#include "postgres.h"
#include "access/amapi.h"
#include "access/htup_details.h"
#include "access/relation.h"
#include "access/reloptions.h"
#include "access/table.h"
#include "access/toast_compression.h"
#include "catalog/dependency.h"
#include "catalog/heap.h"
#include "catalog/index.h"
#include "catalog/namespace.h"
#include "catalog/pg_am.h"
#include "catalog/pg_collation.h"
#include "catalog/pg_constraint.h"
#include "catalog/pg_opclass.h"
#include "catalog/pg_operator.h"
#include "catalog/pg_statistic_ext.h"
#include "catalog/pg_type.h"
#include "commands/comment.h"
#include "commands/defrem.h"
#include "commands/sequence.h"
#include "commands/tablecmds.h"
#include "commands/tablespace.h"
#include "miscadmin.h"
#include "nodes/makefuncs.h"
#include "nodes/nodeFuncs.h"
#include "optimizer/optimizer.h"
#include "parser/analyze.h"
#include "parser/parse_clause.h"
#include "parser/parse_coerce.h"
#include "parser/parse_collate.h"
#include "parser/parse_expr.h"
#include "parser/parse_relation.h"
#include "parser/parse_target.h"
#include "parser/parse_type.h"
#include "parser/parse_utilcmd.h"
#include "parser/parser.h"
#include "rewrite/rewriteManip.h"
#include "utils/acl.h"
#include "utils/builtins.h"
#include "utils/lsyscache.h"
#include "utils/partcache.h"
#include "utils/rel.h"
#include "utils/ruleutils.h"
#include "utils/syscache.h"
#include "utils/typcache.h"

Include dependency graph for parse_utilcmd.c:

Data Structures
struct	CreateStmtContext

struct	CreateSchemaStmtContext

Macros
#define	SUPPORTS_ATTRS(node)

Functions
static void	transformColumnDefinition (CreateStmtContext cxt, ColumnDef column)

static void	transformTableConstraint (CreateStmtContext cxt, Constraint constraint)

static void	transformTableLikeClause (CreateStmtContext cxt, TableLikeClause table_like_clause)

static void	transformOfType (CreateStmtContext cxt, TypeName ofTypename)

static CreateStatsStmt *	generateClonedExtStatsStmt (RangeVar *heapRel, Oid heapRelid, Oid source_statsid)

static List *	get_collation (Oid collation, Oid actual_datatype)

static List *	get_opclass (Oid opclass, Oid actual_datatype)

static void	transformIndexConstraints (CreateStmtContext *cxt)

static IndexStmt *	transformIndexConstraint (Constraint constraint, CreateStmtContext cxt)

static void	transformExtendedStatistics (CreateStmtContext *cxt)

static void	transformFKConstraints (CreateStmtContext *cxt, bool skipValidation, bool isAddConstraint)

static void	transformCheckConstraints (CreateStmtContext *cxt, bool skipValidation)

static void	transformConstraintAttrs (CreateStmtContext cxt, List constraintList)

static void	transformColumnType (CreateStmtContext cxt, ColumnDef column)

static void	setSchemaName (const char context_schema, char *stmt_schema_name)

static void	transformPartitionCmd (CreateStmtContext cxt, PartitionCmd cmd)

static List *	transformPartitionRangeBounds (ParseState pstate, List blist, Relation parent)

static void	validateInfiniteBounds (ParseState pstate, List blist)

static Const *	transformPartitionBoundValue (ParseState pstate, Node val, const char *colName, Oid colType, int32 colTypmod, Oid partCollation)

List *	transformCreateStmt (CreateStmt stmt, const char queryString)

static void	generateSerialExtraStmts (CreateStmtContext cxt, ColumnDef column, Oid seqtypid, List seqoptions, bool for_identity, bool col_exists, char snamespace_p, char *sname_p)

List *	expandTableLikeClause (RangeVar heapRel, TableLikeClause table_like_clause)

IndexStmt *	generateClonedIndexStmt (RangeVar heapRel, Relation source_idx, const AttrMap attmap, Oid *constraintOid)

IndexStmt *	transformIndexStmt (Oid relid, IndexStmt stmt, const char queryString)

CreateStatsStmt *	transformStatsStmt (Oid relid, CreateStatsStmt stmt, const char queryString)

void	transformRuleStmt (RuleStmt stmt, const char queryString, List actions, Node whereClause)

AlterTableStmt *	transformAlterTableStmt (Oid relid, AlterTableStmt stmt, const char queryString, List beforeStmts, List afterStmts)

List *	transformCreateSchemaStmtElements (List schemaElts, const char schemaName)

PartitionBoundSpec *	transformPartitionBound (ParseState pstate, Relation parent, PartitionBoundSpec spec)

Macro Definition Documentation

◆ SUPPORTS_ATTRS

#define SUPPORTS_ATTRS ( node )

Value:

    ((node) != NULL &&                      \
     ((node)->contype == CONSTR_PRIMARY ||  \
      (node)->contype == CONSTR_UNIQUE ||   \
      (node)->contype == CONSTR_EXCLUSION || \
      (node)->contype == CONSTR_FOREIGN))

Function Documentation

◆ expandTableLikeClause()

List* expandTableLikeClause	(	RangeVar *	heapRel,
		TableLikeClause *	table_like_clause
	)

Definition at line 1289 of file parse_utilcmd.c.

 {
     List       *result = NIL;
     List       *atsubcmds = NIL;
     AttrNumber  parent_attno;
     Relation    relation;
     Relation    childrel;
     TupleDesc   tupleDesc;
     TupleConstr *constr;
     AttrMap    *attmap;
     char       *comment;
     bool        at_pushed = false;
     ListCell   *lc;
  
     /*
      * Open the relation referenced by the LIKE clause.  We should still have
      * the table lock obtained by transformTableLikeClause (and this'll throw
      * an assertion failure if not).  Hence, no need to recheck privileges
      * etc.  We must open the rel by OID not name, to be sure we get the same
      * table.
      */
     if (!OidIsValid(table_like_clause->relationOid))
         elog(ERROR, "expandTableLikeClause called on untransformed LIKE clause");
  
     relation = relation_open(table_like_clause->relationOid, NoLock);
  
     tupleDesc = RelationGetDescr(relation);
     constr = tupleDesc->constr;
  
     /*
      * Open the newly-created child relation; we have lock on that too.
      */
     childrel = relation_openrv(heapRel, NoLock);
  
     /*
      * Construct a map from the LIKE relation's attnos to the child rel's.
      * This re-checks type match etc, although it shouldn't be possible to
      * have a failure since both tables are locked.
      */
     attmap = build_attrmap_by_name(RelationGetDescr(childrel),
                                    tupleDesc,
                                    false);
  
     /*
      * Process defaults, if required.
      */
     if ((table_like_clause->options &
          (CREATE_TABLE_LIKE_DEFAULTS | CREATE_TABLE_LIKE_GENERATED)) &&
         constr != NULL)
     {
         for (parent_attno = 1; parent_attno <= tupleDesc->natts;
              parent_attno++)
         {
             Form_pg_attribute attribute = TupleDescAttr(tupleDesc,
                                                         parent_attno - 1);
  
             /*
              * Ignore dropped columns in the parent.
              */
             if (attribute->attisdropped)
                 continue;
  
             /*
              * Copy default, if present and it should be copied.  We have
              * separate options for plain default expressions and GENERATED
              * defaults.
              */
             if (attribute->atthasdef &&
                 (attribute->attgenerated ?
                  (table_like_clause->options & CREATE_TABLE_LIKE_GENERATED) :
                  (table_like_clause->options & CREATE_TABLE_LIKE_DEFAULTS)))
             {
                 Node       *this_default;
                 AlterTableCmd *atsubcmd;
                 bool        found_whole_row;
  
                 this_default = TupleDescGetDefault(tupleDesc, parent_attno);
                 if (this_default == NULL)
                     elog(ERROR, "default expression not found for attribute %d of relation \"%s\"",
                          parent_attno, RelationGetRelationName(relation));
  
                 atsubcmd = makeNode(AlterTableCmd);
                 atsubcmd->subtype = AT_CookedColumnDefault;
                 atsubcmd->num = attmap->attnums[parent_attno - 1];
                 atsubcmd->def = map_variable_attnos(this_default,
                                                     1, 0,
                                                     attmap,
                                                     InvalidOid,
                                                     &found_whole_row);
  
                 /*
                  * Prevent this for the same reason as for constraints below.
                  * Note that defaults cannot contain any vars, so it's OK that
                  * the error message refers to generated columns.
                  */
                 if (found_whole_row)
                     ereport(ERROR,
                             (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
                              errmsg("cannot convert whole-row table reference"),
                              errdetail("Generation expression for column \"%s\" contains a whole-row reference to table \"%s\".",
                                        NameStr(attribute->attname),
                                        RelationGetRelationName(relation))));
  
                 atsubcmds = lappend(atsubcmds, atsubcmd);
             }
         }
     }
  
     /*
      * Copy CHECK constraints if requested, being careful to adjust attribute
      * numbers so they match the child.
      */
     if ((table_like_clause->options & CREATE_TABLE_LIKE_CONSTRAINTS) &&
         constr != NULL)
     {
         int         ccnum;
  
         for (ccnum = 0; ccnum < constr->num_check; ccnum++)
         {
             char       *ccname = constr->check[ccnum].ccname;
             char       *ccbin = constr->check[ccnum].ccbin;
             bool        ccnoinherit = constr->check[ccnum].ccnoinherit;
             Node       *ccbin_node;
             bool        found_whole_row;
             Constraint *n;
             AlterTableCmd *atsubcmd;
  
             ccbin_node = map_variable_attnos(stringToNode(ccbin),
                                              1, 0,
                                              attmap,
                                              InvalidOid, &found_whole_row);
  
             /*
              * We reject whole-row variables because the whole point of LIKE
              * is that the new table's rowtype might later diverge from the
              * parent's.  So, while translation might be possible right now,
              * it wouldn't be possible to guarantee it would work in future.
              */
             if (found_whole_row)
                 ereport(ERROR,
                         (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
                          errmsg("cannot convert whole-row table reference"),
                          errdetail("Constraint \"%s\" contains a whole-row reference to table \"%s\".",
                                    ccname,
                                    RelationGetRelationName(relation))));
  
             n = makeNode(Constraint);
             n->contype = CONSTR_CHECK;
             n->conname = pstrdup(ccname);
             n->location = -1;
             n->is_no_inherit = ccnoinherit;
             n->raw_expr = NULL;
             n->cooked_expr = nodeToString(ccbin_node);
  
             /* We can skip validation, since the new table should be empty. */
             n->skip_validation = true;
             n->initially_valid = true;
  
             atsubcmd = makeNode(AlterTableCmd);
             atsubcmd->subtype = AT_AddConstraint;
             atsubcmd->def = (Node *) n;
             atsubcmds = lappend(atsubcmds, atsubcmd);
  
             /* Copy comment on constraint */
             if ((table_like_clause->options & CREATE_TABLE_LIKE_COMMENTS) &&
                 (comment = GetComment(get_relation_constraint_oid(RelationGetRelid(relation),
                                                                   n->conname, false),
                                       ConstraintRelationId,
                                       0)) != NULL)
             {
                 CommentStmt *stmt = makeNode(CommentStmt);
  
                 stmt->objtype = OBJECT_TABCONSTRAINT;
                 stmt->object = (Node *) list_make3(makeString(heapRel->schemaname),
                                                    makeString(heapRel->relname),
                                                    makeString(n->conname));
                 stmt->comment = comment;
  
                 result = lappend(result, stmt);
             }
         }
     }
  
     /*
      * Copy not-null constraints, too (these do not require any option to have
      * been given).
      */
     foreach(lc, RelationGetNotNullConstraints(RelationGetRelid(relation), false))
     {
         AlterTableCmd *atsubcmd;
  
         atsubcmd = makeNode(AlterTableCmd);
         atsubcmd->subtype = AT_AddConstraint;
         atsubcmd->def = (Node *) lfirst_node(Constraint, lc);
         atsubcmds = lappend(atsubcmds, atsubcmd);
     }
  
     /*
      * If we generated any ALTER TABLE actions above, wrap them into a single
      * ALTER TABLE command.  Stick it at the front of the result, so it runs
      * before any CommentStmts we made above.
      */
     if (atsubcmds)
     {
         AlterTableStmt *atcmd = makeNode(AlterTableStmt);
  
         atcmd->relation = copyObject(heapRel);
         atcmd->cmds = atsubcmds;
         atcmd->objtype = OBJECT_TABLE;
         atcmd->missing_ok = false;
         result = lcons(atcmd, result);
  
         at_pushed = true;
     }
  
     /*
      * Process indexes if required.
      */
     if ((table_like_clause->options & CREATE_TABLE_LIKE_INDEXES) &&
         relation->rd_rel->relhasindex)
     {
         List       *parent_indexes;
         ListCell   *l;
  
         parent_indexes = RelationGetIndexList(relation);
  
         foreach(l, parent_indexes)
         {
             Oid         parent_index_oid = lfirst_oid(l);
             Relation    parent_index;
             IndexStmt  *index_stmt;
  
             parent_index = index_open(parent_index_oid, AccessShareLock);
  
