LCOV - code coverage report
Current view: top level - src/include/access - tableam.h (source / functions) Hit Total Coverage
Test: PostgreSQL 19devel Lines: 143 150 95.3 %
Date: 2025-10-10 18:17:38 Functions: 48 48 100.0 %
Legend: Lines: hit not hit

          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 "commands/vacuum.h"
      24             : #include "executor/tuptable.h"
      25             : #include "storage/read_stream.h"
      26             : #include "utils/rel.h"
      27             : #include "utils/snapshot.h"
      28             : 
      29             : 
      30             : #define DEFAULT_TABLE_ACCESS_METHOD "heap"
      31             : 
      32             : /* GUCs */
      33             : extern PGDLLIMPORT char *default_table_access_method;
      34             : extern PGDLLIMPORT bool synchronize_seqscans;
      35             : 
      36             : 
      37             : /* forward references in this file */
      38             : typedef struct BulkInsertStateData BulkInsertStateData;
      39             : typedef struct IndexInfo IndexInfo;
      40             : typedef struct SampleScanState SampleScanState;
      41             : typedef struct ValidateIndexState ValidateIndexState;
      42             : 
      43             : /*
      44             :  * Bitmask values for the flags argument to the scan_begin callback.
      45             :  */
      46             : typedef enum ScanOptions
      47             : {
      48             :     /* one of SO_TYPE_* may be specified */
      49             :     SO_TYPE_SEQSCAN = 1 << 0,
      50             :     SO_TYPE_BITMAPSCAN = 1 << 1,
      51             :     SO_TYPE_SAMPLESCAN = 1 << 2,
      52             :     SO_TYPE_TIDSCAN = 1 << 3,
      53             :     SO_TYPE_TIDRANGESCAN = 1 << 4,
      54             :     SO_TYPE_ANALYZE = 1 << 5,
      55             : 
      56             :     /* several of SO_ALLOW_* may be specified */
      57             :     /* allow or disallow use of access strategy */
      58             :     SO_ALLOW_STRAT = 1 << 6,
      59             :     /* report location to syncscan logic? */
      60             :     SO_ALLOW_SYNC = 1 << 7,
      61             :     /* verify visibility page-at-a-time? */
      62             :     SO_ALLOW_PAGEMODE = 1 << 8,
      63             : 
      64             :     /* unregister snapshot at scan end? */
      65             :     SO_TEMP_SNAPSHOT = 1 << 9,
      66             : }           ScanOptions;
      67             : 
      68             : /*
      69             :  * Result codes for table_{update,delete,lock_tuple}, and for visibility
      70             :  * routines inside table AMs.
      71             :  */
      72             : typedef enum TM_Result
      73             : {
      74             :     /*
      75             :      * Signals that the action succeeded (i.e. update/delete performed, lock
      76             :      * was acquired)
      77             :      */
      78             :     TM_Ok,
      79             : 
      80             :     /* The affected tuple wasn't visible to the relevant snapshot */
      81             :     TM_Invisible,
      82             : 
      83             :     /* The affected tuple was already modified by the calling backend */
      84             :     TM_SelfModified,
      85             : 
      86             :     /*
      87             :      * The affected tuple was updated by another transaction. This includes
      88             :      * the case where tuple was moved to another partition.
      89             :      */
      90             :     TM_Updated,
      91             : 
      92             :     /* The affected tuple was deleted by another transaction */
      93             :     TM_Deleted,
      94             : 
      95             :     /*
      96             :      * The affected tuple is currently being modified by another session. This
      97             :      * will only be returned if table_(update/delete/lock_tuple) are
      98             :      * instructed not to wait.
      99             :      */
     100             :     TM_BeingModified,
     101             : 
     102             :     /* lock couldn't be acquired, action skipped. Only used by lock_tuple */
     103             :     TM_WouldBlock,
     104             : } TM_Result;
     105             : 
     106             : /*
     107             :  * Result codes for table_update(..., update_indexes*..).
     108             :  * Used to determine which indexes to update.
     109             :  */
     110             : typedef enum TU_UpdateIndexes
     111             : {
     112             :     /* No indexed columns were updated (incl. TID addressing of tuple) */
     113             :     TU_None,
     114             : 
     115             :     /* A non-summarizing indexed column was updated, or the TID has changed */
     116             :     TU_All,
     117             : 
     118             :     /* Only summarized columns were updated, TID is unchanged */
     119             :     TU_Summarizing,
     120             : } TU_UpdateIndexes;
     121             : 
     122             : /*
     123             :  * When table_tuple_update, table_tuple_delete, or table_tuple_lock fail
     124             :  * because the target tuple is already outdated, they fill in this struct to
     125             :  * provide information to the caller about what happened. When those functions
     126             :  * succeed, the contents of this struct should not be relied upon, except for
     127             :  * `traversed`, which may be set in both success and failure cases.
     128             :  *
     129             :  * ctid is the target's ctid link: it is the same as the target's TID if the
     130             :  * target was deleted, or the location of the replacement tuple if the target
     131             :  * was updated.
     132             :  *
     133             :  * xmax is the outdating transaction's XID.  If the caller wants to visit the
     134             :  * replacement tuple, it must check that this matches before believing the
     135             :  * replacement is really a match.  This is InvalidTransactionId if the target
     136             :  * was !LP_NORMAL (expected only for a TID retrieved from syscache).
     137             :  *
     138             :  * cmax is the outdating command's CID, but only when the failure code is
     139             :  * TM_SelfModified (i.e., something in the current transaction outdated the
     140             :  * tuple); otherwise cmax is zero.  (We make this restriction because
     141             :  * HeapTupleHeaderGetCmax doesn't work for tuples outdated in other
     142             :  * transactions.)
     143             :  *
     144             :  * traversed indicates if an update chain was followed in order to try to lock
     145             :  * the target tuple.  (This may be set in both success and failure cases.)
     146             :  */
     147             : typedef struct TM_FailureData
     148             : {
     149             :     ItemPointerData ctid;
     150             :     TransactionId xmax;
     151             :     CommandId   cmax;
     152             :     bool        traversed;
     153             : } TM_FailureData;
     154             : 
     155             : /*
     156             :  * State used when calling table_index_delete_tuples().
     157             :  *
     158             :  * Represents the status of table tuples, referenced by table TID and taken by
     159             :  * index AM from index tuples.  State consists of high level parameters of the
     160             :  * deletion operation, plus two mutable palloc()'d arrays for information
     161             :  * about the status of individual table tuples.  These are conceptually one
     162             :  * single array.  Using two arrays keeps the TM_IndexDelete struct small,
     163             :  * which makes sorting the first array (the deltids array) fast.
     164             :  *
     165             :  * Some index AM callers perform simple index tuple deletion (by specifying
     166             :  * bottomup = false), and include only known-dead deltids.  These known-dead
     167             :  * entries are all marked knowndeletable = true directly (typically these are
     168             :  * TIDs from LP_DEAD-marked index tuples), but that isn't strictly required.
     169             :  *
     170             :  * Callers that specify bottomup = true are "bottom-up index deletion"
     171             :  * callers.  The considerations for the tableam are more subtle with these
     172             :  * callers because they ask the tableam to perform highly speculative work,
     173             :  * and might only expect the tableam to check a small fraction of all entries.
     174             :  * Caller is not allowed to specify knowndeletable = true for any entry
     175             :  * because everything is highly speculative.  Bottom-up caller provides
     176             :  * context and hints to tableam -- see comments below for details on how index
     177             :  * AMs and tableams should coordinate during bottom-up index deletion.
     178             :  *
     179             :  * Simple index deletion callers may ask the tableam to perform speculative
     180             :  * work, too.  This is a little like bottom-up deletion, but not too much.
     181             :  * The tableam will only perform speculative work when it's practically free
     182             :  * to do so in passing for simple deletion caller (while always performing
     183             :  * whatever work is needed to enable knowndeletable/LP_DEAD index tuples to
     184             :  * be deleted within index AM).  This is the real reason why it's possible for
     185             :  * simple index deletion caller to specify knowndeletable = false up front
     186             :  * (this means "check if it's possible for me to delete corresponding index
     187             :  * tuple when it's cheap to do so in passing").  The index AM should only
     188             :  * include "extra" entries for index tuples whose TIDs point to a table block
     189             :  * that tableam is expected to have to visit anyway (in the event of a block
     190             :  * orientated tableam).  The tableam isn't strictly obligated to check these
     191             :  * "extra" TIDs, but a block-based AM should always manage to do so in
     192             :  * practice.
     193             :  *
     194             :  * The final contents of the deltids/status arrays are interesting to callers
     195             :  * that ask tableam to perform speculative work (i.e. when _any_ items have
     196             :  * knowndeletable set to false up front).  These index AM callers will
     197             :  * naturally need to consult final state to determine which index tuples are
     198             :  * in fact deletable.
     199             :  *
     200             :  * The index AM can keep track of which index tuple relates to which deltid by
     201             :  * setting idxoffnum (and/or relying on each entry being uniquely identifiable
     202             :  * using tid), which is important when the final contents of the array will
     203             :  * need to be interpreted -- the array can shrink from initial size after
     204             :  * tableam processing and/or have entries in a new order (tableam may sort
     205             :  * deltids array for its own reasons).  Bottom-up callers may find that final
     206             :  * ndeltids is 0 on return from call to tableam, in which case no index tuple
     207             :  * deletions are possible.  Simple deletion callers can rely on any entries
     208             :  * they know to be deletable appearing in the final array as deletable.
     209             :  */
     210             : typedef struct TM_IndexDelete
     211             : {
     212             :     ItemPointerData tid;        /* table TID from index tuple */
     213             :     int16       id;             /* Offset into TM_IndexStatus array */
     214             : } TM_IndexDelete;
     215             : 
     216             : typedef struct TM_IndexStatus
     217             : {
     218             :     OffsetNumber idxoffnum;     /* Index am page offset number */
     219             :     bool        knowndeletable; /* Currently known to be deletable? */
     220             : 
     221             :     /* Bottom-up index deletion specific fields follow */
     222             :     bool        promising;      /* Promising (duplicate) index tuple? */
     223             :     int16       freespace;      /* Space freed in index if deleted */
     224             : } TM_IndexStatus;
     225             : 
     226             : /*
     227             :  * Index AM/tableam coordination is central to the design of bottom-up index
     228             :  * deletion.  The index AM provides hints about where to look to the tableam
     229             :  * by marking some entries as "promising".  Index AM does this with duplicate
     230             :  * index tuples that are strongly suspected to be old versions left behind by
     231             :  * UPDATEs that did not logically modify indexed values.  Index AM may find it
     232             :  * helpful to only mark entries as promising when they're thought to have been
     233             :  * affected by such an UPDATE in the recent past.
     234             :  *
     235             :  * Bottom-up index deletion casts a wide net at first, usually by including
     236             :  * all TIDs on a target index page.  It is up to the tableam to worry about
     237             :  * the cost of checking transaction status information.  The tableam is in
     238             :  * control, but needs careful guidance from the index AM.  Index AM requests
     239             :  * that bottomupfreespace target be met, while tableam measures progress
     240             :  * towards that goal by tallying the per-entry freespace value for known
     241             :  * deletable entries. (All !bottomup callers can just set these space related
     242             :  * fields to zero.)
     243             :  */
     244             : typedef struct TM_IndexDeleteOp
     245             : {
     246             :     Relation    irel;           /* Target index relation */
     247             :     BlockNumber iblknum;        /* Index block number (for error reports) */
     248             :     bool        bottomup;       /* Bottom-up (not simple) deletion? */
     249             :     int         bottomupfreespace;  /* Bottom-up space target */
     250             : 
     251             :     /* Mutable per-TID information follows (index AM initializes entries) */
     252             :     int         ndeltids;       /* Current # of deltids/status elements */
     253             :     TM_IndexDelete *deltids;
     254             :     TM_IndexStatus *status;
     255             : } TM_IndexDeleteOp;
     256             : 
     257             : /* "options" flag bits for table_tuple_insert */
     258             : /* TABLE_INSERT_SKIP_WAL was 0x0001; RelationNeedsWAL() now governs */
     259             : #define TABLE_INSERT_SKIP_FSM       0x0002
     260             : #define TABLE_INSERT_FROZEN         0x0004
     261             : #define TABLE_INSERT_NO_LOGICAL     0x0008
     262             : 
     263             : /* flag bits for table_tuple_lock */
     264             : /* Follow tuples whose update is in progress if lock modes don't conflict  */
     265             : #define TUPLE_LOCK_FLAG_LOCK_UPDATE_IN_PROGRESS (1 << 0)
     266             : /* Follow update chain and lock latest version of tuple */
     267             : #define TUPLE_LOCK_FLAG_FIND_LAST_VERSION       (1 << 1)
     268             : 
     269             : 
     270             : /* Typedef for callback function for table_index_build_scan */
     271             : typedef void (*IndexBuildCallback) (Relation index,
     272             :                                     ItemPointer tid,
     273             :                                     Datum *values,
     274             :                                     bool *isnull,
     275             :                                     bool tupleIsAlive,
     276             :                                     void *state);
     277             : 
     278             : /*
     279             :  * API struct for a table AM.  Note this must be allocated in a
     280             :  * server-lifetime manner, typically as a static const struct, which then gets
     281             :  * returned by FormData_pg_am.amhandler.
