LCOV - code coverage report
Current view: top level - src/include/access - tableam.h (source / functions) Hit Total Coverage
Test: PostgreSQL 18devel Lines: 145 152 95.4 %
Date: 2025-04-01 15:15:16 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 "executor/tuptable.h"
      24             : #include "storage/read_stream.h"
      25             : #include "utils/rel.h"
      26             : #include "utils/snapshot.h"
      27             : 
      28             : 
      29             : #define DEFAULT_TABLE_ACCESS_METHOD "heap"
      30             : 
      31             : /* GUCs */
      32             : extern PGDLLIMPORT char *default_table_access_method;
      33             : extern PGDLLIMPORT bool synchronize_seqscans;
      34             : 
      35             : 
      36             : struct BulkInsertStateData;
      37             : struct IndexInfo;
      38             : struct SampleScanState;
      39             : struct VacuumParams;
      40             : struct ValidateIndexState;
      41             : 
      42             : /*
      43             :  * Bitmask values for the flags argument to the scan_begin callback.
      44             :  */
      45             : typedef enum ScanOptions
      46             : {
      47             :     /* one of SO_TYPE_* may be specified */
      48             :     SO_TYPE_SEQSCAN = 1 << 0,
      49             :     SO_TYPE_BITMAPSCAN = 1 << 1,
      50             :     SO_TYPE_SAMPLESCAN = 1 << 2,
      51             :     SO_TYPE_TIDSCAN = 1 << 3,
      52             :     SO_TYPE_TIDRANGESCAN = 1 << 4,
      53             :     SO_TYPE_ANALYZE = 1 << 5,
      54             : 
      55             :     /* several of SO_ALLOW_* may be specified */
      56             :     /* allow or disallow use of access strategy */
      57             :     SO_ALLOW_STRAT = 1 << 6,
      58             :     /* report location to syncscan logic? */
      59             :     SO_ALLOW_SYNC = 1 << 7,
      60             :     /* verify visibility page-at-a-time? */
      61             :     SO_ALLOW_PAGEMODE = 1 << 8,
      62             : 
      63             :     /* unregister snapshot at scan end? */
      64             :     SO_TEMP_SNAPSHOT = 1 << 9,
      65             : 
      66             :     /*
      67             :      * At the discretion of the table AM, bitmap table scans may be able to
      68             :      * skip fetching a block from the table if none of the table data is
      69             :      * needed. If table data may be needed, set SO_NEED_TUPLES.
      70             :      */
      71             :     SO_NEED_TUPLES = 1 << 10,
      72             : }           ScanOptions;
      73             : 
      74             : /*
      75             :  * Result codes for table_{update,delete,lock_tuple}, and for visibility
      76             :  * routines inside table AMs.
      77             :  */
      78             : typedef enum TM_Result
      79             : {
      80             :     /*
      81             :      * Signals that the action succeeded (i.e. update/delete performed, lock
      82             :      * was acquired)
      83             :      */
      84             :     TM_Ok,
      85             : 
      86             :     /* The affected tuple wasn't visible to the relevant snapshot */
      87             :     TM_Invisible,
      88             : 
      89             :     /* The affected tuple was already modified by the calling backend */
      90             :     TM_SelfModified,
      91             : 
      92             :     /*
      93             :      * The affected tuple was updated by another transaction. This includes
      94             :      * the case where tuple was moved to another partition.
      95             :      */
      96             :     TM_Updated,
      97             : 
      98             :     /* The affected tuple was deleted by another transaction */
      99             :     TM_Deleted,
     100             : 
     101             :     /*
     102             :      * The affected tuple is currently being modified by another session. This
     103             :      * will only be returned if table_(update/delete/lock_tuple) are
     104             :      * instructed not to wait.
     105             :      */
     106             :     TM_BeingModified,
     107             : 
     108             :     /* lock couldn't be acquired, action skipped. Only used by lock_tuple */
     109             :     TM_WouldBlock,
     110             : } TM_Result;
     111             : 
     112             : /*
     113             :  * Result codes for table_update(..., update_indexes*..).
     114             :  * Used to determine which indexes to update.
     115             :  */
     116             : typedef enum TU_UpdateIndexes
     117             : {
     118             :     /* No indexed columns were updated (incl. TID addressing of tuple) */
     119             :     TU_None,
     120             : 
     121             :     /* A non-summarizing indexed column was updated, or the TID has changed */
     122             :     TU_All,
     123             : 
     124             :     /* Only summarized columns were updated, TID is unchanged */
     125             :     TU_Summarizing,
     126             : } TU_UpdateIndexes;
     127             : 
     128             : /*
     129             :  * When table_tuple_update, table_tuple_delete, or table_tuple_lock fail
     130             :  * because the target tuple is already outdated, they fill in this struct to
     131             :  * provide information to the caller about what happened.
     132             :  *
     133             :  * ctid is the target's ctid link: it is the same as the target's TID if the
     134             :  * target was deleted, or the location of the replacement tuple if the target
     135             :  * was updated.
     136             :  *
     137             :  * xmax is the outdating transaction's XID.  If the caller wants to visit the
     138             :  * replacement tuple, it must check that this matches before believing the
     139             :  * replacement is really a match.  This is InvalidTransactionId if the target
     140             :  * was !LP_NORMAL (expected only for a TID retrieved from syscache).
     141             :  *
     142             :  * cmax is the outdating command's CID, but only when the failure code is
     143             :  * TM_SelfModified (i.e., something in the current transaction outdated the
     144             :  * tuple); otherwise cmax is zero.  (We make this restriction because
     145             :  * HeapTupleHeaderGetCmax doesn't work for tuples outdated in other
     146             :  * transactions.)
     147             :  */
     148             : typedef struct TM_FailureData
     149             : {
     150             :     ItemPointerData ctid;
     151             :     TransactionId xmax;
     152             :     CommandId   cmax;
     153             :     bool        traversed;
     154             : } TM_FailureData;
     155             : 
     156             : /*
     157             :  * State used when calling table_index_delete_tuples().
     158             :  *
     159             :  * Represents the status of table tuples, referenced by table TID and taken by
     160             :  * index AM from index tuples.  State consists of high level parameters of the
     161             :  * deletion operation, plus two mutable palloc()'d arrays for information
     162             :  * about the status of individual table tuples.  These are conceptually one
     163             :  * single array.  Using two arrays keeps the TM_IndexDelete struct small,
     164             :  * which makes sorting the first array (the deltids array) fast.
     165             :  *
     166             :  * Some index AM callers perform simple index tuple deletion (by specifying
     167             :  * bottomup = false), and include only known-dead deltids.  These known-dead
     168             :  * entries are all marked knowndeletable = true directly (typically these are
     169             :  * TIDs from LP_DEAD-marked index tuples), but that isn't strictly required.
     170             :  *
     171             :  * Callers that specify bottomup = true are "bottom-up index deletion"
     172             :  * callers.  The considerations for the tableam are more subtle with these
     173             :  * callers because they ask the tableam to perform highly speculative work,
     174             :  * and might only expect the tableam to check a small fraction of all entries.
     175             :  * Caller is not allowed to specify knowndeletable = true for any entry
     176             :  * because everything is highly speculative.  Bottom-up caller provides
     177             :  * context and hints to tableam -- see comments below for details on how index
     178             :  * AMs and tableams should coordinate during bottom-up index deletion.
     179             :  *
     180             :  * Simple index deletion callers may ask the tableam to perform speculative
     181             :  * work, too.  This is a little like bottom-up deletion, but not too much.
     182             :  * The tableam will only perform speculative work when it's practically free
     183             :  * to do so in passing for simple deletion caller (while always performing
     184             :  * whatever work is needed to enable knowndeletable/LP_DEAD index tuples to
     185             :  * be deleted within index AM).  This is the real reason why it's possible for
     186             :  * simple index deletion caller to specify knowndeletable = false up front
     187             :  * (this means "check if it's possible for me to delete corresponding index
     188             :  * tuple when it's cheap to do so in passing").  The index AM should only
     189             :  * include "extra" entries for index tuples whose TIDs point to a table block
     190             :  * that tableam is expected to have to visit anyway (in the event of a block
     191             :  * orientated tableam).  The tableam isn't strictly obligated to check these
     192             :  * "extra" TIDs, but a block-based AM should always manage to do so in
     193             :  * practice.
     194             :  *
     195             :  * The final contents of the deltids/status arrays are interesting to callers
     196             :  * that ask tableam to perform speculative work (i.e. when _any_ items have
     197             :  * knowndeletable set to false up front).  These index AM callers will
     198             :  * naturally need to consult final state to determine which index tuples are
     199             :  * in fact deletable.
     200             :  *
     201             :  * The index AM can keep track of which index tuple relates to which deltid by
     202             :  * setting idxoffnum (and/or relying on each entry being uniquely identifiable
     203             :  * using tid), which is important when the final contents of the array will
     204             :  * need to be interpreted -- the array can shrink from initial size after
     205             :  * tableam processing and/or have entries in a new order (tableam may sort
     206             :  * deltids array for its own reasons).  Bottom-up callers may find that final
     207             :  * ndeltids is 0 on return from call to tableam, in which case no index tuple
     208             :  * deletions are possible.  Simple deletion callers can rely on any entries
     209             :  * they know to be deletable appearing in the final array as deletable.
     210             :  */
     211             : typedef struct TM_IndexDelete
     212             : {
     213             :     ItemPointerData tid;        /* table TID from index tuple */
     214             :     int16       id;             /* Offset into TM_IndexStatus array */
     215             : } TM_IndexDelete;
     216             : 
     217             : typedef struct TM_IndexStatus
     218             : {
     219             :     OffsetNumber idxoffnum;     /* Index am page offset number */
     220             :     bool        knowndeletable; /* Currently known to be deletable? */
     221             : 
     222             :     /* Bottom-up index deletion specific fields follow */
     223             :     bool        promising;      /* Promising (duplicate) index tuple? */
     224             :     int16       freespace;      /* Space freed in index if deleted */
     225             : } TM_IndexStatus;
     226             : 
     227             : /*
     228             :  * Index AM/tableam coordination is central to the design of bottom-up index
     229             :  * deletion.  The index AM provides hints about where to look to the tableam
     230             :  * by marking some entries as "promising".  Index AM does this with duplicate
     231             :  * index tuples that are strongly suspected to be old versions left behind by
     232             :  * UPDATEs that did not logically modify indexed values.  Index AM may find it
     233             :  * helpful to only mark entries as promising when they're thought to have been
     234             :  * affected by such an UPDATE in the recent past.
     235             :  *
     236             :  * Bottom-up index deletion casts a wide net at first, usually by including
     237             :  * all TIDs on a target index page.  It is up to the tableam to worry about
     238             :  * the cost of checking transaction status information.  The tableam is in
     239             :  * control, but needs careful guidance from the index AM.  Index AM requests
     240             :  * that bottomupfreespace target be met, while tableam measures progress
     241             :  * towards that goal by tallying the per-entry freespace value for known
     242             :  * deletable entries. (All !bottomup callers can just set these space related
     243             :  * fields to zero.)
     244             :  */
     245             : typedef struct TM_IndexDeleteOp
     246             : {
     247             :     Relation    irel;           /* Target index relation */
     248             :     BlockNumber iblknum;        /* Index block number (for error reports) */
     249             :     bool        bottomup;       /* Bottom-up (not simple) deletion? */
     250             :     int         bottomupfreespace;  /* Bottom-up space target */
     251             : 
     252             :     /* Mutable per-TID information follows (index AM initializes entries) */
     253             :     int         ndeltids;       /* Current # of deltids/status elements */
     254             :     TM_IndexDelete *deltids;
     255             :     TM_IndexStatus *status;
     256             : } TM_IndexDeleteOp;
     257             : 
     258             : /* "options" flag bits for table_tuple_insert */
     259             : /* TABLE_INSERT_SKIP_WAL was 0x0001; RelationNeedsWAL() now governs */
     260             : #define TABLE_INSERT_SKIP_FSM       0x0002
     261             : #define TABLE_INSERT_FROZEN         0x0004
     262             : #define TABLE_INSERT_NO_LOGICAL     0x0008
     263             : 
     264             : /* flag bits for table_tuple_lock */
     265             : /* Follow tuples whose update is in progress if lock modes don't conflict  */
     266             : #define TUPLE_LOCK_FLAG_LOCK_UPDATE_IN_PROGRESS (1 << 0)
     267             : /* Follow update chain and lock latest version of tuple */
     268             : #define TUPLE_LOCK_FLAG_FIND_LAST_VERSION       (1 << 1)
     269             : 
     270             : 
     271             : /* Typedef for callback function for table_index_build_scan */
     272             : typedef void (*IndexBuildCallback) (Relation index,
     273             :                                     ItemPointer tid,
     274             :                                     Datum *values,
     275             :                                     bool *isnull,
     276             :                                     bool tupleIsAlive,
     277             :                                     void *state);
     278             : 
     279             : /*
     280             :  * API struct for a table AM.  Note this must be allocated in a
     281             :  * server-lifetime manner, typically as a static const struct, which then gets
     282             :  * returned by FormData_pg_am.amhandler.
