index.c

PostgreSQL7.4.6 for Linux

 * here, we just prepare its input arrays datum[] and nullv[].
 * ----------------
 */
void
FormIndexDatum(IndexInfo *indexInfo,
			   HeapTuple heapTuple,
			   TupleDesc heapDescriptor,
			   EState *estate,
			   Datum *datum,
			   char *nullv)
{
	List	   *indexprs;
	int			i;

	if (indexInfo->ii_Expressions != NIL &&
		indexInfo->ii_ExpressionsState == NIL)
	{
		/* First time through, set up expression evaluation state */
		indexInfo->ii_ExpressionsState = (List *)
			ExecPrepareExpr((Expr *) indexInfo->ii_Expressions,
							estate);
		/* Check caller has set up context correctly */
		Assert(GetPerTupleExprContext(estate)->ecxt_scantuple->val == heapTuple);
	}
	indexprs = indexInfo->ii_ExpressionsState;

	for (i = 0; i < indexInfo->ii_NumIndexAttrs; i++)
	{
		int			keycol = indexInfo->ii_KeyAttrNumbers[i];
		Datum		iDatum;
		bool		isNull;

		if (keycol != 0)
		{
			/*
			 * Plain index column; get the value we need directly from the
			 * heap tuple.
			 */
			iDatum = heap_getattr(heapTuple, keycol, heapDescriptor, &isNull);
		}
		else
		{
			/*
			 * Index expression --- need to evaluate it.
			 */
			if (indexprs == NIL)
				elog(ERROR, "wrong number of index expressions");
			iDatum = ExecEvalExprSwitchContext((ExprState *) lfirst(indexprs),
										  GetPerTupleExprContext(estate),
											   &isNull,
											   NULL);
			indexprs = lnext(indexprs);
		}
		datum[i] = iDatum;
		nullv[i] = (isNull) ? 'n' : ' ';
	}

	if (indexprs != NIL)
		elog(ERROR, "wrong number of index expressions");
}


/* ----------------
 *		set relhasindex of relation's pg_class entry
 *
 * If isprimary is TRUE, we are defining a primary index, so also set
 * relhaspkey to TRUE.	Otherwise, leave relhaspkey alone.
 *
 * If reltoastidxid is not InvalidOid, also set reltoastidxid to that value.
 * This is only used for TOAST relations.
 *
 * NOTE: an important side-effect of this operation is that an SI invalidation
 * message is sent out to all backends --- including me --- causing relcache
 * entries to be flushed or updated with the new hasindex data.  This must
 * happen even if we find that no change is needed in the pg_class row.
 * ----------------
 */
void
setRelhasindex(Oid relid, bool hasindex, bool isprimary, Oid reltoastidxid)
{
	Relation	pg_class;
	HeapTuple	tuple;
	Form_pg_class classtuple;
	bool		dirty = false;
	HeapScanDesc pg_class_scan = NULL;

	/*
	 * Find the tuple to update in pg_class.  In bootstrap mode we can't
	 * use heap_update, so cheat and overwrite the tuple in-place.  In
	 * normal processing, make a copy to scribble on.
	 */
	pg_class = heap_openr(RelationRelationName, RowExclusiveLock);

	if (!IsBootstrapProcessingMode())
	{
		tuple = SearchSysCacheCopy(RELOID,
								   ObjectIdGetDatum(relid),
								   0, 0, 0);
	}
	else
	{
		ScanKeyData key[1];

		ScanKeyEntryInitialize(&key[0], 0,
							   ObjectIdAttributeNumber,
							   F_OIDEQ,
							   ObjectIdGetDatum(relid));

		pg_class_scan = heap_beginscan(pg_class, SnapshotNow, 1, key);
		tuple = heap_getnext(pg_class_scan, ForwardScanDirection);
	}

	if (!HeapTupleIsValid(tuple))
		elog(ERROR, "could not find tuple for relation %u", relid);
	classtuple = (Form_pg_class) GETSTRUCT(tuple);

	/* Apply required updates */

	if (pg_class_scan)
		LockBuffer(pg_class_scan->rs_cbuf, BUFFER_LOCK_EXCLUSIVE);

	if (classtuple->relhasindex != hasindex)
	{
		classtuple->relhasindex = hasindex;
		dirty = true;
	}
	if (isprimary)
	{
		if (!classtuple->relhaspkey)
		{
			classtuple->relhaspkey = true;
			dirty = true;
		}
	}
	if (OidIsValid(reltoastidxid))
	{
		Assert(classtuple->relkind == RELKIND_TOASTVALUE);
		if (classtuple->reltoastidxid != reltoastidxid)
		{
			classtuple->reltoastidxid = reltoastidxid;
			dirty = true;
		}
	}

	if (pg_class_scan)
		LockBuffer(pg_class_scan->rs_cbuf, BUFFER_LOCK_UNLOCK);

	if (pg_class_scan)
	{
		/* Write the modified tuple in-place */
		WriteNoReleaseBuffer(pg_class_scan->rs_cbuf);
		/* Send out shared cache inval if necessary */
		if (!IsBootstrapProcessingMode())
			CacheInvalidateHeapTuple(pg_class, tuple);
		BufferSync();
	}
	else if (dirty)
	{
		simple_heap_update(pg_class, &tuple->t_self, tuple);

