Tom Lane b1577a7c78 New cost model for planning, incorporating a penalty for random page
accesses versus sequential accesses, a (very crude) estimate of the
effects of caching on random page accesses, and cost to evaluate WHERE-
clause expressions.  Export critical parameters for this model as SET
variables.  Also, create SET variables for the planner's enable flags
(enable_seqscan, enable_indexscan, etc) so that these can be controlled
more conveniently than via PGOPTIONS.

Planner now estimates both startup cost (cost before retrieving
first tuple) and total cost of each path, so it can optimize queries
with LIMIT on a reasonable basis by interpolating between these costs.
Same facility is a win for EXISTS(...) subqueries and some other cases.

Redesign pathkey representation to achieve a major speedup in planning
(I saw as much as 5X on a 10-way join); also minor changes in planner
to reduce memory consumption by recycling discarded Path nodes and
not constructing unnecessary lists.

Minor cleanups to display more-plausible costs in some cases in
EXPLAIN output.

Initdb forced by change in interface to index cost estimation
functions.
2000-02-15 20:49:31 +00:00

423 lines
7.5 KiB
C

/*-------------------------------------------------------------------------
*
* print.c
* various print routines (used mostly for debugging)
*
* Portions Copyright (c) 1996-2000, PostgreSQL, Inc
* Portions Copyright (c) 1994, Regents of the University of California
*
*
* IDENTIFICATION
* $Header: /cvsroot/pgsql/src/backend/nodes/print.c,v 1.37 2000/02/15 20:49:12 tgl Exp $
*
* HISTORY
* AUTHOR DATE MAJOR EVENT
* Andrew Yu Oct 26, 1994 file creation
*
*-------------------------------------------------------------------------
*/
#include "postgres.h"
#include "access/printtup.h"
#include "nodes/print.h"
#include "optimizer/clauses.h"
#include "parser/parsetree.h"
#include "utils/lsyscache.h"
static char *plannode_type(Plan *p);
/*
* print
* print contents of Node to stdout
*/
void
print(void *obj)
{
char *s;
s = nodeToString(obj);
printf("%s\n", s);
fflush(stdout);
return;
}
/*
* pretty print hack extraordinaire. -ay 10/94
*/
void
pprint(void *obj)
{
char *s;
int i;
char line[80];
int indentLev;
int j;
s = nodeToString(obj);
indentLev = 0;
i = 0;
for (;;)
{
for (j = 0; j < indentLev * 3; j++)
line[j] = ' ';
for (; j < 75 && s[i] != '\0'; i++, j++)
{
line[j] = s[i];
switch (line[j])
{
case '}':
if (j != indentLev * 3)
{
line[j] = '\0';
printf("%s\n", line);
line[indentLev * 3] = '\0';
printf("%s}\n", line);
}
else
{
line[j] = '\0';
printf("%s}\n", line);
}
indentLev--;
j = indentLev * 3 - 1; /* print the line before :
* and resets */
break;
case ')':
line[j + 1] = '\0';
printf("%s\n", line);
j = indentLev * 3 - 1;
break;
case '{':
indentLev++;
/* !!! FALLS THROUGH */
case ':':
if (j != 0)
{
line[j] = '\0';
printf("%s\n", line);
/* print the line before : and resets */
for (j = 0; j < indentLev * 3; j++)
line[j] = ' ';
}
line[j] = s[i];
break;
}
}
line[j] = '\0';
if (s[i] == '\0')
break;
printf("%s\n", line);
}
if (j != 0)
printf("%s\n", line);
fflush(stdout);
return;
}
/*
* print_rt
* print contents of range table
*/
void
print_rt(List *rtable)
{
List *l;
int i = 1;
printf("resno\trelname(refname)\trelid\tinFromCl\n");
printf("-----\t----------------\t-----\t--------\n");
foreach(l, rtable)
{
RangeTblEntry *rte = lfirst(l);
printf("%d\t%s(%s)\t%u\t%d\t%s\n",
i, rte->relname, rte->ref->relname, rte->relid,
rte->inFromCl,
(rte->inh ? "inh" : ""));
i++;
}
}
/*
* print_expr
* print an expression
*/
void
print_expr(Node *expr, List *rtable)
{
if (expr == NULL)
{
printf("<>");
return;
}
if (IsA(expr, Var))
{
Var *var = (Var *) expr;
RangeTblEntry *rt;
char *relname,
*attname;
switch (var->varno)
{
case INNER:
relname = "INNER";
attname = "?";
break;
case OUTER:
relname = "OUTER";
attname = "?";
break;
default:
{
rt = rt_fetch(var->varno, rtable);
relname = rt->relname;
if (rt->ref && rt->ref->relname)