     283             :  *
     284             :  * In most cases it's not appropriate to call the callbacks directly, use the
     285             :  * table_* wrapper functions instead.
     286             :  *
     287             :  * GetTableAmRoutine() asserts that required callbacks are filled in, remember
     288             :  * to update when adding a callback.
     289             :  */
     290             : typedef struct TableAmRoutine
     291             : {
     292             :     /* this must be set to T_TableAmRoutine */
     293             :     NodeTag     type;
     294             : 
     295             : 
     296             :     /* ------------------------------------------------------------------------
     297             :      * Slot related callbacks.
     298             :      * ------------------------------------------------------------------------
     299             :      */
     300             : 
     301             :     /*
     302             :      * Return slot implementation suitable for storing a tuple of this AM.
     303             :      */
     304             :     const TupleTableSlotOps *(*slot_callbacks) (Relation rel);
     305             : 
     306             : 
     307             :     /* ------------------------------------------------------------------------
     308             :      * Table scan callbacks.
     309             :      * ------------------------------------------------------------------------
     310             :      */
     311             : 
     312             :     /*
     313             :      * Start a scan of `rel`.  The callback has to return a TableScanDesc,
     314             :      * which will typically be embedded in a larger, AM specific, struct.
     315             :      *
     316             :      * If nkeys != 0, the results need to be filtered by those scan keys.
     317             :      *
     318             :      * pscan, if not NULL, will have already been initialized with
     319             :      * parallelscan_initialize(), and has to be for the same relation. Will
     320             :      * only be set coming from table_beginscan_parallel().
     321             :      *
     322             :      * `flags` is a bitmask indicating the type of scan (ScanOptions's
     323             :      * SO_TYPE_*, currently only one may be specified), options controlling
     324             :      * the scan's behaviour (ScanOptions's SO_ALLOW_*, several may be
     325             :      * specified, an AM may ignore unsupported ones) and whether the snapshot
     326             :      * needs to be deallocated at scan_end (ScanOptions's SO_TEMP_SNAPSHOT).
     327             :      */
     328             :     TableScanDesc (*scan_begin) (Relation rel,
     329             :                                  Snapshot snapshot,
     330             :                                  int nkeys, struct ScanKeyData *key,
     331             :                                  ParallelTableScanDesc pscan,
     332             :                                  uint32 flags);
     333             : 
     334             :     /*
     335             :      * Release resources and deallocate scan. If TableScanDesc.temp_snap,
     336             :      * TableScanDesc.rs_snapshot needs to be unregistered.
     337             :      */
     338             :     void        (*scan_end) (TableScanDesc scan);
     339             : 
     340             :     /*
     341             :      * Restart relation scan.  If set_params is set to true, allow_{strat,
     342             :      * sync, pagemode} (see scan_begin) changes should be taken into account.
     343             :      */
     344             :     void        (*scan_rescan) (TableScanDesc scan, struct ScanKeyData *key,
     345             :                                 bool set_params, bool allow_strat,
     346             :                                 bool allow_sync, bool allow_pagemode);
     347             : 
     348             :     /*
     349             :      * Return next tuple from `scan`, store in slot.
     350             :      */
     351             :     bool        (*scan_getnextslot) (TableScanDesc scan,
     352             :                                      ScanDirection direction,
     353             :                                      TupleTableSlot *slot);
     354             : 
     355             :     /*-----------
     356             :      * Optional functions to provide scanning for ranges of ItemPointers.
     357             :      * Implementations must either provide both of these functions, or neither
     358             :      * of them.
     359             :      *
     360             :      * Implementations of scan_set_tidrange must themselves handle
     361             :      * ItemPointers of any value. i.e, they must handle each of the following:
     362             :      *
     363             :      * 1) mintid or maxtid is beyond the end of the table; and
     364             :      * 2) mintid is above maxtid; and
     365             :      * 3) item offset for mintid or maxtid is beyond the maximum offset
     366             :      * allowed by the AM.
     367             :      *
     368             :      * Implementations can assume that scan_set_tidrange is always called
     369             :      * before scan_getnextslot_tidrange or after scan_rescan and before any
     370             :      * further calls to scan_getnextslot_tidrange.
     371             :      */
     372             :     void        (*scan_set_tidrange) (TableScanDesc scan,
     373             :                                       ItemPointer mintid,
     374             :                                       ItemPointer maxtid);
     375             : 
     376             :     /*
     377             :      * Return next tuple from `scan` that's in the range of TIDs defined by
     378             :      * scan_set_tidrange.
     379             :      */
     380             :     bool        (*scan_getnextslot_tidrange) (TableScanDesc scan,
     381             :                                               ScanDirection direction,
     382             :                                               TupleTableSlot *slot);
     383             : 
     384             :     /* ------------------------------------------------------------------------
     385             :      * Parallel table scan related functions.
     386             :      * ------------------------------------------------------------------------
     387             :      */
     388             : 
     389             :     /*
     390             :      * Estimate the size of shared memory needed for a parallel scan of this
     391             :      * relation. The snapshot does not need to be accounted for.
     392             :      */
     393             :     Size        (*parallelscan_estimate) (Relation rel);
     394             : 
     395             :     /*
     396             :      * Initialize ParallelTableScanDesc for a parallel scan of this relation.
     397             :      * `pscan` will be sized according to parallelscan_estimate() for the same
     398             :      * relation.
     399             :      */
     400             :     Size        (*parallelscan_initialize) (Relation rel,
     401             :                                             ParallelTableScanDesc pscan);
     402             : 
     403             :     /*
     404             :      * Reinitialize `pscan` for a new scan. `rel` will be the same relation as
     405             :      * when `pscan` was initialized by parallelscan_initialize.
     406             :      */
     407             :     void        (*parallelscan_reinitialize) (Relation rel,
     408             :                                               ParallelTableScanDesc pscan);
     409             : 
     410             : 
     411             :     /* ------------------------------------------------------------------------
     412             :      * Index Scan Callbacks
     413             :      * ------------------------------------------------------------------------
     414             :      */
     415             : 
     416             :     /*
     417             :      * Prepare to fetch tuples from the relation, as needed when fetching
     418             :      * tuples for an index scan.  The callback has to return an
     419             :      * IndexFetchTableData, which the AM will typically embed in a larger
     420             :      * structure with additional information.
     421             :      *
     422             :      * Tuples for an index scan can then be fetched via index_fetch_tuple.
     423             :      */
     424             :     struct IndexFetchTableData *(*index_fetch_begin) (Relation rel);
     425             : 
     426             :     /*
     427             :      * Reset index fetch. Typically this will release cross index fetch
     428             :      * resources held in IndexFetchTableData.
     429             :      */
     430             :     void        (*index_fetch_reset) (struct IndexFetchTableData *data);
     431             : 
     432             :     /*
     433             :      * Release resources and deallocate index fetch.
     434             :      */
     435             :     void        (*index_fetch_end) (struct IndexFetchTableData *data);
     436             : 
     437             :     /*
     438             :      * Fetch tuple at `tid` into `slot`, after doing a visibility test
     439             :      * according to `snapshot`. If a tuple was found and passed the visibility
     440             :      * test, return true, false otherwise.
     441             :      *
     442             :      * Note that AMs that do not necessarily update indexes when indexed
     443             :      * columns do not change, need to return the current/correct version of
     444             :      * the tuple that is visible to the snapshot, even if the tid points to an
     445             :      * older version of the tuple.
     446             :      *
     447             :      * *call_again is false on the first call to index_fetch_tuple for a tid.
     448             :      * If there potentially is another tuple matching the tid, *call_again
     449             :      * needs to be set to true by index_fetch_tuple, signaling to the caller
     450             :      * that index_fetch_tuple should be called again for the same tid.
     451             :      *
     452             :      * *all_dead, if all_dead is not NULL, should be set to true by
     453             :      * index_fetch_tuple iff it is guaranteed that no backend needs to see
     454             :      * that tuple. Index AMs can use that to avoid returning that tid in
     455             :      * future searches.
     456             :      */
     457             :     bool        (*index_fetch_tuple) (struct IndexFetchTableData *scan,
     458             :                                       ItemPointer tid,
     459             :                                       Snapshot snapshot,
     460             :                                       TupleTableSlot *slot,
     461             :                                       bool *call_again, bool *all_dead);
     462             : 
     463             : 
     464             :     /* ------------------------------------------------------------------------
     465             :      * Callbacks for non-modifying operations on individual tuples
     466             :      * ------------------------------------------------------------------------
     467             :      */
     468             : 
     469             :     /*
     470             :      * Fetch tuple at `tid` into `slot`, after doing a visibility test
     471             :      * according to `snapshot`. If a tuple was found and passed the visibility
     472             :      * test, returns true, false otherwise.
     473             :      */
     474             :     bool        (*tuple_fetch_row_version) (Relation rel,
     475             :                                             ItemPointer tid,
     476             :                                             Snapshot snapshot,
     477             :                                             TupleTableSlot *slot);
     478             : 
     479             :     /*
     480             :      * Is tid valid for a scan of this relation.
     481             :      */
     482             :     bool        (*tuple_tid_valid) (TableScanDesc scan,
     483             :                                     ItemPointer tid);
     484             : 
     485             :     /*
     486             :      * Return the latest version of the tuple at `tid`, by updating `tid` to
     487             :      * point at the newest version.
     488             :      */
     489             :     void        (*tuple_get_latest_tid) (TableScanDesc scan,
     490             :                                          ItemPointer tid);
     491             : 
     492             :     /*
     493             :      * Does the tuple in `slot` satisfy `snapshot`?  The slot needs to be of
     494             :      * the appropriate type for the AM.
     495             :      */
     496             :     bool        (*tuple_satisfies_snapshot) (Relation rel,
     497             :                                              TupleTableSlot *slot,
     498             :                                              Snapshot snapshot);
     499             : 
     500             :     /* see table_index_delete_tuples() */
     501             :     TransactionId (*index_delete_tuples) (Relation rel,
     502             :                                           TM_IndexDeleteOp *delstate);
     503             : 
     504             : 
     505             :     /* ------------------------------------------------------------------------
     506             :      * Manipulations of physical tuples.
     507             :      * ------------------------------------------------------------------------
     508             :      */
     509             : 
     510             :     /* see table_tuple_insert() for reference about parameters */
     511             :     void        (*tuple_insert) (Relation rel, TupleTableSlot *slot,
     512             :                                  CommandId cid, int options,
     513             :                                  struct BulkInsertStateData *bistate);
     514             : 
     515             :     /* see table_tuple_insert_speculative() for reference about parameters */
     516             :     void        (*tuple_insert_speculative) (Relation rel,
     517             :                                              TupleTableSlot *slot,
     518             :                                              CommandId cid,
     519             :                                              int options,
     520             :                                              struct BulkInsertStateData *bistate,
     521             :                                              uint32 specToken);
     522             : 
     523             :     /* see table_tuple_complete_speculative() for reference about parameters */
     524             :     void        (*tuple_complete_speculative) (Relation rel,
     525             :                                                TupleTableSlot *slot,
     526             :                                                uint32 specToken,
     527             :                                                bool succeeded);
     528             : 
     529             :     /* see table_multi_insert() for reference about parameters */
     530             :     void        (*multi_insert) (Relation rel, TupleTableSlot **slots, int nslots,
     531             :                                  CommandId cid, int options, struct BulkInsertStateData *bistate);
     532             : 
     533             :     /* see table_tuple_delete() for reference about parameters */
     534             :     TM_Result   (*tuple_delete) (Relation rel,
     535             :                                  ItemPointer tid,
     536             :                                  CommandId cid,
     537             :                                  Snapshot snapshot,
     538             :                                  Snapshot crosscheck,
     539             :                                  bool wait,
     540             :                                  TM_FailureData *tmfd,
     541             :                                  bool changingPart);
     542             : 
     543             :     /* see table_tuple_update() for reference about parameters */
     544             :     TM_Result   (*tuple_update) (Relation rel,
     545             :                                  ItemPointer otid,
     546             :                                  TupleTableSlot *slot,
     547             :                                  CommandId cid,
     548             :                                  Snapshot snapshot,
     549             :                                  Snapshot crosscheck,
     550             :                                  bool wait,
     551             :                                  TM_FailureData *tmfd,
     552             :                                  LockTupleMode *lockmode,
     553             :                                  TU_UpdateIndexes *update_indexes);
     554             : 
     555             :     /* see table_tuple_lock() for reference about parameters */
     556             :     TM_Result   (*tuple_lock) (Relation rel,
     557             :                                ItemPointer tid,
     558             :                                Snapshot snapshot,
     559             :                                TupleTableSlot *slot,
     560             :                                CommandId cid,
     561             :                                LockTupleMode mode,
     562             :                                LockWaitPolicy wait_policy,
     563             :                                uint8 flags,
     564             :                                TM_FailureData *tmfd);
     565             : 
     566             :     /*
     567             :      * Perform operations necessary to complete insertions made via
     568             :      * tuple_insert and multi_insert with a BulkInsertState specified. In-tree
     569             :      * access methods ceased to use this.
