Line data Source code
1 : /*-------------------------------------------------------------------------
2 : *
3 : * tableam.h
4 : * POSTGRES table access method definitions.
5 : *
6 : *
7 : * Portions Copyright (c) 1996-2025, 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 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_TIDRANGESCAN = 1 << 4,
53 : SO_TYPE_ANALYZE = 1 << 5,
54 :
55 : /* several of SO_ALLOW_* may be specified */
56 : /* allow or disallow use of access strategy */
57 : SO_ALLOW_STRAT = 1 << 6,
58 : /* report location to syncscan logic? */
59 : SO_ALLOW_SYNC = 1 << 7,
60 : /* verify visibility page-at-a-time? */
61 : SO_ALLOW_PAGEMODE = 1 << 8,
62 :
63 : /* unregister snapshot at scan end? */
64 : SO_TEMP_SNAPSHOT = 1 << 9,
65 :
66 : /*
67 : * At the discretion of the table AM, bitmap table scans may be able to
68 : * skip fetching a block from the table if none of the table data is
69 : * needed. If table data may be needed, set SO_NEED_TUPLES.
70 : */
71 : SO_NEED_TUPLES = 1 << 10,
72 : } ScanOptions;
73 :
74 : /*
75 : * Result codes for table_{update,delete,lock_tuple}, and for visibility
76 : * routines inside table AMs.
77 : */
78 : typedef enum TM_Result
79 : {
80 : /*
81 : * Signals that the action succeeded (i.e. update/delete performed, lock
82 : * was acquired)
83 : */
84 : TM_Ok,
85 :
86 : /* The affected tuple wasn't visible to the relevant snapshot */
87 : TM_Invisible,
88 :
89 : /* The affected tuple was already modified by the calling backend */
90 : TM_SelfModified,
91 :
92 : /*
93 : * The affected tuple was updated by another transaction. This includes
94 : * the case where tuple was moved to another partition.
95 : */
96 : TM_Updated,
97 :
98 : /* The affected tuple was deleted by another transaction */
99 : TM_Deleted,
100 :
101 : /*
102 : * The affected tuple is currently being modified by another session. This
103 : * will only be returned if table_(update/delete/lock_tuple) are
104 : * instructed not to wait.
105 : */
106 : TM_BeingModified,
107 :
108 : /* lock couldn't be acquired, action skipped. Only used by lock_tuple */
109 : TM_WouldBlock,
110 : } TM_Result;
111 :
112 : /*
113 : * Result codes for table_update(..., update_indexes*..).
114 : * Used to determine which indexes to update.
115 : */
116 : typedef enum TU_UpdateIndexes
117 : {
118 : /* No indexed columns were updated (incl. TID addressing of tuple) */
119 : TU_None,
120 :
121 : /* A non-summarizing indexed column was updated, or the TID has changed */
122 : TU_All,
123 :
124 : /* Only summarized columns were updated, TID is unchanged */
125 : TU_Summarizing,
126 : } TU_UpdateIndexes;
127 :
128 : /*
129 : * When table_tuple_update, table_tuple_delete, or table_tuple_lock fail
130 : * because the target tuple is already outdated, they fill in this struct to
131 : * provide information to the caller about what happened.
132 : *
133 : * ctid is the target's ctid link: it is the same as the target's TID if the
134 : * target was deleted, or the location of the replacement tuple if the target
135 : * was updated.
136 : *
137 : * xmax is the outdating transaction's XID. If the caller wants to visit the
138 : * replacement tuple, it must check that this matches before believing the
139 : * replacement is really a match. This is InvalidTransactionId if the target
140 : * was !LP_NORMAL (expected only for a TID retrieved from syscache).
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 `blockno` as part of a
784 : * bitmap table scan. `scan` was started via table_beginscan_bm(). Return
785 : * false if the bitmap is exhausted and true otherwise.
786 : *
787 : * This will typically read and pin the target block, and do the necessary
788 : * work to allow scan_bitmap_next_tuple() to return tuples (e.g. it might
789 : * make sense to perform tuple visibility checks at this time).
790 : *
791 : * `lossy_pages` and `exact_pages` are EXPLAIN counters that can be
792 : * incremented by the table AM to indicate whether or not the block's
793 : * representation in the bitmap is lossy.
794 : *
795 : * `recheck` is set by the table AM to indicate whether or not the tuples
796 : * from this block should be rechecked. Tuples from lossy pages will
797 : * always need to be rechecked, but some non-lossy pages' tuples may also
798 : * require recheck.
799 : *
800 : * `blockno` is the current block and is set by the table AM. The table AM
801 : * is responsible for advancing the main iterator, but the bitmap table
802 : * scan code still advances the prefetch iterator. `blockno` is used by
803 : * bitmap table scan code to validate that the prefetch block stays ahead
804 : * of the current block.
805 : *
806 : * XXX: Currently this may only be implemented if the AM uses md.c as its
807 : * storage manager, and uses ItemPointer->ip_blkid in a manner that maps
808 : * blockids directly to the underlying storage. nodeBitmapHeapscan.c
809 : * performs prefetching directly using that interface. This probably
810 : * needs to be rectified at a later point.
811 : *
812 : * XXX: Currently this may only be implemented if the AM uses the
813 : * visibilitymap, as nodeBitmapHeapscan.c unconditionally accesses it to
814 : * perform prefetching. This probably needs to be rectified at a later
815 : * point.
816 : *
817 : * Optional callback, but either both scan_bitmap_next_block and
818 : * scan_bitmap_next_tuple need to exist, or neither.
819 : */
820 : bool (*scan_bitmap_next_block) (TableScanDesc scan,
821 : BlockNumber *blockno,
822 : bool *recheck,
823 : uint64 *lossy_pages,
824 : uint64 *exact_pages);
825 :
826 : /*
827 : * Fetch the next tuple of a bitmap table scan into `slot` and return true
828 : * if a visible tuple was found, false otherwise.
829 : *
830 : * Optional callback, but either both scan_bitmap_next_block and
831 : * scan_bitmap_next_tuple need to exist, or neither.
832 : */
833 : bool (*scan_bitmap_next_tuple) (TableScanDesc scan,
834 : TupleTableSlot *slot);
835 :
836 : /*
837 : * Prepare to fetch tuples from the next block in a sample scan. Return
838 : * false if the sample scan is finished, true otherwise. `scan` was
839 : * started via table_beginscan_sampling().
840 : *
841 : * Typically this will first determine the target block by calling the
842 : * TsmRoutine's NextSampleBlock() callback if not NULL, or alternatively
843 : * perform a sequential scan over all blocks. The determined block is
844 : * then typically read and pinned.
845 : *
846 : * As the TsmRoutine interface is block based, a block needs to be passed
847 : * to NextSampleBlock(). If that's not appropriate for an AM, it
848 : * internally needs to perform mapping between the internal and a block
849 : * based representation.
850 : *
851 : * Note that it's not acceptable to hold deadlock prone resources such as
852 : * lwlocks until scan_sample_next_tuple() has exhausted the tuples on the
853 : * block - the tuple is likely to be returned to an upper query node, and
854 : * the next call could be off a long while. Holding buffer pins and such
855 : * is obviously OK.
856 : *
857 : * Currently it is required to implement this interface, as there's no
858 : * alternative way (contrary e.g. to bitmap scans) to implement sample
859 : * scans. If infeasible to implement, the AM may raise an error.
860 : */
861 : bool (*scan_sample_next_block) (TableScanDesc scan,
862 : struct SampleScanState *scanstate);
863 :
864 : /*
865 : * This callback, only called after scan_sample_next_block has returned
866 : * true, should determine the next tuple to be returned from the selected
867 : * block using the TsmRoutine's NextSampleTuple() callback.
868 : *
869 : * The callback needs to perform visibility checks, and only return
870 : * visible tuples. That obviously can mean calling NextSampleTuple()
871 : * multiple times.
872 : *
873 : * The TsmRoutine interface assumes that there's a maximum offset on a
874 : * given page, so if that doesn't apply to an AM, it needs to emulate that
875 : * assumption somehow.