             /* Build CREATE INDEX statement to recreate the parent_index */
             index_stmt = generateClonedIndexStmt(heapRel,
                                                  parent_index,
                                                  attmap,
                                                  NULL);
  
             /*
              * The PK columns might not yet non-nullable, so make sure they
              * become so.
              */
             if (index_stmt->primary)
             {
                 foreach(lc, index_stmt->indexParams)
                 {
                     IndexElem  *col = lfirst_node(IndexElem, lc);
                     AlterTableCmd *notnullcmd = makeNode(AlterTableCmd);
  
                     notnullcmd->subtype = AT_SetAttNotNull;
                     notnullcmd->name = pstrdup(col->name);
                     /* Luckily we can still add more AT-subcmds here */
                     atsubcmds = lappend(atsubcmds, notnullcmd);
                 }
  
                 /*
                  * If we had already put the AlterTableStmt into the output
                  * list, we don't need to do so again; otherwise do it.
                  */
                 if (!at_pushed)
                 {
                     AlterTableStmt *atcmd = makeNode(AlterTableStmt);
  
                     atcmd->relation = copyObject(heapRel);
                     atcmd->cmds = atsubcmds;
                     atcmd->objtype = OBJECT_TABLE;
                     atcmd->missing_ok = false;
                     result = lcons(atcmd, result);
                 }
             }
  
             /* Copy comment on index, if requested */
             if (table_like_clause->options & CREATE_TABLE_LIKE_COMMENTS)
             {
                 comment = GetComment(parent_index_oid, RelationRelationId, 0);
  
                 /*
                  * We make use of IndexStmt's idxcomment option, so as not to
                  * need to know now what name the index will have.
                  */
                 index_stmt->idxcomment = comment;
             }
  
             result = lappend(result, index_stmt);
  
             index_close(parent_index, AccessShareLock);
         }
     }
  
     /* Done with child rel */
     table_close(childrel, NoLock);
  
     /*
      * Close the parent rel, but keep our AccessShareLock on it until xact
      * commit.  That will prevent someone else from deleting or ALTERing the
      * parent before the child is committed.
      */
     table_close(relation, NoLock);
  
     return result;
 }

References AccessShareLock, AT_AddConstraint, AT_CookedColumnDefault, AT_SetAttNotNull, AttrMap::attnums, build_attrmap_by_name(), ConstrCheck::ccbin, ConstrCheck::ccname, ConstrCheck::ccnoinherit, TupleConstr::check, AlterTableStmt::cmds, comment, Constraint::conname, TupleDescData::constr, CONSTR_CHECK, Constraint::contype, Constraint::cooked_expr, copyObject, CREATE_TABLE_LIKE_COMMENTS, CREATE_TABLE_LIKE_CONSTRAINTS, CREATE_TABLE_LIKE_DEFAULTS, CREATE_TABLE_LIKE_GENERATED, CREATE_TABLE_LIKE_INDEXES, AlterTableCmd::def, elog(), ereport, errcode(), errdetail(), errmsg(), ERROR, generateClonedIndexStmt(), get_relation_constraint_oid(), GetComment(), IndexStmt::idxcomment, index_close(), index_open(), IndexStmt::indexParams, Constraint::initially_valid, InvalidOid, Constraint::is_no_inherit, lappend(), lcons(), lfirst_node, lfirst_oid, list_make3, Constraint::location, makeNode, makeString(), map_variable_attnos(), AlterTableStmt::missing_ok, IndexElem::name, AlterTableCmd::name, NameStr, TupleDescData::natts, NIL, nodeToString(), NoLock, AlterTableCmd::num, TupleConstr::num_check, OBJECT_TABCONSTRAINT, OBJECT_TABLE, AlterTableStmt::objtype, OidIsValid, TableLikeClause::options, IndexStmt::primary, pstrdup(), Constraint::raw_expr, RelationData::rd_rel, AlterTableStmt::relation, relation_open(), relation_openrv(), RelationGetDescr, RelationGetIndexList(), RelationGetNotNullConstraints(), RelationGetRelationName, RelationGetRelid, TableLikeClause::relationOid, RangeVar::relname, RangeVar::schemaname, Constraint::skip_validation, stmt, stringToNode(), AlterTableCmd::subtype, table_close(), TupleDescAttr, and TupleDescGetDefault().

Referenced by ProcessUtilitySlow().

◆ generateClonedExtStatsStmt()

static CreateStatsStmt * generateClonedExtStatsStmt	(	RangeVar *	heapRel,
		Oid	heapRelid,
		Oid	source_statsid
	)

static

Definition at line 1996 of file parse_utilcmd.c.

 {
     HeapTuple   ht_stats;
     Form_pg_statistic_ext statsrec;
     CreateStatsStmt *stats;
     List       *stat_types = NIL;
     List       *def_names = NIL;
     bool        isnull;
     Datum       datum;
     ArrayType  *arr;
     char       *enabled;
     int         i;
  
     Assert(OidIsValid(heapRelid));
     Assert(heapRel != NULL);
  
     /*
      * Fetch pg_statistic_ext tuple of source statistics object.
      */
     ht_stats = SearchSysCache1(STATEXTOID, ObjectIdGetDatum(source_statsid));
     if (!HeapTupleIsValid(ht_stats))
         elog(ERROR, "cache lookup failed for statistics object %u", source_statsid);
     statsrec = (Form_pg_statistic_ext) GETSTRUCT(ht_stats);
  
     /* Determine which statistics types exist */
     datum = SysCacheGetAttrNotNull(STATEXTOID, ht_stats,
                                    Anum_pg_statistic_ext_stxkind);
     arr = DatumGetArrayTypeP(datum);
     if (ARR_NDIM(arr) != 1 ||
         ARR_HASNULL(arr) ||
         ARR_ELEMTYPE(arr) != CHAROID)
         elog(ERROR, "stxkind is not a 1-D char array");
     enabled = (char *) ARR_DATA_PTR(arr);
     for (i = 0; i < ARR_DIMS(arr)[0]; i++)
     {
         if (enabled[i] == STATS_EXT_NDISTINCT)
             stat_types = lappend(stat_types, makeString("ndistinct"));
         else if (enabled[i] == STATS_EXT_DEPENDENCIES)
             stat_types = lappend(stat_types, makeString("dependencies"));
         else if (enabled[i] == STATS_EXT_MCV)
             stat_types = lappend(stat_types, makeString("mcv"));
         else if (enabled[i] == STATS_EXT_EXPRESSIONS)
             /* expression stats are not exposed to users */
             continue;
         else
             elog(ERROR, "unrecognized statistics kind %c", enabled[i]);
     }
  
     /* Determine which columns the statistics are on */
     for (i = 0; i < statsrec->stxkeys.dim1; i++)
     {
         StatsElem  *selem = makeNode(StatsElem);
         AttrNumber  attnum = statsrec->stxkeys.values[i];
  
         selem->name = get_attname(heapRelid, attnum, false);
         selem->expr = NULL;
  
         def_names = lappend(def_names, selem);
     }
  
     /*
      * Now handle expressions, if there are any. The order (with respect to
      * regular attributes) does not really matter for extended stats, so we
      * simply append them after simple column references.
      *
      * XXX Some places during build/estimation treat expressions as if they
      * are before attributes, but for the CREATE command that's entirely
      * irrelevant.
      */
     datum = SysCacheGetAttr(STATEXTOID, ht_stats,
                             Anum_pg_statistic_ext_stxexprs, &isnull);
  
     if (!isnull)
     {
         ListCell   *lc;
         List       *exprs = NIL;
         char       *exprsString;
  
         exprsString = TextDatumGetCString(datum);
         exprs = (List *) stringToNode(exprsString);
  
         foreach(lc, exprs)
         {
             StatsElem  *selem = makeNode(StatsElem);
  
             selem->name = NULL;
             selem->expr = (Node *) lfirst(lc);
  
             def_names = lappend(def_names, selem);
         }
  
         pfree(exprsString);
     }
  
     /* finally, build the output node */
     stats = makeNode(CreateStatsStmt);
     stats->defnames = NULL;
     stats->stat_types = stat_types;
     stats->exprs = def_names;
     stats->relations = list_make1(heapRel);
     stats->stxcomment = NULL;
     stats->transformed = true;  /* don't need transformStatsStmt again */
     stats->if_not_exists = false;
  
     /* Clean up */
     ReleaseSysCache(ht_stats);
  
     return stats;
 }

References ARR_DATA_PTR, ARR_DIMS, ARR_ELEMTYPE, ARR_HASNULL, ARR_NDIM, Assert(), attnum, DatumGetArrayTypeP, CreateStatsStmt::defnames, elog(), ERROR, StatsElem::expr, CreateStatsStmt::exprs, get_attname(), GETSTRUCT, HeapTupleIsValid, i, CreateStatsStmt::if_not_exists, lappend(), lfirst, list_make1, makeNode, makeString(), StatsElem::name, NIL, ObjectIdGetDatum(), OidIsValid, pfree(), CreateStatsStmt::relations, ReleaseSysCache(), SearchSysCache1(), CreateStatsStmt::stat_types, STATEXTOID, stringToNode(), CreateStatsStmt::stxcomment, SysCacheGetAttr(), SysCacheGetAttrNotNull(), TextDatumGetCString, and CreateStatsStmt::transformed.

Referenced by transformTableLikeClause().

◆ generateClonedIndexStmt()

IndexStmt* generateClonedIndexStmt	(	RangeVar *	heapRel,
		Relation	source_idx,
		const AttrMap *	attmap,
		Oid *	constraintOid
	)

Definition at line 1647 of file parse_utilcmd.c.

 {
     Oid         source_relid = RelationGetRelid(source_idx);
     HeapTuple   ht_idxrel;
     HeapTuple   ht_idx;
     HeapTuple   ht_am;
     Form_pg_class idxrelrec;
     Form_pg_index idxrec;
     Form_pg_am  amrec;
     oidvector  *indcollation;
     oidvector  *indclass;
     IndexStmt  *index;
     List       *indexprs;
     ListCell   *indexpr_item;
     Oid         indrelid;
     int         keyno;
     Oid         keycoltype;
     Datum       datum;
     bool        isnull;
  
     if (constraintOid)
         *constraintOid = InvalidOid;
  
     /*
      * Fetch pg_class tuple of source index.  We can't use the copy in the
      * relcache entry because it doesn't include optional fields.
      */
     ht_idxrel = SearchSysCache1(RELOID, ObjectIdGetDatum(source_relid));
     if (!HeapTupleIsValid(ht_idxrel))
         elog(ERROR, "cache lookup failed for relation %u", source_relid);
     idxrelrec = (Form_pg_class) GETSTRUCT(ht_idxrel);
  
     /* Fetch pg_index tuple for source index from relcache entry */
     ht_idx = source_idx->rd_indextuple;
     idxrec = (Form_pg_index) GETSTRUCT(ht_idx);
     indrelid = idxrec->indrelid;
  
     /* Fetch the pg_am tuple of the index' access method */
     ht_am = SearchSysCache1(AMOID, ObjectIdGetDatum(idxrelrec->relam));
     if (!HeapTupleIsValid(ht_am))
         elog(ERROR, "cache lookup failed for access method %u",
              idxrelrec->relam);
     amrec = (Form_pg_am) GETSTRUCT(ht_am);
  
     /* Extract indcollation from the pg_index tuple */
     datum = SysCacheGetAttrNotNull(INDEXRELID, ht_idx,
                                    Anum_pg_index_indcollation);
     indcollation = (oidvector *) DatumGetPointer(datum);
  
     /* Extract indclass from the pg_index tuple */
     datum = SysCacheGetAttrNotNull(INDEXRELID, ht_idx, Anum_pg_index_indclass);
     indclass = (oidvector *) DatumGetPointer(datum);
  
     /* Begin building the IndexStmt */
     index = makeNode(IndexStmt);
     index->relation = heapRel;
     index->accessMethod = pstrdup(NameStr(amrec->amname));
     if (OidIsValid(idxrelrec->reltablespace))
         index->tableSpace = get_tablespace_name(idxrelrec->reltablespace);
     else
         index->tableSpace = NULL;
     index->excludeOpNames = NIL;
     index->idxcomment = NULL;
     index->indexOid = InvalidOid;
     index->oldNumber = InvalidRelFileNumber;
     index->oldCreateSubid = InvalidSubTransactionId;
     index->oldFirstRelfilelocatorSubid = InvalidSubTransactionId;
     index->unique = idxrec->indisunique;
     index->nulls_not_distinct = idxrec->indnullsnotdistinct;
     index->primary = idxrec->indisprimary;
     index->transformed = true;  /* don't need transformIndexStmt */
     index->concurrent = false;
     index->if_not_exists = false;
     index->reset_default_tblspc = false;
  
     /*
      * We don't try to preserve the name of the source index; instead, just
      * let DefineIndex() choose a reasonable name.  (If we tried to preserve
      * the name, we'd get duplicate-relation-name failures unless the source
      * table was in a different schema.)
      */
     index->idxname = NULL;
  