     282             :  *
     283             :  * In most cases it's not appropriate to call the callbacks directly, use the
     284             :  * table_* wrapper functions instead.
     285             :  *
     286             :  * GetTableAmRoutine() asserts that required callbacks are filled in, remember
     287             :  * to update when adding a callback.
     288             :  */
     289             : typedef struct TableAmRoutine
     290             : {
     291             :     /* this must be set to T_TableAmRoutine */
     292             :     NodeTag     type;
     293             : 
     294             : 
     295             :     /* ------------------------------------------------------------------------
     296             :      * Slot related callbacks.
     297             :      * ------------------------------------------------------------------------
     298             :      */
     299             : 
     300             :     /*
     301             :      * Return slot implementation suitable for storing a tuple of this AM.
     302             :      */
     303             :     const TupleTableSlotOps *(*slot_callbacks) (Relation rel);
     304             : 
     305             : 
     306             :     /* ------------------------------------------------------------------------
     307             :      * Table scan callbacks.
     308             :      * ------------------------------------------------------------------------
     309             :      */
     310             : 
     311             :     /*
     312             :      * Start a scan of `rel`.  The callback has to return a TableScanDesc,
     313             :      * which will typically be embedded in a larger, AM specific, struct.
     314             :      *
     315             :      * If nkeys != 0, the results need to be filtered by those scan keys.
     316             :      *
     317             :      * pscan, if not NULL, will have already been initialized with
     318             :      * parallelscan_initialize(), and has to be for the same relation. Will
     319             :      * only be set coming from table_beginscan_parallel().
     320             :      *
     321             :      * `flags` is a bitmask indicating the type of scan (ScanOptions's
     322             :      * SO_TYPE_*, currently only one may be specified), options controlling
     323             :      * the scan's behaviour (ScanOptions's SO_ALLOW_*, several may be
     324             :      * specified, an AM may ignore unsupported ones) and whether the snapshot
     325             :      * needs to be deallocated at scan_end (ScanOptions's SO_TEMP_SNAPSHOT).
     326             :      */
     327             :     TableScanDesc (*scan_begin) (Relation rel,
     328             :                                  Snapshot snapshot,
     329             :                                  int nkeys, ScanKeyData *key,
     330             :                                  ParallelTableScanDesc pscan,
     331             :                                  uint32 flags);
     332             : 
     333             :     /*
     334             :      * Release resources and deallocate scan. If TableScanDesc.temp_snap,
     335             :      * TableScanDesc.rs_snapshot needs to be unregistered.
     336             :      */
     337             :     void        (*scan_end) (TableScanDesc scan);
     338             : 
     339             :     /*
     340             :      * Restart relation scan.  If set_params is set to true, allow_{strat,
     341             :      * sync, pagemode} (see scan_begin) changes should be taken into account.
     342             :      */
     343             :     void        (*scan_rescan) (TableScanDesc scan, ScanKeyData *key,
     344             :                                 bool set_params, bool allow_strat,
     345             :                                 bool allow_sync, bool allow_pagemode);
     346             : 
     347             :     /*
     348             :      * Return next tuple from `scan`, store in slot.
     349             :      */
     350             :     bool        (*scan_getnextslot) (TableScanDesc scan,
     351             :                                      ScanDirection direction,
     352             :                                      TupleTableSlot *slot);
     353             : 
     354             :     /*-----------
     355             :      * Optional functions to provide scanning for ranges of ItemPointers.
     356             :      * Implementations must either provide both of these functions, or neither
     357             :      * of them.
     358             :      *
     359             :      * Implementations of scan_set_tidrange must themselves handle
     360             :      * ItemPointers of any value. i.e, they must handle each of the following:
     361             :      *
     362             :      * 1) mintid or maxtid is beyond the end of the table; and
     363             :      * 2) mintid is above maxtid; and
     364             :      * 3) item offset for mintid or maxtid is beyond the maximum offset
     365             :      * allowed by the AM.
     366             :      *
     367             :      * Implementations can assume that scan_set_tidrange is always called
     368             :      * before scan_getnextslot_tidrange or after scan_rescan and before any
     369             :      * further calls to scan_getnextslot_tidrange.
     370             :      */
     371             :     void        (*scan_set_tidrange) (TableScanDesc scan,
     372             :                                       ItemPointer mintid,
     373             :                                       ItemPointer maxtid);
     374             : 
     375             :     /*
     376             :      * Return next tuple from `scan` that's in the range of TIDs defined by
     377             :      * scan_set_tidrange.
     378             :      */
     379             :     bool        (*scan_getnextslot_tidrange) (TableScanDesc scan,
     380             :                                               ScanDirection direction,
     381             :                                               TupleTableSlot *slot);
     382             : 
     383             :     /* ------------------------------------------------------------------------
     384             :      * Parallel table scan related functions.
     385             :      * ------------------------------------------------------------------------
     386             :      */
     387             : 
     388             :     /*
     389             :      * Estimate the size of shared memory needed for a parallel scan of this
     390             :      * relation. The snapshot does not need to be accounted for.
     391             :      */
     392             :     Size        (*parallelscan_estimate) (Relation rel);
     393             : 
     394             :     /*
     395             :      * Initialize ParallelTableScanDesc for a parallel scan of this relation.
     396             :      * `pscan` will be sized according to parallelscan_estimate() for the same
     397             :      * relation.
     398             :      */
     399             :     Size        (*parallelscan_initialize) (Relation rel,
     400             :                                             ParallelTableScanDesc pscan);
     401             : 
     402             :     /*
     403             :      * Reinitialize `pscan` for a new scan. `rel` will be the same relation as
     404             :      * when `pscan` was initialized by parallelscan_initialize.
     405             :      */
     406             :     void        (*parallelscan_reinitialize) (Relation rel,
     407             :                                               ParallelTableScanDesc pscan);
     408             : 
     409             : 
     410             :     /* ------------------------------------------------------------------------
     411             :      * Index Scan Callbacks
     412             :      * ------------------------------------------------------------------------
     413             :      */
     414             : 
     415             :     /*
     416             :      * Prepare to fetch tuples from the relation, as needed when fetching
     417             :      * tuples for an index scan.  The callback has to return an
     418             :      * IndexFetchTableData, which the AM will typically embed in a larger
     419             :      * structure with additional information.
     420             :      *
     421             :      * Tuples for an index scan can then be fetched via index_fetch_tuple.
     422             :      */
     423             :     struct IndexFetchTableData *(*index_fetch_begin) (Relation rel);
     424             : 
     425             :     /*
     426             :      * Reset index fetch. Typically this will release cross index fetch
     427             :      * resources held in IndexFetchTableData.
     428             :      */
     429             :     void        (*index_fetch_reset) (struct IndexFetchTableData *data);
     430             : 
     431             :     /*
     432             :      * Release resources and deallocate index fetch.
     433             :      */
     434             :     void        (*index_fetch_end) (struct IndexFetchTableData *data);
     435             : 
     436             :     /*
     437             :      * Fetch tuple at `tid` into `slot`, after doing a visibility test
     438             :      * according to `snapshot`. If a tuple was found and passed the visibility
     439             :      * test, return true, false otherwise.
     440             :      *
     441             :      * Note that AMs that do not necessarily update indexes when indexed
     442             :      * columns do not change, need to return the current/correct version of
     443             :      * the tuple that is visible to the snapshot, even if the tid points to an
     444             :      * older version of the tuple.
     445             :      *
     446             :      * *call_again is false on the first call to index_fetch_tuple for a tid.
     447             :      * If there potentially is another tuple matching the tid, *call_again
     448             :      * needs to be set to true by index_fetch_tuple, signaling to the caller
     449             :      * that index_fetch_tuple should be called again for the same tid.
     450             :      *
     451             :      * *all_dead, if all_dead is not NULL, should be set to true by
     452             :      * index_fetch_tuple iff it is guaranteed that no backend needs to see
     453             :      * that tuple. Index AMs can use that to avoid returning that tid in
     454             :      * future searches.
     455             :      */
     456             :     bool        (*index_fetch_tuple) (struct IndexFetchTableData *scan,
     457             :                                       ItemPointer tid,
     458             :                                       Snapshot snapshot,
     459             :                                       TupleTableSlot *slot,
     460             :                                       bool *call_again, bool *all_dead);
     461             : 
     462             : 
     463             :     /* ------------------------------------------------------------------------
     464             :      * Callbacks for non-modifying operations on individual tuples
     465             :      * ------------------------------------------------------------------------
     466             :      */
     467             : 
     468             :     /*
     469             :      * Fetch tuple at `tid` into `slot`, after doing a visibility test
     470             :      * according to `snapshot`. If a tuple was found and passed the visibility
     471             :      * test, returns true, false otherwise.
     472             :      */
     473             :     bool        (*tuple_fetch_row_version) (Relation rel,
     474             :                                             ItemPointer tid,
     475             :                                             Snapshot snapshot,
     476             :                                             TupleTableSlot *slot);
     477             : 
     478             :     /*
     479             :      * Is tid valid for a scan of this relation.
     480             :      */
     481             :     bool        (*tuple_tid_valid) (TableScanDesc scan,
     482             :                                     ItemPointer tid);
     483             : 
     484             :     /*
     485             :      * Return the latest version of the tuple at `tid`, by updating `tid` to
     486             :      * point at the newest version.
     487             :      */
     488             :     void        (*tuple_get_latest_tid) (TableScanDesc scan,
     489             :                                          ItemPointer tid);
     490             : 
     491             :     /*
     492             :      * Does the tuple in `slot` satisfy `snapshot`?  The slot needs to be of
     493             :      * the appropriate type for the AM.
     494             :      */
     495             :     bool        (*tuple_satisfies_snapshot) (Relation rel,
     496             :                                              TupleTableSlot *slot,
     497             :                                              Snapshot snapshot);
     498             : 
     499             :     /* see table_index_delete_tuples() */
     500             :     TransactionId (*index_delete_tuples) (Relation rel,
     501             :                                           TM_IndexDeleteOp *delstate);
     502             : 
     503             : 
     504             :     /* ------------------------------------------------------------------------
     505             :      * Manipulations of physical tuples.