876 : */
877 : bool (*scan_sample_next_tuple) (TableScanDesc scan,
878 : struct SampleScanState *scanstate,
879 : TupleTableSlot *slot);
880 :
881 : } TableAmRoutine;
882 :
883 :
884 : /* ----------------------------------------------------------------------------
885 : * Slot functions.
886 : * ----------------------------------------------------------------------------
887 : */
888 :
889 : /*
890 : * Returns slot callbacks suitable for holding tuples of the appropriate type
891 : * for the relation. Works for tables, views, foreign tables and partitioned
892 : * tables.
893 : */
894 : extern const TupleTableSlotOps *table_slot_callbacks(Relation relation);
895 :
896 : /*
897 : * Returns slot using the callbacks returned by table_slot_callbacks(), and
898 : * registers it on *reglist.
899 : */
900 : extern TupleTableSlot *table_slot_create(Relation relation, List **reglist);
901 :
902 :
903 : /* ----------------------------------------------------------------------------
904 : * Table scan functions.
905 : * ----------------------------------------------------------------------------
906 : */
907 :
908 : /*
909 : * Start a scan of `rel`. Returned tuples pass a visibility test of
910 : * `snapshot`, and if nkeys != 0, the results are filtered by those scan keys.
911 : */
912 : static inline TableScanDesc
913 201840 : table_beginscan(Relation rel, Snapshot snapshot,
914 : int nkeys, struct ScanKeyData *key)
915 : {
916 201840 : uint32 flags = SO_TYPE_SEQSCAN |
917 : SO_ALLOW_STRAT | SO_ALLOW_SYNC | SO_ALLOW_PAGEMODE;
918 :
919 201840 : return rel->rd_tableam->scan_begin(rel, snapshot, nkeys, key, NULL, flags);
920 : }
921 :
922 : /*
923 : * Like table_beginscan(), but for scanning catalog. It'll automatically use a
924 : * snapshot appropriate for scanning catalog relations.
925 : */
926 : extern TableScanDesc table_beginscan_catalog(Relation relation, int nkeys,
927 : struct ScanKeyData *key);
928 :
929 : /*
930 : * Like table_beginscan(), but table_beginscan_strat() offers an extended API
931 : * that lets the caller control whether a nondefault buffer access strategy
932 : * can be used, and whether syncscan can be chosen (possibly resulting in the
933 : * scan not starting from block zero). Both of these default to true with
934 : * plain table_beginscan.
935 : */
936 : static inline TableScanDesc
937 395992 : table_beginscan_strat(Relation rel, Snapshot snapshot,
938 : int nkeys, struct ScanKeyData *key,
939 : bool allow_strat, bool allow_sync)
940 : {
941 395992 : uint32 flags = SO_TYPE_SEQSCAN | SO_ALLOW_PAGEMODE;
942 :
943 395992 : if (allow_strat)
944 395992 : flags |= SO_ALLOW_STRAT;
945 395992 : if (allow_sync)
946 51132 : flags |= SO_ALLOW_SYNC;
947 :
948 395992 : return rel->rd_tableam->scan_begin(rel, snapshot, nkeys, key, NULL, flags);
949 : }
950 :
951 : /*
952 : * table_beginscan_bm is an alternative entry point for setting up a
953 : * TableScanDesc for a bitmap heap scan. Although that scan technology is
954 : * really quite unlike a standard seqscan, there is just enough commonality to
955 : * make it worth using the same data structure.
956 : */
957 : static inline TableScanDesc
958 21322 : table_beginscan_bm(Relation rel, Snapshot snapshot,
959 : int nkeys, struct ScanKeyData *key, bool need_tuple)
960 : {
961 21322 : uint32 flags = SO_TYPE_BITMAPSCAN | SO_ALLOW_PAGEMODE;
962 :
963 21322 : if (need_tuple)
964 18438 : flags |= SO_NEED_TUPLES;
965 :
966 21322 : return rel->rd_tableam->scan_begin(rel, snapshot, nkeys, key,
967 : NULL, flags);
968 : }
969 :
970 : /*
971 : * table_beginscan_sampling is an alternative entry point for setting up a
972 : * TableScanDesc for a TABLESAMPLE scan. As with bitmap scans, it's worth
973 : * using the same data structure although the behavior is rather different.
974 : * In addition to the options offered by table_beginscan_strat, this call
975 : * also allows control of whether page-mode visibility checking is used.
976 : */
977 : static inline TableScanDesc
978 146 : table_beginscan_sampling(Relation rel, Snapshot snapshot,
979 : int nkeys, struct ScanKeyData *key,
980 : bool allow_strat, bool allow_sync,
981 : bool allow_pagemode)
982 : {
983 146 : uint32 flags = SO_TYPE_SAMPLESCAN;
984 :
985 146 : if (allow_strat)
986 134 : flags |= SO_ALLOW_STRAT;
987 146 : if (allow_sync)
988 66 : flags |= SO_ALLOW_SYNC;
989 146 : if (allow_pagemode)
990 122 : flags |= SO_ALLOW_PAGEMODE;
991 :
992 146 : return rel->rd_tableam->scan_begin(rel, snapshot, nkeys, key, NULL, flags);
993 : }
994 :
995 : /*
996 : * table_beginscan_tid is an alternative entry point for setting up a
997 : * TableScanDesc for a Tid scan. As with bitmap scans, it's worth using
998 : * the same data structure although the behavior is rather different.
999 : */
1000 : static inline TableScanDesc
1001 730 : table_beginscan_tid(Relation rel, Snapshot snapshot)
1002 : {
1003 730 : uint32 flags = SO_TYPE_TIDSCAN;
1004 :
1005 730 : return rel->rd_tableam->scan_begin(rel, snapshot, 0, NULL, NULL, flags);
1006 : }
1007 :
1008 : /*
1009 : * table_beginscan_analyze is an alternative entry point for setting up a
1010 : * TableScanDesc for an ANALYZE scan. As with bitmap scans, it's worth using
1011 : * the same data structure although the behavior is rather different.
1012 : */
1013 : static inline TableScanDesc
1014 14898 : table_beginscan_analyze(Relation rel)
1015 : {
1016 14898 : uint32 flags = SO_TYPE_ANALYZE;
1017 :
1018 14898 : return rel->rd_tableam->scan_begin(rel, NULL, 0, NULL, NULL, flags);
1019 : }
1020 :
1021 : /*
1022 : * End relation scan.
1023 : */
1024 : static inline void
1025 702494 : table_endscan(TableScanDesc scan)
1026 : {
1027 702494 : scan->rs_rd->rd_tableam->scan_end(scan);
1028 702494 : }
1029 :
1030 : /*
1031 : * Restart a relation scan.
1032 : */
1033 : static inline void
1034 1058990 : table_rescan(TableScanDesc scan,
1035 : struct ScanKeyData *key)
1036 : {
1037 1058990 : scan->rs_rd->rd_tableam->scan_rescan(scan, key, false, false, false, false);
1038 1058990 : }
1039 :
1040 : /*
1041 : * Restart a relation scan after changing params.
1042 : *
1043 : * This call allows changing the buffer strategy, syncscan, and pagemode
1044 : * options before starting a fresh scan. Note that although the actual use of
1045 : * syncscan might change (effectively, enabling or disabling reporting), the
1046 : * previously selected startblock will be kept.
1047 : */
1048 : static inline void
1049 30 : table_rescan_set_params(TableScanDesc scan, struct ScanKeyData *key,
1050 : bool allow_strat, bool allow_sync, bool allow_pagemode)
1051 : {
1052 30 : scan->rs_rd->rd_tableam->scan_rescan(scan, key, true,
1053 : allow_strat, allow_sync,
1054 : allow_pagemode);
1055 30 : }
1056 :
1057 : /*
1058 : * Return next tuple from `scan`, store in slot.
