2008-10-04 23:56:55 +02:00
|
|
|
/*-------------------------------------------------------------------------
|
|
|
|
*
|
|
|
|
* nodeCtescan.c
|
|
|
|
* routines to handle CteScan nodes.
|
|
|
|
*
|
2018-01-03 05:30:12 +01:00
|
|
|
* Portions Copyright (c) 1996-2018, PostgreSQL Global Development Group
|
2008-10-04 23:56:55 +02:00
|
|
|
* Portions Copyright (c) 1994, Regents of the University of California
|
|
|
|
*
|
|
|
|
*
|
|
|
|
* IDENTIFICATION
|
2010-09-20 22:08:53 +02:00
|
|
|
* src/backend/executor/nodeCtescan.c
|
2008-10-04 23:56:55 +02:00
|
|
|
*
|
|
|
|
*-------------------------------------------------------------------------
|
|
|
|
*/
|
|
|
|
|
|
|
|
#include "postgres.h"
|
|
|
|
|
|
|
|
#include "executor/execdebug.h"
|
|
|
|
#include "executor/nodeCtescan.h"
|
|
|
|
#include "miscadmin.h"
|
|
|
|
|
|
|
|
static TupleTableSlot *CteScanNext(CteScanState *node);
|
|
|
|
|
|
|
|
/* ----------------------------------------------------------------
|
|
|
|
* CteScanNext
|
|
|
|
*
|
|
|
|
* This is a workhorse for ExecCteScan
|
|
|
|
* ----------------------------------------------------------------
|
|
|
|
*/
|
|
|
|
static TupleTableSlot *
|
|
|
|
CteScanNext(CteScanState *node)
|
|
|
|
{
|
|
|
|
EState *estate;
|
|
|
|
ScanDirection dir;
|
|
|
|
bool forward;
|
|
|
|
Tuplestorestate *tuplestorestate;
|
|
|
|
bool eof_tuplestore;
|
|
|
|
TupleTableSlot *slot;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* get state info from node
|
|
|
|
*/
|
|
|
|
estate = node->ss.ps.state;
|
|
|
|
dir = estate->es_direction;
|
|
|
|
forward = ScanDirectionIsForward(dir);
|
|
|
|
tuplestorestate = node->leader->cte_table;
|
|
|
|
tuplestore_select_read_pointer(tuplestorestate, node->readptr);
|
|
|
|
slot = node->ss.ss_ScanTupleSlot;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If we are not at the end of the tuplestore, or are going backwards, try
|
|
|
|
* to fetch a tuple from tuplestore.
|
|
|
|
*/
|
|
|
|
eof_tuplestore = tuplestore_ateof(tuplestorestate);
|
|
|
|
|
|
|
|
if (!forward && eof_tuplestore)
|
|
|
|
{
|
|
|
|
if (!node->leader->eof_cte)
|
|
|
|
{
|
|
|
|
/*
|
|
|
|
* When reversing direction at tuplestore EOF, the first
|
|
|
|
* gettupleslot call will fetch the last-added tuple; but we want
|
|
|
|
* to return the one before that, if possible. So do an extra
|
|
|
|
* fetch.
|
|
|
|
*/
|
|
|
|
if (!tuplestore_advance(tuplestorestate, forward))
|
|
|
|
return NULL; /* the tuplestore must be empty */
|
|
|
|
}
|
|
|
|
eof_tuplestore = false;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If we can fetch another tuple from the tuplestore, return it.
|
2009-03-27 19:30:21 +01:00
|
|
|
*
|
|
|
|
* Note: we have to use copy=true in the tuplestore_gettupleslot call,
|
2009-06-11 16:49:15 +02:00
|
|
|
* because we are sharing the tuplestore with other nodes that might write
|
|
|
|
* into the tuplestore before we get called again.
|
2008-10-04 23:56:55 +02:00
|
|
|
*/
|
|
|
|
if (!eof_tuplestore)
|
|
|
|
{
|
2009-03-27 19:30:21 +01:00
|
|
|
if (tuplestore_gettupleslot(tuplestorestate, forward, true, slot))
|
2008-10-04 23:56:55 +02:00
|
|
|
return slot;
|
|
|
|
if (forward)
|
|
|
|
eof_tuplestore = true;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If necessary, try to fetch another row from the CTE query.
|
|
|
|
*
|
|
|
|
* Note: the eof_cte state variable exists to short-circuit further calls
|
|
|
|
* of the CTE plan. It's not optional, unfortunately, because some plan
|
|
|
|
* node types are not robust about being called again when they've already
|
|
|
|
* returned NULL.