     506             :      * ------------------------------------------------------------------------
     507             :      */
     508             : 
     509             :     /* see table_tuple_insert() for reference about parameters */
     510             :     void        (*tuple_insert) (Relation rel, TupleTableSlot *slot,
     511             :                                  CommandId cid, int options,
     512             :                                  BulkInsertStateData *bistate);
     513             : 
     514             :     /* see table_tuple_insert_speculative() for reference about parameters */
     515             :     void        (*tuple_insert_speculative) (Relation rel,
     516             :                                              TupleTableSlot *slot,
     517             :                                              CommandId cid,
     518             :                                              int options,
     519             :                                              BulkInsertStateData *bistate,
     520             :                                              uint32 specToken);
     521             : 
     522             :     /* see table_tuple_complete_speculative() for reference about parameters */
     523             :     void        (*tuple_complete_speculative) (Relation rel,
     524             :                                                TupleTableSlot *slot,
     525             :                                                uint32 specToken,
     526             :                                                bool succeeded);
     527             : 
     528             :     /* see table_multi_insert() for reference about parameters */
     529             :     void        (*multi_insert) (Relation rel, TupleTableSlot **slots, int nslots,
     530             :                                  CommandId cid, int options, BulkInsertStateData *bistate);
     531             : 
     532             :     /* see table_tuple_delete() for reference about parameters */
     533             :     TM_Result   (*tuple_delete) (Relation rel,
     534             :                                  ItemPointer tid,
     535             :                                  CommandId cid,
     536             :                                  Snapshot snapshot,
     537             :                                  Snapshot crosscheck,
     538             :                                  bool wait,
     539             :                                  TM_FailureData *tmfd,
     540             :                                  bool changingPart);
     541             : 
     542             :     /* see table_tuple_update() for reference about parameters */
     543             :     TM_Result   (*tuple_update) (Relation rel,
     544             :                                  ItemPointer otid,
     545             :                                  TupleTableSlot *slot,
     546             :                                  CommandId cid,
     547             :                                  Snapshot snapshot,
     548             :                                  Snapshot crosscheck,
     549             :                                  bool wait,
     550             :                                  TM_FailureData *tmfd,
     551             :                                  LockTupleMode *lockmode,
     552             :                                  TU_UpdateIndexes *update_indexes);
     553             : 
     554             :     /* see table_tuple_lock() for reference about parameters */
     555             :     TM_Result   (*tuple_lock) (Relation rel,
     556             :                                ItemPointer tid,
     557             :                                Snapshot snapshot,
     558             :                                TupleTableSlot *slot,
     559             :                                CommandId cid,
     560             :                                LockTupleMode mode,
     561             :                                LockWaitPolicy wait_policy,
     562             :                                uint8 flags,
     563             :                                TM_FailureData *tmfd);
     564             : 
     565             :     /*
     566             :      * Perform operations necessary to complete insertions made via
     567             :      * tuple_insert and multi_insert with a BulkInsertState specified. In-tree
     568             :      * access methods ceased to use this.
     569             :      *
     570             :      * Typically callers of tuple_insert and multi_insert will just pass all
     571             :      * the flags that apply to them, and each AM has to decide which of them
     572             :      * make sense for it, and then only take actions in finish_bulk_insert for
     573             :      * those flags, and ignore others.
     574             :      *
     575             :      * Optional callback.
     576             :      */
     577             :     void        (*finish_bulk_insert) (Relation rel, int options);
     578             : 
     579             : 
     580             :     /* ------------------------------------------------------------------------
     581             :      * DDL related functionality.
     582             :      * ------------------------------------------------------------------------
     583             :      */
     584             : 
     585             :     /*
     586             :      * This callback needs to create new relation storage for `rel`, with
     587             :      * appropriate durability behaviour for `persistence`.
     588             :      *
     589             :      * Note that only the subset of the relcache filled by
     590             :      * RelationBuildLocalRelation() can be relied upon and that the relation's
     591             :      * catalog entries will either not yet exist (new relation), or will still
     592             :      * reference the old relfilelocator.
     593             :      *
     594             :      * As output *freezeXid, *minmulti must be set to the values appropriate
     595             :      * for pg_class.{relfrozenxid, relminmxid}. For AMs that don't need those
     596             :      * fields to be filled they can be set to InvalidTransactionId and
     597             :      * InvalidMultiXactId, respectively.
     598             :      *
     599             :      * See also table_relation_set_new_filelocator().
     600             :      */
     601             :     void        (*relation_set_new_filelocator) (Relation rel,
     602             :                                                  const RelFileLocator *newrlocator,
     603             :                                                  char persistence,
     604             :                                                  TransactionId *freezeXid,
     605             :                                                  MultiXactId *minmulti);
     606             : 
     607             :     /*
     608             :      * This callback needs to remove all contents from `rel`'s current
     609             :      * relfilelocator. No provisions for transactional behaviour need to be
     610             :      * made.  Often this can be implemented by truncating the underlying
     611             :      * storage to its minimal size.
     612             :      *
     613             :      * See also table_relation_nontransactional_truncate().
     614             :      */
     615             :     void        (*relation_nontransactional_truncate) (Relation rel);
     616             : 
     617             :     /*
     618             :      * See table_relation_copy_data().
     619             :      *
     620             :      * This can typically be implemented by directly copying the underlying
     621             :      * storage, unless it contains references to the tablespace internally.
     622             :      */
     623             :     void        (*relation_copy_data) (Relation rel,
     624             :                                        const RelFileLocator *newrlocator);
     625             : 
     626             :     /* See table_relation_copy_for_cluster() */
     627             :     void        (*relation_copy_for_cluster) (Relation OldTable,
     628             :                                               Relation NewTable,
     629             :                                               Relation OldIndex,
     630             :                                               bool use_sort,
     631             :                                               TransactionId OldestXmin,
     632             :                                               TransactionId *xid_cutoff,
     633             :                                               MultiXactId *multi_cutoff,
     634             :                                               double *num_tuples,
     635             :                                               double *tups_vacuumed,
     636             :                                               double *tups_recently_dead);
     637             : 
     638             :     /*
     639             :      * React to VACUUM command on the relation. The VACUUM can be triggered by
     640             :      * a user or by autovacuum. The specific actions performed by the AM will
     641             :      * depend heavily on the individual AM.
     642             :      *
     643             :      * On entry a transaction is already established, and the relation is
     644             :      * locked with a ShareUpdateExclusive lock.
     645             :      *
     646             :      * Note that neither VACUUM FULL (and CLUSTER), nor ANALYZE go through
     647             :      * this routine, even if (for ANALYZE) it is part of the same VACUUM
     648             :      * command.
     649             :      *
     650             :      * There probably, in the future, needs to be a separate callback to
     651             :      * integrate with autovacuum's scheduling.
     652             :      */
     653             :     void        (*relation_vacuum) (Relation rel,
     654             :                                     const VacuumParams params,
     655             :                                     BufferAccessStrategy bstrategy);
     656             : 
     657             :     /*
     658             :      * Prepare to analyze block `blockno` of `scan`. The scan has been started
     659             :      * with table_beginscan_analyze().  See also
     660             :      * table_scan_analyze_next_block().
     661             :      *
     662             :      * The callback may acquire resources like locks that are held until
     663             :      * table_scan_analyze_next_tuple() returns false. It e.g. can make sense
     664             :      * to hold a lock until all tuples on a block have been analyzed by
     665             :      * scan_analyze_next_tuple.
     666             :      *
     667             :      * The callback can return false if the block is not suitable for
     668             :      * sampling, e.g. because it's a metapage that could never contain tuples.
     669             :      *
     670             :      * XXX: This obviously is primarily suited for block-based AMs. It's not
     671             :      * clear what a good interface for non block based AMs would be, so there
     672             :      * isn't one yet.
     673             :      */
     674             :     bool        (*scan_analyze_next_block) (TableScanDesc scan,
     675             :                                             ReadStream *stream);
     676             : 
     677             :     /*
     678             :      * See table_scan_analyze_next_tuple().
     679             :      *
     680             :      * Not every AM might have a meaningful concept of dead rows, in which
     681             :      * case it's OK to not increment *deadrows - but note that that may
     682             :      * influence autovacuum scheduling (see comment for relation_vacuum
     683             :      * callback).
     684             :      */
     685             :     bool        (*scan_analyze_next_tuple) (TableScanDesc scan,
     686             :                                             TransactionId OldestXmin,
     687             :                                             double *liverows,
     688             :                                             double *deadrows,
     689             :                                             TupleTableSlot *slot);
     690             : 
     691             :     /* see table_index_build_range_scan for reference about parameters */
     692             :     double      (*index_build_range_scan) (Relation table_rel,
     693             :                                            Relation index_rel,
     694             :                                            IndexInfo *index_info,
     695             :                                            bool allow_sync,
     696             :                                            bool anyvisible,
     697             :                                            bool progress,
     698             :                                            BlockNumber start_blockno,
     699             :                                            BlockNumber numblocks,
     700             :                                            IndexBuildCallback callback,
     701             :                                            void *callback_state,
     702             :                                            TableScanDesc scan);
     703             : 
     704             :     /* see table_index_validate_scan for reference about parameters */
     705             :     void        (*index_validate_scan) (Relation table_rel,
     706             :                                         Relation index_rel,
     707             :                                         IndexInfo *index_info,
     708             :                                         Snapshot snapshot,
     709             :                                         ValidateIndexState *state);
     710             : 
     711             : 
     712             :     /* ------------------------------------------------------------------------
     713             :      * Miscellaneous functions.
     714             :      * ------------------------------------------------------------------------
     715             :      */
     716             : 
     717             :     /*
     718             :      * See table_relation_size().
     719             :      *
     720             :      * Note that currently a few callers use the MAIN_FORKNUM size to figure
     721             :      * out the range of potentially interesting blocks (brin, analyze). It's
     722             :      * probable that we'll need to revise the interface for those at some
     723             :      * point.
     724             :      */
     725             :     uint64      (*relation_size) (Relation rel, ForkNumber forkNumber);
     726             : 
     727             : 
     728             :     /*
     729             :      * This callback should return true if the relation requires a TOAST table
     730             :      * and false if it does not.  It may wish to examine the relation's tuple
     731             :      * descriptor before making a decision, but if it uses some other method
     732             :      * of storing large values (or if it does not support them) it can simply
     733             :      * return false.
     734             :      */
     735             :     bool        (*relation_needs_toast_table) (Relation rel);
     736             : 
     737             :     /*
     738             :      * This callback should return the OID of the table AM that implements
     739             :      * TOAST tables for this AM.  If the relation_needs_toast_table callback
     740             :      * always returns false, this callback is not required.
     741             :      */
     742             :     Oid         (*relation_toast_am) (Relation rel);
     743             : 
     744             :     /*
     745             :      * This callback is invoked when detoasting a value stored in a toast
     746             :      * table implemented by this AM.  See table_relation_fetch_toast_slice()
     747             :      * for more details.
     748             :      */
     749             :     void        (*relation_fetch_toast_slice) (Relation toastrel, Oid valueid,
     750             :                                                int32 attrsize,
     751             :                                                int32 sliceoffset,
     752             :                                                int32 slicelength,
     753             :                                                struct varlena *result);
     754             : 
     755             : 
     756             :     /* ------------------------------------------------------------------------
     757             :      * Planner related functions.
     758             :      * ------------------------------------------------------------------------
     759             :      */
     760             : 
     761             :     /*
     762             :      * See table_relation_estimate_size().
     763             :      *
     764             :      * While block oriented, it shouldn't be too hard for an AM that doesn't
     765             :      * internally use blocks to convert into a usable representation.
     766             :      *
     767             :      * This differs from the relation_size callback by returning size
     768             :      * estimates (both relation size and tuple count) for planning purposes,
     769             :      * rather than returning a currently correct estimate.
     770             :      */
     771             :     void        (*relation_estimate_size) (Relation rel, int32 *attr_widths,
     772             :                                            BlockNumber *pages, double *tuples,
     773             :                                            double *allvisfrac);
     774             : 
     775             : 
     776             :     /* ------------------------------------------------------------------------
     777             :      * Executor related functions.
     778             :      * ------------------------------------------------------------------------
     779             :      */
     780             : 
     781             :     /*
     782             :      * Fetch the next tuple of a bitmap table scan into `slot` and return true
     783             :      * if a visible tuple was found, false otherwise.
     784             :      *
     785             :      * `lossy_pages` is incremented if the bitmap is lossy for the selected
     786             :      * page; otherwise, `exact_pages` is incremented. These are tracked for
     787             :      * display in EXPLAIN ANALYZE output.
     788             :      *
     789             :      * Prefetching additional data from the bitmap is left to the table AM.
     790             :      *
     791             :      * This is an optional callback.
     792             :      */
     793             :     bool        (*scan_bitmap_next_tuple) (TableScanDesc scan,
     794             :                                            TupleTableSlot *slot,
     795             :                                            bool *recheck,
     796             :                                            uint64 *lossy_pages,
     797             :                                            uint64 *exact_pages);
     798             : 
     799             :     /*
     800             :      * Prepare to fetch tuples from the next block in a sample scan. Return
     801             :      * false if the sample scan is finished, true otherwise. `scan` was
     802             :      * started via table_beginscan_sampling().
     803             :      *
     804             :      * Typically this will first determine the target block by calling the
     805             :      * TsmRoutine's NextSampleBlock() callback if not NULL, or alternatively
     806             :      * perform a sequential scan over all blocks.  The determined block is
     807             :      * then typically read and pinned.
     808             :      *
     809             :      * As the TsmRoutine interface is block based, a block needs to be passed
     810             :      * to NextSampleBlock(). If that's not appropriate for an AM, it
     811             :      * internally needs to perform mapping between the internal and a block
     812             :      * based representation.
