next_tuple() for the same block may have returned false.
1913 : *
1914 : * This will call the TsmRoutine's NextSampleTuple() callback.
1915 : */
1916 : static inline bool
1917 169452 : table_scan_sample_next_tuple(TableScanDesc scan,
1918 : struct SampleScanState *scanstate,
1919 : TupleTableSlot *slot)
1920 : {
1921 : /*
1922 : * We don't expect direct calls to table_scan_sample_next_tuple with valid
1923 : * CheckXidAlive for catalog or regular tables. See detailed comments in
1924 : * xact.c where these variables are declared.
1925 : */
1926 169452 : if (unlikely(TransactionIdIsValid(CheckXidAlive) && !bsysscan))
1927 0 : elog(ERROR, "unexpected table_scan_sample_next_tuple call during logical decoding");
1928 169452 : return scan->rs_rd->rd_tableam->scan_sample_next_tuple(scan, scanstate,
1929 : slot);
1930 : }
1931 :
1932 :
1933 : /* ----------------------------------------------------------------------------
1934 : * Functions to make modifications a bit simpler.
1935 : * ----------------------------------------------------------------------------
1936 : */
1937 :
1938 : extern void simple_table_tuple_insert(Relation rel, TupleTableSlot *slot);
1939 : extern void simple_table_tuple_delete(Relation rel, ItemPointer tid,
1940 : Snapshot snapshot);
1941 : extern void simple_table_tuple_update(Relation rel, ItemPointer otid,
1942 : TupleTableSlot *slot, Snapshot snapshot,
1943 : bool *update_indexes);
1944 :
1945 :
1946 : /* ----------------------------------------------------------------------------
1947 : * Helper functions to implement parallel scans for block oriented AMs.
1948 : * ----------------------------------------------------------------------------
1949 : */
1950 :
1951 : extern Size table_block_parallelscan_estimate(Relation rel);
1952 : extern Size table_block_parallelscan_initialize(Relation rel,
1953 : ParallelTableScanDesc pscan);
1954 : extern void table_block_parallelscan_reinitialize(Relation rel,
1955 : ParallelTableScanDesc pscan);
1956 : extern BlockNumber table_block_parallelscan_nextpage(Relation rel,
1957 : ParallelBlockTableScanWorker pbscanwork,
1958 : ParallelBlockTableScanDesc pbscan);
1959 : extern void table_block_parallelscan_startblock_init(Relation rel,
1960 : ParallelBlockTableScanWorker pbscanwork,
1961 : ParallelBlockTableScanDesc pbscan);
1962 :
1963 :
1964 : /* ----------------------------------------------------------------------------
1965 : * Helper functions to implement relation sizing for block oriented AMs.
1966 : * ----------------------------------------------------------------------------
1967 : */
1968 :
1969 : extern uint64 table_block_relation_size(Relation rel, ForkNumber forkNumber);
1970 : extern void table_block_relation_estimate_size(Relation rel,
1971 : int32 *attr_widths,
1972 : BlockNumber *pages,
1973 : double *tuples,
1974 : double *allvisfrac,
1975 : Size overhead_bytes_per_tuple,
1976 : Size usable_bytes_per_page);
1977 :
1978 : /* ----------------------------------------------------------------------------
1979 : * Functions in tableamapi.c
1980 : * ----------------------------------------------------------------------------
1981 : */
1982 :
1983 : extern const TableAmRoutine *GetTableAmRoutine(Oid amhandler);
1984 : extern const TableAmRoutine *GetHeapamTableAmRoutine(void);
1985 : extern bool check_default_table_access_method(char **newval, void **extra,
1986 : GucSource source);
1987 :
1988 : #endif /* TABLEAM_H */
|