     /*
      * If the index is marked PRIMARY or has an exclusion condition, it's
      * certainly from a constraint; else, if it's not marked UNIQUE, it
      * certainly isn't.  If it is or might be from a constraint, we have to
      * fetch the pg_constraint record.
      */
     if (index->primary || index->unique || idxrec->indisexclusion)
     {
         Oid         constraintId = get_index_constraint(source_relid);
  
         if (OidIsValid(constraintId))
         {
             HeapTuple   ht_constr;
             Form_pg_constraint conrec;
  
             if (constraintOid)
                 *constraintOid = constraintId;
  
             ht_constr = SearchSysCache1(CONSTROID,
                                         ObjectIdGetDatum(constraintId));
             if (!HeapTupleIsValid(ht_constr))
                 elog(ERROR, "cache lookup failed for constraint %u",
                      constraintId);
             conrec = (Form_pg_constraint) GETSTRUCT(ht_constr);
  
             index->isconstraint = true;
             index->deferrable = conrec->condeferrable;
             index->initdeferred = conrec->condeferred;
  
             /* If it's an exclusion constraint, we need the operator names */
             if (idxrec->indisexclusion)
             {
                 Datum      *elems;
                 int         nElems;
                 int         i;
  
                 Assert(conrec->contype == CONSTRAINT_EXCLUSION);
                 /* Extract operator OIDs from the pg_constraint tuple */
                 datum = SysCacheGetAttrNotNull(CONSTROID, ht_constr,
                                                Anum_pg_constraint_conexclop);
                 deconstruct_array_builtin(DatumGetArrayTypeP(datum), OIDOID, &elems, NULL, &nElems);
  
                 for (i = 0; i < nElems; i++)
                 {
                     Oid         operid = DatumGetObjectId(elems[i]);
                     HeapTuple   opertup;
                     Form_pg_operator operform;
                     char       *oprname;
                     char       *nspname;
                     List       *namelist;
  
                     opertup = SearchSysCache1(OPEROID,
                                               ObjectIdGetDatum(operid));
                     if (!HeapTupleIsValid(opertup))
                         elog(ERROR, "cache lookup failed for operator %u",
                              operid);
                     operform = (Form_pg_operator) GETSTRUCT(opertup);
                     oprname = pstrdup(NameStr(operform->oprname));
                     /* For simplicity we always schema-qualify the op name */
                     nspname = get_namespace_name(operform->oprnamespace);
                     namelist = list_make2(makeString(nspname),
                                           makeString(oprname));
                     index->excludeOpNames = lappend(index->excludeOpNames,
                                                     namelist);
                     ReleaseSysCache(opertup);
                 }
             }
  
             ReleaseSysCache(ht_constr);
         }
         else
             index->isconstraint = false;
     }
     else
         index->isconstraint = false;
  
     /* Get the index expressions, if any */
     datum = SysCacheGetAttr(INDEXRELID, ht_idx,
                             Anum_pg_index_indexprs, &isnull);
     if (!isnull)
     {
         char       *exprsString;
  
         exprsString = TextDatumGetCString(datum);
         indexprs = (List *) stringToNode(exprsString);
     }
     else
         indexprs = NIL;
  
     /* Build the list of IndexElem */
     index->indexParams = NIL;
     index->indexIncludingParams = NIL;
  
     indexpr_item = list_head(indexprs);
     for (keyno = 0; keyno < idxrec->indnkeyatts; keyno++)
     {
         IndexElem  *iparam;
         AttrNumber  attnum = idxrec->indkey.values[keyno];
         Form_pg_attribute attr = TupleDescAttr(RelationGetDescr(source_idx),
                                                keyno);
         int16       opt = source_idx->rd_indoption[keyno];
  
         iparam = makeNode(IndexElem);
  
         if (AttributeNumberIsValid(attnum))
         {
             /* Simple index column */
             char       *attname;
  
             attname = get_attname(indrelid, attnum, false);
             keycoltype = get_atttype(indrelid, attnum);
  
             iparam->name = attname;
             iparam->expr = NULL;
         }
         else
         {
             /* Expressional index */
             Node       *indexkey;
             bool        found_whole_row;
  
             if (indexpr_item == NULL)
                 elog(ERROR, "too few entries in indexprs list");
             indexkey = (Node *) lfirst(indexpr_item);
             indexpr_item = lnext(indexprs, indexpr_item);
  
             /* Adjust Vars to match new table's column numbering */
             indexkey = map_variable_attnos(indexkey,
                                            1, 0,
                                            attmap,
                                            InvalidOid, &found_whole_row);
  
             /* As in expandTableLikeClause, reject whole-row variables */
             if (found_whole_row)
                 ereport(ERROR,
                         (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
                          errmsg("cannot convert whole-row table reference"),
                          errdetail("Index \"%s\" contains a whole-row table reference.",
                                    RelationGetRelationName(source_idx))));
  
             iparam->name = NULL;
             iparam->expr = indexkey;
  
             keycoltype = exprType(indexkey);
         }
  
         /* Copy the original index column name */
         iparam->indexcolname = pstrdup(NameStr(attr->attname));
  
         /* Add the collation name, if non-default */
         iparam->collation = get_collation(indcollation->values[keyno], keycoltype);
  
         /* Add the operator class name, if non-default */
         iparam->opclass = get_opclass(indclass->values[keyno], keycoltype);
         iparam->opclassopts =
             untransformRelOptions(get_attoptions(source_relid, keyno + 1));
  
         iparam->ordering = SORTBY_DEFAULT;
         iparam->nulls_ordering = SORTBY_NULLS_DEFAULT;
  
         /* Adjust options if necessary */
         if (source_idx->rd_indam->amcanorder)
         {
             /*
              * If it supports sort ordering, copy DESC and NULLS opts. Don't
              * set non-default settings unnecessarily, though, so as to
              * improve the chance of recognizing equivalence to constraint
              * indexes.
              */
             if (opt & INDOPTION_DESC)
             {
                 iparam->ordering = SORTBY_DESC;
                 if ((opt & INDOPTION_NULLS_FIRST) == 0)
                     iparam->nulls_ordering = SORTBY_NULLS_LAST;
             }
             else
             {
                 if (opt & INDOPTION_NULLS_FIRST)
                     iparam->nulls_ordering = SORTBY_NULLS_FIRST;
             }
         }
  
         index->indexParams = lappend(index->indexParams, iparam);
     }
  
     /* Handle included columns separately */
     for (keyno = idxrec->indnkeyatts; keyno < idxrec->indnatts; keyno++)
     {
         IndexElem  *iparam;
         AttrNumber  attnum = idxrec->indkey.values[keyno];
         Form_pg_attribute attr = TupleDescAttr(RelationGetDescr(source_idx),
                                                keyno);
  
         iparam = makeNode(IndexElem);
  
         if (AttributeNumberIsValid(attnum))
         {
             /* Simple index column */
             char       *attname;
  
             attname = get_attname(indrelid, attnum, false);
  
             iparam->name = attname;
             iparam->expr = NULL;
         }
         else
             ereport(ERROR,
                     (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
                      errmsg("expressions are not supported in included columns")));
  
         /* Copy the original index column name */
         iparam->indexcolname = pstrdup(NameStr(attr->attname));
  
         index->indexIncludingParams = lappend(index->indexIncludingParams, iparam);
     }
     /* Copy reloptions if any */
     datum = SysCacheGetAttr(RELOID, ht_idxrel,
                             Anum_pg_class_reloptions, &isnull);
     if (!isnull)
         index->options = untransformRelOptions(datum);
  
     /* If it's a partial index, decompile and append the predicate */
     datum = SysCacheGetAttr(INDEXRELID, ht_idx,
                             Anum_pg_index_indpred, &isnull);
     if (!isnull)
     {
         char       *pred_str;
         Node       *pred_tree;
         bool        found_whole_row;
  
         /* Convert text string to node tree */
         pred_str = TextDatumGetCString(datum);
         pred_tree = (Node *) stringToNode(pred_str);
  
         /* Adjust Vars to match new table's column numbering */
         pred_tree = map_variable_attnos(pred_tree,
                                         1, 0,
                                         attmap,
                                         InvalidOid, &found_whole_row);
  
         /* As in expandTableLikeClause, reject whole-row variables */
         if (found_whole_row)
             ereport(ERROR,
                     (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
                      errmsg("cannot convert whole-row table reference"),
                      errdetail("Index \"%s\" contains a whole-row table reference.",
                                RelationGetRelationName(source_idx))));
  
         index->whereClause = pred_tree;
     }
  
     /* Clean up */
     ReleaseSysCache(ht_idxrel);
     ReleaseSysCache(ht_am);
  
     return index;
 }

References IndexAmRoutine::amcanorder, AMOID, Assert(), attname, attnum, AttributeNumberIsValid, IndexElem::collation, CONSTROID, DatumGetArrayTypeP, DatumGetObjectId(), DatumGetPointer(), deconstruct_array_builtin(), elog(), ereport, errcode(), errdetail(), errmsg(), ERROR, IndexElem::expr, exprType(), get_attname(), get_attoptions(), get_atttype(), get_collation(), get_index_constraint(), get_namespace_name(), get_opclass(), get_tablespace_name(), GETSTRUCT, HeapTupleIsValid, i, IndexElem::indexcolname, INDEXRELID, InvalidOid, InvalidRelFileNumber, InvalidSubTransactionId, lappend(), lfirst, list_head(), list_make2, lnext(), makeNode, makeString(), map_variable_attnos(), IndexElem::name, NameStr, NIL, IndexElem::nulls_ordering, ObjectIdGetDatum(), OidIsValid, IndexElem::opclass, IndexElem::opclassopts, OPEROID, IndexElem::ordering, pstrdup(), RelationData::rd_indam, RelationData::rd_indextuple, RelationData::rd_indoption, RelationGetDescr, RelationGetRelationName, RelationGetRelid, ReleaseSysCache(), RELOID, SearchSysCache1(), SORTBY_DEFAULT, SORTBY_DESC, SORTBY_NULLS_DEFAULT, SORTBY_NULLS_FIRST, SORTBY_NULLS_LAST, stringToNode(), SysCacheGetAttr(), SysCacheGetAttrNotNull(), TextDatumGetCString, TupleDescAttr, untransformRelOptions(), and oidvector::values.

Referenced by AttachPartitionEnsureIndexes(), DefineRelation(), and expandTableLikeClause().

◆ generateSerialExtraStmts()

static void generateSerialExtraStmts	(	CreateStmtContext *	cxt,
		ColumnDef *	column,
		Oid	seqtypid,
		List *	seqoptions,
		bool	for_identity,
		bool	col_exists,
		char **	snamespace_p,
		char **	sname_p
	)

static

Definition at line 370 of file parse_utilcmd.c.

 {
     ListCell   *option;
     DefElem    *nameEl = NULL;
     Oid         snamespaceid;
     char       *snamespace;
     char       *sname;
     CreateSeqStmt *seqstmt;
     AlterSeqStmt *altseqstmt;
     List       *attnamelist;
     int         nameEl_idx = -1;
  
     /* Make a copy of this as we may end up modifying it in the code below */
     seqoptions = list_copy(seqoptions);
  
     /*
      * Determine namespace and name to use for the sequence.
      *
      * First, check if a sequence name was passed in as an option.  This is
      * used by pg_dump.  Else, generate a name.
      *
      * Although we use ChooseRelationName, it's not guaranteed that the
      * selected sequence name won't conflict; given sufficiently long field
      * names, two different serial columns in the same table could be assigned
      * the same sequence name, and we'd not notice since we aren't creating
      * the sequence quite yet.  In practice this seems quite unlikely to be a
      * problem, especially since few people would need two serial columns in
      * one table.
      */
     foreach(option, seqoptions)
     {
         DefElem    *defel = lfirst_node(DefElem, option);
  
         if (strcmp(defel->defname, "sequence_name") == 0)
         {
             if (nameEl)
                 errorConflictingDefElem(defel, cxt->pstate);
             nameEl = defel;
             nameEl_idx = foreach_current_index(option);
         }
     }
  
     if (nameEl)
     {
         RangeVar   *rv = makeRangeVarFromNameList(castNode(List, nameEl->arg));
  
         snamespace = rv->schemaname;
         if (!snamespace)
         {
             /* Given unqualified SEQUENCE NAME, select namespace */
             if (cxt->rel)
                 snamespaceid = RelationGetNamespace(cxt->rel);
             else
                 snamespaceid = RangeVarGetCreationNamespace(cxt->relation);
             snamespace = get_namespace_name(snamespaceid);
         }
         sname = rv->relname;
         /* Remove the SEQUENCE NAME item from seqoptions */
         seqoptions = list_delete_nth_cell(seqoptions, nameEl_idx);
     }
     else
     {
         if (cxt->rel)
             snamespaceid = RelationGetNamespace(cxt->rel);
         else
         {
             snamespaceid = RangeVarGetCreationNamespace(cxt->relation);
             RangeVarAdjustRelationPersistence(cxt->relation, snamespaceid);
         }
         snamespace = get_namespace_name(snamespaceid);
         sname = ChooseRelationName(cxt->relation->relname,
                                    column->colname,
                                    "seq",
                                    snamespaceid,
                                    false);
     }
  
     ereport(DEBUG1,
             (errmsg_internal("%s will create implicit sequence \"%s\" for serial column \"%s.%s\"",
                              cxt->stmtType, sname,
                              cxt->relation->relname, column->colname)));
  