		/* Keep the catalog indexes up to date */
		CatalogUpdateIndexes(pg_class, tuple);
	}
	else
	{
		/* no need to change tuple, but force relcache rebuild anyway */
		CacheInvalidateRelcache(relid);
	}

	if (!pg_class_scan)
		heap_freetuple(tuple);
	else
		heap_endscan(pg_class_scan);

	heap_close(pg_class, RowExclusiveLock);
}

/*
 * setNewRelfilenode		- assign a new relfilenode value to the relation
 *
 * Caller must already hold exclusive lock on the relation.
 */
void
setNewRelfilenode(Relation relation)
{
	Oid			newrelfilenode;
	Relation	pg_class;
	HeapTuple	tuple;
	Form_pg_class rd_rel;
	RelationData workrel;

	/* Can't change relfilenode for nailed tables (indexes ok though) */
	Assert(!relation->rd_isnailed ||
		   relation->rd_rel->relkind == RELKIND_INDEX);
	/* Can't change for shared tables or indexes */
	Assert(!relation->rd_rel->relisshared);

	/* Allocate a new relfilenode */
	newrelfilenode = newoid();

	/*
	 * Find the pg_class tuple for the given relation.  This is not used
	 * during bootstrap, so okay to use heap_update always.
	 */
	pg_class = heap_openr(RelationRelationName, RowExclusiveLock);

	tuple = SearchSysCacheCopy(RELOID,
							   ObjectIdGetDatum(RelationGetRelid(relation)),
							   0, 0, 0);
	if (!HeapTupleIsValid(tuple))
		elog(ERROR, "could not find tuple for relation %u",
			 RelationGetRelid(relation));
	rd_rel = (Form_pg_class) GETSTRUCT(tuple);

	/* create another storage file. Is it a little ugly ? */
	/* NOTE: any conflict in relfilenode value will be caught here */
	memcpy((char *) &workrel, relation, sizeof(RelationData));
	workrel.rd_fd = -1;
	workrel.rd_node.relNode = newrelfilenode;
	heap_storage_create(&workrel);
	smgrclose(DEFAULT_SMGR, &workrel);

	/* schedule unlinking old relfilenode */
	smgrunlink(DEFAULT_SMGR, relation);

	/* update the pg_class row */
	rd_rel->relfilenode = newrelfilenode;
	simple_heap_update(pg_class, &tuple->t_self, tuple);
	CatalogUpdateIndexes(pg_class, tuple);

	heap_freetuple(tuple);

	heap_close(pg_class, RowExclusiveLock);

	/* Make sure the relfilenode change is visible */
	CommandCounterIncrement();
}


/* ----------------
 *		UpdateStats
 *
 * Update pg_class' relpages and reltuples statistics for the given relation
 * (which can be either a table or an index).  Note that this is not used
 * in the context of VACUUM.
 * ----------------
 */
void
UpdateStats(Oid relid, double reltuples)
{
	Relation	whichRel;
	Relation	pg_class;
	HeapTuple	tuple;
	BlockNumber relpages;
	Form_pg_class rd_rel;
	HeapScanDesc pg_class_scan = NULL;
	bool		in_place_upd;

	/*
	 * This routine handles updates for both the heap and index relation
	 * statistics.	In order to guarantee that we're able to *see* the
	 * index relation tuple, we bump the command counter id here.  The
	 * index relation tuple was created in the current transaction.
	 */
	CommandCounterIncrement();

	/*
	 * CommandCounterIncrement() flushes invalid cache entries, including
	 * those for the heap and index relations for which we're updating
	 * statistics.	Now that the cache is flushed, it's safe to open the
	 * relation again.	We need the relation open in order to figure out
	 * how many blocks it contains.
	 */

	/*
	 * Grabbing lock here is probably redundant ...
	 */
	whichRel = relation_open(relid, ShareLock);

	/*
	 * Find the tuple to update in pg_class.  Normally we make a copy of
	 * the tuple using the syscache, modify it, and apply heap_update.
	 * But in bootstrap mode we can't use heap_update, so we cheat and
	 * overwrite the tuple in-place.
	 *
	 * We also must cheat if reindexing pg_class itself, because the
	 * target index may presently not be part of the set of indexes that
	 * CatalogUpdateIndexes would update (see reindex_relation).  In this
	 * case the stats updates will not be WAL-logged and so could be lost
	 * in a crash.  This seems OK considering VACUUM does the same thing.
	 */
	pg_class = heap_openr(RelationRelationName, RowExclusiveLock);

	in_place_upd = IsBootstrapProcessingMode() ||
		ReindexIsProcessingHeap(RelationGetRelid(pg_class));

	if (!in_place_upd)
	{
		tuple = SearchSysCacheCopy(RELOID,
								   ObjectIdGetDatum(relid),
								   0, 0, 0);
	}
	else
	{
		ScanKeyData key[1];