|
|
|
|
*/
|
|
|
|
if (eof_tuplestore && !node->leader->eof_cte)
|
|
|
|
{
|
|
|
|
TupleTableSlot *cteslot;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* We can only get here with forward==true, so no need to worry about
|
|
|
|
* which direction the subplan will go.
|
|
|
|
*/
|
|
|
|
cteslot = ExecProcNode(node->cteplanstate);
|
|
|
|
if (TupIsNull(cteslot))
|
|
|
|
{
|
|
|
|
node->leader->eof_cte = true;
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Append a copy of the returned tuple to tuplestore. NOTE: because
|
|
|
|
* our read pointer is certainly in EOF state, its read position will
|
|
|
|
* move forward over the added tuple. This is what we want. Also,
|
2009-06-11 16:49:15 +02:00
|
|
|
* any other readers will *not* move past the new tuple, which is what
|
|
|
|
* they want.
|
2008-10-04 23:56:55 +02:00
|
|
|
*/
|
|
|
|
tuplestore_puttupleslot(tuplestorestate, cteslot);
|
|
|
|
|
|
|
|
/*
|
2009-06-11 16:49:15 +02:00
|
|
|
* We MUST copy the CTE query's output tuple into our own slot. This
|
|
|
|
* is because other CteScan nodes might advance the CTE query before
|
|
|
|
* we are called again, and our output tuple must stay stable over
|
|
|
|
* that.
|
2008-10-04 23:56:55 +02:00
|
|
|
*/
|
|
|
|
return ExecCopySlot(slot, cteslot);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Nothing left ...
|
|
|
|
*/
|
|
|
|
return ExecClearTuple(slot);
|
|
|
|
}
|
|
|
|
|
Re-implement EvalPlanQual processing to improve its performance and eliminate
a lot of strange behaviors that occurred in join cases. We now identify the
"current" row for every joined relation in UPDATE, DELETE, and SELECT FOR
UPDATE/SHARE queries. If an EvalPlanQual recheck is necessary, we jam the
appropriate row into each scan node in the rechecking plan, forcing it to emit
only that one row. The former behavior could rescan the whole of each joined
relation for each recheck, which was terrible for performance, and what's much
worse could result in duplicated output tuples.
Also, the original implementation of EvalPlanQual could not re-use the recheck
execution tree --- it had to go through a full executor init and shutdown for
every row to be tested. To avoid this overhead, I've associated a special
runtime Param with each LockRows or ModifyTable plan node, and arranged to
make every scan node below such a node depend on that Param. Thus, by
signaling a change in that Param, the EPQ machinery can just rescan the
already-built test plan.
This patch also adds a prohibition on set-returning functions in the
targetlist of SELECT FOR UPDATE/SHARE. This is needed to avoid the
duplicate-output-tuple problem. It seems fairly reasonable since the
other restrictions on SELECT FOR UPDATE are meant to ensure that there
is a unique correspondence between source tuples and result tuples,
which an output SRF destroys as much as anything else does.
2009-10-26 03:26:45 +01:00
|
|
|
/*
|
|
|
|
* CteScanRecheck -- access method routine to recheck a tuple in EvalPlanQual
|
|
|
|
*/
|
|
|
|
static bool
|
|
|
|
CteScanRecheck(CteScanState *node, TupleTableSlot *slot)
|
|
|
|
{
|
|
|
|
/* nothing to check */
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
2008-10-04 23:56:55 +02:00
|
|
|
/* ----------------------------------------------------------------
|
|
|
|
* ExecCteScan(node)
|
|
|
|
*
|
|
|
|
* Scans the CTE sequentially and returns the next qualifying tuple.
|
Re-implement EvalPlanQual processing to improve its performance and eliminate
a lot of strange behaviors that occurred in join cases. We now identify the
"current" row for every joined relation in UPDATE, DELETE, and SELECT FOR
UPDATE/SHARE queries. If an EvalPlanQual recheck is necessary, we jam the
appropriate row into each scan node in the rechecking plan, forcing it to emit
only that one row. The former behavior could rescan the whole of each joined
relation for each recheck, which was terrible for performance, and what's much
worse could result in duplicated output tuples.
Also, the original implementation of EvalPlanQual could not re-use the recheck
execution tree --- it had to go through a full executor init and shutdown for
every row to be tested. To avoid this overhead, I've associated a special
runtime Param with each LockRows or ModifyTable plan node, and arranged to
make every scan node below such a node depend on that Param. Thus, by
signaling a change in that Param, the EPQ machinery can just rescan the
already-built test plan.