1059 : */
1060 : static inline bool
1061 87208690 : table_scan_getnextslot(TableScanDesc sscan, ScanDirection direction, TupleTableSlot *slot)
1062 : {
1063 87208690 : slot->tts_tableOid = RelationGetRelid(sscan->rs_rd);
1064 :
1065 : /* We don't expect actual scans using NoMovementScanDirection */
1066 : Assert(direction == ForwardScanDirection ||
1067 : direction == BackwardScanDirection);
1068 :
1069 : /*
1070 : * We don't expect direct calls to table_scan_getnextslot with valid
1071 : * CheckXidAlive for catalog or regular tables. See detailed comments in
1072 : * xact.c where these variables are declared.
1073 : */
1074 87208690 : if (unlikely(TransactionIdIsValid(CheckXidAlive) && !bsysscan))
1075 0 : elog(ERROR, "unexpected table_scan_getnextslot call during logical decoding");
1076 :
1077 87208690 : return sscan->rs_rd->rd_tableam->scan_getnextslot(sscan, direction, slot);
1078 : }
1079 :
1080 : /* ----------------------------------------------------------------------------
1081 : * TID Range scanning related functions.
1082 : * ----------------------------------------------------------------------------
1083 : */
1084 :
1085 : /*
1086 : * table_beginscan_tidrange is the entry point for setting up a TableScanDesc
1087 : * for a TID range scan.
1088 : */
1089 : static inline TableScanDesc
1090 112 : table_beginscan_tidrange(Relation rel, Snapshot snapshot,
1091 : ItemPointer mintid,
1092 : ItemPointer maxtid)
1093 : {
1094 : TableScanDesc sscan;
1095 112 : uint32 flags = SO_TYPE_TIDRANGESCAN | SO_ALLOW_PAGEMODE;
1096 :
1097 112 : sscan = rel->rd_tableam->scan_begin(rel, snapshot, 0, NULL, NULL, flags);
1098 :
1099 : /* Set the range of TIDs to scan */
1100 112 : sscan->rs_rd->rd_tableam->scan_set_tidrange(sscan, mintid, maxtid);
1101 :
1102 112 : return sscan;
1103 : }
1104 :
1105 : /*
1106 : * table_rescan_tidrange resets the scan position and sets the minimum and
1107 : * maximum TID range to scan for a TableScanDesc created by
1108 : * table_beginscan_tidrange.
1109 : */
1110 : static inline void
1111 66 : table_rescan_tidrange(TableScanDesc sscan, ItemPointer mintid,
1112 : ItemPointer maxtid)
1113 : {
1114 : /* Ensure table_beginscan_tidrange() was used. */
1115 : Assert((sscan->rs_flags & SO_TYPE_TIDRANGESCAN) != 0);
1116 :
1117 66 : sscan->rs_rd->rd_tableam->scan_rescan(sscan, NULL, false, false, false, false);
1118 66 : sscan->rs_rd->rd_tableam->scan_set_tidrange(sscan, mintid, maxtid);
1119 66 : }
1120 :
1121 : /*
1122 : * Fetch the next tuple from `sscan` for a TID range scan created by
1123 : * table_beginscan_tidrange(). Stores the tuple in `slot` and returns true,
1124 : * or returns false if no more tuples exist in the range.
1125 : */
1126 : static inline bool
1127 5940 : table_scan_getnextslot_tidrange(TableScanDesc sscan, ScanDirection direction,
1128 : TupleTableSlot *slot)
1129 : {
1130 : /* Ensure table_beginscan_tidrange() was used. */
1131 : Assert((sscan->rs_flags & SO_TYPE_TIDRANGESCAN) != 0);
1132 :
1133 : /* We don't expect actual scans using NoMovementScanDirection */
1134 : Assert(direction == ForwardScanDirection ||
1135 : direction == BackwardScanDirection);
1136 :
1137 5940 : return sscan->rs_rd->rd_tableam->scan_getnextslot_tidrange(sscan,
1138 : direction,
1139 : slot);
1140 : }
1141 :
1142 :
1143 : /* ----------------------------------------------------------------------------
1144 : * Parallel table scan related functions.
1145 : * ----------------------------------------------------------------------------
1146 : */
1147 :
1148 : /*
1149 : * Estimate the size of shared memory needed for a parallel scan of this
1150 : * relation.
1151 : */
1152 : extern Size table_parallelscan_estimate(Relation rel, Snapshot snapshot);
1153 :
1154 : /*
1155 : * Initialize ParallelTableScanDesc for a parallel scan of this
1156 : * relation. `pscan` needs to be sized according to parallelscan_estimate()
1157 : * for the same relation. Call this just once in the leader process; then,
1158 : * individual workers attach via table_beginscan_parallel.
1159 : */
1160 : extern void table_parallelscan_initialize(Relation rel,
1161 : ParallelTableScanDesc pscan,
1162 : Snapshot snapshot);
1163 :
1164 : /*
1165 : * Begin a parallel scan. `pscan` needs to have been initialized with
1166 : * table_parallelscan_initialize(), for the same relation. The initialization
1167 : * does not need to have happened in this backend.
1168 : *
1169 : * Caller must hold a suitable lock on the relation.
1170 : */
1171 : extern TableScanDesc table_beginscan_parallel(Relation relation,
1172 : ParallelTableScanDesc pscan);
1173 :
1174 : /*
1175 : * Restart a parallel scan. Call this in the leader process. Caller is
1176 : * responsible for making sure that all workers have finished the scan
1177 : * beforehand.
1178 : */
1179 : static inline void
1180 228 : table_parallelscan_reinitialize(Relation rel, ParallelTableScanDesc pscan)
1181 : {
1182 228 : rel->rd_tableam->parallelscan_reinitialize(rel, pscan);
1183 228 : }
1184 :
1185 :
1186 : /* ----------------------------------------------------------------------------
1187 : * Index scan related functions.
1188 : * ----------------------------------------------------------------------------
1189 : */
1190 :
1191 : /*
1192 : * Prepare to fetch tuples from the relation, as needed when fetching tuples
1193 : * for an index scan.
1194 : *
1195 : * Tuples for an index scan can then be fetched via table_index_fetch_tuple().
1196 : */
1197 : static inline IndexFetchTableData *
1198 24705274 : table_index_fetch_begin(Relation rel)
1199 : {
1200 24705274 : return rel->rd_tableam->index_fetch_begin(rel);
1201 : }
1202 :
1203 : /*
1204 : * Reset index fetch. Typically this will release cross index fetch resources
1205 : * held in IndexFetchTableData.
1206 : */
1207 : static inline void
1208 20304984 : table_index_fetch_reset(struct IndexFetchTableData *scan)
1209 : {
1210 20304984 : scan->rel->rd_tableam->index_fetch_reset(scan);
1211 20304984 : }
1212 :
1213 : /*
1214 : * Release resources and deallocate index fetch.
1215 : */
1216 : static inline void
1217 24703570 : table_index_fetch_end(struct IndexFetchTableData *scan)
1218 : {
1219 24703570 : scan->rel->rd_tableam->index_fetch_end(scan);
1220 24703570 : }
1221 :
1222 : /*
1223 : * Fetches, as part of an index scan, tuple at `tid` into `slot`, after doing
1224 : * a visibility test according to `snapshot`. If a tuple was found and passed
1225 : * the visibility test, returns true, false otherwise. Note that *tid may be
1226 : * modified when we return true (see later remarks on multiple row versions
1227 : * reachable via a single index entry).
1228 : *
1229 : * *call_again needs to be false on the first call to table_index_fetch_tuple() for
1230 : * a tid. If there potentially is another tuple matching the tid, *call_again
1231 : * will be set to true, signaling that table_index_fetch_tuple() should be called
1232 : * again for the same tid.
1233 : *
1234 : * *all_dead, if all_dead is not NULL, will be set to true by
1235 : * table_index_fetch_tuple() iff it is guaranteed that no backend needs to see
1236 : * that tuple. Index AMs can use that to avoid returning that tid in future
1237 : * searches.
1238 : *
1239 : * The difference between this function and table_tuple_fetch_row_version()
1240 : * is that this function returns the currently visible version of a row if
1241 : * the AM supports storing multiple row versions reachable via a single index
1242 : * entry (like heap's HOT). Whereas table_tuple_fetch_row_version() only
1243 : * evaluates the tuple exactly at `tid`. Outside of index entry ->table tuple
1244 : * lookups, table_tuple_fetch_row_version() is what's usually needed.