     /*
      * Build a CREATE SEQUENCE command to create the sequence object, and add
      * it to the list of things to be done before this CREATE/ALTER TABLE.
      */
     seqstmt = makeNode(CreateSeqStmt);
     seqstmt->for_identity = for_identity;
     seqstmt->sequence = makeRangeVar(snamespace, sname, -1);
     seqstmt->sequence->relpersistence = cxt->relation->relpersistence;
     seqstmt->options = seqoptions;
  
     /*
      * If a sequence data type was specified, add it to the options.  Prepend
      * to the list rather than append; in case a user supplied their own AS
      * clause, the "redundant options" error will point to their occurrence,
      * not our synthetic one.
      */
     if (seqtypid)
         seqstmt->options = lcons(makeDefElem("as",
                                              (Node *) makeTypeNameFromOid(seqtypid, -1),
                                              -1),
                                  seqstmt->options);
  
     /*
      * If this is ALTER ADD COLUMN, make sure the sequence will be owned by
      * the table's owner.  The current user might be someone else (perhaps a
      * superuser, or someone who's only a member of the owning role), but the
      * SEQUENCE OWNED BY mechanisms will bleat unless table and sequence have
      * exactly the same owning role.
      */
     if (cxt->rel)
         seqstmt->ownerId = cxt->rel->rd_rel->relowner;
     else
         seqstmt->ownerId = InvalidOid;
  
     cxt->blist = lappend(cxt->blist, seqstmt);
  
     /*
      * Store the identity sequence name that we decided on.  ALTER TABLE ...
      * ADD COLUMN ... IDENTITY needs this so that it can fill the new column
      * with values from the sequence, while the association of the sequence
      * with the table is not set until after the ALTER TABLE.
      */
     column->identitySequence = seqstmt->sequence;
  
     /*
      * Build an ALTER SEQUENCE ... OWNED BY command to mark the sequence as
      * owned by this column, and add it to the appropriate list of things to
      * be done along with this CREATE/ALTER TABLE.  In a CREATE or ALTER ADD
      * COLUMN, it must be done after the statement because we don't know the
      * column's attnum yet.  But if we do have the attnum (in AT_AddIdentity),
      * we can do the marking immediately, which improves some ALTER TABLE
      * behaviors.
      */
     altseqstmt = makeNode(AlterSeqStmt);
     altseqstmt->sequence = makeRangeVar(snamespace, sname, -1);
     attnamelist = list_make3(makeString(snamespace),
                              makeString(cxt->relation->relname),
                              makeString(column->colname));
     altseqstmt->options = list_make1(makeDefElem("owned_by",
                                                  (Node *) attnamelist, -1));
     altseqstmt->for_identity = for_identity;
  
     if (col_exists)
         cxt->blist = lappend(cxt->blist, altseqstmt);
     else
         cxt->alist = lappend(cxt->alist, altseqstmt);
  
     if (snamespace_p)
         *snamespace_p = snamespace;
     if (sname_p)
         *sname_p = sname;
 }

References CreateStmtContext::alist, DefElem::arg, CreateStmtContext::blist, castNode, ChooseRelationName(), ColumnDef::colname, DEBUG1, DefElem::defname, ereport, errmsg_internal(), errorConflictingDefElem(), CreateSeqStmt::for_identity, AlterSeqStmt::for_identity, foreach_current_index, get_namespace_name(), ColumnDef::identitySequence, InvalidOid, lappend(), lcons(), lfirst_node, list_copy(), list_delete_nth_cell(), list_make1, list_make3, makeDefElem(), makeNode, makeRangeVar(), makeRangeVarFromNameList(), makeString(), makeTypeNameFromOid(), CreateSeqStmt::options, AlterSeqStmt::options, CreateSeqStmt::ownerId, CreateStmtContext::pstate, RangeVarAdjustRelationPersistence(), RangeVarGetCreationNamespace(), RelationData::rd_rel, CreateStmtContext::rel, CreateStmtContext::relation, RelationGetNamespace, RangeVar::relname, RangeVar::relpersistence, RangeVar::schemaname, CreateSeqStmt::sequence, AlterSeqStmt::sequence, and CreateStmtContext::stmtType.

Referenced by transformAlterTableStmt(), transformColumnDefinition(), and transformTableLikeClause().

◆ get_collation()

static List * get_collation	(	Oid	collation,
		Oid	actual_datatype
	)

static

Definition at line 2114 of file parse_utilcmd.c.

 {
     List       *result;
     HeapTuple   ht_coll;
     Form_pg_collation coll_rec;
     char       *nsp_name;
     char       *coll_name;
  
     if (!OidIsValid(collation))
         return NIL;             /* easy case */
     if (collation == get_typcollation(actual_datatype))
         return NIL;             /* just let it default */
  
     ht_coll = SearchSysCache1(COLLOID, ObjectIdGetDatum(collation));
     if (!HeapTupleIsValid(ht_coll))
         elog(ERROR, "cache lookup failed for collation %u", collation);
     coll_rec = (Form_pg_collation) GETSTRUCT(ht_coll);
  
     /* For simplicity, we always schema-qualify the name */
     nsp_name = get_namespace_name(coll_rec->collnamespace);
     coll_name = pstrdup(NameStr(coll_rec->collname));
     result = list_make2(makeString(nsp_name), makeString(coll_name));
  
     ReleaseSysCache(ht_coll);
     return result;
 }

References COLLOID, elog(), ERROR, get_namespace_name(), get_typcollation(), GETSTRUCT, HeapTupleIsValid, list_make2, makeString(), NameStr, NIL, ObjectIdGetDatum(), OidIsValid, pstrdup(), ReleaseSysCache(), and SearchSysCache1().

Referenced by generateClonedIndexStmt().

◆ get_opclass()

static List * get_opclass	(	Oid	opclass,
		Oid	actual_datatype
	)

static

Definition at line 2148 of file parse_utilcmd.c.

 {
     List       *result = NIL;
     HeapTuple   ht_opc;
     Form_pg_opclass opc_rec;
  
     ht_opc = SearchSysCache1(CLAOID, ObjectIdGetDatum(opclass));
     if (!HeapTupleIsValid(ht_opc))
         elog(ERROR, "cache lookup failed for opclass %u", opclass);
     opc_rec = (Form_pg_opclass) GETSTRUCT(ht_opc);
  
     if (GetDefaultOpClass(actual_datatype, opc_rec->opcmethod) != opclass)
     {
         /* For simplicity, we always schema-qualify the name */
         char       *nsp_name = get_namespace_name(opc_rec->opcnamespace);
         char       *opc_name = pstrdup(NameStr(opc_rec->opcname));
  
         result = list_make2(makeString(nsp_name), makeString(opc_name));
     }
  
     ReleaseSysCache(ht_opc);
     return result;
 }

References CLAOID, elog(), ERROR, get_namespace_name(), GetDefaultOpClass(), GETSTRUCT, HeapTupleIsValid, list_make2, makeString(), NameStr, NIL, ObjectIdGetDatum(), pstrdup(), ReleaseSysCache(), and SearchSysCache1().

Referenced by generateClonedIndexStmt().

◆ setSchemaName()

static void setSchemaName	(	const char *	context_schema,
		char **	stmt_schema_name
	)

static

Definition at line 4037 of file parse_utilcmd.c.

 {
     if (*stmt_schema_name == NULL)
         *stmt_schema_name = unconstify(char *, context_schema);
     else if (strcmp(context_schema, *stmt_schema_name) != 0)
         ereport(ERROR,
                 (errcode(ERRCODE_INVALID_SCHEMA_DEFINITION),
                  errmsg("CREATE specifies a schema (%s) "
                         "different from the one being created (%s)",
                         *stmt_schema_name, context_schema)));
 }

References ereport, errcode(), errmsg(), ERROR, and unconstify.

Referenced by transformCreateSchemaStmtElements().

◆ transformAlterTableStmt()

AlterTableStmt* transformAlterTableStmt	(	Oid	relid,
		AlterTableStmt *	stmt,
		const char *	queryString,
		List **	beforeStmts,
		List **	afterStmts
	)

Definition at line 3394 of file parse_utilcmd.c.

 {
     Relation    rel;
     TupleDesc   tupdesc;
     ParseState *pstate;
     CreateStmtContext cxt;
     List       *save_alist;
     ListCell   *lcmd,
                *l;
     List       *newcmds = NIL;
     bool        skipValidation = true;
     AlterTableCmd *newcmd;
     ParseNamespaceItem *nsitem;
  
     /* Caller is responsible for locking the relation */
     rel = relation_open(relid, NoLock);
     tupdesc = RelationGetDescr(rel);
  
     /* Set up pstate */
     pstate = make_parsestate(NULL);
     pstate->p_sourcetext = queryString;
     nsitem = addRangeTableEntryForRelation(pstate,
                                            rel,
                                            AccessShareLock,
                                            NULL,
                                            false,
                                            true);
     addNSItemToQuery(pstate, nsitem, false, true, true);
  
     /* Set up CreateStmtContext */
     cxt.pstate = pstate;
     if (rel->rd_rel->relkind == RELKIND_FOREIGN_TABLE)
     {
         cxt.stmtType = "ALTER FOREIGN TABLE";
         cxt.isforeign = true;
     }
     else
     {
         cxt.stmtType = "ALTER TABLE";
         cxt.isforeign = false;
     }
     cxt.relation = stmt->relation;
     cxt.rel = rel;
     cxt.inhRelations = NIL;
     cxt.isalter = true;
     cxt.columns = NIL;
     cxt.ckconstraints = NIL;
     cxt.nnconstraints = NIL;
     cxt.fkconstraints = NIL;
     cxt.ixconstraints = NIL;
     cxt.likeclauses = NIL;
     cxt.extstats = NIL;
     cxt.blist = NIL;
     cxt.alist = NIL;
     cxt.pkey = NULL;
     cxt.ispartitioned = (rel->rd_rel->relkind == RELKIND_PARTITIONED_TABLE);
     cxt.partbound = NULL;
     cxt.ofType = false;
  
     /*
      * Transform ALTER subcommands that need it (most don't).  These largely
      * re-use code from CREATE TABLE.
      */
     foreach(lcmd, stmt->cmds)
     {
         AlterTableCmd *cmd = (AlterTableCmd *) lfirst(lcmd);
  
         switch (cmd->subtype)
         {
             case AT_AddColumn:
                 {
                     ColumnDef  *def = castNode(ColumnDef, cmd->def);
  
                     transformColumnDefinition(&cxt, def);
  
                     /*
                      * If the column has a non-null default, we can't skip
                      * validation of foreign keys.
                      */
                     if (def->raw_default != NULL)
                         skipValidation = false;
  
                     /*
                      * All constraints are processed in other ways. Remove the
                      * original list
                      */
                     def->constraints = NIL;
  
                     newcmds = lappend(newcmds, cmd);
                     break;
                 }
  
             case AT_AddConstraint:
  
                 /*
                  * The original AddConstraint cmd node doesn't go to newcmds
                  */
                 if (IsA(cmd->def, Constraint))
                 {
                     transformTableConstraint(&cxt, (Constraint *) cmd->def);
                     if (((Constraint *) cmd->def)->contype == CONSTR_FOREIGN)
                         skipValidation = false;
                 }
                 else
                     elog(ERROR, "unrecognized node type: %d",
                          (int) nodeTag(cmd->def));
                 break;
  
             case AT_AlterColumnType:
                 {
                     ColumnDef  *def = castNode(ColumnDef, cmd->def);
                     AttrNumber  attnum;
  
                     /*
                      * For ALTER COLUMN TYPE, transform the USING clause if
                      * one was specified.
                      */
                     if (def->raw_default)
                     {
                         def->cooked_default =
                             transformExpr(pstate, def->raw_default,
                                           EXPR_KIND_ALTER_COL_TRANSFORM);
                     }
  
                     /*
                      * For identity column, create ALTER SEQUENCE command to
                      * change the data type of the sequence.
                      */
                     attnum = get_attnum(relid, cmd->name);
                     if (attnum == InvalidAttrNumber)
                         ereport(ERROR,
                                 (errcode(ERRCODE_UNDEFINED_COLUMN),
                                  errmsg("column \"%s\" of relation \"%s\" does not exist",
                                         cmd->name, RelationGetRelationName(rel))));
  
                     if (attnum > 0 &&
                         TupleDescAttr(tupdesc, attnum - 1)->attidentity)
                     {
                         Oid         seq_relid = getIdentitySequence(relid, attnum, false);
                         Oid         typeOid = typenameTypeId(pstate, def->typeName);
                         AlterSeqStmt *altseqstmt = makeNode(AlterSeqStmt);
  