This patch also adds a prohibition on set-returning functions in the
targetlist of SELECT FOR UPDATE/SHARE. This is needed to avoid the
duplicate-output-tuple problem. It seems fairly reasonable since the
other restrictions on SELECT FOR UPDATE are meant to ensure that there
is a unique correspondence between source tuples and result tuples,
which an output SRF destroys as much as anything else does.
2009-10-26 03:26:45 +01:00
|
|
|
* We call the ExecScan() routine and pass it the appropriate
|
|
|
|
* access method functions.
|
2008-10-04 23:56:55 +02:00
|
|
|
* ----------------------------------------------------------------
|
|
|
|
*/
|
2017-07-17 09:33:49 +02:00
|
|
|
static TupleTableSlot *
|
|
|
|
ExecCteScan(PlanState *pstate)
|
2008-10-04 23:56:55 +02:00
|
|
|
{
|
2017-07-17 09:33:49 +02:00
|
|
|
CteScanState *node = castNode(CteScanState, pstate);
|
|
|
|
|
Re-implement EvalPlanQual processing to improve its performance and eliminate
a lot of strange behaviors that occurred in join cases. We now identify the
"current" row for every joined relation in UPDATE, DELETE, and SELECT FOR
UPDATE/SHARE queries. If an EvalPlanQual recheck is necessary, we jam the
appropriate row into each scan node in the rechecking plan, forcing it to emit
only that one row. The former behavior could rescan the whole of each joined
relation for each recheck, which was terrible for performance, and what's much
worse could result in duplicated output tuples.
Also, the original implementation of EvalPlanQual could not re-use the recheck
execution tree --- it had to go through a full executor init and shutdown for
every row to be tested. To avoid this overhead, I've associated a special
runtime Param with each LockRows or ModifyTable plan node, and arranged to
make every scan node below such a node depend on that Param. Thus, by
signaling a change in that Param, the EPQ machinery can just rescan the
already-built test plan.
This patch also adds a prohibition on set-returning functions in the
targetlist of SELECT FOR UPDATE/SHARE. This is needed to avoid the
duplicate-output-tuple problem. It seems fairly reasonable since the
other restrictions on SELECT FOR UPDATE are meant to ensure that there
is a unique correspondence between source tuples and result tuples,
which an output SRF destroys as much as anything else does.
2009-10-26 03:26:45 +01:00
|
|
|
return ExecScan(&node->ss,
|
|
|
|
(ExecScanAccessMtd) CteScanNext,
|
|
|
|
(ExecScanRecheckMtd) CteScanRecheck);
|
2008-10-04 23:56:55 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/* ----------------------------------------------------------------
|
|
|
|
* ExecInitCteScan
|
|
|
|
* ----------------------------------------------------------------
|
|
|
|
*/
|
|
|
|
CteScanState *
|
|
|
|
ExecInitCteScan(CteScan *node, EState *estate, int eflags)
|
|
|
|
{
|
|
|
|
CteScanState *scanstate;
|
|
|
|
ParamExecData *prmdata;
|
|
|
|
|
|
|
|
/* check for unsupported flags */
|
|
|
|
Assert(!(eflags & EXEC_FLAG_MARK));
|
|
|
|
|
|
|
|
/*
|
|
|
|
* For the moment we have to force the tuplestore to allow REWIND, because
|
|
|
|
* we might be asked to rescan the CTE even though upper levels didn't
|
|
|
|
* tell us to be prepared to do it efficiently. Annoying, since this
|
|
|
|
* prevents truncation of the tuplestore. XXX FIXME
|
|
|
|
*/
|
|
|
|
eflags |= EXEC_FLAG_REWIND;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* CteScan should not have any children.
|
|
|
|
*/
|
|
|
|
Assert(outerPlan(node) == NULL);
|
|
|
|
Assert(innerPlan(node) == NULL);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* create new CteScanState for node
|
|
|
|
*/
|
|
|
|
scanstate = makeNode(CteScanState);
|
|
|
|
scanstate->ss.ps.plan = (Plan *) node;
|
|
|
|
scanstate->ss.ps.state = estate;
|
2017-07-17 09:33:49 +02:00
|
|
|
scanstate->ss.ps.ExecProcNode = ExecCteScan;
|
2008-10-04 23:56:55 +02:00
|
|
|
scanstate->eflags = eflags;
|
|
|
|
scanstate->cte_table = NULL;
|
|
|
|
scanstate->eof_cte = false;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Find the already-initialized plan for the CTE query.