     570             :      *
     571             :      * Typically callers of tuple_insert and multi_insert will just pass all
     572             :      * the flags that apply to them, and each AM has to decide which of them
     573             :      * make sense for it, and then only take actions in finish_bulk_insert for
     574             :      * those flags, and ignore others.
     575             :      *
     576             :      * Optional callback.
     577             :      */
     578             :     void        (*finish_bulk_insert) (Relation rel, int options);
     579             : 
     580             : 
     581             :     /* ------------------------------------------------------------------------
     582             :      * DDL related functionality.
     583             :      * ------------------------------------------------------------------------
     584             :      */
     585             : 
     586             :     /*
     587             :      * This callback needs to create new relation storage for `rel`, with
     588             :      * appropriate durability behaviour for `persistence`.
     589             :      *
     590             :      * Note that only the subset of the relcache filled by
     591             :      * RelationBuildLocalRelation() can be relied upon and that the relation's
     592             :      * catalog entries will either not yet exist (new relation), or will still
     593             :      * reference the old relfilelocator.
     594             :      *
     595             :      * As output *freezeXid, *minmulti must be set to the values appropriate
     596             :      * for pg_class.{relfrozenxid, relminmxid}. For AMs that don't need those
     597             :      * fields to be filled they can be set to InvalidTransactionId and
     598             :      * InvalidMultiXactId, respectively.
     599             :      *
     600             :      * See also table_relation_set_new_filelocator().
     601             :      */
     602             :     void        (*relation_set_new_filelocator) (Relation rel,
     603             :                                                  const RelFileLocator *newrlocator,
     604             :                                                  char persistence,
     605             :                                                  TransactionId *freezeXid,
     606             :                                                  MultiXactId *minmulti);
     607             : 
     608             :     /*
     609             :      * This callback needs to remove all contents from `rel`'s current
     610             :      * relfilelocator. No provisions for transactional behaviour need to be
     611             :      * made.  Often this can be implemented by truncating the underlying
     612             :      * storage to its minimal size.
     613             :      *
     614             :      * See also table_relation_nontransactional_truncate().
     615             :      */
     616             :     void        (*relation_nontransactional_truncate) (Relation rel);
     617             : 
     618             :     /*
     619             :      * See table_relation_copy_data().
     620             :      *
     621             :      * This can typically be implemented by directly copying the underlying
     622             :      * storage, unless it contains references to the tablespace internally.
     623             :      */
     624             :     void        (*relation_copy_data) (Relation rel,
     625             :                                        const RelFileLocator *newrlocator);
     626             : 
     627             :     /* See table_relation_copy_for_cluster() */
     628             :     void        (*relation_copy_for_cluster) (Relation OldTable,
     629             :                                               Relation NewTable,
     630             :                                               Relation OldIndex,
     631             :                                               bool use_sort,
     632             :                                               TransactionId OldestXmin,
     633             :                                               TransactionId *xid_cutoff,
     634             :                                               MultiXactId *multi_cutoff,
     635             :                                               double *num_tuples,
     636             :                                               double *tups_vacuumed,
     637             :                                               double *tups_recently_dead);
     638             : 
     639             :     /*
     640             :      * React to VACUUM command on the relation. The VACUUM can be triggered by
     641             :      * a user or by autovacuum. The specific actions performed by the AM will
     642             :      * depend heavily on the individual AM.
     643             :      *
     644             :      * On entry a transaction is already established, and the relation is
     645             :      * locked with a ShareUpdateExclusive lock.
     646             :      *
     647             :      * Note that neither VACUUM FULL (and CLUSTER), nor ANALYZE go through
     648             :      * this routine, even if (for ANALYZE) it is part of the same VACUUM
     649             :      * command.
     650             :      *
     651             :      * There probably, in the future, needs to be a separate callback to
     652             :      * integrate with autovacuum's scheduling.
     653             :      */
     654             :     void        (*relation_vacuum) (Relation rel,
     655             :                                     struct VacuumParams *params,
     656             :                                     BufferAccessStrategy bstrategy);
     657             : 
     658             :     /*
     659             :      * Prepare to analyze block `blockno` of `scan`. The scan has been started
     660             :      * with table_beginscan_analyze().  See also
     661             :      * table_scan_analyze_next_block().
     662             :      *
     663             :      * The callback may acquire resources like locks that are held until
     664             :      * table_scan_analyze_next_tuple() returns false. It e.g. can make sense
     665             :      * to hold a lock until all tuples on a block have been analyzed by
     666             :      * scan_analyze_next_tuple.
     667             :      *
     668             :      * The callback can return false if the block is not suitable for
     669             :      * sampling, e.g. because it's a metapage that could never contain tuples.
     670             :      *
     671             :      * XXX: This obviously is primarily suited for block-based AMs. It's not
     672             :      * clear what a good interface for non block based AMs would be, so there
     673             :      * isn't one yet.
     674             :      */
     675             :     bool        (*scan_analyze_next_block) (TableScanDesc scan,
     676             :                                             ReadStream *stream);
     677             : 
     678             :     /*
     679             :      * See table_scan_analyze_next_tuple().
     680             :      *
     681             :      * Not every AM might have a meaningful concept of dead rows, in which
     682             :      * case it's OK to not increment *deadrows - but note that that may
     683             :      * influence autovacuum scheduling (see comment for relation_vacuum
     684             :      * callback).
     685             :      */
     686             :     bool        (*scan_analyze_next_tuple) (TableScanDesc scan,
     687             :                                             TransactionId OldestXmin,
     688             :                                             double *liverows,
     689             :                                             double *deadrows,
     690             :                                             TupleTableSlot *slot);
     691             : 
     692             :     /* see table_index_build_range_scan for reference about parameters */
     693             :     double      (*index_build_range_scan) (Relation table_rel,
     694             :                                            Relation index_rel,
     695             :                                            struct IndexInfo *index_info,
     696             :                                            bool allow_sync,
     697             :                                            bool anyvisible,
     698             :                                            bool progress,
     699             :                                            BlockNumber start_blockno,
     700             :                                            BlockNumber numblocks,
     701             :                                            IndexBuildCallback callback,
     702             :                                            void *callback_state,
     703             :                                            TableScanDesc scan);
     704             : 
     705             :     /* see table_index_validate_scan for reference about parameters */
     706             :     void        (*index_validate_scan) (Relation table_rel,
     707             :                                         Relation index_rel,
     708             :                                         struct IndexInfo *index_info,
     709             :                                         Snapshot snapshot,
     710             :                                         struct ValidateIndexState *state);
     711             : 
     712             : 
     713             :     /* ------------------------------------------------------------------------
     714             :      * Miscellaneous functions.
     715             :      * ------------------------------------------------------------------------
     716             :      */
     717             : 
     718             :     /*
     719             :      * See table_relation_size().
     720             :      *
     721             :      * Note that currently a few callers use the MAIN_FORKNUM size to figure
     722             :      * out the range of potentially interesting blocks (brin, analyze). It's
     723             :      * probable that we'll need to revise the interface for those at some
     724             :      * point.
     725             :      */
     726             :     uint64      (*relation_size) (Relation rel, ForkNumber forkNumber);
     727             : 
     728             : 
     729             :     /*
     730             :      * This callback should return true if the relation requires a TOAST table
     731             :      * and false if it does not.  It may wish to examine the relation's tuple
     732             :      * descriptor before making a decision, but if it uses some other method
     733             :      * of storing large values (or if it does not support them) it can simply
     734             :      * return false.
     735             :      */
     736             :     bool        (*relation_needs_toast_table) (Relation rel);
     737             : 
     738             :     /*
     739             :      * This callback should return the OID of the table AM that implements
     740             :      * TOAST tables for this AM.  If the relation_needs_toast_table callback
     741             :      * always returns false, this callback is not required.
     742             :      */
     743             :     Oid         (*relation_toast_am) (Relation rel);
     744             : 
     745             :     /*
     746             :      * This callback is invoked when detoasting a value stored in a toast
     747             :      * table implemented by this AM.  See table_relation_fetch_toast_slice()
     748             :      * for more details.
     749             :      */
     750             :     void        (*relation_fetch_toast_slice) (Relation toastrel, Oid valueid,
     751             :                                                int32 attrsize,
     752             :                                                int32 sliceoffset,
     753             :                                                int32 slicelength,
     754             :                                                struct varlena *result);
     755             : 
     756             : 
     757             :     /* ------------------------------------------------------------------------
     758             :      * Planner related functions.
     759             :      * ------------------------------------------------------------------------
     760             :      */
     761             : 
     762             :     /*
     763             :      * See table_relation_estimate_size().
     764             :      *
     765             :      * While block oriented, it shouldn't be too hard for an AM that doesn't
     766             :      * internally use blocks to convert into a usable representation.
     767             :      *
     768             :      * This differs from the relation_size callback by returning size
     769             :      * estimates (both relation size and tuple count) for planning purposes,
     770             :      * rather than returning a currently correct estimate.
     771             :      */
     772             :     void        (*relation_estimate_size) (Relation rel, int32 *attr_widths,
     773             :                                            BlockNumber *pages, double *tuples,
     774             :                                            double *allvisfrac);
     775             : 
     776             : 
     777             :     /* ------------------------------------------------------------------------
     778             :      * Executor related functions.
     779             :      * ------------------------------------------------------------------------
     780             :      */
     781             : 
     782             :     /*
     783             :      * Fetch the next tuple of a bitmap table scan into `slot` and return true
     784             :      * if a visible tuple was found, false otherwise.
     785             :      *
     786             :      * `lossy_pages` is incremented if the bitmap is lossy for the selected
     787             :      * page; otherwise, `exact_pages` is incremented. These are tracked for
     788             :      * display in EXPLAIN ANALYZE output.
     789             :      *
     790             :      * Prefetching additional data from the bitmap is left to the table AM.
     791             :      *
     792             :      * This is an optional callback.
     793             :      */
     794             :     bool        (*scan_bitmap_next_tuple) (TableScanDesc scan,
     795             :                                            TupleTableSlot *slot,
     796             :                                            bool *recheck,
     797             :                                            uint64 *lossy_pages,
     798             :                                            uint64 *exact_pages);
     799             : 
     800             :     /*
     801             :      * Prepare to fetch tuples from the next block in a sample scan. Return
     802             :      * false if the sample scan is finished, true otherwise. `scan` was
     803             :      * started via table_beginscan_sampling().
     804             :      *
     805             :      * Typically this will first determine the target block by calling the
     806             :      * TsmRoutine's NextSampleBlock() callback if not NULL, or alternatively
     807             :      * perform a sequential scan over all blocks.  The determined block is
     808             :      * then typically read and pinned.
     809             :      *
     810             :      * As the TsmRoutine interface is block based, a block needs to be passed
     811             :      * to NextSampleBlock(). If that's not appropriate for an AM, it
     812             :      * internally needs to perform mapping between the internal and a block
     813             :      * based representation.