                         altseqstmt->sequence = makeRangeVar(get_namespace_name(get_rel_namespace(seq_relid)),
                                                             get_rel_name(seq_relid),
                                                             -1);
                         altseqstmt->options = list_make1(makeDefElem("as", (Node *) makeTypeNameFromOid(typeOid, -1), -1));
                         altseqstmt->for_identity = true;
                         cxt.blist = lappend(cxt.blist, altseqstmt);
                     }
  
                     newcmds = lappend(newcmds, cmd);
                     break;
                 }
  
             case AT_AddIdentity:
                 {
                     Constraint *def = castNode(Constraint, cmd->def);
                     ColumnDef  *newdef = makeNode(ColumnDef);
                     AttrNumber  attnum;
  
                     newdef->colname = cmd->name;
                     newdef->identity = def->generated_when;
                     cmd->def = (Node *) newdef;
  
                     attnum = get_attnum(relid, cmd->name);
                     if (attnum == InvalidAttrNumber)
                         ereport(ERROR,
                                 (errcode(ERRCODE_UNDEFINED_COLUMN),
                                  errmsg("column \"%s\" of relation \"%s\" does not exist",
                                         cmd->name, RelationGetRelationName(rel))));
  
                     generateSerialExtraStmts(&cxt, newdef,
                                              get_atttype(relid, attnum),
                                              def->options, true, true,
                                              NULL, NULL);
  
                     newcmds = lappend(newcmds, cmd);
                     break;
                 }
  
             case AT_SetIdentity:
                 {
                     /*
                      * Create an ALTER SEQUENCE statement for the internal
                      * sequence of the identity column.
                      */
                     ListCell   *lc;
                     List       *newseqopts = NIL;
                     List       *newdef = NIL;
                     AttrNumber  attnum;
                     Oid         seq_relid;
  
                     /*
                      * Split options into those handled by ALTER SEQUENCE and
                      * those for ALTER TABLE proper.
                      */
                     foreach(lc, castNode(List, cmd->def))
                     {
                         DefElem    *def = lfirst_node(DefElem, lc);
  
                         if (strcmp(def->defname, "generated") == 0)
                             newdef = lappend(newdef, def);
                         else
                             newseqopts = lappend(newseqopts, def);
                     }
  
                     attnum = get_attnum(relid, cmd->name);
                     if (attnum == InvalidAttrNumber)
                         ereport(ERROR,
                                 (errcode(ERRCODE_UNDEFINED_COLUMN),
                                  errmsg("column \"%s\" of relation \"%s\" does not exist",
                                         cmd->name, RelationGetRelationName(rel))));
  
                     seq_relid = getIdentitySequence(relid, attnum, true);
  
                     if (seq_relid)
                     {
                         AlterSeqStmt *seqstmt;
  
                         seqstmt = makeNode(AlterSeqStmt);
                         seqstmt->sequence = makeRangeVar(get_namespace_name(get_rel_namespace(seq_relid)),
                                                          get_rel_name(seq_relid), -1);
                         seqstmt->options = newseqopts;
                         seqstmt->for_identity = true;
                         seqstmt->missing_ok = false;
  
                         cxt.blist = lappend(cxt.blist, seqstmt);
                     }
  
                     /*
                      * If column was not an identity column, we just let the
                      * ALTER TABLE command error out later.  (There are cases
                      * this fails to cover, but we'll need to restructure
                      * where creation of the sequence dependency linkage
                      * happens before we can fix it.)
                      */
  
                     cmd->def = (Node *) newdef;
                     newcmds = lappend(newcmds, cmd);
                     break;
                 }
  
             case AT_AttachPartition:
             case AT_DetachPartition:
                 {
                     PartitionCmd *partcmd = (PartitionCmd *) cmd->def;
  
                     transformPartitionCmd(&cxt, partcmd);
                     /* assign transformed value of the partition bound */
                     partcmd->bound = cxt.partbound;
                 }
  
                 newcmds = lappend(newcmds, cmd);
                 break;
  
             default:
  
                 /*
                  * Currently, we shouldn't actually get here for subcommand
                  * types that don't require transformation; but if we do, just
                  * emit them unchanged.
                  */
                 newcmds = lappend(newcmds, cmd);
                 break;
         }
     }
  
     /*
      * Transfer anything we already have in cxt.alist into save_alist, to keep
      * it separate from the output of transformIndexConstraints.
      */
     save_alist = cxt.alist;
     cxt.alist = NIL;
  
     /* Postprocess constraints */
     transformIndexConstraints(&cxt);
     transformFKConstraints(&cxt, skipValidation, true);
     transformCheckConstraints(&cxt, false);
  
     /*
      * Push any index-creation commands into the ALTER, so that they can be
      * scheduled nicely by tablecmds.c.  Note that tablecmds.c assumes that
      * the IndexStmt attached to an AT_AddIndex or AT_AddIndexConstraint
      * subcommand has already been through transformIndexStmt.
      */
     foreach(l, cxt.alist)
     {
         Node       *istmt = (Node *) lfirst(l);
  
         /*
          * We assume here that cxt.alist contains only IndexStmts and possibly
          * AT_SetAttNotNull statements generated from primary key constraints.
          * We absorb the subcommands of the latter directly.
          */
         if (IsA(istmt, IndexStmt))
         {
             IndexStmt  *idxstmt = (IndexStmt *) istmt;
  
             idxstmt = transformIndexStmt(relid, idxstmt, queryString);
             newcmd = makeNode(AlterTableCmd);
             newcmd->subtype = OidIsValid(idxstmt->indexOid) ? AT_AddIndexConstraint : AT_AddIndex;
             newcmd->def = (Node *) idxstmt;
             newcmds = lappend(newcmds, newcmd);
         }
         else if (IsA(istmt, AlterTableStmt))
         {
             AlterTableStmt *alterstmt = (AlterTableStmt *) istmt;
  
             newcmds = list_concat(newcmds, alterstmt->cmds);
         }
         else
             elog(ERROR, "unexpected stmt type %d", (int) nodeTag(istmt));
     }
     cxt.alist = NIL;
  
     /* Append any CHECK, NOT NULL or FK constraints to the commands list */
     foreach(l, cxt.ckconstraints)
     {
         newcmd = makeNode(AlterTableCmd);
         newcmd->subtype = AT_AddConstraint;
         newcmd->def = (Node *) lfirst_node(Constraint, l);
         newcmds = lappend(newcmds, newcmd);
     }
     foreach(l, cxt.nnconstraints)
     {
         newcmd = makeNode(AlterTableCmd);
         newcmd->subtype = AT_AddConstraint;
         newcmd->def = (Node *) lfirst_node(Constraint, l);
         newcmds = lappend(newcmds, newcmd);
     }
     foreach(l, cxt.fkconstraints)
     {
         newcmd = makeNode(AlterTableCmd);
         newcmd->subtype = AT_AddConstraint;
         newcmd->def = (Node *) lfirst_node(Constraint, l);
         newcmds = lappend(newcmds, newcmd);
     }
  
     /* Append extended statistics objects */
     transformExtendedStatistics(&cxt);
  
     /* Close rel */
     relation_close(rel, NoLock);
  
     /*
      * Output results.
      */
     stmt->cmds = newcmds;
  
     *beforeStmts = cxt.blist;
     *afterStmts = list_concat(cxt.alist, save_alist);
  
     return stmt;
 }

Referenced by ATParseTransformCmd(), and ATPostAlterTypeParse().

◆ transformCheckConstraints()

static void transformCheckConstraints	(	CreateStmtContext *	cxt,
		bool	skipValidation
	)

static

Definition at line 2818 of file parse_utilcmd.c.

 {
     ListCell   *ckclist;
  
     if (cxt->ckconstraints == NIL)
         return;
  
     /*
      * If creating a new table (but not a foreign table), we can safely skip
      * validation of check constraints, and nonetheless mark them valid. (This
      * will override any user-supplied NOT VALID flag.)
      */
     if (skipValidation)
     {
         foreach(ckclist, cxt->ckconstraints)
         {
             Constraint *constraint = (Constraint *) lfirst(ckclist);
  
             constraint->skip_validation = true;
             constraint->initially_valid = true;
         }
     }
 }

References CreateStmtContext::ckconstraints, Constraint::initially_valid, lfirst, NIL, and Constraint::skip_validation.

Referenced by transformAlterTableStmt(), and transformCreateStmt().

◆ transformColumnDefinition()

static void transformColumnDefinition	(	CreateStmtContext *	cxt,
		ColumnDef *	column
	)

static

Definition at line 534 of file parse_utilcmd.c.

 {
     bool        is_serial;
     bool        saw_nullable;
     bool        saw_default;
     bool        saw_identity;
     bool        saw_generated;
     bool        need_notnull = false;
     ListCell   *clist;
  
     cxt->columns = lappend(cxt->columns, column);
  
     /* Check for SERIAL pseudo-types */
     is_serial = false;
     if (column->typeName
         && list_length(column->typeName->names) == 1
         && !column->typeName->pct_type)
     {
         char       *typname = strVal(linitial(column->typeName->names));
  
         if (strcmp(typname, "smallserial") == 0 ||
             strcmp(typname, "serial2") == 0)
         {
             is_serial = true;
             column->typeName->names = NIL;
             column->typeName->typeOid = INT2OID;
         }
         else if (strcmp(typname, "serial") == 0 ||
                  strcmp(typname, "serial4") == 0)
         {
             is_serial = true;
             column->typeName->names = NIL;
             column->typeName->typeOid = INT4OID;
         }
         else if (strcmp(typname, "bigserial") == 0 ||
                  strcmp(typname, "serial8") == 0)
         {
             is_serial = true;
             column->typeName->names = NIL;
             column->typeName->typeOid = INT8OID;
         }
  
         /*
          * We have to reject "serial[]" explicitly, because once we've set
          * typeid, LookupTypeName won't notice arrayBounds.  We don't need any
          * special coding for serial(typmod) though.
          */
         if (is_serial && column->typeName->arrayBounds != NIL)
             ereport(ERROR,
                     (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
                      errmsg("array of serial is not implemented"),
                      parser_errposition(cxt->pstate,
                                         column->typeName->location)));
     }
  
     /* Do necessary work on the column type declaration */
     if (column->typeName)
         transformColumnType(cxt, column);
  
     /* Special actions for SERIAL pseudo-types */
     if (is_serial)
     {
         char       *snamespace;
         char       *sname;
         char       *qstring;
         A_Const    *snamenode;
         TypeCast   *castnode;
         FuncCall   *funccallnode;
         Constraint *constraint;
  
         generateSerialExtraStmts(cxt, column,
                                  column->typeName->typeOid, NIL,
                                  false, false,
                                  &snamespace, &sname);
  
         /*
          * Create appropriate constraints for SERIAL.  We do this in full,
          * rather than shortcutting, so that we will detect any conflicting
          * constraints the user wrote (like a different DEFAULT).
          *
          * Create an expression tree representing the function call
          * nextval('sequencename').  We cannot reduce the raw tree to cooked
          * form until after the sequence is created, but there's no need to do
          * so.
          */
         qstring = quote_qualified_identifier(snamespace, sname);
         snamenode = makeNode(A_Const);
         snamenode->val.node.type = T_String;
         snamenode->val.sval.sval = qstring;
         snamenode->location = -1;
         castnode = makeNode(TypeCast);
         castnode->typeName = SystemTypeName("regclass");
         castnode->arg = (Node *) snamenode;
         castnode->location = -1;
         funccallnode = makeFuncCall(SystemFuncName("nextval"),
                                     list_make1(castnode),
                                     COERCE_EXPLICIT_CALL,
                                     -1);
         constraint = makeNode(Constraint);
         constraint->contype = CONSTR_DEFAULT;
         constraint->location = -1;
         constraint->raw_expr = (Node *) funccallnode;
         constraint->cooked_expr = NULL;
         column->constraints = lappend(column->constraints, constraint);
  
         /* have a not-null constraint added later */
         need_notnull = true;
     }
  
     /* Process column constraints, if any... */
     transformConstraintAttrs(cxt, column->constraints);
  
     saw_nullable = false;
     saw_default = false;
     saw_identity = false;
     saw_generated = false;
  
     foreach(clist, column->constraints)
     {
         Constraint *constraint = lfirst_node(Constraint, clist);
  
         switch (constraint->contype)
         {
             case CONSTR_NULL:
                 if ((saw_nullable && column->is_not_null) || need_notnull)
                     ereport(ERROR,
                             (errcode(ERRCODE_SYNTAX_ERROR),
                              errmsg("conflicting NULL/NOT NULL declarations for column \"%s\" of table \"%s\"",
                                     column->colname, cxt->relation->relname),
                              parser_errposition(cxt->pstate,
                                                 constraint->location)));
                 column->is_not_null = false;
                 saw_nullable = true;
                 break;
  
             case CONSTR_NOTNULL:
  