1245 : */
1246 : static inline bool
1247 35030928 : table_index_fetch_tuple(struct IndexFetchTableData *scan,
1248 : ItemPointer tid,
1249 : Snapshot snapshot,
1250 : TupleTableSlot *slot,
1251 : bool *call_again, bool *all_dead)
1252 : {
1253 : /*
1254 : * We don't expect direct calls to table_index_fetch_tuple with valid
1255 : * CheckXidAlive for catalog or regular tables. See detailed comments in
1256 : * xact.c where these variables are declared.
1257 : */
1258 35030928 : if (unlikely(TransactionIdIsValid(CheckXidAlive) && !bsysscan))
1259 0 : elog(ERROR, "unexpected table_index_fetch_tuple call during logical decoding");
1260 :
1261 35030928 : return scan->rel->rd_tableam->index_fetch_tuple(scan, tid, snapshot,
1262 : slot, call_again,
1263 : all_dead);
1264 : }
1265 :
1266 : /*
1267 : * This is a convenience wrapper around table_index_fetch_tuple() which
1268 : * returns whether there are table tuple items corresponding to an index
1269 : * entry. This likely is only useful to verify if there's a conflict in a
1270 : * unique index.
1271 : */
1272 : extern bool table_index_fetch_tuple_check(Relation rel,
1273 : ItemPointer tid,
1274 : Snapshot snapshot,
1275 : bool *all_dead);
1276 :
1277 :
1278 : /* ------------------------------------------------------------------------
1279 : * Functions for non-modifying operations on individual tuples
1280 : * ------------------------------------------------------------------------
1281 : */
1282 :
1283 :
1284 : /*
1285 : * Fetch tuple at `tid` into `slot`, after doing a visibility test according to
1286 : * `snapshot`. If a tuple was found and passed the visibility test, returns
1287 : * true, false otherwise.
1288 : *
1289 : * See table_index_fetch_tuple's comment about what the difference between
1290 : * these functions is. It is correct to use this function outside of index
1291 : * entry->table tuple lookups.
1292 : */
1293 : static inline bool
1294 348004 : table_tuple_fetch_row_version(Relation rel,
1295 : ItemPointer tid,
1296 : Snapshot snapshot,
1297 : TupleTableSlot *slot)
1298 : {
1299 : /*
1300 : * We don't expect direct calls to table_tuple_fetch_row_version with
1301 : * valid CheckXidAlive for catalog or regular tables. See detailed
1302 : * comments in xact.c where these variables are declared.
1303 : */
1304 348004 : if (unlikely(TransactionIdIsValid(CheckXidAlive) && !bsysscan))
1305 0 : elog(ERROR, "unexpected table_tuple_fetch_row_version call during logical decoding");
1306 :
1307 348004 : return rel->rd_tableam->tuple_fetch_row_version(rel, tid, snapshot, slot);
1308 : }
1309 :
1310 : /*
1311 : * Verify that `tid` is a potentially valid tuple identifier. That doesn't
1312 : * mean that the pointed to row needs to exist or be visible, but that
1313 : * attempting to fetch the row (e.g. with table_tuple_get_latest_tid() or
1314 : * table_tuple_fetch_row_version()) should not error out if called with that
1315 : * tid.
1316 : *
1317 : * `scan` needs to have been started via table_beginscan().
1318 : */
1319 : static inline bool
1320 394 : table_tuple_tid_valid(TableScanDesc scan, ItemPointer tid)
1321 : {
1322 394 : return scan->rs_rd->rd_tableam->tuple_tid_valid(scan, tid);
1323 : }
1324 :
1325 : /*
1326 : * Return the latest version of the tuple at `tid`, by updating `tid` to
1327 : * point at the newest version.
1328 : */
1329 : extern void table_tuple_get_latest_tid(TableScanDesc scan, ItemPointer tid);
1330 :
1331 : /*
1332 : * Return true iff tuple in slot satisfies the snapshot.
1333 : *
1334 : * This assumes the slot's tuple is valid, and of the appropriate type for the
1335 : * AM.
1336 : *
1337 : * Some AMs might modify the data underlying the tuple as a side-effect. If so
1338 : * they ought to mark the relevant buffer dirty.
1339 : */
1340 : static inline bool
1341 225728 : table_tuple_satisfies_snapshot(Relation rel, TupleTableSlot *slot,
1342 : Snapshot snapshot)
1343 : {
1344 225728 : return rel->rd_tableam->tuple_satisfies_snapshot(rel, slot, snapshot);
1345 : }
1346 :
1347 : /*
1348 : * Determine which index tuples are safe to delete based on their table TID.
1349 : *
1350 : * Determines which entries from index AM caller's TM_IndexDeleteOp state
1351 : * point to vacuumable table tuples. Entries that are found by tableam to be
1352 : * vacuumable are naturally safe for index AM to delete, and so get directly
1353 : * marked as deletable. See comments above TM_IndexDelete and comments above
1354 : * TM_IndexDeleteOp for full details.
1355 : *
1356 : * Returns a snapshotConflictHorizon transaction ID that caller places in
1357 : * its index deletion WAL record. This might be used during subsequent REDO
1358 : * of the WAL record when in Hot Standby mode -- a recovery conflict for the
1359 : * index deletion operation might be required on the standby.
1360 : */
1361 : static inline TransactionId
1362 11048 : table_index_delete_tuples(Relation rel, TM_IndexDeleteOp *delstate)
1363 : {
1364 11048 : return rel->rd_tableam->index_delete_tuples(rel, delstate);
1365 : }
1366 :
1367 :
1368 : /* ----------------------------------------------------------------------------
1369 : * Functions for manipulations of physical tuples.
1370 : * ----------------------------------------------------------------------------
1371 : */
1372 :
1373 : /*
1374 : * Insert a tuple from a slot into table AM routine.
1375 : *
1376 : * The options bitmask allows the caller to specify options that may change the
1377 : * behaviour of the AM. The AM will ignore options that it does not support.
1378 : *
1379 : * If the TABLE_INSERT_SKIP_FSM option is specified, AMs are free to not reuse
1380 : * free space in the relation. This can save some cycles when we know the
1381 : * relation is new and doesn't contain useful amounts of free space.
1382 : * TABLE_INSERT_SKIP_FSM is commonly passed directly to
1383 : * RelationGetBufferForTuple. See that method for more information.
1384 : *
1385 : * TABLE_INSERT_FROZEN should only be specified for inserts into
1386 : * relation storage created during the current subtransaction and when
1387 : * there are no prior snapshots or pre-existing portals open.
1388 : * This causes rows to be frozen, which is an MVCC violation and
1389 : * requires explicit options chosen by user.
1390 : *
1391 : * TABLE_INSERT_NO_LOGICAL force-disables the emitting of logical decoding
1392 : * information for the tuple. This should solely be used during table rewrites
1393 : * where RelationIsLogicallyLogged(relation) is not yet accurate for the new
1394 : * relation.
1395 : *
1396 : * Note that most of these options will be applied when inserting into the
1397 : * heap's TOAST table, too, if the tuple requires any out-of-line data.
1398 : *
1399 : * The BulkInsertState object (if any; bistate can be NULL for default
1400 : * behavior) is also just passed through to RelationGetBufferForTuple. If
1401 : * `bistate` is provided, table_finish_bulk_insert() needs to be called.
1402 : *
1403 : * On return the slot's tts_tid and tts_tableOid are updated to reflect the
1404 : * insertion. But note that any toasting of fields within the slot is NOT
1405 : * reflected in the slots contents.