		ScanKeyEntryInitialize(&key[0], 0,
							   ObjectIdAttributeNumber,
							   F_OIDEQ,
							   ObjectIdGetDatum(relid));

		pg_class_scan = heap_beginscan(pg_class, SnapshotNow, 1, key);
		tuple = heap_getnext(pg_class_scan, ForwardScanDirection);
	}

	if (!HeapTupleIsValid(tuple))
		elog(ERROR, "could not find tuple for relation %u", relid);
	rd_rel = (Form_pg_class) GETSTRUCT(tuple);

	/*
	 * Figure values to insert.
	 *
	 * If we found zero tuples in the scan, do NOT believe it; instead put a
	 * bogus estimate into the statistics fields.  Otherwise, the common
	 * pattern "CREATE TABLE; CREATE INDEX; insert data" leaves the table
	 * with zero size statistics until a VACUUM is done.  The optimizer
	 * will generate very bad plans if the stats claim the table is empty
	 * when it is actually sizable.  See also CREATE TABLE in heap.c.
	 *
	 * Note: this path is also taken during bootstrap, because bootstrap.c
	 * passes reltuples = 0 after loading a table.	We have to estimate
	 * some number for reltuples based on the actual number of pages.
	 */
	relpages = RelationGetNumberOfBlocks(whichRel);

	if (reltuples == 0)
	{
		if (relpages == 0)
		{
			/* Bogus defaults for a virgin table, same as heap.c */
			reltuples = 1000;
			relpages = 10;
		}
		else if (whichRel->rd_rel->relkind == RELKIND_INDEX && relpages <= 2)
		{
			/* Empty index, leave bogus defaults in place */
			reltuples = 1000;
		}
		else
			reltuples = ((double) relpages) * NTUPLES_PER_PAGE(whichRel->rd_rel->relnatts);
	}

	/*
	 * Update statistics in pg_class, if they changed.	(Avoiding an
	 * unnecessary update is not just a tiny performance improvement; it
	 * also reduces the window wherein concurrent CREATE INDEX commands
	 * may conflict.)
	 */
	if (rd_rel->relpages != (int32) relpages ||
		rd_rel->reltuples != (float4) reltuples)
	{
		if (in_place_upd)
		{
			/* Bootstrap or reindex case: overwrite fields in place. */
			LockBuffer(pg_class_scan->rs_cbuf, BUFFER_LOCK_EXCLUSIVE);
			rd_rel->relpages = (int32) relpages;
			rd_rel->reltuples = (float4) reltuples;
			LockBuffer(pg_class_scan->rs_cbuf, BUFFER_LOCK_UNLOCK);
			WriteNoReleaseBuffer(pg_class_scan->rs_cbuf);
			if (!IsBootstrapProcessingMode())
				CacheInvalidateHeapTuple(pg_class, tuple);
		}
		else
		{
			/* During normal processing, must work harder. */
			rd_rel->relpages = (int32) relpages;
			rd_rel->reltuples = (float4) reltuples;
			simple_heap_update(pg_class, &tuple->t_self, tuple);
			CatalogUpdateIndexes(pg_class, tuple);
		}
	}

	if (!pg_class_scan)
		heap_freetuple(tuple);
	else
		heap_endscan(pg_class_scan);

	/*
	 * We shouldn't have to do this, but we do...  Modify the reldesc in
	 * place with the new values so that the cache contains the latest
	 * copy.  (XXX is this really still necessary?	The relcache will get
	 * fixed at next CommandCounterIncrement, so why bother here?)
	 */
	whichRel->rd_rel->relpages = (int32) relpages;
	whichRel->rd_rel->reltuples = (float4) reltuples;

	heap_close(pg_class, RowExclusiveLock);
	relation_close(whichRel, NoLock);
}


/*
 * index_build - invoke access-method-specific index build procedure
 */
void
index_build(Relation heapRelation,
			Relation indexRelation,
			IndexInfo *indexInfo)
{
	RegProcedure procedure;

	/*
	 * sanity checks
	 */
	Assert(RelationIsValid(indexRelation));
	Assert(PointerIsValid(indexRelation->rd_am));

	procedure = indexRelation->rd_am->ambuild;
	Assert(RegProcedureIsValid(procedure));

	/*
	 * Call the access method's build procedure
	 */
	OidFunctionCall3(procedure,
					 PointerGetDatum(heapRelation),
					 PointerGetDatum(indexRelation),
					 PointerGetDatum(indexInfo));
}


/*
 * IndexBuildHeapScan - scan the heap relation to find tuples to be indexed
 *
 * This is called back from an access-method-specific index build procedure
 * after the AM has done whatever setup it needs.  The parent heap relation
 * is scanned to find tuples that should be entered into the index.  Each
 * such tuple is passed to the AM's callback routine, which does the right
 * things to add it to the new index.  After we return, the AM's index
 * build procedure does whatever cleanup is needed; in particular, it should
 * close the heap and index relations.
 *
 * The total count of heap tuples is returned.	This is for updating pg_class
 * statistics.	(It's annoying not to be able to do that here, but we can't
 * do it until after the relation is closed.)  Note that the index AM itself
 * must keep track of the number of index tuples; we don't do so here because
 * the AM might reject some of the tuples for its own reasons, such as being