                 /*
                  * Disallow conflicting [NOT] NULL markings
                  */
                 if (saw_nullable && !column->is_not_null)
                     ereport(ERROR,
                             (errcode(ERRCODE_SYNTAX_ERROR),
                              errmsg("conflicting NULL/NOT NULL declarations for column \"%s\" of table \"%s\"",
                                     column->colname, cxt->relation->relname),
                              parser_errposition(cxt->pstate,
                                                 constraint->location)));
                 /* Ignore redundant NOT NULL markings */
  
                 /*
                  * If this is the first time we see this column being marked
                  * not null, add the constraint entry; and get rid of any
                  * previous markings to mark the column NOT NULL.
                  */
                 if (!column->is_not_null)
                 {
                     column->is_not_null = true;
                     saw_nullable = true;
  
                     constraint->keys = list_make1(makeString(column->colname));
                     cxt->nnconstraints = lappend(cxt->nnconstraints, constraint);
  
                     /* Don't need this anymore, if we had it */
                     need_notnull = false;
                 }
  
                 break;
  
             case CONSTR_DEFAULT:
                 if (saw_default)
                     ereport(ERROR,
                             (errcode(ERRCODE_SYNTAX_ERROR),
                              errmsg("multiple default values specified for column \"%s\" of table \"%s\"",
                                     column->colname, cxt->relation->relname),
                              parser_errposition(cxt->pstate,
                                                 constraint->location)));
                 column->raw_default = constraint->raw_expr;
                 Assert(constraint->cooked_expr == NULL);
                 saw_default = true;
                 break;
  
             case CONSTR_IDENTITY:
                 {
                     Type        ctype;
                     Oid         typeOid;
  
                     if (cxt->ofType)
                         ereport(ERROR,
                                 (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
                                  errmsg("identity columns are not supported on typed tables")));
                     if (cxt->partbound)
                         ereport(ERROR,
                                 (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
                                  errmsg("identity columns are not supported on partitions")));
  
                     ctype = typenameType(cxt->pstate, column->typeName, NULL);
                     typeOid = ((Form_pg_type) GETSTRUCT(ctype))->oid;
                     ReleaseSysCache(ctype);
  
                     if (saw_identity)
                         ereport(ERROR,
                                 (errcode(ERRCODE_SYNTAX_ERROR),
                                  errmsg("multiple identity specifications for column \"%s\" of table \"%s\"",
                                         column->colname, cxt->relation->relname),
                                  parser_errposition(cxt->pstate,
                                                     constraint->location)));
  
                     generateSerialExtraStmts(cxt, column,
                                              typeOid, constraint->options,
                                              true, false,
                                              NULL, NULL);
  
                     column->identity = constraint->generated_when;
                     saw_identity = true;
  
                     /*
                      * Identity columns are always NOT NULL, but we may have a
                      * constraint already.
                      */
                     if (!saw_nullable)
                         need_notnull = true;
                     else if (!column->is_not_null)
                         ereport(ERROR,
                                 (errcode(ERRCODE_SYNTAX_ERROR),
                                  errmsg("conflicting NULL/NOT NULL declarations for column \"%s\" of table \"%s\"",
                                         column->colname, cxt->relation->relname),
                                  parser_errposition(cxt->pstate,
                                                     constraint->location)));
                     break;
                 }
  
             case CONSTR_GENERATED:
                 if (cxt->ofType)
                     ereport(ERROR,
                             (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
                              errmsg("generated columns are not supported on typed tables")));
                 if (saw_generated)
                     ereport(ERROR,
                             (errcode(ERRCODE_SYNTAX_ERROR),
                              errmsg("multiple generation clauses specified for column \"%s\" of table \"%s\"",
                                     column->colname, cxt->relation->relname),
                              parser_errposition(cxt->pstate,
                                                 constraint->location)));
                 column->generated = ATTRIBUTE_GENERATED_STORED;
                 column->raw_default = constraint->raw_expr;
                 Assert(constraint->cooked_expr == NULL);
                 saw_generated = true;
                 break;
  
             case CONSTR_CHECK:
                 cxt->ckconstraints = lappend(cxt->ckconstraints, constraint);
                 break;
  
             case CONSTR_PRIMARY:
                 if (cxt->isforeign)
                     ereport(ERROR,
                             (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
                              errmsg("primary key constraints are not supported on foreign tables"),
                              parser_errposition(cxt->pstate,
                                                 constraint->location)));
                 /* FALL THRU */
  
             case CONSTR_UNIQUE:
                 if (cxt->isforeign)
                     ereport(ERROR,
                             (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
                              errmsg("unique constraints are not supported on foreign tables"),
                              parser_errposition(cxt->pstate,
                                                 constraint->location)));
                 if (constraint->keys == NIL)
                     constraint->keys = list_make1(makeString(column->colname));
                 cxt->ixconstraints = lappend(cxt->ixconstraints, constraint);
                 break;
  
             case CONSTR_EXCLUSION:
                 /* grammar does not allow EXCLUDE as a column constraint */
                 elog(ERROR, "column exclusion constraints are not supported");
                 break;
  
             case CONSTR_FOREIGN:
                 if (cxt->isforeign)
                     ereport(ERROR,
                             (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
                              errmsg("foreign key constraints are not supported on foreign tables"),
                              parser_errposition(cxt->pstate,
                                                 constraint->location)));
  
                 /*
                  * Fill in the current attribute's name and throw it into the
                  * list of FK constraints to be processed later.
                  */
                 constraint->fk_attrs = list_make1(makeString(column->colname));
                 cxt->fkconstraints = lappend(cxt->fkconstraints, constraint);
                 break;
  
             case CONSTR_ATTR_DEFERRABLE:
             case CONSTR_ATTR_NOT_DEFERRABLE:
             case CONSTR_ATTR_DEFERRED:
             case CONSTR_ATTR_IMMEDIATE:
                 /* transformConstraintAttrs took care of these */
                 break;
  
             default:
                 elog(ERROR, "unrecognized constraint type: %d",
                      constraint->contype);
                 break;
         }
  
         if (saw_default && saw_identity)
             ereport(ERROR,
                     (errcode(ERRCODE_SYNTAX_ERROR),
                      errmsg("both default and identity specified for column \"%s\" of table \"%s\"",
                             column->colname, cxt->relation->relname),
                      parser_errposition(cxt->pstate,
                                         constraint->location)));
  
         if (saw_default && saw_generated)
             ereport(ERROR,
                     (errcode(ERRCODE_SYNTAX_ERROR),
                      errmsg("both default and generation expression specified for column \"%s\" of table \"%s\"",
                             column->colname, cxt->relation->relname),
                      parser_errposition(cxt->pstate,
                                         constraint->location)));
  
         if (saw_identity && saw_generated)
             ereport(ERROR,
                     (errcode(ERRCODE_SYNTAX_ERROR),
                      errmsg("both identity and generation expression specified for column \"%s\" of table \"%s\"",
                             column->colname, cxt->relation->relname),
                      parser_errposition(cxt->pstate,
                                         constraint->location)));
     }
  
     /*
      * If we need a not-null constraint for SERIAL or IDENTITY, and one was
      * not explicitly specified, add one now.
      */
     if (need_notnull && !(saw_nullable && column->is_not_null))
     {
         Constraint *notnull;
  
         column->is_not_null = true;
  
         notnull = makeNode(Constraint);
         notnull->contype = CONSTR_NOTNULL;
         notnull->conname = NULL;
         notnull->deferrable = false;
         notnull->initdeferred = false;
         notnull->location = -1;
         notnull->keys = list_make1(makeString(column->colname));
         notnull->skip_validation = false;
         notnull->initially_valid = true;
  
         cxt->nnconstraints = lappend(cxt->nnconstraints, notnull);
     }
  
     /*
      * If needed, generate ALTER FOREIGN TABLE ALTER COLUMN statement to add
      * per-column foreign data wrapper options to this column after creation.
      */
     if (column->fdwoptions != NIL)
     {
         AlterTableStmt *stmt;
         AlterTableCmd *cmd;
  
         cmd = makeNode(AlterTableCmd);
         cmd->subtype = AT_AlterColumnGenericOptions;
         cmd->name = column->colname;
         cmd->def = (Node *) column->fdwoptions;
         cmd->behavior = DROP_RESTRICT;
         cmd->missing_ok = false;
  
         stmt = makeNode(AlterTableStmt);
         stmt->relation = cxt->relation;
         stmt->cmds = NIL;
         stmt->objtype = OBJECT_FOREIGN_TABLE;
         stmt->cmds = lappend(stmt->cmds, cmd);
  
         cxt->alist = lappend(cxt->alist, stmt);
     }
 }

Referenced by transformAlterTableStmt(), and transformCreateStmt().

◆ transformColumnType()

static void transformColumnType	(	CreateStmtContext *	cxt,
		ColumnDef *	column
	)

static

Definition at line 3877 of file parse_utilcmd.c.

 {
     /*
      * All we really need to do here is verify that the type is valid,
      * including any collation spec that might be present.
      */
     Type        ctype = typenameType(cxt->pstate, column->typeName, NULL);
  
     if (column->collClause)
     {
         Form_pg_type typtup = (Form_pg_type) GETSTRUCT(ctype);
  
         LookupCollation(cxt->pstate,
                         column->collClause->collname,
                         column->collClause->location);
         /* Complain if COLLATE is applied to an uncollatable type */
         if (!OidIsValid(typtup->typcollation))
             ereport(ERROR,
                     (errcode(ERRCODE_DATATYPE_MISMATCH),
                      errmsg("collations are not supported by type %s",
                             format_type_be(typtup->oid)),
                      parser_errposition(cxt->pstate,
                                         column->collClause->location)));
     }
  
     ReleaseSysCache(ctype);
 }

References ColumnDef::collClause, CollateClause::collname, ereport, errcode(), errmsg(), ERROR, format_type_be(), GETSTRUCT, CollateClause::location, LookupCollation(), OidIsValid, parser_errposition(), CreateStmtContext::pstate, ReleaseSysCache(), ColumnDef::typeName, and typenameType().

Referenced by transformColumnDefinition().

◆ transformConstraintAttrs()

static void transformConstraintAttrs	(	CreateStmtContext *	cxt,
		List *	constraintList
	)

static

Definition at line 3762 of file parse_utilcmd.c.

 {
     Constraint *lastprimarycon = NULL;
     bool        saw_deferrability = false;
     bool        saw_initially = false;
     ListCell   *clist;
  
 #define SUPPORTS_ATTRS(node)                \
     ((node) != NULL &&                      \
      ((node)->contype == CONSTR_PRIMARY ||  \
       (node)->contype == CONSTR_UNIQUE ||   \
       (node)->contype == CONSTR_EXCLUSION || \
       (node)->contype == CONSTR_FOREIGN))
  
     foreach(clist, constraintList)
     {
         Constraint *con = (Constraint *) lfirst(clist);
  
         if (!IsA(con, Constraint))
             elog(ERROR, "unrecognized node type: %d",
                  (int) nodeTag(con));
         switch (con->contype)
         {
             case CONSTR_ATTR_DEFERRABLE:
                 if (!SUPPORTS_ATTRS(lastprimarycon))
                     ereport(ERROR,
                             (errcode(ERRCODE_SYNTAX_ERROR),
                              errmsg("misplaced DEFERRABLE clause"),
                              parser_errposition(cxt->pstate, con->location)));
                 if (saw_deferrability)
                     ereport(ERROR,
                             (errcode(ERRCODE_SYNTAX_ERROR),
                              errmsg("multiple DEFERRABLE/NOT DEFERRABLE clauses not allowed"),
                              parser_errposition(cxt->pstate, con->location)));
                 saw_deferrability = true;
                 lastprimarycon->deferrable = true;
                 break;
  
             case CONSTR_ATTR_NOT_DEFERRABLE:
                 if (!SUPPORTS_ATTRS(lastprimarycon))
                     ereport(ERROR,
                             (errcode(ERRCODE_SYNTAX_ERROR),
                              errmsg("misplaced NOT DEFERRABLE clause"),
                              parser_errposition(cxt->pstate, con->location)));
                 if (saw_deferrability)
                     ereport(ERROR,
                             (errcode(ERRCODE_SYNTAX_ERROR),
                              errmsg("multiple DEFERRABLE/NOT DEFERRABLE clauses not allowed"),
                              parser_errposition(cxt->pstate, con->location)));
                 saw_deferrability = true;
                 lastprimarycon->deferrable = false;
                 if (saw_initially &&
                     lastprimarycon->initdeferred)
                     ereport(ERROR,
                             (errcode(ERRCODE_SYNTAX_ERROR),
                              errmsg("constraint declared INITIALLY DEFERRED must be DEFERRABLE"),
                              parser_errposition(cxt->pstate, con->location)));
                 break;
  
             case CONSTR_ATTR_DEFERRED:
                 if (!SUPPORTS_ATTRS(lastprimarycon))
                     ereport(ERROR,
                             (errcode(ERRCODE_SYNTAX_ERROR),
                              errmsg("misplaced INITIALLY DEFERRED clause"),
                              parser_errposition(cxt->pstate, con->location)));
                 if (saw_initially)
                     ereport(ERROR,
                             (errcode(ERRCODE_SYNTAX_ERROR),
                              errmsg("multiple INITIALLY IMMEDIATE/DEFERRED clauses not allowed"),
                              parser_errposition(cxt->pstate, con->location)));
                 saw_initially = true;
                 lastprimarycon->initdeferred = true;
  