     813             :      *
     814             :      * Note that it's not acceptable to hold deadlock prone resources such as
     815             :      * lwlocks until scan_sample_next_tuple() has exhausted the tuples on the
     816             :      * block - the tuple is likely to be returned to an upper query node, and
     817             :      * the next call could be off a long while. Holding buffer pins and such
     818             :      * is obviously OK.
     819             :      *
     820             :      * Currently it is required to implement this interface, as there's no
     821             :      * alternative way (contrary e.g. to bitmap scans) to implement sample
     822             :      * scans. If infeasible to implement, the AM may raise an error.
     823             :      */
     824             :     bool        (*scan_sample_next_block) (TableScanDesc scan,
     825             :                                            SampleScanState *scanstate);
     826             : 
     827             :     /*
     828             :      * This callback, only called after scan_sample_next_block has returned
     829             :      * true, should determine the next tuple to be returned from the selected
     830             :      * block using the TsmRoutine's NextSampleTuple() callback.
     831             :      *
     832             :      * The callback needs to perform visibility checks, and only return
     833             :      * visible tuples. That obviously can mean calling NextSampleTuple()
     834             :      * multiple times.
     835             :      *
     836             :      * The TsmRoutine interface assumes that there's a maximum offset on a
     837             :      * given page, so if that doesn't apply to an AM, it needs to emulate that
     838             :      * assumption somehow.
     839             :      */
     840             :     bool        (*scan_sample_next_tuple) (TableScanDesc scan,
     841             :                                            SampleScanState *scanstate,
     842             :                                            TupleTableSlot *slot);
     843             : 
     844             : } TableAmRoutine;
     845             : 
     846             : 
     847             : /* ----------------------------------------------------------------------------
     848             :  * Slot functions.
     849             :  * ----------------------------------------------------------------------------
     850             :  */
     851             : 
     852             : /*
     853             :  * Returns slot callbacks suitable for holding tuples of the appropriate type
     854             :  * for the relation.  Works for tables, views, foreign tables and partitioned
     855             :  * tables.
     856             :  */
     857             : extern const TupleTableSlotOps *table_slot_callbacks(Relation relation);
     858             : 
     859             : /*
     860             :  * Returns slot using the callbacks returned by table_slot_callbacks(), and
     861             :  * registers it on *reglist.
     862             :  */
     863             : extern TupleTableSlot *table_slot_create(Relation relation, List **reglist);
     864             : 
     865             : 
     866             : /* ----------------------------------------------------------------------------
     867             :  * Table scan functions.
     868             :  * ----------------------------------------------------------------------------
     869             :  */
     870             : 
     871             : /*
     872             :  * Start a scan of `rel`. Returned tuples pass a visibility test of
     873             :  * `snapshot`, and if nkeys != 0, the results are filtered by those scan keys.
     874             :  */
     875             : static inline TableScanDesc
     876      205268 : table_beginscan(Relation rel, Snapshot snapshot,
     877             :                 int nkeys, ScanKeyData *key)
     878             : {
     879      205268 :     uint32      flags = SO_TYPE_SEQSCAN |
     880             :         SO_ALLOW_STRAT | SO_ALLOW_SYNC | SO_ALLOW_PAGEMODE;
     881             : 
     882      205268 :     return rel->rd_tableam->scan_begin(rel, snapshot, nkeys, key, NULL, flags);
     883             : }
     884             : 
     885             : /*
     886             :  * Like table_beginscan(), but for scanning catalog. It'll automatically use a
     887             :  * snapshot appropriate for scanning catalog relations.
     888             :  */
     889             : extern TableScanDesc table_beginscan_catalog(Relation relation, int nkeys,
     890             :                                              ScanKeyData *key);
     891             : 
     892             : /*
     893             :  * Like table_beginscan(), but table_beginscan_strat() offers an extended API
     894             :  * that lets the caller control whether a nondefault buffer access strategy
     895             :  * can be used, and whether syncscan can be chosen (possibly resulting in the
     896             :  * scan not starting from block zero).  Both of these default to true with
     897             :  * plain table_beginscan.
     898             :  */
     899             : static inline TableScanDesc
     900      447412 : table_beginscan_strat(Relation rel, Snapshot snapshot,
     901             :                       int nkeys, ScanKeyData *key,
     902             :                       bool allow_strat, bool allow_sync)
     903             : {
     904      447412 :     uint32      flags = SO_TYPE_SEQSCAN | SO_ALLOW_PAGEMODE;
     905             : 
     906      447412 :     if (allow_strat)
     907      447412 :         flags |= SO_ALLOW_STRAT;
     908      447412 :     if (allow_sync)
     909       53476 :         flags |= SO_ALLOW_SYNC;
     910             : 
     911      447412 :     return rel->rd_tableam->scan_begin(rel, snapshot, nkeys, key, NULL, flags);
     912             : }
     913             : 
     914             : /*
     915             :  * table_beginscan_bm is an alternative entry point for setting up a
     916             :  * TableScanDesc for a bitmap heap scan.  Although that scan technology is
     917             :  * really quite unlike a standard seqscan, there is just enough commonality to
     918             :  * make it worth using the same data structure.
     919             :  */
     920             : static inline TableScanDesc
     921       16292 : table_beginscan_bm(Relation rel, Snapshot snapshot,
     922             :                    int nkeys, ScanKeyData *key)
     923             : {
     924       16292 :     uint32      flags = SO_TYPE_BITMAPSCAN | SO_ALLOW_PAGEMODE;
     925             : 
     926       16292 :     return rel->rd_tableam->scan_begin(rel, snapshot, nkeys, key,
     927             :                                        NULL, flags);
     928             : }
     929             : 
     930             : /*
     931             :  * table_beginscan_sampling is an alternative entry point for setting up a
     932             :  * TableScanDesc for a TABLESAMPLE scan.  As with bitmap scans, it's worth
     933             :  * using the same data structure although the behavior is rather different.
     934             :  * In addition to the options offered by table_beginscan_strat, this call
     935             :  * also allows control of whether page-mode visibility checking is used.
     936             :  */
     937             : static inline TableScanDesc
     938         146 : table_beginscan_sampling(Relation rel, Snapshot snapshot,
     939             :                          int nkeys, ScanKeyData *key,
     940             :                          bool allow_strat, bool allow_sync,
     941             :                          bool allow_pagemode)
     942             : {
     943         146 :     uint32      flags = SO_TYPE_SAMPLESCAN;
     944             : 
     945         146 :     if (allow_strat)
     946         134 :         flags |= SO_ALLOW_STRAT;
     947         146 :     if (allow_sync)
     948          66 :         flags |= SO_ALLOW_SYNC;
     949         146 :     if (allow_pagemode)
     950         122 :         flags |= SO_ALLOW_PAGEMODE;
     951             : 
     952         146 :     return rel->rd_tableam->scan_begin(rel, snapshot, nkeys, key, NULL, flags);
     953             : }
     954             : 
     955             : /*
     956             :  * table_beginscan_tid is an alternative entry point for setting up a
     957             :  * TableScanDesc for a Tid scan. As with bitmap scans, it's worth using
     958             :  * the same data structure although the behavior is rather different.
     959             :  */
     960             : static inline TableScanDesc
     961         786 : table_beginscan_tid(Relation rel, Snapshot snapshot)
     962             : {
     963         786 :     uint32      flags = SO_TYPE_TIDSCAN;
     964             : 
     965         786 :     return rel->rd_tableam->scan_begin(rel, snapshot, 0, NULL, NULL, flags);
     966             : }
     967             : 
     968             : /*
     969             :  * table_beginscan_analyze is an alternative entry point for setting up a
     970             :  * TableScanDesc for an ANALYZE scan.  As with bitmap scans, it's worth using
     971             :  * the same data structure although the behavior is rather different.
     972             :  */
     973             : static inline TableScanDesc
     974       17290 : table_beginscan_analyze(Relation rel)
     975             : {
     976       17290 :     uint32      flags = SO_TYPE_ANALYZE;
     977             : 
     978       17290 :     return rel->rd_tableam->scan_begin(rel, NULL, 0, NULL, NULL, flags);
     979             : }
     980             : 
     981             : /*
     982             :  * End relation scan.
     983             :  */
     984             : static inline void
     985      764984 : table_endscan(TableScanDesc scan)
     986             : {
     987      764984 :     scan->rs_rd->rd_tableam->scan_end(scan);
     988      764984 : }
     989             : 
     990             : /*
     991             :  * Restart a relation scan.
     992             :  */
     993             : static inline void
     994     1224900 : table_rescan(TableScanDesc scan, ScanKeyData *key)
     995             : {
     996     1224900 :     scan->rs_rd->rd_tableam->scan_rescan(scan, key, false, false, false, false);
     997     1224900 : }
     998             : 
     999             : /*
    1000             :  * Restart a relation scan after changing params.
    1001             :  *
    1002             :  * This call allows changing the buffer strategy, syncscan, and pagemode
    1003             :  * options before starting a fresh scan.  Note that although the actual use of
    1004             :  * syncscan might change (effectively, enabling or disabling reporting), the
    1005             :  * previously selected startblock will be kept.
    1006             :  */
    1007             : static inline void
    1008          30 : table_rescan_set_params(TableScanDesc scan, ScanKeyData *key,
    1009             :                         bool allow_strat, bool allow_sync, bool allow_pagemode)
    1010             : {
    1011          30 :     scan->rs_rd->rd_tableam->scan_rescan(scan, key, true,
    1012             :                                          allow_strat, allow_sync,
    1013             :                                          allow_pagemode);
    1014          30 : }
    1015             : 
    1016             : /*
    1017             :  * Return next tuple from `scan`, store in slot.
    1018             :  */
    1019             : static inline bool
    1020    95726642 : table_scan_getnextslot(TableScanDesc sscan, ScanDirection direction, TupleTableSlot *slot)
    1021             : {
    1022    95726642 :     slot->tts_tableOid = RelationGetRelid(sscan->rs_rd);
    1023             : 
    1024             :     /* We don't expect actual scans using NoMovementScanDirection */
    1025             :     Assert(direction == ForwardScanDirection ||
    1026             :            direction == BackwardScanDirection);
    1027             : 
    1028             :     /*
    1029             :      * We don't expect direct calls to table_scan_getnextslot with valid
    1030             :      * CheckXidAlive for catalog or regular tables.  See detailed comments in
    1031             :      * xact.c where these variables are declared.
    1032             :      */
    1033    95726642 :     if (unlikely(TransactionIdIsValid(CheckXidAlive) && !bsysscan))
    1034           0 :         elog(ERROR, "unexpected table_scan_getnextslot call during logical decoding");
    1035             : 
    1036    95726642 :     return sscan->rs_rd->rd_tableam->scan_getnextslot(sscan, direction, slot);
    1037             : }
    1038             : 
    1039             : /* ----------------------------------------------------------------------------
    1040             :  * TID Range scanning related functions.
    1041             :  * ----------------------------------------------------------------------------
    1042             :  */
    1043             : 
    1044             : /*
    1045             :  * table_beginscan_tidrange is the entry point for setting up a TableScanDesc
    1046             :  * for a TID range scan.
    1047             :  */
    1048             : static inline TableScanDesc
    1049        1854 : table_beginscan_tidrange(Relation rel, Snapshot snapshot,
    1050             :                          ItemPointer mintid,
    1051             :                          ItemPointer maxtid)
    1052             : {
    1053             :     TableScanDesc sscan;
    1054        1854 :     uint32      flags = SO_TYPE_TIDRANGESCAN | SO_ALLOW_PAGEMODE;
    1055             : 
    1056        1854 :     sscan = rel->rd_tableam->scan_begin(rel, snapshot, 0, NULL, NULL, flags);
    1057             : 
    1058             :     /* Set the range of TIDs to scan */
    1059        1854 :     sscan->rs_rd->rd_tableam->scan_set_tidrange(sscan, mintid, maxtid);
    1060             : 
    1061        1854 :     return sscan;
    1062             : }
    1063             : 
    1064             : /*
    1065             :  * table_rescan_tidrange resets the scan position and sets the minimum and
    1066             :  * maximum TID range to scan for a TableScanDesc created by
    1067             :  * table_beginscan_tidrange.