|
|
|
|
*/
|
|
|
|
scanstate->cteplanstate = (PlanState *) list_nth(estate->es_subplanstates,
|
|
|
|
node->ctePlanId - 1);
|
|
|
|
|
|
|
|
/*
|
2009-06-11 16:49:15 +02:00
|
|
|
* The Param slot associated with the CTE query is used to hold a pointer
|
|
|
|
* to the CteState of the first CteScan node that initializes for this
|
|
|
|
* CTE. This node will be the one that holds the shared state for all the
|
2012-08-16 01:01:16 +02:00
|
|
|
* CTEs, particularly the shared tuplestore.
|
2008-10-04 23:56:55 +02:00
|
|
|
*/
|
|
|
|
prmdata = &(estate->es_param_exec_vals[node->cteParam]);
|
|
|
|
Assert(prmdata->execPlan == NULL);
|
|
|
|
Assert(!prmdata->isnull);
|
2017-01-27 01:47:03 +01:00
|
|
|
scanstate->leader = castNode(CteScanState, DatumGetPointer(prmdata->value));
|
2008-10-04 23:56:55 +02:00
|
|
|
if (scanstate->leader == NULL)
|
|
|
|
{
|
|
|
|
/* I am the leader */
|
|
|
|
prmdata->value = PointerGetDatum(scanstate);
|
|
|
|
scanstate->leader = scanstate;
|
|
|
|
scanstate->cte_table = tuplestore_begin_heap(true, false, work_mem);
|
|
|
|
tuplestore_set_eflags(scanstate->cte_table, scanstate->eflags);
|
|
|
|
scanstate->readptr = 0;
|
|
|
|
}
|
|
|
|
else
|
|
|
|
{
|
|
|
|
/* Not the leader */
|
2016-09-22 17:34:44 +02:00
|
|
|
/* Create my own read pointer, and ensure it is at start */
|
2008-10-04 23:56:55 +02:00
|
|
|
scanstate->readptr =
|
|
|
|
tuplestore_alloc_read_pointer(scanstate->leader->cte_table,
|
|
|
|
scanstate->eflags);
|
2016-09-22 17:34:44 +02:00
|
|
|
tuplestore_select_read_pointer(scanstate->leader->cte_table,
|
|
|
|
scanstate->readptr);
|
|
|
|
tuplestore_rescan(scanstate->leader->cte_table);
|
2008-10-04 23:56:55 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Miscellaneous initialization
|
|
|
|
*
|
|
|
|
* create expression context for node
|
|
|
|
*/
|
|
|
|
ExecAssignExprContext(estate, &scanstate->ss.ps);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* The scan tuple type (ie, the rowtype we expect to find in the work
|
|
|
|
* table) is the same as the result rowtype of the CTE query.
|
|
|
|
*/
|
2018-02-17 06:17:38 +01:00
|
|
|
ExecInitScanTupleSlot(estate, &scanstate->ss,
|
|
|
|
ExecGetResultType(scanstate->cteplanstate));
|
2008-10-04 23:56:55 +02:00
|
|
|
|
|
|
|
/*
|
2018-02-17 06:17:38 +01:00
|
|
|
* Initialize result slot, type and projection.
|
2008-10-04 23:56:55 +02:00
|
|
|
*/
|
2018-02-17 06:17:38 +01:00
|
|
|
ExecInitResultTupleSlotTL(estate, &scanstate->ss.ps);
|
2008-10-04 23:56:55 +02:00
|
|
|
ExecAssignScanProjectionInfo(&scanstate->ss);
|
|
|
|
|
2018-02-17 06:17:38 +01:00
|
|
|
/*
|
|
|
|
* initialize child expressions
|
|
|
|
*/
|
|
|
|
scanstate->ss.ps.qual =
|
|
|
|
ExecInitQual(node->scan.plan.qual, (PlanState *) scanstate);
|
|
|
|
|
2008-10-04 23:56:55 +02:00
|
|
|
return scanstate;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* ----------------------------------------------------------------
|
|
|
|
* ExecEndCteScan
|
|
|
|
*
|
|
|
|
* frees any storage allocated through C routines.
|
|
|
|
* ----------------------------------------------------------------
|
|
|
|
*/
|
|
|
|
void
|
|
|
|
ExecEndCteScan(CteScanState *node)
|
|
|
|
{
|
|
|
|
/*
|
|
|
|
* Free exprcontext
|
|
|
|
*/
|
|
|
|
ExecFreeExprContext(&node->ss.ps);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* clean out the tuple table
|
|
|
|
*/
|
|
|
|
ExecClearTuple(node->ss.ps.ps_ResultTupleSlot);
|
|
|
|
ExecClearTuple(node->ss.ss_ScanTupleSlot);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If I am the leader, free the tuplestore.