                 /*
                  * If only INITIALLY DEFERRED appears, assume DEFERRABLE
                  */
                 if (!saw_deferrability)
                     lastprimarycon->deferrable = true;
                 else if (!lastprimarycon->deferrable)
                     ereport(ERROR,
                             (errcode(ERRCODE_SYNTAX_ERROR),
                              errmsg("constraint declared INITIALLY DEFERRED must be DEFERRABLE"),
                              parser_errposition(cxt->pstate, con->location)));
                 break;
  
             case CONSTR_ATTR_IMMEDIATE:
                 if (!SUPPORTS_ATTRS(lastprimarycon))
                     ereport(ERROR,
                             (errcode(ERRCODE_SYNTAX_ERROR),
                              errmsg("misplaced INITIALLY IMMEDIATE clause"),
                              parser_errposition(cxt->pstate, con->location)));
                 if (saw_initially)
                     ereport(ERROR,
                             (errcode(ERRCODE_SYNTAX_ERROR),
                              errmsg("multiple INITIALLY IMMEDIATE/DEFERRED clauses not allowed"),
                              parser_errposition(cxt->pstate, con->location)));
                 saw_initially = true;
                 lastprimarycon->initdeferred = false;
                 break;
  
             default:
                 /* Otherwise it's not an attribute */
                 lastprimarycon = con;
                 /* reset flags for new primary node */
                 saw_deferrability = false;
                 saw_initially = false;
                 break;
         }
     }
 }

References CONSTR_ATTR_DEFERRABLE, CONSTR_ATTR_DEFERRED, CONSTR_ATTR_IMMEDIATE, CONSTR_ATTR_NOT_DEFERRABLE, Constraint::contype, Constraint::deferrable, elog(), ereport, errcode(), errmsg(), ERROR, Constraint::initdeferred, IsA, lfirst, Constraint::location, nodeTag, parser_errposition(), CreateStmtContext::pstate, and SUPPORTS_ATTRS.

Referenced by transformColumnDefinition().

◆ transformCreateSchemaStmtElements()

List* transformCreateSchemaStmtElements	(	List *	schemaElts,
		const char *	schemaName
	)

Definition at line 3934 of file parse_utilcmd.c.

 {
     CreateSchemaStmtContext cxt;
     List       *result;
     ListCell   *elements;
  
     cxt.schemaname = schemaName;
     cxt.sequences = NIL;
     cxt.tables = NIL;
     cxt.views = NIL;
     cxt.indexes = NIL;
     cxt.triggers = NIL;
     cxt.grants = NIL;
  
     /*
      * Run through each schema element in the schema element list. Separate
      * statements by type, and do preliminary analysis.
      */
     foreach(elements, schemaElts)
     {
         Node       *element = lfirst(elements);
  
         switch (nodeTag(element))
         {
             case T_CreateSeqStmt:
                 {
                     CreateSeqStmt *elp = (CreateSeqStmt *) element;
  
                     setSchemaName(cxt.schemaname, &elp->sequence->schemaname);
                     cxt.sequences = lappend(cxt.sequences, element);
                 }
                 break;
  
             case T_CreateStmt:
                 {
                     CreateStmt *elp = (CreateStmt *) element;
  
                     setSchemaName(cxt.schemaname, &elp->relation->schemaname);
  
                     /*
                      * XXX todo: deal with constraints
                      */
                     cxt.tables = lappend(cxt.tables, element);
                 }
                 break;
  
             case T_ViewStmt:
                 {
                     ViewStmt   *elp = (ViewStmt *) element;
  
                     setSchemaName(cxt.schemaname, &elp->view->schemaname);
  
                     /*
                      * XXX todo: deal with references between views
                      */
                     cxt.views = lappend(cxt.views, element);
                 }
                 break;
  
             case T_IndexStmt:
                 {
                     IndexStmt  *elp = (IndexStmt *) element;
  
                     setSchemaName(cxt.schemaname, &elp->relation->schemaname);
                     cxt.indexes = lappend(cxt.indexes, element);
                 }
                 break;
  
             case T_CreateTrigStmt:
                 {
                     CreateTrigStmt *elp = (CreateTrigStmt *) element;
  
                     setSchemaName(cxt.schemaname, &elp->relation->schemaname);
                     cxt.triggers = lappend(cxt.triggers, element);
                 }
                 break;
  
             case T_GrantStmt:
                 cxt.grants = lappend(cxt.grants, element);
                 break;
  
             default:
                 elog(ERROR, "unrecognized node type: %d",
                      (int) nodeTag(element));
         }
     }
  
     result = NIL;
     result = list_concat(result, cxt.sequences);
     result = list_concat(result, cxt.tables);
     result = list_concat(result, cxt.views);
     result = list_concat(result, cxt.indexes);
     result = list_concat(result, cxt.triggers);
     result = list_concat(result, cxt.grants);
  
     return result;
 }

References element(), elog(), ERROR, CreateSchemaStmtContext::grants, CreateSchemaStmtContext::indexes, lappend(), lfirst, list_concat(), NIL, nodeTag, CreateStmt::relation, CreateTrigStmt::relation, IndexStmt::relation, CreateSchemaStmtContext::schemaname, RangeVar::schemaname, CreateSeqStmt::sequence, CreateSchemaStmtContext::sequences, setSchemaName(), CreateSchemaStmtContext::tables, CreateSchemaStmtContext::triggers, ViewStmt::view, and CreateSchemaStmtContext::views.

Referenced by CreateSchemaCommand().

◆ transformCreateStmt()

List* transformCreateStmt	(	CreateStmt *	stmt,
		const char *	queryString
	)

Definition at line 164 of file parse_utilcmd.c.

 {
     ParseState *pstate;
     CreateStmtContext cxt;
     List       *result;
     List       *save_alist;
     ListCell   *elements;
     Oid         namespaceid;
     Oid         existing_relid;
     ParseCallbackState pcbstate;
  
     /* Set up pstate */
     pstate = make_parsestate(NULL);
     pstate->p_sourcetext = queryString;
  
     /*
      * Look up the creation namespace.  This also checks permissions on the
      * target namespace, locks it against concurrent drops, checks for a
      * preexisting relation in that namespace with the same name, and updates
      * stmt->relation->relpersistence if the selected namespace is temporary.
      */
     setup_parser_errposition_callback(&pcbstate, pstate,
                                       stmt->relation->location);
     namespaceid =
         RangeVarGetAndCheckCreationNamespace(stmt->relation, NoLock,
                                              &existing_relid);
     cancel_parser_errposition_callback(&pcbstate);
  
     /*
      * If the relation already exists and the user specified "IF NOT EXISTS",
      * bail out with a NOTICE.
      */
     if (stmt->if_not_exists && OidIsValid(existing_relid))
     {
         /*
          * If we are in an extension script, insist that the pre-existing
          * object be a member of the extension, to avoid security risks.
          */
         ObjectAddress address;
  
         ObjectAddressSet(address, RelationRelationId, existing_relid);
         checkMembershipInCurrentExtension(&address);
  
         /* OK to skip */
         ereport(NOTICE,
                 (errcode(ERRCODE_DUPLICATE_TABLE),
                  errmsg("relation \"%s\" already exists, skipping",
                         stmt->relation->relname)));
         return NIL;
     }
  
     /*
      * If the target relation name isn't schema-qualified, make it so.  This
      * prevents some corner cases in which added-on rewritten commands might
      * think they should apply to other relations that have the same name and
      * are earlier in the search path.  But a local temp table is effectively
      * specified to be in pg_temp, so no need for anything extra in that case.
      */
     if (stmt->relation->schemaname == NULL
         && stmt->relation->relpersistence != RELPERSISTENCE_TEMP)
         stmt->relation->schemaname = get_namespace_name(namespaceid);
  
     /* Set up CreateStmtContext */
     cxt.pstate = pstate;
     if (IsA(stmt, CreateForeignTableStmt))
     {
         cxt.stmtType = "CREATE FOREIGN TABLE";
         cxt.isforeign = true;
     }
     else
     {
         cxt.stmtType = "CREATE TABLE";
         cxt.isforeign = false;
     }
     cxt.relation = stmt->relation;
     cxt.rel = NULL;
     cxt.inhRelations = stmt->inhRelations;
     cxt.isalter = false;
     cxt.columns = NIL;
     cxt.ckconstraints = NIL;
     cxt.nnconstraints = NIL;
     cxt.fkconstraints = NIL;
     cxt.ixconstraints = NIL;
     cxt.likeclauses = NIL;
     cxt.extstats = NIL;
     cxt.blist = NIL;
     cxt.alist = NIL;
     cxt.pkey = NULL;
     cxt.ispartitioned = stmt->partspec != NULL;
     cxt.partbound = stmt->partbound;
     cxt.ofType = (stmt->ofTypename != NULL);
  
     Assert(!stmt->ofTypename || !stmt->inhRelations);   /* grammar enforces */
  
     if (stmt->ofTypename)
         transformOfType(&cxt, stmt->ofTypename);
  
     if (stmt->partspec)
     {
         if (stmt->inhRelations && !stmt->partbound)
             ereport(ERROR,
                     (errcode(ERRCODE_INVALID_OBJECT_DEFINITION),
                      errmsg("cannot create partitioned table as inheritance child")));
     }
  
     /*
      * Run through each primary element in the table creation clause. Separate
      * column defs from constraints, and do preliminary analysis.
      */
     foreach(elements, stmt->tableElts)
     {
         Node       *element = lfirst(elements);
  
         switch (nodeTag(element))
         {
             case T_ColumnDef:
                 transformColumnDefinition(&cxt, (ColumnDef *) element);
                 break;
  
             case T_Constraint:
                 transformTableConstraint(&cxt, (Constraint *) element);
                 break;
  
             case T_TableLikeClause:
                 transformTableLikeClause(&cxt, (TableLikeClause *) element);
                 break;
  
             default:
                 elog(ERROR, "unrecognized node type: %d",
                      (int) nodeTag(element));
                 break;
         }
     }
  
     /*
      * Transfer anything we already have in cxt.alist into save_alist, to keep
      * it separate from the output of transformIndexConstraints.  (This may
      * not be necessary anymore, but we'll keep doing it to preserve the
      * historical order of execution of the alist commands.)
      */
     save_alist = cxt.alist;
     cxt.alist = NIL;
  
     Assert(stmt->constraints == NIL);
  
     /*
      * Postprocess constraints that give rise to index definitions.
      */
     transformIndexConstraints(&cxt);
  
     /*
      * Re-consideration of LIKE clauses should happen after creation of
      * indexes, but before creation of foreign keys.  This order is critical
      * because a LIKE clause may attempt to create a primary key.  If there's
      * also a pkey in the main CREATE TABLE list, creation of that will not
      * check for a duplicate at runtime (since index_check_primary_key()
      * expects that we rejected dups here).  Creation of the LIKE-generated
      * pkey behaves like ALTER TABLE ADD, so it will check, but obviously that
      * only works if it happens second.  On the other hand, we want to make
      * pkeys before foreign key constraints, in case the user tries to make a
      * self-referential FK.
      */
     cxt.alist = list_concat(cxt.alist, cxt.likeclauses);
  
     /*
      * Postprocess foreign-key constraints.
      */
     transformFKConstraints(&cxt, true, false);
  
     /*
      * Postprocess check constraints.
      *
      * For regular tables all constraints can be marked valid immediately,
      * because the table is new therefore empty. Not so for foreign tables.
      */
     transformCheckConstraints(&cxt, !cxt.isforeign);
  
     /*
      * Postprocess extended statistics.
      */
     transformExtendedStatistics(&cxt);
  
     /*
      * Output results.
      */
     stmt->tableElts = cxt.columns;
     stmt->constraints = cxt.ckconstraints;
     stmt->nnconstraints = cxt.nnconstraints;
  
     result = lappend(cxt.blist, stmt);
     result = list_concat(result, cxt.alist);
     result = list_concat(result, save_alist);
  
     return result;
 }

Referenced by ProcessUtilitySlow().

◆ transformExtendedStatistics()

static void transformExtendedStatistics ( CreateStmtContext * cxt )

static

Definition at line 2803 of file parse_utilcmd.c.

 {
     cxt->alist = list_concat(cxt->alist, cxt->extstats);
 }

References CreateStmtContext::alist, CreateStmtContext::extstats, and list_concat().

Referenced by transformAlterTableStmt(), and transformCreateStmt().

◆ transformFKConstraints()

static void transformFKConstraints	(	CreateStmtContext *	cxt,
		bool	skipValidation,
		bool	isAddConstraint
	)

static

Definition at line 2847 of file parse_utilcmd.c.