     814             :      *
     815             :      * Note that it's not acceptable to hold deadlock prone resources such as
     816             :      * lwlocks until scan_sample_next_tuple() has exhausted the tuples on the
     817             :      * block - the tuple is likely to be returned to an upper query node, and
     818             :      * the next call could be off a long while. Holding buffer pins and such
     819             :      * is obviously OK.
     820             :      *
     821             :      * Currently it is required to implement this interface, as there's no
     822             :      * alternative way (contrary e.g. to bitmap scans) to implement sample
     823             :      * scans. If infeasible to implement, the AM may raise an error.
     824             :      */
     825             :     bool        (*scan_sample_next_block) (TableScanDesc scan,
     826             :                                            struct SampleScanState *scanstate);
     827             : 
     828             :     /*
     829             :      * This callback, only called after scan_sample_next_block has returned
     830             :      * true, should determine the next tuple to be returned from the selected
     831             :      * block using the TsmRoutine's NextSampleTuple() callback.
     832             :      *
     833             :      * The callback needs to perform visibility checks, and only return
     834             :      * visible tuples. That obviously can mean calling NextSampleTuple()
     835             :      * multiple times.
     836             :      *
     837             :      * The TsmRoutine interface assumes that there's a maximum offset on a
     838             :      * given page, so if that doesn't apply to an AM, it needs to emulate that
     839             :      * assumption somehow.
     840             :      */
     841             :     bool        (*scan_sample_next_tuple) (TableScanDesc scan,
     842             :                                            struct SampleScanState *scanstate,
     843             :                                            TupleTableSlot *slot);
     844             : 
     845             : } TableAmRoutine;
     846             : 
     847             : 
     848             : /* ----------------------------------------------------------------------------
     849             :  * Slot functions.
     850             :  * ----------------------------------------------------------------------------
     851             :  */
     852             : 
     853             : /*
     854             :  * Returns slot callbacks suitable for holding tuples of the appropriate type
     855             :  * for the relation.  Works for tables, views, foreign tables and partitioned
     856             :  * tables.
     857             :  */
     858             : extern const TupleTableSlotOps *table_slot_callbacks(Relation relation);
     859             : 
     860             : /*
     861             :  * Returns slot using the callbacks returned by table_slot_callbacks(), and
     862             :  * registers it on *reglist.
     863             :  */
     864             : extern TupleTableSlot *table_slot_create(Relation relation, List **reglist);
     865             : 
     866             : 
     867             : /* ----------------------------------------------------------------------------
     868             :  * Table scan functions.
     869             :  * ----------------------------------------------------------------------------
     870             :  */
     871             : 
     872             : /*
     873             :  * Start a scan of `rel`. Returned tuples pass a visibility test of
     874             :  * `snapshot`, and if nkeys != 0, the results are filtered by those scan keys.
     875             :  */
     876             : static inline TableScanDesc
     877      195796 : table_beginscan(Relation rel, Snapshot snapshot,
     878             :                 int nkeys, struct ScanKeyData *key)
     879             : {
     880      195796 :     uint32      flags = SO_TYPE_SEQSCAN |
     881             :         SO_ALLOW_STRAT | SO_ALLOW_SYNC | SO_ALLOW_PAGEMODE;
     882             : 
     883      195796 :     return rel->rd_tableam->scan_begin(rel, snapshot, nkeys, key, NULL, flags);
     884             : }
     885             : 
     886             : /*
     887             :  * Like table_beginscan(), but for scanning catalog. It'll automatically use a
     888             :  * snapshot appropriate for scanning catalog relations.
     889             :  */
     890             : extern TableScanDesc table_beginscan_catalog(Relation relation, int nkeys,
     891             :                                              struct ScanKeyData *key);
     892             : 
     893             : /*
     894             :  * Like table_beginscan(), but table_beginscan_strat() offers an extended API
     895             :  * that lets the caller control whether a nondefault buffer access strategy
     896             :  * can be used, and whether syncscan can be chosen (possibly resulting in the
     897             :  * scan not starting from block zero).  Both of these default to true with
     898             :  * plain table_beginscan.
     899             :  */
     900             : static inline TableScanDesc
     901      412328 : table_beginscan_strat(Relation rel, Snapshot snapshot,
     902             :                       int nkeys, struct ScanKeyData *key,
     903             :                       bool allow_strat, bool allow_sync)
     904             : {
     905      412328 :     uint32      flags = SO_TYPE_SEQSCAN | SO_ALLOW_PAGEMODE;
     906             : 
     907      412328 :     if (allow_strat)
     908      412328 :         flags |= SO_ALLOW_STRAT;
     909      412328 :     if (allow_sync)
     910       51222 :         flags |= SO_ALLOW_SYNC;
     911             : 
     912      412328 :     return rel->rd_tableam->scan_begin(rel, snapshot, nkeys, key, NULL, flags);
     913             : }
     914             : 
     915             : /*
     916             :  * table_beginscan_bm is an alternative entry point for setting up a
     917             :  * TableScanDesc for a bitmap heap scan.  Although that scan technology is
     918             :  * really quite unlike a standard seqscan, there is just enough commonality to
     919             :  * make it worth using the same data structure.
     920             :  */
     921             : static inline TableScanDesc
     922       21352 : table_beginscan_bm(Relation rel, Snapshot snapshot,
     923             :                    int nkeys, struct ScanKeyData *key, bool need_tuple)
     924             : {
     925       21352 :     uint32      flags = SO_TYPE_BITMAPSCAN | SO_ALLOW_PAGEMODE;
     926             : 
     927       21352 :     if (need_tuple)
     928       18464 :         flags |= SO_NEED_TUPLES;
     929             : 
     930       21352 :     return rel->rd_tableam->scan_begin(rel, snapshot, nkeys, key,
     931             :                                        NULL, flags);
     932             : }
     933             : 
     934             : /*
     935             :  * table_beginscan_sampling is an alternative entry point for setting up a
     936             :  * TableScanDesc for a TABLESAMPLE scan.  As with bitmap scans, it's worth
     937             :  * using the same data structure although the behavior is rather different.
     938             :  * In addition to the options offered by table_beginscan_strat, this call
     939             :  * also allows control of whether page-mode visibility checking is used.
     940             :  */
     941             : static inline TableScanDesc
     942         146 : table_beginscan_sampling(Relation rel, Snapshot snapshot,
     943             :                          int nkeys, struct ScanKeyData *key,
     944             :                          bool allow_strat, bool allow_sync,
     945             :                          bool allow_pagemode)
     946             : {
     947         146 :     uint32      flags = SO_TYPE_SAMPLESCAN;
     948             : 
     949         146 :     if (allow_strat)
     950         134 :         flags |= SO_ALLOW_STRAT;
     951         146 :     if (allow_sync)
     952          66 :         flags |= SO_ALLOW_SYNC;
     953         146 :     if (allow_pagemode)
     954         122 :         flags |= SO_ALLOW_PAGEMODE;
     955             : 
     956         146 :     return rel->rd_tableam->scan_begin(rel, snapshot, nkeys, key, NULL, flags);
     957             : }
     958             : 
     959             : /*
     960             :  * table_beginscan_tid is an alternative entry point for setting up a
     961             :  * TableScanDesc for a Tid scan. As with bitmap scans, it's worth using
     962             :  * the same data structure although the behavior is rather different.
     963             :  */
     964             : static inline TableScanDesc
     965         730 : table_beginscan_tid(Relation rel, Snapshot snapshot)
     966             : {
     967         730 :     uint32      flags = SO_TYPE_TIDSCAN;
     968             : 
     969         730 :     return rel->rd_tableam->scan_begin(rel, snapshot, 0, NULL, NULL, flags);
     970             : }
     971             : 
     972             : /*
     973             :  * table_beginscan_analyze is an alternative entry point for setting up a
     974             :  * TableScanDesc for an ANALYZE scan.  As with bitmap scans, it's worth using
     975             :  * the same data structure although the behavior is rather different.
     976             :  */
     977             : static inline TableScanDesc
     978       16188 : table_beginscan_analyze(Relation rel)
     979             : {
     980       16188 :     uint32      flags = SO_TYPE_ANALYZE;
     981             : 
     982       16188 :     return rel->rd_tableam->scan_begin(rel, NULL, 0, NULL, NULL, flags);
     983             : }
     984             : 
     985             : /*
     986             :  * End relation scan.
     987             :  */
     988             : static inline void
     989      714958 : table_endscan(TableScanDesc scan)
     990             : {
     991      714958 :     scan->rs_rd->rd_tableam->scan_end(scan);
     992      714958 : }
     993             : 
     994             : /*
     995             :  * Restart a relation scan.
     996             :  */
     997             : static inline void
     998     1173128 : table_rescan(TableScanDesc scan,
     999             :              struct ScanKeyData *key)
    1000             : {
    1001     1173128 :     scan->rs_rd->rd_tableam->scan_rescan(scan, key, false, false, false, false);
    1002     1173128 : }
    1003             : 
    1004             : /*
    1005             :  * Restart a relation scan after changing params.
    1006             :  *
    1007             :  * This call allows changing the buffer strategy, syncscan, and pagemode
    1008             :  * options before starting a fresh scan.  Note that although the actual use of
    1009             :  * syncscan might change (effectively, enabling or disabling reporting), the
    1010             :  * previously selected startblock will be kept.
    1011             :  */
    1012             : static inline void
    1013          30 : table_rescan_set_params(TableScanDesc scan, struct ScanKeyData *key,
    1014             :                         bool allow_strat, bool allow_sync, bool allow_pagemode)
    1015             : {
    1016          30 :     scan->rs_rd->rd_tableam->scan_rescan(scan, key, true,
    1017             :                                          allow_strat, allow_sync,
    1018             :                                          allow_pagemode);
    1019          30 : }
    1020             : 
    1021             : /*
    1022             :  * Return next tuple from `scan`, store in slot.
    1023             :  */
    1024             : static inline bool
    1025    89912780 : table_scan_getnextslot(TableScanDesc sscan, ScanDirection direction, TupleTableSlot *slot)
    1026             : {
    1027    89912780 :     slot->tts_tableOid = RelationGetRelid(sscan->rs_rd);
    1028             : 
    1029             :     /* We don't expect actual scans using NoMovementScanDirection */
    1030             :     Assert(direction == ForwardScanDirection ||
    1031             :            direction == BackwardScanDirection);
    1032             : 
    1033             :     /*
    1034             :      * We don't expect direct calls to table_scan_getnextslot with valid
    1035             :      * CheckXidAlive for catalog or regular tables.  See detailed comments in
    1036             :      * xact.c where these variables are declared.
    1037             :      */
    1038    89912780 :     if (unlikely(TransactionIdIsValid(CheckXidAlive) && !bsysscan))
    1039           0 :         elog(ERROR, "unexpected table_scan_getnextslot call during logical decoding");
    1040             : 
    1041    89912780 :     return sscan->rs_rd->rd_tableam->scan_getnextslot(sscan, direction, slot);
    1042             : }
    1043             : 
    1044             : /* ----------------------------------------------------------------------------
    1045             :  * TID Range scanning related functions.
    1046             :  * ----------------------------------------------------------------------------
    1047             :  */
    1048             : 
    1049             : /*
    1050             :  * table_beginscan_tidrange is the entry point for setting up a TableScanDesc
    1051             :  * for a TID range scan.
    1052             :  */
    1053             : static inline TableScanDesc
    1054        1846 : table_beginscan_tidrange(Relation rel, Snapshot snapshot,
    1055             :                          ItemPointer mintid,
    1056             :                          ItemPointer maxtid)
    1057             : {
    1058             :     TableScanDesc sscan;
    1059        1846 :     uint32      flags = SO_TYPE_TIDRANGESCAN | SO_ALLOW_PAGEMODE;
    1060             : 
    1061        1846 :     sscan = rel->rd_tableam->scan_begin(rel, snapshot, 0, NULL, NULL, flags);
    1062             : 
    1063             :     /* Set the range of TIDs to scan */
    1064        1846 :     sscan->rs_rd->rd_tableam->scan_set_tidrange(sscan, mintid, maxtid);
    1065             : 
    1066        1846 :     return sscan;
    1067             : }
    1068             : 
    1069             : /*
    1070             :  * table_rescan_tidrange resets the scan position and sets the minimum and
    1071             :  * maximum TID range to scan for a TableScanDesc created by
    1072             :  * table_beginscan_tidrange.