    1068             :  */
    1069             : static inline void
    1070          66 : table_rescan_tidrange(TableScanDesc sscan, ItemPointer mintid,
    1071             :                       ItemPointer maxtid)
    1072             : {
    1073             :     /* Ensure table_beginscan_tidrange() was used. */
    1074             :     Assert((sscan->rs_flags & SO_TYPE_TIDRANGESCAN) != 0);
    1075             : 
    1076          66 :     sscan->rs_rd->rd_tableam->scan_rescan(sscan, NULL, false, false, false, false);
    1077          66 :     sscan->rs_rd->rd_tableam->scan_set_tidrange(sscan, mintid, maxtid);
    1078          66 : }
    1079             : 
    1080             : /*
    1081             :  * Fetch the next tuple from `sscan` for a TID range scan created by
    1082             :  * table_beginscan_tidrange().  Stores the tuple in `slot` and returns true,
    1083             :  * or returns false if no more tuples exist in the range.
    1084             :  */
    1085             : static inline bool
    1086        8602 : table_scan_getnextslot_tidrange(TableScanDesc sscan, ScanDirection direction,
    1087             :                                 TupleTableSlot *slot)
    1088             : {
    1089             :     /* Ensure table_beginscan_tidrange() was used. */
    1090             :     Assert((sscan->rs_flags & SO_TYPE_TIDRANGESCAN) != 0);
    1091             : 
    1092             :     /* We don't expect actual scans using NoMovementScanDirection */
    1093             :     Assert(direction == ForwardScanDirection ||
    1094             :            direction == BackwardScanDirection);
    1095             : 
    1096        8602 :     return sscan->rs_rd->rd_tableam->scan_getnextslot_tidrange(sscan,
    1097             :                                                                direction,
    1098             :                                                                slot);
    1099             : }
    1100             : 
    1101             : 
    1102             : /* ----------------------------------------------------------------------------
    1103             :  * Parallel table scan related functions.
    1104             :  * ----------------------------------------------------------------------------
    1105             :  */
    1106             : 
    1107             : /*
    1108             :  * Estimate the size of shared memory needed for a parallel scan of this
    1109             :  * relation.
    1110             :  */
    1111             : extern Size table_parallelscan_estimate(Relation rel, Snapshot snapshot);
    1112             : 
    1113             : /*
    1114             :  * Initialize ParallelTableScanDesc for a parallel scan of this
    1115             :  * relation. `pscan` needs to be sized according to parallelscan_estimate()
    1116             :  * for the same relation.  Call this just once in the leader process; then,
    1117             :  * individual workers attach via table_beginscan_parallel.
    1118             :  */
    1119             : extern void table_parallelscan_initialize(Relation rel,
    1120             :                                           ParallelTableScanDesc pscan,
    1121             :                                           Snapshot snapshot);
    1122             : 
    1123             : /*
    1124             :  * Begin a parallel scan. `pscan` needs to have been initialized with
    1125             :  * table_parallelscan_initialize(), for the same relation. The initialization
    1126             :  * does not need to have happened in this backend.
    1127             :  *
    1128             :  * Caller must hold a suitable lock on the relation.
    1129             :  */
    1130             : extern TableScanDesc table_beginscan_parallel(Relation relation,
    1131             :                                               ParallelTableScanDesc pscan);
    1132             : 
    1133             : /*
    1134             :  * Restart a parallel scan.  Call this in the leader process.  Caller is
    1135             :  * responsible for making sure that all workers have finished the scan
    1136             :  * beforehand.
    1137             :  */
    1138             : static inline void
    1139         228 : table_parallelscan_reinitialize(Relation rel, ParallelTableScanDesc pscan)
    1140             : {
    1141         228 :     rel->rd_tableam->parallelscan_reinitialize(rel, pscan);
    1142         228 : }
    1143             : 
    1144             : 
    1145             : /* ----------------------------------------------------------------------------
    1146             :  *  Index scan related functions.
    1147             :  * ----------------------------------------------------------------------------
    1148             :  */
    1149             : 
    1150             : /*
    1151             :  * Prepare to fetch tuples from the relation, as needed when fetching tuples
    1152             :  * for an index scan.
    1153             :  *
    1154             :  * Tuples for an index scan can then be fetched via table_index_fetch_tuple().
    1155             :  */
    1156             : static inline IndexFetchTableData *
    1157    27033836 : table_index_fetch_begin(Relation rel)
    1158             : {
    1159    27033836 :     return rel->rd_tableam->index_fetch_begin(rel);
    1160             : }
    1161             : 
    1162             : /*
    1163             :  * Reset index fetch. Typically this will release cross index fetch resources
    1164             :  * held in IndexFetchTableData.
    1165             :  */
    1166             : static inline void
    1167    23321410 : table_index_fetch_reset(struct IndexFetchTableData *scan)
    1168             : {
    1169    23321410 :     scan->rel->rd_tableam->index_fetch_reset(scan);
    1170    23321410 : }
    1171             : 
    1172             : /*
    1173             :  * Release resources and deallocate index fetch.
    1174             :  */
    1175             : static inline void
    1176    27032042 : table_index_fetch_end(struct IndexFetchTableData *scan)
    1177             : {
    1178    27032042 :     scan->rel->rd_tableam->index_fetch_end(scan);
    1179    27032042 : }
    1180             : 
    1181             : /*
    1182             :  * Fetches, as part of an index scan, tuple at `tid` into `slot`, after doing
    1183             :  * a visibility test according to `snapshot`. If a tuple was found and passed
    1184             :  * the visibility test, returns true, false otherwise. Note that *tid may be
    1185             :  * modified when we return true (see later remarks on multiple row versions
    1186             :  * reachable via a single index entry).
    1187             :  *
    1188             :  * *call_again needs to be false on the first call to table_index_fetch_tuple() for
    1189             :  * a tid. If there potentially is another tuple matching the tid, *call_again
    1190             :  * will be set to true, signaling that table_index_fetch_tuple() should be called
    1191             :  * again for the same tid.
    1192             :  *
    1193             :  * *all_dead, if all_dead is not NULL, will be set to true by
    1194             :  * table_index_fetch_tuple() iff it is guaranteed that no backend needs to see
    1195             :  * that tuple. Index AMs can use that to avoid returning that tid in future
    1196             :  * searches.
    1197             :  *
    1198             :  * The difference between this function and table_tuple_fetch_row_version()
    1199             :  * is that this function returns the currently visible version of a row if
    1200             :  * the AM supports storing multiple row versions reachable via a single index
    1201             :  * entry (like heap's HOT). Whereas table_tuple_fetch_row_version() only
    1202             :  * evaluates the tuple exactly at `tid`. Outside of index entry ->table tuple
    1203             :  * lookups, table_tuple_fetch_row_version() is what's usually needed.
    1204             :  */
    1205             : static inline bool
    1206    38510984 : table_index_fetch_tuple(struct IndexFetchTableData *scan,
    1207             :                         ItemPointer tid,
    1208             :                         Snapshot snapshot,
    1209             :                         TupleTableSlot *slot,
    1210             :                         bool *call_again, bool *all_dead)
    1211             : {
    1212             :     /*
    1213             :      * We don't expect direct calls to table_index_fetch_tuple with valid
    1214             :      * CheckXidAlive for catalog or regular tables.  See detailed comments in
    1215             :      * xact.c where these variables are declared.
    1216             :      */
    1217    38510984 :     if (unlikely(TransactionIdIsValid(CheckXidAlive) && !bsysscan))
    1218           0 :         elog(ERROR, "unexpected table_index_fetch_tuple call during logical decoding");
    1219             : 
    1220    38510984 :     return scan->rel->rd_tableam->index_fetch_tuple(scan, tid, snapshot,
    1221             :                                                     slot, call_again,
    1222             :                                                     all_dead);
    1223             : }
    1224             : 
    1225             : /*
    1226             :  * This is a convenience wrapper around table_index_fetch_tuple() which
    1227             :  * returns whether there are table tuple items corresponding to an index
    1228             :  * entry.  This likely is only useful to verify if there's a conflict in a
    1229             :  * unique index.
    1230             :  */
    1231             : extern bool table_index_fetch_tuple_check(Relation rel,
    1232             :                                           ItemPointer tid,
    1233             :                                           Snapshot snapshot,
    1234             :                                           bool *all_dead);
    1235             : 
    1236             : 
    1237             : /* ------------------------------------------------------------------------
    1238             :  * Functions for non-modifying operations on individual tuples
    1239             :  * ------------------------------------------------------------------------
    1240             :  */
    1241             : 
    1242             : 
    1243             : /*
    1244             :  * Fetch tuple at `tid` into `slot`, after doing a visibility test according to
    1245             :  * `snapshot`. If a tuple was found and passed the visibility test, returns
    1246             :  * true, false otherwise.
    1247             :  *
    1248             :  * See table_index_fetch_tuple's comment about what the difference between
    1249             :  * these functions is. It is correct to use this function outside of index
    1250             :  * entry->table tuple lookups.
    1251             :  */
    1252             : static inline bool
    1253      355330 : table_tuple_fetch_row_version(Relation rel,
    1254             :                               ItemPointer tid,
    1255             :                               Snapshot snapshot,
    1256             :                               TupleTableSlot *slot)
    1257             : {
    1258             :     /*
    1259             :      * We don't expect direct calls to table_tuple_fetch_row_version with
    1260             :      * valid CheckXidAlive for catalog or regular tables.  See detailed
    1261             :      * comments in xact.c where these variables are declared.
    1262             :      */
    1263      355330 :     if (unlikely(TransactionIdIsValid(CheckXidAlive) && !bsysscan))
    1264           0 :         elog(ERROR, "unexpected table_tuple_fetch_row_version call during logical decoding");
    1265             : 
    1266      355330 :     return rel->rd_tableam->tuple_fetch_row_version(rel, tid, snapshot, slot);
    1267             : }
    1268             : 
    1269             : /*
    1270             :  * Verify that `tid` is a potentially valid tuple identifier. That doesn't
    1271             :  * mean that the pointed to row needs to exist or be visible, but that
    1272             :  * attempting to fetch the row (e.g. with table_tuple_get_latest_tid() or
    1273             :  * table_tuple_fetch_row_version()) should not error out if called with that
    1274             :  * tid.
    1275             :  *
    1276             :  * `scan` needs to have been started via table_beginscan().
    1277             :  */
    1278             : static inline bool
    1279         454 : table_tuple_tid_valid(TableScanDesc scan, ItemPointer tid)
    1280             : {
    1281         454 :     return scan->rs_rd->rd_tableam->tuple_tid_valid(scan, tid);
    1282             : }
    1283             : 
    1284             : /*
    1285             :  * Return the latest version of the tuple at `tid`, by updating `tid` to
    1286             :  * point at the newest version.
    1287             :  */
    1288             : extern void table_tuple_get_latest_tid(TableScanDesc scan, ItemPointer tid);
    1289             : 
    1290             : /*
    1291             :  * Return true iff tuple in slot satisfies the snapshot.
    1292             :  *
    1293             :  * This assumes the slot's tuple is valid, and of the appropriate type for the
    1294             :  * AM.
    1295             :  *
    1296             :  * Some AMs might modify the data underlying the tuple as a side-effect. If so
    1297             :  * they ought to mark the relevant buffer dirty.
    1298             :  */
    1299             : static inline bool
    1300      239654 : table_tuple_satisfies_snapshot(Relation rel, TupleTableSlot *slot,
    1301             :                                Snapshot snapshot)
    1302             : {
    1303      239654 :     return rel->rd_tableam->tuple_satisfies_snapshot(rel, slot, snapshot);
    1304             : }
    1305             : 
    1306             : /*
    1307             :  * Determine which index tuples are safe to delete based on their table TID.
    1308             :  *
    1309             :  * Determines which entries from index AM caller's TM_IndexDeleteOp state
    1310             :  * point to vacuumable table tuples.  Entries that are found by tableam to be
    1311             :  * vacuumable are naturally safe for index AM to delete, and so get directly
    1312             :  * marked as deletable.  See comments above TM_IndexDelete and comments above
    1313             :  * TM_IndexDeleteOp for full details.
    1314             :  *
    1315             :  * Returns a snapshotConflictHorizon transaction ID that caller places in
    1316             :  * its index deletion WAL record.  This might be used during subsequent REDO
    1317             :  * of the WAL record when in Hot Standby mode -- a recovery conflict for the
    1318             :  * index deletion operation might be required on the standby.
    1319             :  */
    1320             : static inline TransactionId
    1321       11540 : table_index_delete_tuples(Relation rel, TM_IndexDeleteOp *delstate)
    1322             : {
    1323       11540 :     return rel->rd_tableam->index_delete_tuples(rel, delstate);
    1324             : }
    1325             : 
    1326             : 
    1327             : /* ----------------------------------------------------------------------------
    1328             :  *  Functions for manipulations of physical tuples.