|
|
|
|
*/
|
|
|
|
if (node->leader == node)
|
2012-08-16 01:01:16 +02:00
|
|
|
{
|
2008-10-04 23:56:55 +02:00
|
|
|
tuplestore_end(node->cte_table);
|
2012-08-16 01:01:16 +02:00
|
|
|
node->cte_table = NULL;
|
|
|
|
}
|
2008-10-04 23:56:55 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/* ----------------------------------------------------------------
|
2010-07-12 19:01:06 +02:00
|
|
|
* ExecReScanCteScan
|
2008-10-04 23:56:55 +02:00
|
|
|
*
|
|
|
|
* Rescans the relation.
|
|
|
|
* ----------------------------------------------------------------
|
|
|
|
*/
|
|
|
|
void
|
2010-07-12 19:01:06 +02:00
|
|
|
ExecReScanCteScan(CteScanState *node)
|
2008-10-04 23:56:55 +02:00
|
|
|
{
|
2008-10-23 17:29:23 +02:00
|
|
|
Tuplestorestate *tuplestorestate = node->leader->cte_table;
|
2008-10-04 23:56:55 +02:00
|
|
|
|
|
|
|
ExecClearTuple(node->ss.ps.ps_ResultTupleSlot);
|
Re-implement EvalPlanQual processing to improve its performance and eliminate
a lot of strange behaviors that occurred in join cases. We now identify the
"current" row for every joined relation in UPDATE, DELETE, and SELECT FOR
UPDATE/SHARE queries. If an EvalPlanQual recheck is necessary, we jam the
appropriate row into each scan node in the rechecking plan, forcing it to emit
only that one row. The former behavior could rescan the whole of each joined
relation for each recheck, which was terrible for performance, and what's much
worse could result in duplicated output tuples.
Also, the original implementation of EvalPlanQual could not re-use the recheck
execution tree --- it had to go through a full executor init and shutdown for
every row to be tested. To avoid this overhead, I've associated a special
runtime Param with each LockRows or ModifyTable plan node, and arranged to
make every scan node below such a node depend on that Param. Thus, by
signaling a change in that Param, the EPQ machinery can just rescan the
already-built test plan.
This patch also adds a prohibition on set-returning functions in the
targetlist of SELECT FOR UPDATE/SHARE. This is needed to avoid the
duplicate-output-tuple problem. It seems fairly reasonable since the
other restrictions on SELECT FOR UPDATE are meant to ensure that there
is a unique correspondence between source tuples and result tuples,
which an output SRF destroys as much as anything else does.
2009-10-26 03:26:45 +01:00
|
|
|
|
|
|
|
ExecScanReScan(&node->ss);
|
2008-10-04 23:56:55 +02:00
|
|
|
|
2012-08-16 01:01:16 +02:00
|
|
|
/*
|
|
|
|
* Clear the tuplestore if a new scan of the underlying CTE is required.
|
|
|
|
* This implicitly resets all the tuplestore's read pointers. Note that
|
|
|
|
* multiple CTE nodes might redundantly clear the tuplestore; that's OK,
|
|
|
|
* and not unduly expensive. We'll stop taking this path as soon as
|
|
|
|
* somebody has attempted to read something from the underlying CTE
|
|
|
|
* (thereby causing its chgParam to be cleared).
|
|
|
|
*/
|
|
|
|
if (node->leader->cteplanstate->chgParam != NULL)
|
2008-10-04 23:56:55 +02:00
|
|
|
{
|
2012-08-16 01:01:16 +02:00
|
|
|
tuplestore_clear(tuplestorestate);
|
|
|
|
node->leader->eof_cte = false;
|
2008-10-04 23:56:55 +02:00
|
|
|
}
|
|
|
|
else
|
|
|
|
{
|
2012-08-16 01:01:16 +02:00
|
|
|
/*
|
|
|
|
* Else, just rewind my own pointer. Either the underlying CTE
|
|
|
|
* doesn't need a rescan (and we can re-read what's in the tuplestore
|
|
|
|
* now), or somebody else already took care of it.
|
|
|
|
*/
|
2008-10-04 23:56:55 +02:00
|
|
|
tuplestore_select_read_pointer(tuplestorestate, node->readptr);
|
|
|
|
tuplestore_rescan(tuplestorestate);
|
|
|
|
}
|
|
|
|
}
|