    1073             :  */
    1074             : static inline void
    1075          66 : table_rescan_tidrange(TableScanDesc sscan, ItemPointer mintid,
    1076             :                       ItemPointer maxtid)
    1077             : {
    1078             :     /* Ensure table_beginscan_tidrange() was used. */
    1079             :     Assert((sscan->rs_flags & SO_TYPE_TIDRANGESCAN) != 0);
    1080             : 
    1081          66 :     sscan->rs_rd->rd_tableam->scan_rescan(sscan, NULL, false, false, false, false);
    1082          66 :     sscan->rs_rd->rd_tableam->scan_set_tidrange(sscan, mintid, maxtid);
    1083          66 : }
    1084             : 
    1085             : /*
    1086             :  * Fetch the next tuple from `sscan` for a TID range scan created by
    1087             :  * table_beginscan_tidrange().  Stores the tuple in `slot` and returns true,
    1088             :  * or returns false if no more tuples exist in the range.
    1089             :  */
    1090             : static inline bool
    1091        7686 : table_scan_getnextslot_tidrange(TableScanDesc sscan, ScanDirection direction,
    1092             :                                 TupleTableSlot *slot)
    1093             : {
    1094             :     /* Ensure table_beginscan_tidrange() was used. */
    1095             :     Assert((sscan->rs_flags & SO_TYPE_TIDRANGESCAN) != 0);
    1096             : 
    1097             :     /* We don't expect actual scans using NoMovementScanDirection */
    1098             :     Assert(direction == ForwardScanDirection ||
    1099             :            direction == BackwardScanDirection);
    1100             : 
    1101        7686 :     return sscan->rs_rd->rd_tableam->scan_getnextslot_tidrange(sscan,
    1102             :                                                                direction,
    1103             :                                                                slot);
    1104             : }
    1105             : 
    1106             : 
    1107             : /* ----------------------------------------------------------------------------
    1108             :  * Parallel table scan related functions.
    1109             :  * ----------------------------------------------------------------------------
    1110             :  */
    1111             : 
    1112             : /*
    1113             :  * Estimate the size of shared memory needed for a parallel scan of this
    1114             :  * relation.
    1115             :  */
    1116             : extern Size table_parallelscan_estimate(Relation rel, Snapshot snapshot);
    1117             : 
    1118             : /*
    1119             :  * Initialize ParallelTableScanDesc for a parallel scan of this
    1120             :  * relation. `pscan` needs to be sized according to parallelscan_estimate()
    1121             :  * for the same relation.  Call this just once in the leader process; then,
    1122             :  * individual workers attach via table_beginscan_parallel.
    1123             :  */
    1124             : extern void table_parallelscan_initialize(Relation rel,
    1125             :                                           ParallelTableScanDesc pscan,
    1126             :                                           Snapshot snapshot);
    1127             : 
    1128             : /*
    1129             :  * Begin a parallel scan. `pscan` needs to have been initialized with
    1130             :  * table_parallelscan_initialize(), for the same relation. The initialization
    1131             :  * does not need to have happened in this backend.
    1132             :  *
    1133             :  * Caller must hold a suitable lock on the relation.
    1134             :  */
    1135             : extern TableScanDesc table_beginscan_parallel(Relation relation,
    1136             :                                               ParallelTableScanDesc pscan);
    1137             : 
    1138             : /*
    1139             :  * Restart a parallel scan.  Call this in the leader process.  Caller is
    1140             :  * responsible for making sure that all workers have finished the scan
    1141             :  * beforehand.
    1142             :  */
    1143             : static inline void
    1144         228 : table_parallelscan_reinitialize(Relation rel, ParallelTableScanDesc pscan)
    1145             : {
    1146         228 :     rel->rd_tableam->parallelscan_reinitialize(rel, pscan);
    1147         228 : }
    1148             : 
    1149             : 
    1150             : /* ----------------------------------------------------------------------------
    1151             :  *  Index scan related functions.
    1152             :  * ----------------------------------------------------------------------------
    1153             :  */
    1154             : 
    1155             : /*
    1156             :  * Prepare to fetch tuples from the relation, as needed when fetching tuples
    1157             :  * for an index scan.
    1158             :  *
    1159             :  * Tuples for an index scan can then be fetched via table_index_fetch_tuple().
    1160             :  */
    1161             : static inline IndexFetchTableData *
    1162    25191188 : table_index_fetch_begin(Relation rel)
    1163             : {
    1164    25191188 :     return rel->rd_tableam->index_fetch_begin(rel);
    1165             : }
    1166             : 
    1167             : /*
    1168             :  * Reset index fetch. Typically this will release cross index fetch resources
    1169             :  * held in IndexFetchTableData.
    1170             :  */
    1171             : static inline void
    1172    20969342 : table_index_fetch_reset(struct IndexFetchTableData *scan)
    1173             : {
    1174    20969342 :     scan->rel->rd_tableam->index_fetch_reset(scan);
    1175    20969342 : }
    1176             : 
    1177             : /*
    1178             :  * Release resources and deallocate index fetch.
    1179             :  */
    1180             : static inline void
    1181    25189502 : table_index_fetch_end(struct IndexFetchTableData *scan)
    1182             : {
    1183    25189502 :     scan->rel->rd_tableam->index_fetch_end(scan);
    1184    25189502 : }
    1185             : 
    1186             : /*
    1187             :  * Fetches, as part of an index scan, tuple at `tid` into `slot`, after doing
    1188             :  * a visibility test according to `snapshot`. If a tuple was found and passed
    1189             :  * the visibility test, returns true, false otherwise. Note that *tid may be
    1190             :  * modified when we return true (see later remarks on multiple row versions
    1191             :  * reachable via a single index entry).
    1192             :  *
    1193             :  * *call_again needs to be false on the first call to table_index_fetch_tuple() for
    1194             :  * a tid. If there potentially is another tuple matching the tid, *call_again
    1195             :  * will be set to true, signaling that table_index_fetch_tuple() should be called
    1196             :  * again for the same tid.
    1197             :  *
    1198             :  * *all_dead, if all_dead is not NULL, will be set to true by
    1199             :  * table_index_fetch_tuple() iff it is guaranteed that no backend needs to see
    1200             :  * that tuple. Index AMs can use that to avoid returning that tid in future
    1201             :  * searches.
    1202             :  *
    1203             :  * The difference between this function and table_tuple_fetch_row_version()
    1204             :  * is that this function returns the currently visible version of a row if
    1205             :  * the AM supports storing multiple row versions reachable via a single index
    1206             :  * entry (like heap's HOT). Whereas table_tuple_fetch_row_version() only
    1207             :  * evaluates the tuple exactly at `tid`. Outside of index entry ->table tuple
    1208             :  * lookups, table_tuple_fetch_row_version() is what's usually needed.
    1209             :  */
    1210             : static inline bool
    1211    35841366 : table_index_fetch_tuple(struct IndexFetchTableData *scan,
    1212             :                         ItemPointer tid,
    1213             :                         Snapshot snapshot,
    1214             :                         TupleTableSlot *slot,
    1215             :                         bool *call_again, bool *all_dead)
    1216             : {
    1217             :     /*
    1218             :      * We don't expect direct calls to table_index_fetch_tuple with valid
    1219             :      * CheckXidAlive for catalog or regular tables.  See detailed comments in
    1220             :      * xact.c where these variables are declared.
    1221             :      */
    1222    35841366 :     if (unlikely(TransactionIdIsValid(CheckXidAlive) && !bsysscan))
    1223           0 :         elog(ERROR, "unexpected table_index_fetch_tuple call during logical decoding");
    1224             : 
    1225    35841366 :     return scan->rel->rd_tableam->index_fetch_tuple(scan, tid, snapshot,
    1226             :                                                     slot, call_again,
    1227             :                                                     all_dead);
    1228             : }
    1229             : 
    1230             : /*
    1231             :  * This is a convenience wrapper around table_index_fetch_tuple() which
    1232             :  * returns whether there are table tuple items corresponding to an index
    1233             :  * entry.  This likely is only useful to verify if there's a conflict in a
    1234             :  * unique index.
    1235             :  */
    1236             : extern bool table_index_fetch_tuple_check(Relation rel,
    1237             :                                           ItemPointer tid,
    1238             :                                           Snapshot snapshot,
    1239             :                                           bool *all_dead);
    1240             : 
    1241             : 
    1242             : /* ------------------------------------------------------------------------
    1243             :  * Functions for non-modifying operations on individual tuples
    1244             :  * ------------------------------------------------------------------------
    1245             :  */
    1246             : 
    1247             : 
    1248             : /*
    1249             :  * Fetch tuple at `tid` into `slot`, after doing a visibility test according to
    1250             :  * `snapshot`. If a tuple was found and passed the visibility test, returns
    1251             :  * true, false otherwise.
    1252             :  *
    1253             :  * See table_index_fetch_tuple's comment about what the difference between
    1254             :  * these functions is. It is correct to use this function outside of index
    1255             :  * entry->table tuple lookups.
    1256             :  */
    1257             : static inline bool
    1258      352828 : table_tuple_fetch_row_version(Relation rel,
    1259             :                               ItemPointer tid,
    1260             :                               Snapshot snapshot,
    1261             :                               TupleTableSlot *slot)
    1262             : {
    1263             :     /*
    1264             :      * We don't expect direct calls to table_tuple_fetch_row_version with
    1265             :      * valid CheckXidAlive for catalog or regular tables.  See detailed
    1266             :      * comments in xact.c where these variables are declared.
    1267             :      */
    1268      352828 :     if (unlikely(TransactionIdIsValid(CheckXidAlive) && !bsysscan))
    1269           0 :         elog(ERROR, "unexpected table_tuple_fetch_row_version call during logical decoding");
    1270             : 
    1271      352828 :     return rel->rd_tableam->tuple_fetch_row_version(rel, tid, snapshot, slot);
    1272             : }
    1273             : 
    1274             : /*
    1275             :  * Verify that `tid` is a potentially valid tuple identifier. That doesn't
    1276             :  * mean that the pointed to row needs to exist or be visible, but that
    1277             :  * attempting to fetch the row (e.g. with table_tuple_get_latest_tid() or
    1278             :  * table_tuple_fetch_row_version()) should not error out if called with that
    1279             :  * tid.
    1280             :  *
    1281             :  * `scan` needs to have been started via table_beginscan().
    1282             :  */
    1283             : static inline bool
    1284         394 : table_tuple_tid_valid(TableScanDesc scan, ItemPointer tid)
    1285             : {
    1286         394 :     return scan->rs_rd->rd_tableam->tuple_tid_valid(scan, tid);
    1287             : }
    1288             : 
    1289             : /*
    1290             :  * Return the latest version of the tuple at `tid`, by updating `tid` to
    1291             :  * point at the newest version.
    1292             :  */
    1293             : extern void table_tuple_get_latest_tid(TableScanDesc scan, ItemPointer tid);
    1294             : 
    1295             : /*
    1296             :  * Return true iff tuple in slot satisfies the snapshot.
    1297             :  *
    1298             :  * This assumes the slot's tuple is valid, and of the appropriate type for the
    1299             :  * AM.
    1300             :  *
    1301             :  * Some AMs might modify the data underlying the tuple as a side-effect. If so
    1302             :  * they ought to mark the relevant buffer dirty.
    1303             :  */
    1304             : static inline bool
    1305      227058 : table_tuple_satisfies_snapshot(Relation rel, TupleTableSlot *slot,
    1306             :                                Snapshot snapshot)
    1307             : {
    1308      227058 :     return rel->rd_tableam->tuple_satisfies_snapshot(rel, slot, snapshot);
    1309             : }
    1310             : 
    1311             : /*
    1312             :  * Determine which index tuples are safe to delete based on their table TID.
    1313             :  *
    1314             :  * Determines which entries from index AM caller's TM_IndexDeleteOp state
    1315             :  * point to vacuumable table tuples.  Entries that are found by tableam to be
    1316             :  * vacuumable are naturally safe for index AM to delete, and so get directly
    1317             :  * marked as deletable.  See comments above TM_IndexDelete and comments above
    1318             :  * TM_IndexDeleteOp for full details.
    1319             :  *
    1320             :  * Returns a snapshotConflictHorizon transaction ID that caller places in
    1321             :  * its index deletion WAL record.  This might be used during subsequent REDO
    1322             :  * of the WAL record when in Hot Standby mode -- a recovery conflict for the
    1323             :  * index deletion operation might be required on the standby.