1406 : */
1407 : static inline void
1408 13945398 : table_tuple_insert(Relation rel, TupleTableSlot *slot, CommandId cid,
1409 : int options, struct BulkInsertStateData *bistate)
1410 : {
1411 13945398 : rel->rd_tableam->tuple_insert(rel, slot, cid, options,
1412 : bistate);
1413 13945358 : }
1414 :
1415 : /*
1416 : * Perform a "speculative insertion". These can be backed out afterwards
1417 : * without aborting the whole transaction. Other sessions can wait for the
1418 : * speculative insertion to be confirmed, turning it into a regular tuple, or
1419 : * aborted, as if it never existed. Speculatively inserted tuples behave as
1420 : * "value locks" of short duration, used to implement INSERT .. ON CONFLICT.
1421 : *
1422 : * A transaction having performed a speculative insertion has to either abort,
1423 : * or finish the speculative insertion with
1424 : * table_tuple_complete_speculative(succeeded = ...).
1425 : */
1426 : static inline void
1427 4128 : table_tuple_insert_speculative(Relation rel, TupleTableSlot *slot,
1428 : CommandId cid, int options,
1429 : struct BulkInsertStateData *bistate,
1430 : uint32 specToken)
1431 : {
1432 4128 : rel->rd_tableam->tuple_insert_speculative(rel, slot, cid, options,
1433 : bistate, specToken);
1434 4128 : }
1435 :
1436 : /*
1437 : * Complete "speculative insertion" started in the same transaction. If
1438 : * succeeded is true, the tuple is fully inserted, if false, it's removed.
1439 : */
1440 : static inline void
1441 4122 : table_tuple_complete_speculative(Relation rel, TupleTableSlot *slot,
1442 : uint32 specToken, bool succeeded)
1443 : {
1444 4122 : rel->rd_tableam->tuple_complete_speculative(rel, slot, specToken,
1445 : succeeded);
1446 4122 : }
1447 :
1448 : /*
1449 : * Insert multiple tuples into a table.
1450 : *
1451 : * This is like table_tuple_insert(), but inserts multiple tuples in one
1452 : * operation. That's often faster than calling table_tuple_insert() in a loop,
1453 : * because e.g. the AM can reduce WAL logging and page locking overhead.
1454 : *
1455 : * Except for taking `nslots` tuples as input, and an array of TupleTableSlots
1456 : * in `slots`, the parameters for table_multi_insert() are the same as for
1457 : * table_tuple_insert().
1458 : *
1459 : * Note: this leaks memory into the current memory context. You can create a
1460 : * temporary context before calling this, if that's a problem.
1461 : */
1462 : static inline void
1463 2418 : table_multi_insert(Relation rel, TupleTableSlot **slots, int nslots,
1464 : CommandId cid, int options, struct BulkInsertStateData *bistate)
1465 : {
1466 2418 : rel->rd_tableam->multi_insert(rel, slots, nslots,
1467 : cid, options, bistate);
1468 2418 : }
1469 :
1470 : /*
1471 : * Delete a tuple.
1472 : *
1473 : * NB: do not call this directly unless prepared to deal with
1474 : * concurrent-update conditions. Use simple_table_tuple_delete instead.
1475 : *
1476 : * Input parameters:
1477 : * relation - table to be modified (caller must hold suitable lock)
1478 : * tid - TID of tuple to be deleted
1479 : * cid - delete command ID (used for visibility test, and stored into
1480 : * cmax if successful)
1481 : * crosscheck - if not InvalidSnapshot, also check tuple against this
1482 : * wait - true if should wait for any conflicting update to commit/abort
1483 : * Output parameters:
1484 : * tmfd - filled in failure cases (see below)
1485 : * changingPart - true iff the tuple is being moved to another partition
1486 : * table due to an update of the partition key. Otherwise, false.
1487 : *
1488 : * Normal, successful return value is TM_Ok, which means we did actually
1489 : * delete it. Failure return codes are TM_SelfModified, TM_Updated, and
1490 : * TM_BeingModified (the last only possible if wait == false).
1491 : *
1492 : * In the failure cases, the routine fills *tmfd with the tuple's t_ctid,
1493 : * t_xmax, and, if possible, t_cmax. See comments for struct
1494 : * TM_FailureData for additional info.
1495 : */
1496 : static inline TM_Result
1497 1720580 : table_tuple_delete(Relation rel, ItemPointer tid, CommandId cid,
1498 : Snapshot snapshot, Snapshot crosscheck, bool wait,
1499 : TM_FailureData *tmfd, bool changingPart)
1500 : {
1501 1720580 : return rel->rd_tableam->tuple_delete(rel, tid, cid,
1502 : snapshot, crosscheck,
1503 : wait, tmfd, changingPart);
1504 : }
1505 :
1506 : /*
1507 : * Update a tuple.
1508 : *
1509 : * NB: do not call this directly unless you are prepared to deal with
1510 : * concurrent-update conditions. Use simple_table_tuple_update instead.
1511 : *
1512 : * Input parameters:
1513 : * relation - table to be modified (caller must hold suitable lock)
1514 : * otid - TID of old tuple to be replaced
1515 : * slot - newly constructed tuple data to store
1516 : * cid - update command ID (used for visibility test, and stored into
1517 : * cmax/cmin if successful)
1518 : * crosscheck - if not InvalidSnapshot, also check old tuple against this
1519 : * wait - true if should wait for any conflicting update to commit/abort
1520 : * Output parameters:
1521 : * tmfd - filled in failure cases (see below)
1522 : * lockmode - filled with lock mode acquired on tuple
1523 : * update_indexes - in success cases this is set to true if new index entries
1524 : * are required for this tuple
1525 : *
1526 : * Normal, successful return value is TM_Ok, which means we did actually
1527 : * update it. Failure return codes are TM_SelfModified, TM_Updated, and
1528 : * TM_BeingModified (the last only possible if wait == false).
1529 : *
1530 : * On success, the slot's tts_tid and tts_tableOid are updated to match the new
1531 : * stored tuple; in particular, slot->tts_tid is set to the TID where the
1532 : * new tuple was inserted, and its HEAP_ONLY_TUPLE flag is set iff a HOT
1533 : * update was done. However, any TOAST changes in the new tuple's
1534 : * data are not reflected into *newtup.
1535 : *
1536 : * In the failure cases, the routine fills *tmfd with the tuple's t_ctid,
1537 : * t_xmax, and, if possible, t_cmax. See comments for struct TM_FailureData
1538 : * for additional info.
1539 : */
1540 : static inline TM_Result
1541 380172 : table_tuple_update(Relation rel, ItemPointer otid, TupleTableSlot *slot,
1542 : CommandId cid, Snapshot snapshot, Snapshot crosscheck,
1543 : bool wait, TM_FailureData *tmfd, LockTupleMode *lockmode,
1544 : TU_UpdateIndexes *update_indexes)
1545 : {
1546 380172 : return rel->rd_tableam->tuple_update(rel, otid, slot,
1547 : cid, snapshot, crosscheck,
1548 : wait, tmfd,
1549 : lockmode, update_indexes);
1550 : }
1551 :
1552 : /*
1553 : * Lock a tuple in the specified mode.
1554 : *
1555 : * Input parameters:
1556 : * relation: relation containing tuple (caller must hold suitable lock)
1557 : * tid: TID of tuple to lock
1558 : * snapshot: snapshot to use for visibility determinations
1559 : * cid: current command ID (used for visibility test, and stored into
1560 : * tuple's cmax if lock is successful)
1561 : * mode: lock mode desired
1562 : * wait_policy: what to do if tuple lock is not available
1563 : * flags:
1564 : * If TUPLE_LOCK_FLAG_LOCK_UPDATE_IN_PROGRESS, follow the update chain to
1565 : * also lock descendant tuples if lock modes don't conflict.
1566 : * If TUPLE_LOCK_FLAG_FIND_LAST_VERSION, follow the update chain and lock
1567 : * latest version.
1568 : *
1569 : * Output parameters:
1570 : * *slot: contains the target tuple
1571 : * *tmfd: filled in failure cases (see below)
1572 : *
1573 : * Function result may be:
1574 : * TM_Ok: lock was successfully acquired
1575 : * TM_Invisible: lock failed because tuple was never visible to us
1576 : * TM_SelfModified: lock failed because tuple updated by self
1577 : * TM_Updated: lock failed because tuple updated by other xact
1578 : * TM_Deleted: lock failed because tuple deleted by other xact
1579 : * TM_WouldBlock: lock couldn't be acquired and wait_policy is skip
1580 : *
1581 : * In the failure cases other than TM_Invisible and TM_Deleted, the routine
1582 : * fills *tmfd with the tuple's t_ctid, t_xmax, and, if possible, t_cmax. See
1583 : * comments for struct TM_FailureData for additional info.