    1329             :  * ----------------------------------------------------------------------------
    1330             :  */
    1331             : 
    1332             : /*
    1333             :  * Insert a tuple from a slot into table AM routine.
    1334             :  *
    1335             :  * The options bitmask allows the caller to specify options that may change the
    1336             :  * behaviour of the AM. The AM will ignore options that it does not support.
    1337             :  *
    1338             :  * If the TABLE_INSERT_SKIP_FSM option is specified, AMs are free to not reuse
    1339             :  * free space in the relation. This can save some cycles when we know the
    1340             :  * relation is new and doesn't contain useful amounts of free space.
    1341             :  * TABLE_INSERT_SKIP_FSM is commonly passed directly to
    1342             :  * RelationGetBufferForTuple. See that method for more information.
    1343             :  *
    1344             :  * TABLE_INSERT_FROZEN should only be specified for inserts into
    1345             :  * relation storage created during the current subtransaction and when
    1346             :  * there are no prior snapshots or pre-existing portals open.
    1347             :  * This causes rows to be frozen, which is an MVCC violation and
    1348             :  * requires explicit options chosen by user.
    1349             :  *
    1350             :  * TABLE_INSERT_NO_LOGICAL force-disables the emitting of logical decoding
    1351             :  * information for the tuple. This should solely be used during table rewrites
    1352             :  * where RelationIsLogicallyLogged(relation) is not yet accurate for the new
    1353             :  * relation.
    1354             :  *
    1355             :  * Note that most of these options will be applied when inserting into the
    1356             :  * heap's TOAST table, too, if the tuple requires any out-of-line data.
    1357             :  *
    1358             :  * The BulkInsertState object (if any; bistate can be NULL for default
    1359             :  * behavior) is also just passed through to RelationGetBufferForTuple. If
    1360             :  * `bistate` is provided, table_finish_bulk_insert() needs to be called.
    1361             :  *
    1362             :  * On return the slot's tts_tid and tts_tableOid are updated to reflect the
    1363             :  * insertion. But note that any toasting of fields within the slot is NOT
    1364             :  * reflected in the slots contents.
    1365             :  */
    1366             : static inline void
    1367    14814170 : table_tuple_insert(Relation rel, TupleTableSlot *slot, CommandId cid,
    1368             :                    int options, BulkInsertStateData *bistate)
    1369             : {
    1370    14814170 :     rel->rd_tableam->tuple_insert(rel, slot, cid, options,
    1371             :                                   bistate);
    1372    14814138 : }
    1373             : 
    1374             : /*
    1375             :  * Perform a "speculative insertion". These can be backed out afterwards
    1376             :  * without aborting the whole transaction.  Other sessions can wait for the
    1377             :  * speculative insertion to be confirmed, turning it into a regular tuple, or
    1378             :  * aborted, as if it never existed.  Speculatively inserted tuples behave as
    1379             :  * "value locks" of short duration, used to implement INSERT .. ON CONFLICT.
    1380             :  *
    1381             :  * A transaction having performed a speculative insertion has to either abort,
    1382             :  * or finish the speculative insertion with
    1383             :  * table_tuple_complete_speculative(succeeded = ...).
    1384             :  */
    1385             : static inline void
    1386        4140 : table_tuple_insert_speculative(Relation rel, TupleTableSlot *slot,
    1387             :                                CommandId cid, int options,
    1388             :                                BulkInsertStateData *bistate,
    1389             :                                uint32 specToken)
    1390             : {
    1391        4140 :     rel->rd_tableam->tuple_insert_speculative(rel, slot, cid, options,
    1392             :                                               bistate, specToken);
    1393        4140 : }
    1394             : 
    1395             : /*
    1396             :  * Complete "speculative insertion" started in the same transaction. If
    1397             :  * succeeded is true, the tuple is fully inserted, if false, it's removed.
    1398             :  */
    1399             : static inline void
    1400        4134 : table_tuple_complete_speculative(Relation rel, TupleTableSlot *slot,
    1401             :                                  uint32 specToken, bool succeeded)
    1402             : {
    1403        4134 :     rel->rd_tableam->tuple_complete_speculative(rel, slot, specToken,
    1404             :                                                 succeeded);
    1405        4134 : }
    1406             : 
    1407             : /*
    1408             :  * Insert multiple tuples into a table.
    1409             :  *
    1410             :  * This is like table_tuple_insert(), but inserts multiple tuples in one
    1411             :  * operation. That's often faster than calling table_tuple_insert() in a loop,
    1412             :  * because e.g. the AM can reduce WAL logging and page locking overhead.
    1413             :  *
    1414             :  * Except for taking `nslots` tuples as input, and an array of TupleTableSlots
    1415             :  * in `slots`, the parameters for table_multi_insert() are the same as for
    1416             :  * table_tuple_insert().
    1417             :  *
    1418             :  * Note: this leaks memory into the current memory context. You can create a
    1419             :  * temporary context before calling this, if that's a problem.
    1420             :  */
    1421             : static inline void
    1422        2444 : table_multi_insert(Relation rel, TupleTableSlot **slots, int nslots,
    1423             :                    CommandId cid, int options, BulkInsertStateData *bistate)
    1424             : {
    1425        2444 :     rel->rd_tableam->multi_insert(rel, slots, nslots,
    1426             :                                   cid, options, bistate);
    1427        2444 : }
    1428             : 
    1429             : /*
    1430             :  * Delete a tuple.
    1431             :  *
    1432             :  * NB: do not call this directly unless prepared to deal with
    1433             :  * concurrent-update conditions.  Use simple_table_tuple_delete instead.
    1434             :  *
    1435             :  * Input parameters:
    1436             :  *  rel - table to be modified (caller must hold suitable lock)
    1437             :  *  tid - TID of tuple to be deleted
    1438             :  *  cid - delete command ID (used for visibility test, and stored into
    1439             :  *      cmax if successful)
    1440             :  *  crosscheck - if not InvalidSnapshot, also check tuple against this
    1441             :  *  wait - true if should wait for any conflicting update to commit/abort
    1442             :  *  changingPart - true iff the tuple is being moved to another partition
    1443             :  *      table due to an update of the partition key. Otherwise, false.
    1444             :  *
    1445             :  * Output parameters:
    1446             :  *  tmfd - filled in failure cases (see below)
    1447             :  *
    1448             :  * Normal, successful return value is TM_Ok, which means we did actually
    1449             :  * delete it.  Failure return codes are TM_SelfModified, TM_Updated, and
    1450             :  * TM_BeingModified (the last only possible if wait == false).
    1451             :  *
    1452             :  * In the failure cases, the routine fills *tmfd with the tuple's t_ctid,
    1453             :  * t_xmax, and, if possible, t_cmax.  See comments for struct
    1454             :  * TM_FailureData for additional info.
    1455             :  */
    1456             : static inline TM_Result
    1457     1731896 : table_tuple_delete(Relation rel, ItemPointer tid, CommandId cid,
    1458             :                    Snapshot snapshot, Snapshot crosscheck, bool wait,
    1459             :                    TM_FailureData *tmfd, bool changingPart)
    1460             : {
    1461     1731896 :     return rel->rd_tableam->tuple_delete(rel, tid, cid,
    1462             :                                          snapshot, crosscheck,
    1463             :                                          wait, tmfd, changingPart);
    1464             : }
    1465             : 
    1466             : /*
    1467             :  * Update a tuple.
    1468             :  *
    1469             :  * NB: do not call this directly unless you are prepared to deal with
    1470             :  * concurrent-update conditions.  Use simple_table_tuple_update instead.
    1471             :  *
    1472             :  * Input parameters:
    1473             :  *  rel - table to be modified (caller must hold suitable lock)
    1474             :  *  otid - TID of old tuple to be replaced
    1475             :  *  cid - update command ID (used for visibility test, and stored into
    1476             :  *      cmax/cmin if successful)
    1477             :  *  crosscheck - if not InvalidSnapshot, also check old tuple against this
    1478             :  *  wait - true if should wait for any conflicting update to commit/abort
    1479             :  *
    1480             :  * Output parameters:
    1481             :  *  slot - newly constructed tuple data to store
    1482             :  *  tmfd - filled in failure cases (see below)
    1483             :  *  lockmode - filled with lock mode acquired on tuple
    1484             :  *  update_indexes - in success cases this is set to true if new index entries
    1485             :  *      are required for this tuple
    1486             :  *
    1487             :  * Normal, successful return value is TM_Ok, which means we did actually
    1488             :  * update it.  Failure return codes are TM_SelfModified, TM_Updated, and
    1489             :  * TM_BeingModified (the last only possible if wait == false).
    1490             :  *
    1491             :  * On success, the slot's tts_tid and tts_tableOid are updated to match the new
    1492             :  * stored tuple; in particular, slot->tts_tid is set to the TID where the
    1493             :  * new tuple was inserted, and its HEAP_ONLY_TUPLE flag is set iff a HOT
    1494             :  * update was done.  However, any TOAST changes in the new tuple's
    1495             :  * data are not reflected into *newtup.
    1496             :  *
    1497             :  * In the failure cases, the routine fills *tmfd with the tuple's t_ctid,
    1498             :  * t_xmax, and, if possible, t_cmax.  See comments for struct TM_FailureData
    1499             :  * for additional info.
    1500             :  */
    1501             : static inline TM_Result
    1502      387280 : table_tuple_update(Relation rel, ItemPointer otid, TupleTableSlot *slot,
    1503             :                    CommandId cid, Snapshot snapshot, Snapshot crosscheck,
    1504             :                    bool wait, TM_FailureData *tmfd, LockTupleMode *lockmode,
    1505             :                    TU_UpdateIndexes *update_indexes)
    1506             : {
    1507      387280 :     return rel->rd_tableam->tuple_update(rel, otid, slot,
    1508             :                                          cid, snapshot, crosscheck,
    1509             :                                          wait, tmfd,
    1510             :                                          lockmode, update_indexes);
    1511             : }
    1512             : 
    1513             : /*
    1514             :  * Lock a tuple in the specified mode.
    1515             :  *
    1516             :  * Input parameters:
    1517             :  *  rel: relation containing tuple (caller must hold suitable lock)
    1518             :  *  tid: TID of tuple to lock (updated if an update chain was followed)
    1519             :  *  snapshot: snapshot to use for visibility determinations
    1520             :  *  cid: current command ID (used for visibility test, and stored into
    1521             :  *      tuple's cmax if lock is successful)
    1522             :  *  mode: lock mode desired
    1523             :  *  wait_policy: what to do if tuple lock is not available
    1524             :  *  flags:
    1525             :  *      If TUPLE_LOCK_FLAG_LOCK_UPDATE_IN_PROGRESS, follow the update chain to
    1526             :  *      also lock descendant tuples if lock modes don't conflict.
    1527             :  *      If TUPLE_LOCK_FLAG_FIND_LAST_VERSION, follow the update chain and lock
    1528             :  *      latest version.
    1529             :  *
    1530             :  * Output parameters:
    1531             :  *  *slot: contains the target tuple
    1532             :  *  *tmfd: filled in failure cases (see below)
    1533             :  *
    1534             :  * Function result may be:
    1535             :  *  TM_Ok: lock was successfully acquired
    1536             :  *  TM_Invisible: lock failed because tuple was never visible to us
    1537             :  *  TM_SelfModified: lock failed because tuple updated by self
    1538             :  *  TM_Updated: lock failed because tuple updated by other xact
    1539             :  *  TM_Deleted: lock failed because tuple deleted by other xact
    1540             :  *  TM_WouldBlock: lock couldn't be acquired and wait_policy is skip
    1541             :  *
    1542             :  * In the failure cases other than TM_Invisible and TM_Deleted, the routine
    1543             :  * fills *tmfd with the tuple's t_ctid, t_xmax, and, if possible, t_cmax.
    1544             :  * Additionally, in both success and failure cases, tmfd->traversed is set if
    1545             :  * an update chain was followed.  See comments for struct TM_FailureData for
    1546             :  * additional info.