    1324             :  */
    1325             : static inline TransactionId
    1326       11164 : table_index_delete_tuples(Relation rel, TM_IndexDeleteOp *delstate)
    1327             : {
    1328       11164 :     return rel->rd_tableam->index_delete_tuples(rel, delstate);
    1329             : }
    1330             : 
    1331             : 
    1332             : /* ----------------------------------------------------------------------------
    1333             :  *  Functions for manipulations of physical tuples.
    1334             :  * ----------------------------------------------------------------------------
    1335             :  */
    1336             : 
    1337             : /*
    1338             :  * Insert a tuple from a slot into table AM routine.
    1339             :  *
    1340             :  * The options bitmask allows the caller to specify options that may change the
    1341             :  * behaviour of the AM. The AM will ignore options that it does not support.
    1342             :  *
    1343             :  * If the TABLE_INSERT_SKIP_FSM option is specified, AMs are free to not reuse
    1344             :  * free space in the relation. This can save some cycles when we know the
    1345             :  * relation is new and doesn't contain useful amounts of free space.
    1346             :  * TABLE_INSERT_SKIP_FSM is commonly passed directly to
    1347             :  * RelationGetBufferForTuple. See that method for more information.
    1348             :  *
    1349             :  * TABLE_INSERT_FROZEN should only be specified for inserts into
    1350             :  * relation storage created during the current subtransaction and when
    1351             :  * there are no prior snapshots or pre-existing portals open.
    1352             :  * This causes rows to be frozen, which is an MVCC violation and
    1353             :  * requires explicit options chosen by user.
    1354             :  *
    1355             :  * TABLE_INSERT_NO_LOGICAL force-disables the emitting of logical decoding
    1356             :  * information for the tuple. This should solely be used during table rewrites
    1357             :  * where RelationIsLogicallyLogged(relation) is not yet accurate for the new
    1358             :  * relation.
    1359             :  *
    1360             :  * Note that most of these options will be applied when inserting into the
    1361             :  * heap's TOAST table, too, if the tuple requires any out-of-line data.
    1362             :  *
    1363             :  * The BulkInsertState object (if any; bistate can be NULL for default
    1364             :  * behavior) is also just passed through to RelationGetBufferForTuple. If
    1365             :  * `bistate` is provided, table_finish_bulk_insert() needs to be called.
    1366             :  *
    1367             :  * On return the slot's tts_tid and tts_tableOid are updated to reflect the
    1368             :  * insertion. But note that any toasting of fields within the slot is NOT
    1369             :  * reflected in the slots contents.
    1370             :  */
    1371             : static inline void
    1372    14155098 : table_tuple_insert(Relation rel, TupleTableSlot *slot, CommandId cid,
    1373             :                    int options, struct BulkInsertStateData *bistate)
    1374             : {
    1375    14155098 :     rel->rd_tableam->tuple_insert(rel, slot, cid, options,
    1376             :                                   bistate);
    1377    14155064 : }
    1378             : 
    1379             : /*
    1380             :  * Perform a "speculative insertion". These can be backed out afterwards
    1381             :  * without aborting the whole transaction.  Other sessions can wait for the
    1382             :  * speculative insertion to be confirmed, turning it into a regular tuple, or
    1383             :  * aborted, as if it never existed.  Speculatively inserted tuples behave as
    1384             :  * "value locks" of short duration, used to implement INSERT .. ON CONFLICT.
    1385             :  *
    1386             :  * A transaction having performed a speculative insertion has to either abort,
    1387             :  * or finish the speculative insertion with
    1388             :  * table_tuple_complete_speculative(succeeded = ...).
    1389             :  */
    1390             : static inline void
    1391        4128 : table_tuple_insert_speculative(Relation rel, TupleTableSlot *slot,
    1392             :                                CommandId cid, int options,
    1393             :                                struct BulkInsertStateData *bistate,
    1394             :                                uint32 specToken)
    1395             : {
    1396        4128 :     rel->rd_tableam->tuple_insert_speculative(rel, slot, cid, options,
    1397             :                                               bistate, specToken);
    1398        4128 : }
    1399             : 
    1400             : /*
    1401             :  * Complete "speculative insertion" started in the same transaction. If
    1402             :  * succeeded is true, the tuple is fully inserted, if false, it's removed.
    1403             :  */
    1404             : static inline void
    1405        4122 : table_tuple_complete_speculative(Relation rel, TupleTableSlot *slot,
    1406             :                                  uint32 specToken, bool succeeded)
    1407             : {
    1408        4122 :     rel->rd_tableam->tuple_complete_speculative(rel, slot, specToken,
    1409             :                                                 succeeded);
    1410        4122 : }
    1411             : 
    1412             : /*
    1413             :  * Insert multiple tuples into a table.
    1414             :  *
    1415             :  * This is like table_tuple_insert(), but inserts multiple tuples in one
    1416             :  * operation. That's often faster than calling table_tuple_insert() in a loop,
    1417             :  * because e.g. the AM can reduce WAL logging and page locking overhead.
    1418             :  *
    1419             :  * Except for taking `nslots` tuples as input, and an array of TupleTableSlots
    1420             :  * in `slots`, the parameters for table_multi_insert() are the same as for
    1421             :  * table_tuple_insert().
    1422             :  *
    1423             :  * Note: this leaks memory into the current memory context. You can create a
    1424             :  * temporary context before calling this, if that's a problem.
    1425             :  */
    1426             : static inline void
    1427        2444 : table_multi_insert(Relation rel, TupleTableSlot **slots, int nslots,
    1428             :                    CommandId cid, int options, struct BulkInsertStateData *bistate)
    1429             : {
    1430        2444 :     rel->rd_tableam->multi_insert(rel, slots, nslots,
    1431             :                                   cid, options, bistate);
    1432        2444 : }
    1433             : 
    1434             : /*
    1435             :  * Delete a tuple.
    1436             :  *
    1437             :  * NB: do not call this directly unless prepared to deal with
    1438             :  * concurrent-update conditions.  Use simple_table_tuple_delete instead.
    1439             :  *
    1440             :  * Input parameters:
    1441             :  *  relation - table to be modified (caller must hold suitable lock)
    1442             :  *  tid - TID of tuple to be deleted
    1443             :  *  cid - delete command ID (used for visibility test, and stored into
    1444             :  *      cmax if successful)
    1445             :  *  crosscheck - if not InvalidSnapshot, also check tuple against this
    1446             :  *  wait - true if should wait for any conflicting update to commit/abort
    1447             :  * Output parameters:
    1448             :  *  tmfd - filled in failure cases (see below)
    1449             :  *  changingPart - true iff the tuple is being moved to another partition
    1450             :  *      table due to an update of the partition key. Otherwise, false.
    1451             :  *
    1452             :  * Normal, successful return value is TM_Ok, which means we did actually
    1453             :  * delete it.  Failure return codes are TM_SelfModified, TM_Updated, and
    1454             :  * TM_BeingModified (the last only possible if wait == false).
    1455             :  *
    1456             :  * In the failure cases, the routine fills *tmfd with the tuple's t_ctid,
    1457             :  * t_xmax, and, if possible, t_cmax.  See comments for struct
    1458             :  * TM_FailureData for additional info.
    1459             :  */
    1460             : static inline TM_Result
    1461     1726496 : table_tuple_delete(Relation rel, ItemPointer tid, CommandId cid,
    1462             :                    Snapshot snapshot, Snapshot crosscheck, bool wait,
    1463             :                    TM_FailureData *tmfd, bool changingPart)
    1464             : {
    1465     1726496 :     return rel->rd_tableam->tuple_delete(rel, tid, cid,
    1466             :                                          snapshot, crosscheck,
    1467             :                                          wait, tmfd, changingPart);
    1468             : }
    1469             : 
    1470             : /*
    1471             :  * Update a tuple.
    1472             :  *
    1473             :  * NB: do not call this directly unless you are prepared to deal with
    1474             :  * concurrent-update conditions.  Use simple_table_tuple_update instead.
    1475             :  *
    1476             :  * Input parameters:
    1477             :  *  relation - table to be modified (caller must hold suitable lock)
    1478             :  *  otid - TID of old tuple to be replaced
    1479             :  *  slot - newly constructed tuple data to store
    1480             :  *  cid - update command ID (used for visibility test, and stored into
    1481             :  *      cmax/cmin if successful)
    1482             :  *  crosscheck - if not InvalidSnapshot, also check old tuple against this
    1483             :  *  wait - true if should wait for any conflicting update to commit/abort
    1484             :  * Output parameters:
    1485             :  *  tmfd - filled in failure cases (see below)
    1486             :  *  lockmode - filled with lock mode acquired on tuple
    1487             :  *  update_indexes - in success cases this is set to true if new index entries
    1488             :  *      are required for this tuple
    1489             :  *
    1490             :  * Normal, successful return value is TM_Ok, which means we did actually
    1491             :  * update it.  Failure return codes are TM_SelfModified, TM_Updated, and
    1492             :  * TM_BeingModified (the last only possible if wait == false).
    1493             :  *
    1494             :  * On success, the slot's tts_tid and tts_tableOid are updated to match the new
    1495             :  * stored tuple; in particular, slot->tts_tid is set to the TID where the
    1496             :  * new tuple was inserted, and its HEAP_ONLY_TUPLE flag is set iff a HOT
    1497             :  * update was done.  However, any TOAST changes in the new tuple's
    1498             :  * data are not reflected into *newtup.
    1499             :  *
    1500             :  * In the failure cases, the routine fills *tmfd with the tuple's t_ctid,
    1501             :  * t_xmax, and, if possible, t_cmax.  See comments for struct TM_FailureData
    1502             :  * for additional info.
    1503             :  */
    1504             : static inline TM_Result
    1505      385484 : table_tuple_update(Relation rel, ItemPointer otid, TupleTableSlot *slot,
    1506             :                    CommandId cid, Snapshot snapshot, Snapshot crosscheck,
    1507             :                    bool wait, TM_FailureData *tmfd, LockTupleMode *lockmode,
    1508             :                    TU_UpdateIndexes *update_indexes)
    1509             : {
    1510      385484 :     return rel->rd_tableam->tuple_update(rel, otid, slot,
    1511             :                                          cid, snapshot, crosscheck,
    1512             :                                          wait, tmfd,
    1513             :                                          lockmode, update_indexes);
    1514             : }
    1515             : 
    1516             : /*
    1517             :  * Lock a tuple in the specified mode.
    1518             :  *
    1519             :  * Input parameters:
    1520             :  *  relation: relation containing tuple (caller must hold suitable lock)
    1521             :  *  tid: TID of tuple to lock
    1522             :  *  snapshot: snapshot to use for visibility determinations
    1523             :  *  cid: current command ID (used for visibility test, and stored into
    1524             :  *      tuple's cmax if lock is successful)
    1525             :  *  mode: lock mode desired
    1526             :  *  wait_policy: what to do if tuple lock is not available
    1527             :  *  flags:
    1528             :  *      If TUPLE_LOCK_FLAG_LOCK_UPDATE_IN_PROGRESS, follow the update chain to
    1529             :  *      also lock descendant tuples if lock modes don't conflict.
    1530             :  *      If TUPLE_LOCK_FLAG_FIND_LAST_VERSION, follow the update chain and lock
    1531             :  *      latest version.
    1532             :  *
    1533             :  * Output parameters:
    1534             :  *  *slot: contains the target tuple
    1535             :  *  *tmfd: filled in failure cases (see below)
    1536             :  *
    1537             :  * Function result may be:
    1538             :  *  TM_Ok: lock was successfully acquired
    1539             :  *  TM_Invisible: lock failed because tuple was never visible to us
    1540             :  *  TM_SelfModified: lock failed because tuple updated by self
    1541             :  *  TM_Updated: lock failed because tuple updated by other xact
    1542             :  *  TM_Deleted: lock failed because tuple deleted by other xact
    1543             :  *  TM_WouldBlock: lock couldn't be acquired and wait_policy is skip
    1544             :  *
    1545             :  * In the failure cases other than TM_Invisible and TM_Deleted, the routine
    1546             :  * fills *tmfd with the tuple's t_ctid, t_xmax, and, if possible, t_cmax.  See
    1547             :  * comments for struct TM_FailureData for additional info.