1584 : */
1585 : static inline TM_Result
1586 169530 : table_tuple_lock(Relation rel, ItemPointer tid, Snapshot snapshot,
1587 : TupleTableSlot *slot, CommandId cid, LockTupleMode mode,
1588 : LockWaitPolicy wait_policy, uint8 flags,
1589 : TM_FailureData *tmfd)
1590 : {
1591 169530 : return rel->rd_tableam->tuple_lock(rel, tid, snapshot, slot,
1592 : cid, mode, wait_policy,
1593 : flags, tmfd);
1594 : }
1595 :
1596 : /*
1597 : * Perform operations necessary to complete insertions made via
1598 : * tuple_insert and multi_insert with a BulkInsertState specified.
1599 : */
1600 : static inline void
1601 3898 : table_finish_bulk_insert(Relation rel, int options)
1602 : {
1603 : /* optional callback */
1604 3898 : if (rel->rd_tableam && rel->rd_tableam->finish_bulk_insert)
1605 0 : rel->rd_tableam->finish_bulk_insert(rel, options);
1606 3898 : }
1607 :
1608 :
1609 : /* ------------------------------------------------------------------------
1610 : * DDL related functionality.
1611 : * ------------------------------------------------------------------------
1612 : */
1613 :
1614 : /*
1615 : * Create storage for `rel` in `newrlocator`, with persistence set to
1616 : * `persistence`.
1617 : *
1618 : * This is used both during relation creation and various DDL operations to
1619 : * create new rel storage that can be filled from scratch. When creating
1620 : * new storage for an existing relfilelocator, this should be called before the
1621 : * relcache entry has been updated.
1622 : *
1623 : * *freezeXid, *minmulti are set to the xid / multixact horizon for the table
1624 : * that pg_class.{relfrozenxid, relminmxid} have to be set to.
1625 : */
1626 : static inline void
1627 61866 : table_relation_set_new_filelocator(Relation rel,
1628 : const RelFileLocator *newrlocator,
1629 : char persistence,
1630 : TransactionId *freezeXid,
1631 : MultiXactId *minmulti)
1632 : {
1633 61866 : rel->rd_tableam->relation_set_new_filelocator(rel, newrlocator,
1634 : persistence, freezeXid,
1635 : minmulti);
1636 61866 : }
1637 :
1638 : /*
1639 : * Remove all table contents from `rel`, in a non-transactional manner.
1640 : * Non-transactional meaning that there's no need to support rollbacks. This
1641 : * commonly only is used to perform truncations for relation storage created in
1642 : * the current transaction.
1643 : */
1644 : static inline void
1645 576 : table_relation_nontransactional_truncate(Relation rel)
1646 : {
1647 576 : rel->rd_tableam->relation_nontransactional_truncate(rel);
1648 576 : }
1649 :
1650 : /*
1651 : * Copy data from `rel` into the new relfilelocator `newrlocator`. The new
1652 : * relfilelocator may not have storage associated before this function is
1653 : * called. This is only supposed to be used for low level operations like
1654 : * changing a relation's tablespace.
1655 : */
1656 : static inline void
1657 98 : table_relation_copy_data(Relation rel, const RelFileLocator *newrlocator)
1658 : {
1659 98 : rel->rd_tableam->relation_copy_data(rel, newrlocator);
1660 98 : }
1661 :
1662 : /*
1663 : * Copy data from `OldTable` into `NewTable`, as part of a CLUSTER or VACUUM
1664 : * FULL.
1665 : *
1666 : * Additional Input parameters:
1667 : * - use_sort - if true, the table contents are sorted appropriate for
1668 : * `OldIndex`; if false and OldIndex is not InvalidOid, the data is copied
1669 : * in that index's order; if false and OldIndex is InvalidOid, no sorting is
1670 : * performed
1671 : * - OldIndex - see use_sort
1672 : * - OldestXmin - computed by vacuum_get_cutoffs(), even when
1673 : * not needed for the relation's AM
1674 : * - *xid_cutoff - ditto
1675 : * - *multi_cutoff - ditto
1676 : *
1677 : * Output parameters:
1678 : * - *xid_cutoff - rel's new relfrozenxid value, may be invalid
1679 : * - *multi_cutoff - rel's new relminmxid value, may be invalid
1680 : * - *tups_vacuumed - stats, for logging, if appropriate for AM
1681 : * - *tups_recently_dead - stats, for logging, if appropriate for AM
1682 : */
1683 : static inline void
1684 554 : table_relation_copy_for_cluster(Relation OldTable, Relation NewTable,
1685 : Relation OldIndex,
1686 : bool use_sort,
1687 : TransactionId OldestXmin,
1688 : TransactionId *xid_cutoff,
1689 : MultiXactId *multi_cutoff,
1690 : double *num_tuples,
1691 : double *tups_vacuumed,
1692 : double *tups_recently_dead)
1693 : {
1694 554 : OldTable->rd_tableam->relation_copy_for_cluster(OldTable, NewTable, OldIndex,
1695 : use_sort, OldestXmin,
1696 : xid_cutoff, multi_cutoff,
1697 : num_tuples, tups_vacuumed,
1698 : tups_recently_dead);
1699 554 : }
1700 :
1701 : /*
1702 : * Perform VACUUM on the relation. The VACUUM can be triggered by a user or by
1703 : * autovacuum. The specific actions performed by the AM will depend heavily on
1704 : * the individual AM.
1705 : *
1706 : * On entry a transaction needs to already been established, and the
1707 : * table is locked with a ShareUpdateExclusive lock.
1708 : *
1709 : * Note that neither VACUUM FULL (and CLUSTER), nor ANALYZE go through this
1710 : * routine, even if (for ANALYZE) it is part of the same VACUUM command.
1711 : */
1712 : static inline void
1713 117624 : table_relation_vacuum(Relation rel, struct VacuumParams *params,
1714 : BufferAccessStrategy bstrategy)
1715 : {
1716 117624 : rel->rd_tableam->relation_vacuum(rel, params, bstrategy);
1717 117624 : }
1718 :
1719 : /*
1720 : * Prepare to analyze the next block in the read stream. The scan needs to
1721 : * have been started with table_beginscan_analyze(). Note that this routine
1722 : * might acquire resources like locks that are held until
1723 : * table_scan_analyze_next_tuple() returns false.
1724 : *
1725 : * Returns false if block is unsuitable for sampling, true otherwise.
1726 : */
1727 : static inline bool
1728 131506 : table_scan_analyze_next_block(TableScanDesc scan, ReadStream *stream)
1729 : {
1730 131506 : return scan->rs_rd->rd_tableam->scan_analyze_next_block(scan, stream);
1731 : }
1732 :
1733 : /*
1734 : * Iterate over tuples in the block selected with
1735 : * table_scan_analyze_next_block() (which needs to have returned true, and
1736 : * this routine may not have returned false for the same block before). If a
1737 : * tuple that's suitable for sampling is found, true is returned and a tuple
1738 : * is stored in `slot`.
1739 : *
1740 : * *liverows and *deadrows are incremented according to the encountered
1741 : * tuples.
1742 : */
1743 : static inline bool
1744 9278584 : table_scan_analyze_next_tuple(TableScanDesc scan, TransactionId OldestXmin,
1745 : double *liverows, double *deadrows,
1746 : TupleTableSlot *slot)
1747 : {
1748 9278584 : return scan->rs_rd->rd_tableam->scan_analyze_next_tuple(scan, OldestXmin,
1749 : liverows, deadrows,
1750 : slot);
1751 : }
1752 :
1753 : /*
1754 : * table_index_build_scan - scan the table to find tuples to be indexed
1755 : *
1756 : * This is called back from an access-method-specific index build procedure
1757 : * after the AM has done whatever setup it needs. The parent table relation
1758 : * is scanned to find tuples that should be entered into the index. Each
1759 : * such tuple is passed to the AM's callback routine, which does the right
1760 : * things to add it to the new index. After we return, the AM's index
1761 : * build procedure does whatever cleanup it needs.