    1547             :  */
    1548             : static inline TM_Result
    1549      169636 : table_tuple_lock(Relation rel, ItemPointer tid, Snapshot snapshot,
    1550             :                  TupleTableSlot *slot, CommandId cid, LockTupleMode mode,
    1551             :                  LockWaitPolicy wait_policy, uint8 flags,
    1552             :                  TM_FailureData *tmfd)
    1553             : {
    1554      169636 :     return rel->rd_tableam->tuple_lock(rel, tid, snapshot, slot,
    1555             :                                        cid, mode, wait_policy,
    1556             :                                        flags, tmfd);
    1557             : }
    1558             : 
    1559             : /*
    1560             :  * Perform operations necessary to complete insertions made via
    1561             :  * tuple_insert and multi_insert with a BulkInsertState specified.
    1562             :  */
    1563             : static inline void
    1564        4178 : table_finish_bulk_insert(Relation rel, int options)
    1565             : {
    1566             :     /* optional callback */
    1567        4178 :     if (rel->rd_tableam && rel->rd_tableam->finish_bulk_insert)
    1568           0 :         rel->rd_tableam->finish_bulk_insert(rel, options);
    1569        4178 : }
    1570             : 
    1571             : 
    1572             : /* ------------------------------------------------------------------------
    1573             :  * DDL related functionality.
    1574             :  * ------------------------------------------------------------------------
    1575             :  */
    1576             : 
    1577             : /*
    1578             :  * Create storage for `rel` in `newrlocator`, with persistence set to
    1579             :  * `persistence`.
    1580             :  *
    1581             :  * This is used both during relation creation and various DDL operations to
    1582             :  * create new rel storage that can be filled from scratch.  When creating
    1583             :  * new storage for an existing relfilelocator, this should be called before the
    1584             :  * relcache entry has been updated.
    1585             :  *
    1586             :  * *freezeXid, *minmulti are set to the xid / multixact horizon for the table
    1587             :  * that pg_class.{relfrozenxid, relminmxid} have to be set to.
    1588             :  */
    1589             : static inline void
    1590       64818 : table_relation_set_new_filelocator(Relation rel,
    1591             :                                    const RelFileLocator *newrlocator,
    1592             :                                    char persistence,
    1593             :                                    TransactionId *freezeXid,
    1594             :                                    MultiXactId *minmulti)
    1595             : {
    1596       64818 :     rel->rd_tableam->relation_set_new_filelocator(rel, newrlocator,
    1597             :                                                   persistence, freezeXid,
    1598             :                                                   minmulti);
    1599       64818 : }
    1600             : 
    1601             : /*
    1602             :  * Remove all table contents from `rel`, in a non-transactional manner.
    1603             :  * Non-transactional meaning that there's no need to support rollbacks. This
    1604             :  * commonly only is used to perform truncations for relation storage created in
    1605             :  * the current transaction.
    1606             :  */
    1607             : static inline void
    1608         624 : table_relation_nontransactional_truncate(Relation rel)
    1609             : {
    1610         624 :     rel->rd_tableam->relation_nontransactional_truncate(rel);
    1611         624 : }
    1612             : 
    1613             : /*
    1614             :  * Copy data from `rel` into the new relfilelocator `newrlocator`. The new
    1615             :  * relfilelocator may not have storage associated before this function is
    1616             :  * called. This is only supposed to be used for low level operations like
    1617             :  * changing a relation's tablespace.
    1618             :  */
    1619             : static inline void
    1620          98 : table_relation_copy_data(Relation rel, const RelFileLocator *newrlocator)
    1621             : {
    1622          98 :     rel->rd_tableam->relation_copy_data(rel, newrlocator);
    1623          98 : }
    1624             : 
    1625             : /*
    1626             :  * Copy data from `OldTable` into `NewTable`, as part of a CLUSTER or VACUUM
    1627             :  * FULL.
    1628             :  *
    1629             :  * Additional Input parameters:
    1630             :  * - use_sort - if true, the table contents are sorted appropriate for
    1631             :  *   `OldIndex`; if false and OldIndex is not InvalidOid, the data is copied
    1632             :  *   in that index's order; if false and OldIndex is InvalidOid, no sorting is
    1633             :  *   performed
    1634             :  * - OldIndex - see use_sort
    1635             :  * - OldestXmin - computed by vacuum_get_cutoffs(), even when
    1636             :  *   not needed for the relation's AM
    1637             :  * - *xid_cutoff - ditto
    1638             :  * - *multi_cutoff - ditto
    1639             :  *
    1640             :  * Output parameters:
    1641             :  * - *xid_cutoff - rel's new relfrozenxid value, may be invalid
    1642             :  * - *multi_cutoff - rel's new relminmxid value, may be invalid
    1643             :  * - *tups_vacuumed - stats, for logging, if appropriate for AM
    1644             :  * - *tups_recently_dead - stats, for logging, if appropriate for AM
    1645             :  */
    1646             : static inline void
    1647         570 : table_relation_copy_for_cluster(Relation OldTable, Relation NewTable,
    1648             :                                 Relation OldIndex,
    1649             :                                 bool use_sort,
    1650             :                                 TransactionId OldestXmin,
    1651             :                                 TransactionId *xid_cutoff,
    1652             :                                 MultiXactId *multi_cutoff,
    1653             :                                 double *num_tuples,
    1654             :                                 double *tups_vacuumed,
    1655             :                                 double *tups_recently_dead)
    1656             : {
    1657         570 :     OldTable->rd_tableam->relation_copy_for_cluster(OldTable, NewTable, OldIndex,
    1658             :                                                     use_sort, OldestXmin,
    1659             :                                                     xid_cutoff, multi_cutoff,
    1660             :                                                     num_tuples, tups_vacuumed,
    1661             :                                                     tups_recently_dead);
    1662         570 : }
    1663             : 
    1664             : /*
    1665             :  * Perform VACUUM on the relation. The VACUUM can be triggered by a user or by
    1666             :  * autovacuum. The specific actions performed by the AM will depend heavily on
    1667             :  * the individual AM.
    1668             :  *
    1669             :  * On entry a transaction needs to already been established, and the
    1670             :  * table is locked with a ShareUpdateExclusive lock.
    1671             :  *
    1672             :  * Note that neither VACUUM FULL (and CLUSTER), nor ANALYZE go through this
    1673             :  * routine, even if (for ANALYZE) it is part of the same VACUUM command.
    1674             :  */
    1675             : static inline void
    1676      242196 : table_relation_vacuum(Relation rel, const VacuumParams params,
    1677             :                       BufferAccessStrategy bstrategy)
    1678             : {
    1679      242196 :     rel->rd_tableam->relation_vacuum(rel, params, bstrategy);
    1680      242194 : }
    1681             : 
    1682             : /*
    1683             :  * Prepare to analyze the next block in the read stream. The scan needs to
    1684             :  * have been  started with table_beginscan_analyze().  Note that this routine
    1685             :  * might acquire resources like locks that are held until
    1686             :  * table_scan_analyze_next_tuple() returns false.
    1687             :  *
    1688             :  * Returns false if block is unsuitable for sampling, true otherwise.
    1689             :  */
    1690             : static inline bool
    1691      151088 : table_scan_analyze_next_block(TableScanDesc scan, ReadStream *stream)
    1692             : {
    1693      151088 :     return scan->rs_rd->rd_tableam->scan_analyze_next_block(scan, stream);
    1694             : }
    1695             : 
    1696             : /*
    1697             :  * Iterate over tuples in the block selected with
    1698             :  * table_scan_analyze_next_block() (which needs to have returned true, and
    1699             :  * this routine may not have returned false for the same block before). If a
    1700             :  * tuple that's suitable for sampling is found, true is returned and a tuple
    1701             :  * is stored in `slot`.
    1702             :  *
    1703             :  * *liverows and *deadrows are incremented according to the encountered
    1704             :  * tuples.
    1705             :  */
    1706             : static inline bool
    1707    10628128 : table_scan_analyze_next_tuple(TableScanDesc scan, TransactionId OldestXmin,
    1708             :                               double *liverows, double *deadrows,
    1709             :                               TupleTableSlot *slot)
    1710             : {
    1711    10628128 :     return scan->rs_rd->rd_tableam->scan_analyze_next_tuple(scan, OldestXmin,
    1712             :                                                             liverows, deadrows,
    1713             :                                                             slot);
    1714             : }
    1715             : 
    1716             : /*
    1717             :  * table_index_build_scan - scan the table to find tuples to be indexed
    1718             :  *
    1719             :  * This is called back from an access-method-specific index build procedure
    1720             :  * after the AM has done whatever setup it needs.  The parent table relation
    1721             :  * is scanned to find tuples that should be entered into the index.  Each
    1722             :  * such tuple is passed to the AM's callback routine, which does the right
    1723             :  * things to add it to the new index.  After we return, the AM's index
    1724             :  * build procedure does whatever cleanup it needs.
    1725             :  *
    1726             :  * The total count of live tuples is returned.  This is for updating pg_class
    1727             :  * statistics.  (It's annoying not to be able to do that here, but we want to
    1728             :  * merge that update with others; see index_update_stats.)  Note that the
    1729             :  * index AM itself must keep track of the number of index tuples; we don't do
    1730             :  * so here because the AM might reject some of the tuples for its own reasons,
    1731             :  * such as being unable to store NULLs.
    1732             :  *
    1733             :  * If 'progress', the PROGRESS_SCAN_BLOCKS_TOTAL counter is updated when
    1734             :  * starting the scan, and PROGRESS_SCAN_BLOCKS_DONE is updated as we go along.
    1735             :  *
    1736             :  * A side effect is to set indexInfo->ii_BrokenHotChain to true if we detect
    1737             :  * any potentially broken HOT chains.  Currently, we set this if there are any
    1738             :  * RECENTLY_DEAD or DELETE_IN_PROGRESS entries in a HOT chain, without trying
    1739             :  * very hard to detect whether they're really incompatible with the chain tip.
    1740             :  * This only really makes sense for heap AM, it might need to be generalized
    1741             :  * for other AMs later.
    1742             :  */
    1743             : static inline double
    1744       53778 : table_index_build_scan(Relation table_rel,
    1745             :                        Relation index_rel,
    1746             :                        IndexInfo *index_info,
    1747             :                        bool allow_sync,
    1748             :                        bool progress,
    1749             :                        IndexBuildCallback callback,
    1750             :                        void *callback_state,
    1751             :                        TableScanDesc scan)
    1752             : {
    1753       53778 :     return table_rel->rd_tableam->index_build_range_scan(table_rel,
    1754             :                                                          index_rel,
    1755             :                                                          index_info,
    1756             :                                                          allow_sync,
    1757             :                                                          false,
    1758             :                                                          progress,
    1759             :                                                          0,
    1760             :                                                          InvalidBlockNumber,
    1761             :                                                          callback,
    1762             :                                                          callback_state,
    1763             :                                                          scan);
    1764             : }
    1765             : 
    1766             : /*
    1767             :  * As table_index_build_scan(), except that instead of scanning the complete
    1768             :  * table, only the given number of blocks are scanned.  Scan to end-of-rel can
    1769             :  * be signaled by passing InvalidBlockNumber as numblocks.  Note that
    1770             :  * restricting the range to scan cannot be done when requesting syncscan.
    1771             :  *
    1772             :  * When "anyvisible" mode is requested, all tuples visible to any transaction
    1773             :  * are indexed and counted as live, including those inserted or deleted by
    1774             :  * transactions that are still in progress.
    1775             :  */
    1776             : static inline double
    1777        2938 : table_index_build_range_scan(Relation table_rel,
    1778             :                              Relation index_rel,
    1779             :                              IndexInfo *index_info,
    1780             :                              bool allow_sync,
    1781             :                              bool anyvisible,
    1782             :                              bool progress,
    1783             :                              BlockNumber start_blockno,
    1784             :                              BlockNumber numblocks,
    1785             :                              IndexBuildCallback callback,
    1786             :                              void *callback_state,
    1787             :                              TableScanDesc scan)
    1788             : {
    1789        2938 :     return table_rel->rd_tableam->index_build_range_scan(table_rel,
    1790             :                                                          index_rel,
    1791             :                                                          index_info,
    1792             :                                                          allow_sync,
    1793             :                                                          anyvisible,
    1794             :                                                          progress,
    1795             :                                                          start_blockno,
    1796             :                                                          numblocks,
    1797             :                                                          callback,
    1798             :                                                          callback_state,
    1799             :                                                          scan);
    1800             : }
    1801             : 
    1802             : /*
    1803             :  * table_index_validate_scan - second table scan for concurrent index build
    1804             :  *
    1805             :  * See validate_index() for an explanation.