relname = rt->ref->relname; /* table renamed */
attname = get_attname(rt->relid, var->varattno);
}
break;
}
printf("%s.%s", relname, attname);
}
else if (IsA(expr, Expr))
{
Expr *e = (Expr *) expr;
if (is_opclause(expr))
{
char *opname;
print_expr((Node *) get_leftop(e), rtable);
opname = get_opname(((Oper *) e->oper)->opno);
printf(" %s ", ((opname != NULL) ? opname : "(invalid operator)"));
print_expr((Node *) get_rightop(e), rtable);
}
else
printf("an expr");
}
else
printf("not an expr");
}
/*
* print_pathkeys -
* pathkeys list of list of PathKeyItems
*/
void
print_pathkeys(List *pathkeys, List *rtable)
{
List *i,
*k;
printf("(");
foreach(i, pathkeys)
{
List *pathkey = lfirst(i);
printf("(");
foreach(k, pathkey)
{
PathKeyItem *item = lfirst(k);
print_expr(item->key, rtable);
if (lnext(k))
printf(", ");
}
printf(") ");
if (lnext(i))
printf(", ");
}
printf(")\n");
}
/*
* print_tl
* print targetlist in a more legible way.
*/
void
print_tl(List *tlist, List *rtable)
{
List *tl;
printf("(\n");
foreach(tl, tlist)
{
TargetEntry *tle = lfirst(tl);
printf("\t%d %s\t", tle->resdom->resno, tle->resdom->resname);
if (tle->resdom->reskey != 0)
printf("(%d):\t", tle->resdom->reskey);
else
printf(" :\t");
print_expr(tle->expr, rtable);
printf("\n");
}
printf(")\n");
}
/*
* print_slot
* print out the tuple with the given TupleTableSlot
*/
void
print_slot(TupleTableSlot *slot)
{
if (!slot->val)
{
printf("tuple is null.\n");
return;
}
if (!slot->ttc_tupleDescriptor)
{
printf("no tuple descriptor.\n");
return;
}
debugtup(slot->val, slot->ttc_tupleDescriptor, NULL);
}
static char *
plannode_type(Plan *p)
{
switch (nodeTag(p))
{
case T_Plan:
return "PLAN";
break;
case T_Result:
return "RESULT";
break;
case T_Append:
return "APPEND";
break;
case T_Scan:
return "SCAN";
break;
case T_SeqScan:
return "SEQSCAN";
break;
case T_IndexScan:
return "INDEXSCAN";
break;
case T_Join:
return "JOIN";
break;
case T_NestLoop:
return "NESTLOOP";
break;
case T_MergeJoin:
return "MERGEJOIN";
break;
case T_HashJoin:
return "HASHJOIN";
break;
case T_Noname:
return "NONAME";
break;
case T_Material:
return "MATERIAL";
break;
case T_Sort:
return "SORT";
break;
case T_Agg:
return "AGG";
break;
case T_Unique:
return "UNIQUE";
break;
case T_Hash:
return "HASH";
break;
case T_Choose:
return "CHOOSE";
break;
case T_Group:
return "GROUP";
break;
case T_TidScan:
return "TIDSCAN";
break;
default:
return "UNKNOWN";
break;
}
}
/*
prints the ascii description of the plan nodes
does this recursively by doing a depth-first traversal of the
plan tree. for SeqScan and IndexScan, the name of the table is also
printed out
*/
void
print_plan_recursive(Plan *p, Query *parsetree, int indentLevel, char *label)
{
int i;
char extraInfo[100];
if (!p)
return;
for (i = 0; i < indentLevel; i++)
printf(" ");
printf("%s%s :c=%.2f..%.2f :r=%.0f :w=%d ", label, plannode_type(p),
p->startup_cost, p->total_cost,
p->plan_rows, p->plan_width);
if (IsA(p, Scan) ||IsA(p, SeqScan))
{
RangeTblEntry *rte;
rte = rt_fetch(((Scan *) p)->scanrelid, parsetree->rtable);
StrNCpy(extraInfo, rte->relname, NAMEDATALEN);
}
else if (IsA(p, IndexScan))
{
StrNCpy(extraInfo,
((RangeTblEntry *) (nth(((IndexScan *) p)->scan.scanrelid - 1,
parsetree->rtable)))->relname,
NAMEDATALEN);
}
else
extraInfo[0] = '\0';
if (extraInfo[0] != '\0')
printf(" ( %s )\n", extraInfo);
else
printf("\n");
print_plan_recursive(p->lefttree, parsetree, indentLevel + 3, "l: ");
print_plan_recursive(p->righttree, parsetree, indentLevel + 3, "r: ");
if (nodeTag(p) == T_Append)
{
List *lst;
int whichplan = 0;
Append *appendplan = (Append *) p;
foreach(lst, appendplan->appendplans)
{
Plan *subnode = (Plan *) lfirst(lst);
/*
* I don't think we need to fiddle with the range table here,
* bjm
*/
print_plan_recursive(subnode, parsetree, indentLevel + 3, "a: ");
whichplan++;
}
}
}
/* print_plan
prints just the plan node types */
void
print_plan(Plan *p, Query *parsetree)
{
print_plan_recursive(p, parsetree, 0, "");
}