1762 : *
1763 : * The total count of live tuples is returned. This is for updating pg_class
1764 : * statistics. (It's annoying not to be able to do that here, but we want to
1765 : * merge that update with others; see index_update_stats.) Note that the
1766 : * index AM itself must keep track of the number of index tuples; we don't do
1767 : * so here because the AM might reject some of the tuples for its own reasons,
1768 : * such as being unable to store NULLs.
1769 : *
1770 : * If 'progress', the PROGRESS_SCAN_BLOCKS_TOTAL counter is updated when
1771 : * starting the scan, and PROGRESS_SCAN_BLOCKS_DONE is updated as we go along.
1772 : *
1773 : * A side effect is to set indexInfo->ii_BrokenHotChain to true if we detect
1774 : * any potentially broken HOT chains. Currently, we set this if there are any
1775 : * RECENTLY_DEAD or DELETE_IN_PROGRESS entries in a HOT chain, without trying
1776 : * very hard to detect whether they're really incompatible with the chain tip.
1777 : * This only really makes sense for heap AM, it might need to be generalized
1778 : * for other AMs later.
1779 : */
1780 : static inline double
1781 51382 : table_index_build_scan(Relation table_rel,
1782 : Relation index_rel,
1783 : struct IndexInfo *index_info,
1784 : bool allow_sync,
1785 : bool progress,
1786 : IndexBuildCallback callback,
1787 : void *callback_state,
1788 : TableScanDesc scan)
1789 : {
1790 51382 : return table_rel->rd_tableam->index_build_range_scan(table_rel,
1791 : index_rel,
1792 : index_info,
1793 : allow_sync,
1794 : false,
1795 : progress,
1796 : 0,
1797 : InvalidBlockNumber,
1798 : callback,
1799 : callback_state,
1800 : scan);
1801 : }
1802 :
1803 : /*
1804 : * As table_index_build_scan(), except that instead of scanning the complete
1805 : * table, only the given number of blocks are scanned. Scan to end-of-rel can
1806 : * be signaled by passing InvalidBlockNumber as numblocks. Note that
1807 : * restricting the range to scan cannot be done when requesting syncscan.
1808 : *
1809 : * When "anyvisible" mode is requested, all tuples visible to any transaction
1810 : * are indexed and counted as live, including those inserted or deleted by
1811 : * transactions that are still in progress.
1812 : */
1813 : static inline double
1814 2934 : table_index_build_range_scan(Relation table_rel,
1815 : Relation index_rel,
1816 : struct IndexInfo *index_info,
1817 : bool allow_sync,
1818 : bool anyvisible,
1819 : bool progress,
1820 : BlockNumber start_blockno,
1821 : BlockNumber numblocks,
1822 : IndexBuildCallback callback,
1823 : void *callback_state,
1824 : TableScanDesc scan)
1825 : {
1826 2934 : return table_rel->rd_tableam->index_build_range_scan(table_rel,
1827 : index_rel,
1828 : index_info,
1829 : allow_sync,
1830 : anyvisible,
1831 : progress,
1832 : start_blockno,
1833 : numblocks,
1834 : callback,
1835 : callback_state,
1836 : scan);
1837 : }
1838 :
1839 : /*
1840 : * table_index_validate_scan - second table scan for concurrent index build
1841 : *
1842 : * See validate_index() for an explanation.
1843 : */
1844 : static inline void
1845 628 : table_index_validate_scan(Relation table_rel,
1846 : Relation index_rel,
1847 : struct IndexInfo *index_info,
1848 : Snapshot snapshot,
1849 : struct ValidateIndexState *state)
1850 : {
1851 628 : table_rel->rd_tableam->index_validate_scan(table_rel,
1852 : index_rel,
1853 : index_info,
1854 : snapshot,
1855 : state);
1856 628 : }
1857 :
1858 :
1859 : /* ----------------------------------------------------------------------------
1860 : * Miscellaneous functionality
1861 : * ----------------------------------------------------------------------------
1862 : */
1863 :
1864 : /*
1865 : * Return the current size of `rel` in bytes. If `forkNumber` is
1866 : * InvalidForkNumber, return the relation's overall size, otherwise the size
1867 : * for the indicated fork.
1868 : *
1869 : * Note that the overall size might not be the equivalent of the sum of sizes
1870 : * for the individual forks for some AMs, e.g. because the AMs storage does
1871 : * not neatly map onto the builtin types of forks.
1872 : */
1873 : static inline uint64
1874 2533354 : table_relation_size(Relation rel, ForkNumber forkNumber)
1875 : {
1876 2533354 : return rel->rd_tableam->relation_size(rel, forkNumber);
1877 : }
1878 :
1879 : /*
1880 : * table_relation_needs_toast_table - does this relation need a toast table?
1881 : */
1882 : static inline bool
1883 42202 : table_relation_needs_toast_table(Relation rel)
1884 : {
1885 42202 : return rel->rd_tableam->relation_needs_toast_table(rel);
1886 : }
1887 :
1888 : /*
1889 : * Return the OID of the AM that should be used to implement the TOAST table
1890 : * for this relation.
1891 : */
1892 : static inline Oid
1893 16830 : table_relation_toast_am(Relation rel)
1894 : {
1895 16830 : return rel->rd_tableam->relation_toast_am(rel);
1896 : }
1897 :
1898 : /*
1899 : * Fetch all or part of a TOAST value from a TOAST table.
1900 : *
1901 : * If this AM is never used to implement a TOAST table, then this callback
1902 : * is not needed. But, if toasted values are ever stored in a table of this
1903 : * type, then you will need this callback.
1904 : *
1905 : * toastrel is the relation in which the toasted value is stored.
1906 : *
1907 : * valueid identifies which toast value is to be fetched. For the heap,
1908 : * this corresponds to the values stored in the chunk_id column.
1909 : *
1910 : * attrsize is the total size of the toast value to be fetched.
1911 : *
1912 : * sliceoffset is the offset within the toast value of the first byte that
1913 : * should be fetched.
1914 : *
1915 : * slicelength is the number of bytes from the toast value that should be
1916 : * fetched.
1917 : *
1918 : * result is caller-allocated space into which the fetched bytes should be
1919 : * stored.
1920 : */
1921 : static inline void
1922 20456 : table_relation_fetch_toast_slice(Relation toastrel, Oid valueid,
1923 : int32 attrsize, int32 sliceoffset,
1924 : int32 slicelength, struct varlena *result)
1925 : {
1926 20456 : toastrel->rd_tableam->relation_fetch_toast_slice(toastrel, valueid,
1927 : attrsize,
1928 : sliceoffset, slicelength,
1929 : result);
1930 20456 : }
1931 :
1932 :
1933 : /* ----------------------------------------------------------------------------
1934 : * Planner related functionality
1935 : * ----------------------------------------------------------------------------
1936 : */
1937 :
1938 : /*
1939 : * Estimate the current size of the relation, as an AM specific workhorse for
1940 : * estimate_rel_size(). Look there for an explanation of the parameters.
1941 : */
1942 : static inline void
1943 508334 : table_relation_estimate_size(Relation rel, int32 *attr_widths,
1944 : BlockNumber *pages, double *tuples,
1945 : double *allvisfrac)
1946 : {
1947 508334 : rel->rd_tableam->relation_estimate_size(rel, attr_widths, pages, tuples,
1948 : allvisfrac);
1949 508334 : }
1950 :
1951 :
1952 : /* ----------------------------------------------------------------------------
1953 : * Executor related functionality
1954 : * ----------------------------------------------------------------------------
1955 : */
1956 :
1957 : /*
1958 : * Prepare to fetch / check / return tuples as part of a bitmap table scan.
1959 : * `scan` needs to have been started via table_beginscan_bm(). Returns false
1960 : * if there are no more blocks in the bitmap, true otherwise.