    1806             :  */
    1807             : static inline void
    1808         682 : table_index_validate_scan(Relation table_rel,
    1809             :                           Relation index_rel,
    1810             :                           IndexInfo *index_info,
    1811             :                           Snapshot snapshot,
    1812             :                           ValidateIndexState *state)
    1813             : {
    1814         682 :     table_rel->rd_tableam->index_validate_scan(table_rel,
    1815             :                                                index_rel,
    1816             :                                                index_info,
    1817             :                                                snapshot,
    1818             :                                                state);
    1819         682 : }
    1820             : 
    1821             : 
    1822             : /* ----------------------------------------------------------------------------
    1823             :  * Miscellaneous functionality
    1824             :  * ----------------------------------------------------------------------------
    1825             :  */
    1826             : 
    1827             : /*
    1828             :  * Return the current size of `rel` in bytes. If `forkNumber` is
    1829             :  * InvalidForkNumber, return the relation's overall size, otherwise the size
    1830             :  * for the indicated fork.
    1831             :  *
    1832             :  * Note that the overall size might not be the equivalent of the sum of sizes
    1833             :  * for the individual forks for some AMs, e.g. because the AMs storage does
    1834             :  * not neatly map onto the builtin types of forks.
    1835             :  */
    1836             : static inline uint64
    1837     2844440 : table_relation_size(Relation rel, ForkNumber forkNumber)
    1838             : {
    1839     2844440 :     return rel->rd_tableam->relation_size(rel, forkNumber);
    1840             : }
    1841             : 
    1842             : /*
    1843             :  * table_relation_needs_toast_table - does this relation need a toast table?
    1844             :  */
    1845             : static inline bool
    1846       44600 : table_relation_needs_toast_table(Relation rel)
    1847             : {
    1848       44600 :     return rel->rd_tableam->relation_needs_toast_table(rel);
    1849             : }
    1850             : 
    1851             : /*
    1852             :  * Return the OID of the AM that should be used to implement the TOAST table
    1853             :  * for this relation.
    1854             :  */
    1855             : static inline Oid
    1856       17594 : table_relation_toast_am(Relation rel)
    1857             : {
    1858       17594 :     return rel->rd_tableam->relation_toast_am(rel);
    1859             : }
    1860             : 
    1861             : /*
    1862             :  * Fetch all or part of a TOAST value from a TOAST table.
    1863             :  *
    1864             :  * If this AM is never used to implement a TOAST table, then this callback
    1865             :  * is not needed. But, if toasted values are ever stored in a table of this
    1866             :  * type, then you will need this callback.
    1867             :  *
    1868             :  * toastrel is the relation in which the toasted value is stored.
    1869             :  *
    1870             :  * valueid identifies which toast value is to be fetched. For the heap,
    1871             :  * this corresponds to the values stored in the chunk_id column.
    1872             :  *
    1873             :  * attrsize is the total size of the toast value to be fetched.
    1874             :  *
    1875             :  * sliceoffset is the offset within the toast value of the first byte that
    1876             :  * should be fetched.
    1877             :  *
    1878             :  * slicelength is the number of bytes from the toast value that should be
    1879             :  * fetched.
    1880             :  *
    1881             :  * result is caller-allocated space into which the fetched bytes should be
    1882             :  * stored.
    1883             :  */
    1884             : static inline void
    1885       23370 : table_relation_fetch_toast_slice(Relation toastrel, Oid valueid,
    1886             :                                  int32 attrsize, int32 sliceoffset,
    1887             :                                  int32 slicelength, struct varlena *result)
    1888             : {
    1889       23370 :     toastrel->rd_tableam->relation_fetch_toast_slice(toastrel, valueid,
    1890             :                                                      attrsize,
    1891             :                                                      sliceoffset, slicelength,
    1892             :                                                      result);
    1893       23370 : }
    1894             : 
    1895             : 
    1896             : /* ----------------------------------------------------------------------------
    1897             :  * Planner related functionality
    1898             :  * ----------------------------------------------------------------------------
    1899             :  */
    1900             : 
    1901             : /*
    1902             :  * Estimate the current size of the relation, as an AM specific workhorse for
    1903             :  * estimate_rel_size(). Look there for an explanation of the parameters.
    1904             :  */
    1905             : static inline void
    1906      450626 : table_relation_estimate_size(Relation rel, int32 *attr_widths,
    1907             :                              BlockNumber *pages, double *tuples,
    1908             :                              double *allvisfrac)
    1909             : {
    1910      450626 :     rel->rd_tableam->relation_estimate_size(rel, attr_widths, pages, tuples,
    1911             :                                             allvisfrac);
    1912      450626 : }
    1913             : 
    1914             : 
    1915             : /* ----------------------------------------------------------------------------
    1916             :  * Executor related functionality
    1917             :  * ----------------------------------------------------------------------------
    1918             :  */
    1919             : 
    1920             : /*
    1921             :  * Fetch / check / return tuples as part of a bitmap table scan. `scan` needs
    1922             :  * to have been started via table_beginscan_bm(). Fetch the next tuple of a
    1923             :  * bitmap table scan into `slot` and return true if a visible tuple was found,
    1924             :  * false otherwise.
    1925             :  *
    1926             :  * `recheck` is set by the table AM to indicate whether or not the tuple in
    1927             :  * `slot` should be rechecked. Tuples from lossy pages will always need to be
    1928             :  * rechecked, but some non-lossy pages' tuples may also require recheck.
    1929             :  *
    1930             :  * `lossy_pages` is incremented if the block's representation in the bitmap is
    1931             :  * lossy; otherwise, `exact_pages` is incremented.
    1932             :  */
    1933             : static inline bool
    1934     6689572 : table_scan_bitmap_next_tuple(TableScanDesc scan,
    1935             :                              TupleTableSlot *slot,
    1936             :                              bool *recheck,
    1937             :                              uint64 *lossy_pages,
    1938             :                              uint64 *exact_pages)
    1939             : {
    1940             :     /*
    1941             :      * We don't expect direct calls to table_scan_bitmap_next_tuple with valid
    1942             :      * CheckXidAlive for catalog or regular tables.  See detailed comments in
    1943             :      * xact.c where these variables are declared.
    1944             :      */
    1945     6689572 :     if (unlikely(TransactionIdIsValid(CheckXidAlive) && !bsysscan))
    1946           0 :         elog(ERROR, "unexpected table_scan_bitmap_next_tuple call during logical decoding");
    1947             : 
    1948     6689572 :     return scan->rs_rd->rd_tableam->scan_bitmap_next_tuple(scan,
    1949             :                                                            slot,
    1950             :                                                            recheck,
    1951             :                                                            lossy_pages,
    1952             :                                                            exact_pages);
    1953             : }
    1954             : 
    1955             : /*
    1956             :  * Prepare to fetch tuples from the next block in a sample scan. Returns false
    1957             :  * if the sample scan is finished, true otherwise. `scan` needs to have been
    1958             :  * started via table_beginscan_sampling().
    1959             :  *
    1960             :  * This will call the TsmRoutine's NextSampleBlock() callback if necessary
    1961             :  * (i.e. NextSampleBlock is not NULL), or perform a sequential scan over the
    1962             :  * underlying relation.
    1963             :  */
    1964             : static inline bool
    1965       12912 : table_scan_sample_next_block(TableScanDesc scan,
    1966             :                              SampleScanState *scanstate)
    1967             : {
    1968             :     /*
    1969             :      * We don't expect direct calls to table_scan_sample_next_block with valid
    1970             :      * CheckXidAlive for catalog or regular tables.  See detailed comments in
    1971             :      * xact.c where these variables are declared.
    1972             :      */
    1973       12912 :     if (unlikely(TransactionIdIsValid(CheckXidAlive) && !bsysscan))
    1974           0 :         elog(ERROR, "unexpected table_scan_sample_next_block call during logical decoding");
    1975       12912 :     return scan->rs_rd->rd_tableam->scan_sample_next_block(scan, scanstate);
    1976             : }
    1977             : 
    1978             : /*
    1979             :  * Fetch the next sample tuple into `slot` and return true if a visible tuple
    1980             :  * was found, false otherwise. table_scan_sample_next_block() needs to
    1981             :  * previously have selected a block (i.e. returned true), and no previous
    1982             :  * table_scan_sample_next_tuple() for the same block may have returned false.
    1983             :  *
    1984             :  * This will call the TsmRoutine's NextSampleTuple() callback.
    1985             :  */
    1986             : static inline bool
    1987      253896 : table_scan_sample_next_tuple(TableScanDesc scan,
    1988             :                              SampleScanState *scanstate,
    1989             :                              TupleTableSlot *slot)
    1990             : {
    1991             :     /*
    1992             :      * We don't expect direct calls to table_scan_sample_next_tuple with valid
    1993             :      * CheckXidAlive for catalog or regular tables.  See detailed comments in
    1994             :      * xact.c where these variables are declared.
    1995             :      */
    1996      253896 :     if (unlikely(TransactionIdIsValid(CheckXidAlive) && !bsysscan))
    1997           0 :         elog(ERROR, "unexpected table_scan_sample_next_tuple call during logical decoding");
    1998      253896 :     return scan->rs_rd->rd_tableam->scan_sample_next_tuple(scan, scanstate,
    1999             :                                                            slot);
    2000             : }
    2001             : 
    2002             : 
    2003             : /* ----------------------------------------------------------------------------
    2004             :  * Functions to make modifications a bit simpler.
    2005             :  * ----------------------------------------------------------------------------
    2006             :  */
    2007             : 
    2008             : extern void simple_table_tuple_insert(Relation rel, TupleTableSlot *slot);
    2009             : extern void simple_table_tuple_delete(Relation rel, ItemPointer tid,
    2010             :                                       Snapshot snapshot);
    2011             : extern void simple_table_tuple_update(Relation rel, ItemPointer otid,
    2012             :                                       TupleTableSlot *slot, Snapshot snapshot,
    2013             :                                       TU_UpdateIndexes *update_indexes);
    2014             : 
    2015             : 
    2016             : /* ----------------------------------------------------------------------------
    2017             :  * Helper functions to implement parallel scans for block oriented AMs.
    2018             :  * ----------------------------------------------------------------------------
    2019             :  */
    2020             : 
    2021             : extern Size table_block_parallelscan_estimate(Relation rel);
    2022             : extern Size table_block_parallelscan_initialize(Relation rel,
    2023             :                                                 ParallelTableScanDesc pscan);
    2024             : extern void table_block_parallelscan_reinitialize(Relation rel,
    2025             :                                                   ParallelTableScanDesc pscan);
    2026             : extern BlockNumber table_block_parallelscan_nextpage(Relation rel,
    2027             :                                                      ParallelBlockTableScanWorker pbscanwork,
    2028             :                                                      ParallelBlockTableScanDesc pbscan);
    2029             : extern void table_block_parallelscan_startblock_init(Relation rel,
    2030             :                                                      ParallelBlockTableScanWorker pbscanwork,
    2031             :                                                      ParallelBlockTableScanDesc pbscan);
    2032             : 
    2033             : 
    2034             : /* ----------------------------------------------------------------------------
    2035             :  * Helper functions to implement relation sizing for block oriented AMs.
    2036             :  * ----------------------------------------------------------------------------
    2037             :  */
    2038             : 
    2039             : extern uint64 table_block_relation_size(Relation rel, ForkNumber forkNumber);
    2040             : extern void table_block_relation_estimate_size(Relation rel,
    2041             :                                                int32 *attr_widths,
    2042             :                                                BlockNumber *pages,
    2043             :                                                double *tuples,
    2044             :                                                double *allvisfrac,
    2045             :                                                Size overhead_bytes_per_tuple,
    2046             :                                                Size usable_bytes_per_page);
    2047             : 
    2048             : /* ----------------------------------------------------------------------------
    2049             :  * Functions in tableamapi.c
    2050             :  * ----------------------------------------------------------------------------
    2051             :  */
    2052             : 
    2053             : extern const TableAmRoutine *GetTableAmRoutine(Oid amhandler);
    2054             : 
    2055             : /* ----------------------------------------------------------------------------
    2056             :  * Functions in heapam_handler.c
    2057             :  * ----------------------------------------------------------------------------
    2058             :  */
    2059             : 
    2060             : extern const TableAmRoutine *GetHeapamTableAmRoutine(void);
    2061             : 
    2062             : #endif                          /* TABLEAM_H */

Generated by: LCOV version 1.16