    1548             :  */
    1549             : static inline TM_Result
    1550      169384 : table_tuple_lock(Relation rel, ItemPointer tid, Snapshot snapshot,
    1551             :                  TupleTableSlot *slot, CommandId cid, LockTupleMode mode,
    1552             :                  LockWaitPolicy wait_policy, uint8 flags,
    1553             :                  TM_FailureData *tmfd)
    1554             : {
    1555      169384 :     return rel->rd_tableam->tuple_lock(rel, tid, snapshot, slot,
    1556             :                                        cid, mode, wait_policy,
    1557             :                                        flags, tmfd);
    1558             : }
    1559             : 
    1560             : /*
    1561             :  * Perform operations necessary to complete insertions made via
    1562             :  * tuple_insert and multi_insert with a BulkInsertState specified.
    1563             :  */
    1564             : static inline void
    1565        3996 : table_finish_bulk_insert(Relation rel, int options)
    1566             : {
    1567             :     /* optional callback */
    1568        3996 :     if (rel->rd_tableam && rel->rd_tableam->finish_bulk_insert)
    1569           0 :         rel->rd_tableam->finish_bulk_insert(rel, options);
    1570        3996 : }
    1571             : 
    1572             : 
    1573             : /* ------------------------------------------------------------------------
    1574             :  * DDL related functionality.
    1575             :  * ------------------------------------------------------------------------
    1576             :  */
    1577             : 
    1578             : /*
    1579             :  * Create storage for `rel` in `newrlocator`, with persistence set to
    1580             :  * `persistence`.
    1581             :  *
    1582             :  * This is used both during relation creation and various DDL operations to
    1583             :  * create new rel storage that can be filled from scratch.  When creating
    1584             :  * new storage for an existing relfilelocator, this should be called before the
    1585             :  * relcache entry has been updated.
    1586             :  *
    1587             :  * *freezeXid, *minmulti are set to the xid / multixact horizon for the table
    1588             :  * that pg_class.{relfrozenxid, relminmxid} have to be set to.
    1589             :  */
    1590             : static inline void
    1591       62222 : table_relation_set_new_filelocator(Relation rel,
    1592             :                                    const RelFileLocator *newrlocator,
    1593             :                                    char persistence,
    1594             :                                    TransactionId *freezeXid,
    1595             :                                    MultiXactId *minmulti)
    1596             : {
    1597       62222 :     rel->rd_tableam->relation_set_new_filelocator(rel, newrlocator,
    1598             :                                                   persistence, freezeXid,
    1599             :                                                   minmulti);
    1600       62222 : }
    1601             : 
    1602             : /*
    1603             :  * Remove all table contents from `rel`, in a non-transactional manner.
    1604             :  * Non-transactional meaning that there's no need to support rollbacks. This
    1605             :  * commonly only is used to perform truncations for relation storage created in
    1606             :  * the current transaction.
    1607             :  */
    1608             : static inline void
    1609         576 : table_relation_nontransactional_truncate(Relation rel)
    1610             : {
    1611         576 :     rel->rd_tableam->relation_nontransactional_truncate(rel);
    1612         576 : }
    1613             : 
    1614             : /*
    1615             :  * Copy data from `rel` into the new relfilelocator `newrlocator`. The new
    1616             :  * relfilelocator may not have storage associated before this function is
    1617             :  * called. This is only supposed to be used for low level operations like
    1618             :  * changing a relation's tablespace.
    1619             :  */
    1620             : static inline void
    1621          98 : table_relation_copy_data(Relation rel, const RelFileLocator *newrlocator)
    1622             : {
    1623          98 :     rel->rd_tableam->relation_copy_data(rel, newrlocator);
    1624          98 : }
    1625             : 
    1626             : /*
    1627             :  * Copy data from `OldTable` into `NewTable`, as part of a CLUSTER or VACUUM
    1628             :  * FULL.
    1629             :  *
    1630             :  * Additional Input parameters:
    1631             :  * - use_sort - if true, the table contents are sorted appropriate for
    1632             :  *   `OldIndex`; if false and OldIndex is not InvalidOid, the data is copied
    1633             :  *   in that index's order; if false and OldIndex is InvalidOid, no sorting is
    1634             :  *   performed
    1635             :  * - OldIndex - see use_sort
    1636             :  * - OldestXmin - computed by vacuum_get_cutoffs(), even when
    1637             :  *   not needed for the relation's AM
    1638             :  * - *xid_cutoff - ditto
    1639             :  * - *multi_cutoff - ditto
    1640             :  *
    1641             :  * Output parameters:
    1642             :  * - *xid_cutoff - rel's new relfrozenxid value, may be invalid
    1643             :  * - *multi_cutoff - rel's new relminmxid value, may be invalid
    1644             :  * - *tups_vacuumed - stats, for logging, if appropriate for AM
    1645             :  * - *tups_recently_dead - stats, for logging, if appropriate for AM
    1646             :  */
    1647             : static inline void
    1648         562 : table_relation_copy_for_cluster(Relation OldTable, Relation NewTable,
    1649             :                                 Relation OldIndex,
    1650             :                                 bool use_sort,
    1651             :                                 TransactionId OldestXmin,
    1652             :                                 TransactionId *xid_cutoff,
    1653             :                                 MultiXactId *multi_cutoff,
    1654             :                                 double *num_tuples,
    1655             :                                 double *tups_vacuumed,
    1656             :                                 double *tups_recently_dead)
    1657             : {
    1658         562 :     OldTable->rd_tableam->relation_copy_for_cluster(OldTable, NewTable, OldIndex,
    1659             :                                                     use_sort, OldestXmin,
    1660             :                                                     xid_cutoff, multi_cutoff,
    1661             :                                                     num_tuples, tups_vacuumed,
    1662             :                                                     tups_recently_dead);
    1663         562 : }
    1664             : 
    1665             : /*
    1666             :  * Perform VACUUM on the relation. The VACUUM can be triggered by a user or by
    1667             :  * autovacuum. The specific actions performed by the AM will depend heavily on
    1668             :  * the individual AM.
    1669             :  *
    1670             :  * On entry a transaction needs to already been established, and the
    1671             :  * table is locked with a ShareUpdateExclusive lock.
    1672             :  *
    1673             :  * Note that neither VACUUM FULL (and CLUSTER), nor ANALYZE go through this
    1674             :  * routine, even if (for ANALYZE) it is part of the same VACUUM command.
    1675             :  */
    1676             : static inline void
    1677      117782 : table_relation_vacuum(Relation rel, struct VacuumParams *params,
    1678             :                       BufferAccessStrategy bstrategy)
    1679             : {
    1680      117782 :     rel->rd_tableam->relation_vacuum(rel, params, bstrategy);
    1681      117782 : }
    1682             : 
    1683             : /*
    1684             :  * Prepare to analyze the next block in the read stream. The scan needs to
    1685             :  * have been  started with table_beginscan_analyze().  Note that this routine
    1686             :  * might acquire resources like locks that are held until
    1687             :  * table_scan_analyze_next_tuple() returns false.
    1688             :  *
    1689             :  * Returns false if block is unsuitable for sampling, true otherwise.
    1690             :  */
    1691             : static inline bool
    1692      144890 : table_scan_analyze_next_block(TableScanDesc scan, ReadStream *stream)
    1693             : {
    1694      144890 :     return scan->rs_rd->rd_tableam->scan_analyze_next_block(scan, stream);
    1695             : }
    1696             : 
    1697             : /*
    1698             :  * Iterate over tuples in the block selected with
    1699             :  * table_scan_analyze_next_block() (which needs to have returned true, and
    1700             :  * this routine may not have returned false for the same block before). If a
    1701             :  * tuple that's suitable for sampling is found, true is returned and a tuple
    1702             :  * is stored in `slot`.
    1703             :  *
    1704             :  * *liverows and *deadrows are incremented according to the encountered
    1705             :  * tuples.
    1706             :  */
    1707             : static inline bool
    1708    10285642 : table_scan_analyze_next_tuple(TableScanDesc scan, TransactionId OldestXmin,
    1709             :                               double *liverows, double *deadrows,
    1710             :                               TupleTableSlot *slot)
    1711             : {
    1712    10285642 :     return scan->rs_rd->rd_tableam->scan_analyze_next_tuple(scan, OldestXmin,
    1713             :                                                             liverows, deadrows,
    1714             :                                                             slot);
    1715             : }
    1716             : 
    1717             : /*
    1718             :  * table_index_build_scan - scan the table to find tuples to be indexed
    1719             :  *
    1720             :  * This is called back from an access-method-specific index build procedure
    1721             :  * after the AM has done whatever setup it needs.  The parent table relation
    1722             :  * is scanned to find tuples that should be entered into the index.  Each
    1723             :  * such tuple is passed to the AM's callback routine, which does the right
    1724             :  * things to add it to the new index.  After we return, the AM's index
    1725             :  * build procedure does whatever cleanup it needs.
    1726             :  *
    1727             :  * The total count of live tuples is returned.  This is for updating pg_class
    1728             :  * statistics.  (It's annoying not to be able to do that here, but we want to
    1729             :  * merge that update with others; see index_update_stats.)  Note that the
    1730             :  * index AM itself must keep track of the number of index tuples; we don't do
    1731             :  * so here because the AM might reject some of the tuples for its own reasons,
    1732             :  * such as being unable to store NULLs.
    1733             :  *
    1734             :  * If 'progress', the PROGRESS_SCAN_BLOCKS_TOTAL counter is updated when
    1735             :  * starting the scan, and PROGRESS_SCAN_BLOCKS_DONE is updated as we go along.
    1736             :  *
    1737             :  * A side effect is to set indexInfo->ii_BrokenHotChain to true if we detect
    1738             :  * any potentially broken HOT chains.  Currently, we set this if there are any
    1739             :  * RECENTLY_DEAD or DELETE_IN_PROGRESS entries in a HOT chain, without trying
    1740             :  * very hard to detect whether they're really incompatible with the chain tip.
    1741             :  * This only really makes sense for heap AM, it might need to be generalized
    1742             :  * for other AMs later.
    1743             :  */
    1744             : static inline double
    1745       51568 : table_index_build_scan(Relation table_rel,
    1746             :                        Relation index_rel,
    1747             :                        struct IndexInfo *index_info,
    1748             :                        bool allow_sync,
    1749             :                        bool progress,
    1750             :                        IndexBuildCallback callback,
    1751             :                        void *callback_state,
    1752             :                        TableScanDesc scan)
    1753             : {
    1754       51568 :     return table_rel->rd_tableam->index_build_range_scan(table_rel,
    1755             :                                                          index_rel,
    1756             :                                                          index_info,
    1757             :                                                          allow_sync,
    1758             :                                                          false,
    1759             :                                                          progress,
    1760             :                                                          0,
    1761             :                                                          InvalidBlockNumber,
    1762             :                                                          callback,
    1763             :                                                          callback_state,
    1764             :                                                          scan);
    1765             : }
    1766             : 
    1767             : /*
    1768             :  * As table_index_build_scan(), except that instead of scanning the complete
    1769             :  * table, only the given number of blocks are scanned.  Scan to end-of-rel can
    1770             :  * be signaled by passing InvalidBlockNumber as numblocks.  Note that
    1771             :  * restricting the range to scan cannot be done when requesting syncscan.
    1772             :  *
    1773             :  * When "anyvisible" mode is requested, all tuples visible to any transaction
    1774             :  * are indexed and counted as live, including those inserted or deleted by
    1775             :  * transactions that are still in progress.
    1776             :  */
    1777             : static inline double
    1778        2938 : table_index_build_range_scan(Relation table_rel,
    1779             :                              Relation index_rel,
    1780             :                              struct IndexInfo *index_info,
    1781             :                              bool allow_sync,
    1782             :                              bool anyvisible,
    1783             :                              bool progress,
    1784             :                              BlockNumber start_blockno,
    1785             :                              BlockNumber numblocks,
    1786             :                              IndexBuildCallback callback,
    1787             :                              void *callback_state,
    1788             :                              TableScanDesc scan)
    1789             : {
    1790        2938 :     return table_rel->rd_tableam->index_build_range_scan(table_rel,
    1791             :                                                          index_rel,
    1792             :                                                          index_info,
    1793             :                                                          allow_sync,
    1794             :                                                          anyvisible,
    1795             :                                                          progress,
    1796             :                                                          start_blockno,
    1797             :                                                          numblocks,
    1798             :                                                          callback,
    1799             :                                                          callback_state,
    1800             :                                                          scan);
    1801             : }
    1802             : 
    1803             : /*
    1804             :  * table_index_validate_scan - second table scan for concurrent index build
    1805             :  *
    1806             :  * See validate_index() for an explanation.