1961 : *
1962 : * `lossy_pages` and `exact_pages` are EXPLAIN counters that can be
1963 : * incremented by the table AM to indicate whether or not the block's
1964 : * representation in the bitmap is lossy.
1965 : *
1966 : * `recheck` is set by the table AM to indicate whether or not the tuples
1967 : * from this block should be rechecked.
1968 : *
1969 : * `blockno` is the current block and is set by the table AM and is used by
1970 : * bitmap table scan code to validate that the prefetch block stays ahead of
1971 : * the current block.
1972 : *
1973 : * Note, this is an optionally implemented function, therefore should only be
1974 : * used after verifying the presence (at plan time or such).
1975 : */
1976 : static inline bool
1977 418684 : table_scan_bitmap_next_block(TableScanDesc scan,
1978 : BlockNumber *blockno,
1979 : bool *recheck,
1980 : uint64 *lossy_pages,
1981 : uint64 *exact_pages)
1982 : {
1983 : /*
1984 : * We don't expect direct calls to table_scan_bitmap_next_block with valid
1985 : * CheckXidAlive for catalog or regular tables. See detailed comments in
1986 : * xact.c where these variables are declared.
1987 : */
1988 418684 : if (unlikely(TransactionIdIsValid(CheckXidAlive) && !bsysscan))
1989 0 : elog(ERROR, "unexpected table_scan_bitmap_next_block call during logical decoding");
1990 :
1991 418684 : return scan->rs_rd->rd_tableam->scan_bitmap_next_block(scan,
1992 : blockno, recheck,
1993 : lossy_pages,
1994 : exact_pages);
1995 : }
1996 :
1997 : /*
1998 : * Fetch the next tuple of a bitmap table scan into `slot` and return true if
1999 : * a visible tuple was found, false otherwise.
2000 : * table_scan_bitmap_next_block() needs to previously have selected a
2001 : * block (i.e. returned true), and no previous
2002 : * table_scan_bitmap_next_tuple() for the same block may have
2003 : * returned false.
2004 : */
2005 : static inline bool
2006 6826844 : table_scan_bitmap_next_tuple(TableScanDesc scan,
2007 : TupleTableSlot *slot)
2008 : {
2009 : /*
2010 : * We don't expect direct calls to table_scan_bitmap_next_tuple with valid
2011 : * CheckXidAlive for catalog or regular tables. See detailed comments in
2012 : * xact.c where these variables are declared.
2013 : */
2014 6826844 : if (unlikely(TransactionIdIsValid(CheckXidAlive) && !bsysscan))
2015 0 : elog(ERROR, "unexpected table_scan_bitmap_next_tuple call during logical decoding");
2016 :
2017 6826844 : return scan->rs_rd->rd_tableam->scan_bitmap_next_tuple(scan,
2018 : slot);
2019 : }
2020 :
2021 : /*
2022 : * Prepare to fetch tuples from the next block in a sample scan. Returns false
2023 : * if the sample scan is finished, true otherwise. `scan` needs to have been
2024 : * started via table_beginscan_sampling().
2025 : *
2026 : * This will call the TsmRoutine's NextSampleBlock() callback if necessary
2027 : * (i.e. NextSampleBlock is not NULL), or perform a sequential scan over the
2028 : * underlying relation.
2029 : */
2030 : static inline bool
2031 12912 : table_scan_sample_next_block(TableScanDesc scan,
2032 : struct SampleScanState *scanstate)
2033 : {
2034 : /*
2035 : * We don't expect direct calls to table_scan_sample_next_block with valid
2036 : * CheckXidAlive for catalog or regular tables. See detailed comments in
2037 : * xact.c where these variables are declared.
2038 : */
2039 12912 : if (unlikely(TransactionIdIsValid(CheckXidAlive) && !bsysscan))
2040 0 : elog(ERROR, "unexpected table_scan_sample_next_block call during logical decoding");
2041 12912 : return scan->rs_rd->rd_tableam->scan_sample_next_block(scan, scanstate);
2042 : }
2043 :
2044 : /*
2045 : * Fetch the next sample tuple into `slot` and return true if a visible tuple
2046 : * was found, false otherwise. table_scan_sample_next_block() needs to
2047 : * previously have selected a block (i.e. returned true), and no previous
2048 : * table_scan_sample_next_tuple() for the same block may have returned false.
2049 : *
2050 : * This will call the TsmRoutine's NextSampleTuple() callback.
2051 : */
2052 : static inline bool
2053 253896 : table_scan_sample_next_tuple(TableScanDesc scan,
2054 : struct SampleScanState *scanstate,
2055 : TupleTableSlot *slot)
2056 : {
2057 : /*
2058 : * We don't expect direct calls to table_scan_sample_next_tuple with valid
2059 : * CheckXidAlive for catalog or regular tables. See detailed comments in
2060 : * xact.c where these variables are declared.
2061 : */
2062 253896 : if (unlikely(TransactionIdIsValid(CheckXidAlive) && !bsysscan))
2063 0 : elog(ERROR, "unexpected table_scan_sample_next_tuple call during logical decoding");
2064 253896 : return scan->rs_rd->rd_tableam->scan_sample_next_tuple(scan, scanstate,
2065 : slot);
2066 : }
2067 :
2068 :
2069 : /* ----------------------------------------------------------------------------
2070 : * Functions to make modifications a bit simpler.
2071 : * ----------------------------------------------------------------------------
2072 : */
2073 :
2074 : extern void simple_table_tuple_insert(Relation rel, TupleTableSlot *slot);
2075 : extern void simple_table_tuple_delete(Relation rel, ItemPointer tid,
2076 : Snapshot snapshot);
2077 : extern void simple_table_tuple_update(Relation rel, ItemPointer otid,
2078 : TupleTableSlot *slot, Snapshot snapshot,
2079 : TU_UpdateIndexes *update_indexes);
2080 :
2081 :
2082 : /* ----------------------------------------------------------------------------
2083 : * Helper functions to implement parallel scans for block oriented AMs.
2084 : * ----------------------------------------------------------------------------
2085 : */
2086 :
2087 : extern Size table_block_parallelscan_estimate(Relation rel);
2088 : extern Size table_block_parallelscan_initialize(Relation rel,
2089 : ParallelTableScanDesc pscan);
2090 : extern void table_block_parallelscan_reinitialize(Relation rel,
2091 : ParallelTableScanDesc pscan);
2092 : extern BlockNumber table_block_parallelscan_nextpage(Relation rel,
2093 : ParallelBlockTableScanWorker pbscanwork,
2094 : ParallelBlockTableScanDesc pbscan);
2095 : extern void table_block_parallelscan_startblock_init(Relation rel,
2096 : ParallelBlockTableScanWorker pbscanwork,
2097 : ParallelBlockTableScanDesc pbscan);
2098 :
2099 :
2100 : /* ----------------------------------------------------------------------------
2101 : * Helper functions to implement relation sizing for block oriented AMs.
2102 : * ----------------------------------------------------------------------------
2103 : */
2104 :
2105 : extern uint64 table_block_relation_size(Relation rel, ForkNumber forkNumber);
2106 : extern void table_block_relation_estimate_size(Relation rel,
2107 : int32 *attr_widths,
2108 : BlockNumber *pages,
2109 : double *tuples,
2110 : double *allvisfrac,
2111 : Size overhead_bytes_per_tuple,
2112 : Size usable_bytes_per_page);
2113 :
2114 : /* ----------------------------------------------------------------------------
2115 : * Functions in tableamapi.c
2116 : * ----------------------------------------------------------------------------
2117 : */
2118 :
2119 : extern const TableAmRoutine *GetTableAmRoutine(Oid amhandler);
2120 :
2121 : /* ----------------------------------------------------------------------------
2122 : * Functions in heapam_handler.c
2123 : * ----------------------------------------------------------------------------
2124 : */
2125 :
2126 : extern const TableAmRoutine *GetHeapamTableAmRoutine(void);
2127 :
2128 : #endif /* TABLEAM_H */
|