    1807             :  */
    1808             : static inline void
    1809         714 : table_index_validate_scan(Relation table_rel,
    1810             :                           Relation index_rel,
    1811             :                           struct IndexInfo *index_info,
    1812             :                           Snapshot snapshot,
    1813             :                           struct ValidateIndexState *state)
    1814             : {
    1815         714 :     table_rel->rd_tableam->index_validate_scan(table_rel,
    1816             :                                                index_rel,
    1817             :                                                index_info,
    1818             :                                                snapshot,
    1819             :                                                state);
    1820         714 : }
    1821             : 
    1822             : 
    1823             : /* ----------------------------------------------------------------------------
    1824             :  * Miscellaneous functionality
    1825             :  * ----------------------------------------------------------------------------
    1826             :  */
    1827             : 
    1828             : /*
    1829             :  * Return the current size of `rel` in bytes. If `forkNumber` is
    1830             :  * InvalidForkNumber, return the relation's overall size, otherwise the size
    1831             :  * for the indicated fork.
    1832             :  *
    1833             :  * Note that the overall size might not be the equivalent of the sum of sizes
    1834             :  * for the individual forks for some AMs, e.g. because the AMs storage does
    1835             :  * not neatly map onto the builtin types of forks.
    1836             :  */
    1837             : static inline uint64
    1838     2587312 : table_relation_size(Relation rel, ForkNumber forkNumber)
    1839             : {
    1840     2587312 :     return rel->rd_tableam->relation_size(rel, forkNumber);
    1841             : }
    1842             : 
    1843             : /*
    1844             :  * table_relation_needs_toast_table - does this relation need a toast table?
    1845             :  */
    1846             : static inline bool
    1847       42662 : table_relation_needs_toast_table(Relation rel)
    1848             : {
    1849       42662 :     return rel->rd_tableam->relation_needs_toast_table(rel);
    1850             : }
    1851             : 
    1852             : /*
    1853             :  * Return the OID of the AM that should be used to implement the TOAST table
    1854             :  * for this relation.
    1855             :  */
    1856             : static inline Oid
    1857       16966 : table_relation_toast_am(Relation rel)
    1858             : {
    1859       16966 :     return rel->rd_tableam->relation_toast_am(rel);
    1860             : }
    1861             : 
    1862             : /*
    1863             :  * Fetch all or part of a TOAST value from a TOAST table.
    1864             :  *
    1865             :  * If this AM is never used to implement a TOAST table, then this callback
    1866             :  * is not needed. But, if toasted values are ever stored in a table of this
    1867             :  * type, then you will need this callback.
    1868             :  *
    1869             :  * toastrel is the relation in which the toasted value is stored.
    1870             :  *
    1871             :  * valueid identifies which toast value is to be fetched. For the heap,
    1872             :  * this corresponds to the values stored in the chunk_id column.
    1873             :  *
    1874             :  * attrsize is the total size of the toast value to be fetched.
    1875             :  *
    1876             :  * sliceoffset is the offset within the toast value of the first byte that
    1877             :  * should be fetched.
    1878             :  *
    1879             :  * slicelength is the number of bytes from the toast value that should be
    1880             :  * fetched.
    1881             :  *
    1882             :  * result is caller-allocated space into which the fetched bytes should be
    1883             :  * stored.
    1884             :  */
    1885             : static inline void
    1886       21196 : table_relation_fetch_toast_slice(Relation toastrel, Oid valueid,
    1887             :                                  int32 attrsize, int32 sliceoffset,
    1888             :                                  int32 slicelength, struct varlena *result)
    1889             : {
    1890       21196 :     toastrel->rd_tableam->relation_fetch_toast_slice(toastrel, valueid,
    1891             :                                                      attrsize,
    1892             :                                                      sliceoffset, slicelength,
    1893             :                                                      result);
    1894       21196 : }
    1895             : 
    1896             : 
    1897             : /* ----------------------------------------------------------------------------
    1898             :  * Planner related functionality
    1899             :  * ----------------------------------------------------------------------------
    1900             :  */
    1901             : 
    1902             : /*
    1903             :  * Estimate the current size of the relation, as an AM specific workhorse for
    1904             :  * estimate_rel_size(). Look there for an explanation of the parameters.
    1905             :  */
    1906             : static inline void
    1907      428882 : table_relation_estimate_size(Relation rel, int32 *attr_widths,
    1908             :                              BlockNumber *pages, double *tuples,
    1909             :                              double *allvisfrac)
    1910             : {
    1911      428882 :     rel->rd_tableam->relation_estimate_size(rel, attr_widths, pages, tuples,
    1912             :                                             allvisfrac);
    1913      428882 : }
    1914             : 
    1915             : 
    1916             : /* ----------------------------------------------------------------------------
    1917             :  * Executor related functionality
    1918             :  * ----------------------------------------------------------------------------
    1919             :  */
    1920             : 
    1921             : /*
    1922             :  * Fetch / check / return tuples as part of a bitmap table scan. `scan` needs
    1923             :  * to have been started via table_beginscan_bm(). Fetch the next tuple of a
    1924             :  * bitmap table scan into `slot` and return true if a visible tuple was found,
    1925             :  * false otherwise.
    1926             :  *
    1927             :  * `recheck` is set by the table AM to indicate whether or not the tuple in
    1928             :  * `slot` should be rechecked. Tuples from lossy pages will always need to be
    1929             :  * rechecked, but some non-lossy pages' tuples may also require recheck.
    1930             :  *
    1931             :  * `lossy_pages` is incremented if the block's representation in the bitmap is
    1932             :  * lossy; otherwise, `exact_pages` is incremented.
    1933             :  */
    1934             : static inline bool
    1935     6652700 : table_scan_bitmap_next_tuple(TableScanDesc scan,
    1936             :                              TupleTableSlot *slot,
    1937             :                              bool *recheck,
    1938             :                              uint64 *lossy_pages,
    1939             :                              uint64 *exact_pages)
    1940             : {
    1941             :     /*
    1942             :      * We don't expect direct calls to table_scan_bitmap_next_tuple with valid
    1943             :      * CheckXidAlive for catalog or regular tables.  See detailed comments in
    1944             :      * xact.c where these variables are declared.
    1945             :      */
    1946     6652700 :     if (unlikely(TransactionIdIsValid(CheckXidAlive) && !bsysscan))
    1947           0 :         elog(ERROR, "unexpected table_scan_bitmap_next_tuple call during logical decoding");
    1948             : 
    1949     6652700 :     return scan->rs_rd->rd_tableam->scan_bitmap_next_tuple(scan,
    1950             :                                                            slot,
    1951             :                                                            recheck,
    1952             :                                                            lossy_pages,
    1953             :                                                            exact_pages);
    1954             : }
    1955             : 
    1956             : /*
    1957             :  * Prepare to fetch tuples from the next block in a sample scan. Returns false
    1958             :  * if the sample scan is finished, true otherwise. `scan` needs to have been
    1959             :  * started via table_beginscan_sampling().
    1960             :  *
    1961             :  * This will call the TsmRoutine's NextSampleBlock() callback if necessary
    1962             :  * (i.e. NextSampleBlock is not NULL), or perform a sequential scan over the
    1963             :  * underlying relation.
    1964             :  */
    1965             : static inline bool
    1966       12908 : table_scan_sample_next_block(TableScanDesc scan,
    1967             :                              struct SampleScanState *scanstate)
    1968             : {
    1969             :     /*
    1970             :      * We don't expect direct calls to table_scan_sample_next_block with valid
    1971             :      * CheckXidAlive for catalog or regular tables.  See detailed comments in
    1972             :      * xact.c where these variables are declared.
    1973             :      */
    1974       12908 :     if (unlikely(TransactionIdIsValid(CheckXidAlive) && !bsysscan))
    1975           0 :         elog(ERROR, "unexpected table_scan_sample_next_block call during logical decoding");
    1976       12908 :     return scan->rs_rd->rd_tableam->scan_sample_next_block(scan, scanstate);
    1977             : }
    1978             : 
    1979             : /*
    1980             :  * Fetch the next sample tuple into `slot` and return true if a visible tuple
    1981             :  * was found, false otherwise. table_scan_sample_next_block() needs to
    1982             :  * previously have selected a block (i.e. returned true), and no previous
    1983             :  * table_scan_sample_next_tuple() for the same block may have returned false.
    1984             :  *
    1985             :  * This will call the TsmRoutine's NextSampleTuple() callback.
    1986             :  */
    1987             : static inline bool
    1988      253892 : table_scan_sample_next_tuple(TableScanDesc scan,
    1989             :                              struct SampleScanState *scanstate,
    1990             :                              TupleTableSlot *slot)
    1991             : {
    1992             :     /*
    1993             :      * We don't expect direct calls to table_scan_sample_next_tuple with valid
    1994             :      * CheckXidAlive for catalog or regular tables.  See detailed comments in
    1995             :      * xact.c where these variables are declared.
    1996             :      */
    1997      253892 :     if (unlikely(TransactionIdIsValid(CheckXidAlive) && !bsysscan))
    1998           0 :         elog(ERROR, "unexpected table_scan_sample_next_tuple call during logical decoding");
    1999      253892 :     return scan->rs_rd->rd_tableam->scan_sample_next_tuple(scan, scanstate,
    2000             :                                                            slot);
    2001             : }
    2002             : 
    2003             : 
    2004             : /* ----------------------------------------------------------------------------
    2005             :  * Functions to make modifications a bit simpler.
    2006             :  * ----------------------------------------------------------------------------
    2007             :  */
    2008             : 
    2009             : extern void simple_table_tuple_insert(Relation rel, TupleTableSlot *slot);
    2010             : extern void simple_table_tuple_delete(Relation rel, ItemPointer tid,
    2011             :                                       Snapshot snapshot);
    2012             : extern void simple_table_tuple_update(Relation rel, ItemPointer otid,
    2013             :                                       TupleTableSlot *slot, Snapshot snapshot,
    2014             :                                       TU_UpdateIndexes *update_indexes);
    2015             : 
    2016             : 
    2017             : /* ----------------------------------------------------------------------------
    2018             :  * Helper functions to implement parallel scans for block oriented AMs.
    2019             :  * ----------------------------------------------------------------------------
    2020             :  */
    2021             : 
    2022             : extern Size table_block_parallelscan_estimate(Relation rel);
    2023             : extern Size table_block_parallelscan_initialize(Relation rel,
    2024             :                                                 ParallelTableScanDesc pscan);
    2025             : extern void table_block_parallelscan_reinitialize(Relation rel,
    2026             :                                                   ParallelTableScanDesc pscan);
    2027             : extern BlockNumber table_block_parallelscan_nextpage(Relation rel,
    2028             :                                                      ParallelBlockTableScanWorker pbscanwork,
    2029             :                                                      ParallelBlockTableScanDesc pbscan);
    2030             : extern void table_block_parallelscan_startblock_init(Relation rel,
    2031             :                                                      ParallelBlockTableScanWorker pbscanwork,
    2032             :                                                      ParallelBlockTableScanDesc pbscan);
    2033             : 
    2034             : 
    2035             : /* ----------------------------------------------------------------------------
    2036             :  * Helper functions to implement relation sizing for block oriented AMs.
    2037             :  * ----------------------------------------------------------------------------
    2038             :  */
    2039             : 
    2040             : extern uint64 table_block_relation_size(Relation rel, ForkNumber forkNumber);
    2041             : extern void table_block_relation_estimate_size(Relation rel,
    2042             :                                                int32 *attr_widths,
    2043             :                                                BlockNumber *pages,
    2044             :                                                double *tuples,
    2045             :                                                double *allvisfrac,
    2046             :                                                Size overhead_bytes_per_tuple,
    2047             :                                                Size usable_bytes_per_page);
    2048             : 
    2049             : /* ----------------------------------------------------------------------------
    2050             :  * Functions in tableamapi.c
    2051             :  * ----------------------------------------------------------------------------
    2052             :  */
    2053             : 
    2054             : extern const TableAmRoutine *GetTableAmRoutine(Oid amhandler);
    2055             : 
    2056             : /* ----------------------------------------------------------------------------
    2057             :  * Functions in heapam_handler.c
    2058             :  * ----------------------------------------------------------------------------
    2059             :  */
    2060             : 
    2061             : extern const TableAmRoutine *GetHeapamTableAmRoutine(void);
    2062             : 
    2063             : #endif                          /* TABLEAM_H */

Generated by: LCOV version 1.14