 {
     ListCell   *fkclist;
  
     if (cxt->fkconstraints == NIL)
         return;
  
     /*
      * If CREATE TABLE or adding a column with NULL default, we can safely
      * skip validation of FK constraints, and nonetheless mark them valid.
      * (This will override any user-supplied NOT VALID flag.)
      */
     if (skipValidation)
     {
         foreach(fkclist, cxt->fkconstraints)
         {
             Constraint *constraint = (Constraint *) lfirst(fkclist);
  
             constraint->skip_validation = true;
             constraint->initially_valid = true;
         }
     }
  
     /*
      * For CREATE TABLE or ALTER TABLE ADD COLUMN, gin up an ALTER TABLE ADD
      * CONSTRAINT command to execute after the basic command is complete. (If
      * called from ADD CONSTRAINT, that routine will add the FK constraints to
      * its own subcommand list.)
      *
      * Note: the ADD CONSTRAINT command must also execute after any index
      * creation commands.  Thus, this should run after
      * transformIndexConstraints, so that the CREATE INDEX commands are
      * already in cxt->alist.  See also the handling of cxt->likeclauses.
      */
     if (!isAddConstraint)
     {
         AlterTableStmt *alterstmt = makeNode(AlterTableStmt);
  
         alterstmt->relation = cxt->relation;
         alterstmt->cmds = NIL;
         alterstmt->objtype = OBJECT_TABLE;
  
         foreach(fkclist, cxt->fkconstraints)
         {
             Constraint *constraint = (Constraint *) lfirst(fkclist);
             AlterTableCmd *altercmd = makeNode(AlterTableCmd);
  
             altercmd->subtype = AT_AddConstraint;
             altercmd->name = NULL;
             altercmd->def = (Node *) constraint;
             alterstmt->cmds = lappend(alterstmt->cmds, altercmd);
         }
  
         cxt->alist = lappend(cxt->alist, alterstmt);
     }
 }

References CreateStmtContext::alist, AT_AddConstraint, AlterTableStmt::cmds, AlterTableCmd::def, CreateStmtContext::fkconstraints, Constraint::initially_valid, lappend(), lfirst, makeNode, AlterTableCmd::name, NIL, OBJECT_TABLE, AlterTableStmt::objtype, CreateStmtContext::relation, AlterTableStmt::relation, Constraint::skip_validation, and AlterTableCmd::subtype.

Referenced by transformAlterTableStmt(), and transformCreateStmt().

◆ transformIndexConstraint()

static IndexStmt * transformIndexConstraint	(	Constraint *	constraint,
		CreateStmtContext *	cxt
	)

static

Definition at line 2281 of file parse_utilcmd.c.

 {
     IndexStmt  *index;
     List       *notnullcmds = NIL;
     ListCell   *lc;
  
     index = makeNode(IndexStmt);
  
     index->unique = (constraint->contype != CONSTR_EXCLUSION);
     index->primary = (constraint->contype == CONSTR_PRIMARY);
     if (index->primary)
     {
         if (cxt->pkey != NULL)
             ereport(ERROR,
                     (errcode(ERRCODE_INVALID_TABLE_DEFINITION),
                      errmsg("multiple primary keys for table \"%s\" are not allowed",
                             cxt->relation->relname),
                      parser_errposition(cxt->pstate, constraint->location)));
         cxt->pkey = index;
  
         /*
          * In ALTER TABLE case, a primary index might already exist, but
          * DefineIndex will check for it.
          */
     }
     index->nulls_not_distinct = constraint->nulls_not_distinct;
     index->isconstraint = true;
     index->deferrable = constraint->deferrable;
     index->initdeferred = constraint->initdeferred;
  
     if (constraint->conname != NULL)
         index->idxname = pstrdup(constraint->conname);
     else
         index->idxname = NULL;  /* DefineIndex will choose name */
  
     index->relation = cxt->relation;
     index->accessMethod = constraint->access_method ? constraint->access_method : DEFAULT_INDEX_TYPE;
     index->options = constraint->options;
     index->tableSpace = constraint->indexspace;
     index->whereClause = constraint->where_clause;
     index->indexParams = NIL;
     index->indexIncludingParams = NIL;
     index->excludeOpNames = NIL;
     index->idxcomment = NULL;
     index->indexOid = InvalidOid;
     index->oldNumber = InvalidRelFileNumber;
     index->oldCreateSubid = InvalidSubTransactionId;
     index->oldFirstRelfilelocatorSubid = InvalidSubTransactionId;
     index->transformed = false;
     index->concurrent = false;
     index->if_not_exists = false;
     index->reset_default_tblspc = constraint->reset_default_tblspc;
  
     /*
      * If it's ALTER TABLE ADD CONSTRAINT USING INDEX, look up the index and
      * verify it's usable, then extract the implied column name list.  (We
      * will not actually need the column name list at runtime, but we need it
      * now to check for duplicate column entries below.)
      */
     if (constraint->indexname != NULL)
     {
         char       *index_name = constraint->indexname;
         Relation    heap_rel = cxt->rel;
         Oid         index_oid;
         Relation    index_rel;
         Form_pg_index index_form;
         oidvector  *indclass;
         Datum       indclassDatum;
         int         i;
  
         /* Grammar should not allow this with explicit column list */
         Assert(constraint->keys == NIL);
  
         /* Grammar should only allow PRIMARY and UNIQUE constraints */
         Assert(constraint->contype == CONSTR_PRIMARY ||
                constraint->contype == CONSTR_UNIQUE);
  
         /* Must be ALTER, not CREATE, but grammar doesn't enforce that */
         if (!cxt->isalter)
             ereport(ERROR,
                     (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
                      errmsg("cannot use an existing index in CREATE TABLE"),
                      parser_errposition(cxt->pstate, constraint->location)));
  
         /* Look for the index in the same schema as the table */
         index_oid = get_relname_relid(index_name, RelationGetNamespace(heap_rel));
  
         if (!OidIsValid(index_oid))
             ereport(ERROR,
                     (errcode(ERRCODE_UNDEFINED_OBJECT),
                      errmsg("index \"%s\" does not exist", index_name),
                      parser_errposition(cxt->pstate, constraint->location)));
  
         /* Open the index (this will throw an error if it is not an index) */
         index_rel = index_open(index_oid, AccessShareLock);
         index_form = index_rel->rd_index;
  
         /* Check that it does not have an associated constraint already */
         if (OidIsValid(get_index_constraint(index_oid)))
             ereport(ERROR,
                     (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
                      errmsg("index \"%s\" is already associated with a constraint",
                             index_name),
                      parser_errposition(cxt->pstate, constraint->location)));
  
         /* Perform validity checks on the index */
         if (index_form->indrelid != RelationGetRelid(heap_rel))
             ereport(ERROR,
                     (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
                      errmsg("index \"%s\" does not belong to table \"%s\"",
                             index_name, RelationGetRelationName(heap_rel)),
                      parser_errposition(cxt->pstate, constraint->location)));
  
         if (!index_form->indisvalid)
             ereport(ERROR,
                     (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
                      errmsg("index \"%s\" is not valid", index_name),
                      parser_errposition(cxt->pstate, constraint->location)));
  
         if (!index_form->indisunique)
             ereport(ERROR,
                     (errcode(ERRCODE_WRONG_OBJECT_TYPE),
                      errmsg("\"%s\" is not a unique index", index_name),
                      errdetail("Cannot create a primary key or unique constraint using such an index."),
                      parser_errposition(cxt->pstate, constraint->location)));
  
         if (RelationGetIndexExpressions(index_rel) != NIL)
             ereport(ERROR,
                     (errcode(ERRCODE_WRONG_OBJECT_TYPE),
                      errmsg("index \"%s\" contains expressions", index_name),
                      errdetail("Cannot create a primary key or unique constraint using such an index."),
                      parser_errposition(cxt->pstate, constraint->location)));
  
         if (RelationGetIndexPredicate(index_rel) != NIL)
             ereport(ERROR,
                     (errcode(ERRCODE_WRONG_OBJECT_TYPE),
                      errmsg("\"%s\" is a partial index", index_name),
                      errdetail("Cannot create a primary key or unique constraint using such an index."),
                      parser_errposition(cxt->pstate, constraint->location)));
  
         /*
          * It's probably unsafe to change a deferred index to non-deferred. (A
          * non-constraint index couldn't be deferred anyway, so this case
          * should never occur; no need to sweat, but let's check it.)
          */
         if (!index_form->indimmediate && !constraint->deferrable)
             ereport(ERROR,
                     (errcode(ERRCODE_WRONG_OBJECT_TYPE),
                      errmsg("\"%s\" is a deferrable index", index_name),
                      errdetail("Cannot create a non-deferrable constraint using a deferrable index."),
                      parser_errposition(cxt->pstate, constraint->location)));
  
         /*
          * Insist on it being a btree.  That's the only kind that supports
          * uniqueness at the moment anyway; but we must have an index that
          * exactly matches what you'd get from plain ADD CONSTRAINT syntax,
          * else dump and reload will produce a different index (breaking
          * pg_upgrade in particular).
          */
         if (index_rel->rd_rel->relam != get_index_am_oid(DEFAULT_INDEX_TYPE, false))
             ereport(ERROR,
                     (errcode(ERRCODE_WRONG_OBJECT_TYPE),
                      errmsg("index \"%s\" is not a btree", index_name),
                      parser_errposition(cxt->pstate, constraint->location)));
  
         /* Must get indclass the hard way */
         indclassDatum = SysCacheGetAttrNotNull(INDEXRELID,
                                                index_rel->rd_indextuple,
                                                Anum_pg_index_indclass);
         indclass = (oidvector *) DatumGetPointer(indclassDatum);
  
         for (i = 0; i < index_form->indnatts; i++)
         {
             int16       attnum = index_form->indkey.values[i];
             const FormData_pg_attribute *attform;
             char       *attname;
             Oid         defopclass;
  
             /*
              * We shouldn't see attnum == 0 here, since we already rejected
              * expression indexes.  If we do, SystemAttributeDefinition will
              * throw an error.
              */
             if (attnum > 0)
             {
                 Assert(attnum <= heap_rel->rd_att->natts);
                 attform = TupleDescAttr(heap_rel->rd_att, attnum - 1);
             }
             else
                 attform = SystemAttributeDefinition(attnum);
             attname = pstrdup(NameStr(attform->attname));
  
             if (i < index_form->indnkeyatts)
             {
                 /*
                  * Insist on default opclass, collation, and sort options.
                  * While the index would still work as a constraint with
                  * non-default settings, it might not provide exactly the same
                  * uniqueness semantics as you'd get from a normally-created
                  * constraint; and there's also the dump/reload problem
                  * mentioned above.
                  */
                 Datum       attoptions =
                     get_attoptions(RelationGetRelid(index_rel), i + 1);
  
                 defopclass = GetDefaultOpClass(attform->atttypid,
                                                index_rel->rd_rel->relam);
                 if (indclass->values[i] != defopclass ||
                     attform->attcollation != index_rel->rd_indcollation[i] ||
                     attoptions != (Datum) 0 ||
                     index_rel->rd_indoption[i] != 0)
                     ereport(ERROR,
                             (errcode(ERRCODE_WRONG_OBJECT_TYPE),
                              errmsg("index \"%s\" column number %d does not have default sorting behavior", index_name, i + 1),
                              errdetail("Cannot create a primary key or unique constraint using such an index."),
                              parser_errposition(cxt->pstate, constraint->location)));
  
                 constraint->keys = lappend(constraint->keys, makeString(attname));
             }
             else
                 constraint->including = lappend(constraint->including, makeString(attname));
         }
  
         /* Close the index relation but keep the lock */
         relation_close(index_rel, NoLock);
  
         index->indexOid = index_oid;
     }
  
     /*
      * If it's an EXCLUDE constraint, the grammar returns a list of pairs of
      * IndexElems and operator names.  We have to break that apart into
      * separate lists.
      */
     if (constraint->contype == CONSTR_EXCLUSION)
     {
         foreach(lc, constraint->exclusions)
         {
             List       *pair = (List *) lfirst(lc);
             IndexElem  *elem;
             List       *opname;
  
             Assert(list_length(pair) == 2);
             elem = linitial_node(IndexElem, pair);
             opname = lsecond_node(List, pair);
  
             index->indexParams = lappend(index->indexParams, elem);
             index->excludeOpNames = lappend(index->excludeOpNames, opname);
         }
     }
  
     /*
      * For UNIQUE and PRIMARY KEY, we just have a list of column names.
      *
      * Make sure referenced keys exist.  If we are making a PRIMARY KEY index,
      * also make sure they are not-null.
      */
     else
     {
         foreach(lc, constraint->keys)
         {
             char       *key = strVal(lfirst(lc));
             bool        found = false;
             ColumnDef  *column = NULL;
             ListCell   *columns;
             IndexElem  *iparam;
  
             /* Make sure referenced column exists. */
             foreach(columns, cxt->columns)
             {
                 column = lfirst_node(ColumnDef, columns);
                 if (strcmp(column->colname, key) == 0)
                 {
                     found = true;
                     break;
                 }
             }
             if (found)
             {
                 /*
                  * column is defined in the new table.  For PRIMARY KEY, we
                  * can apply the not-null constraint cheaply here ... unless
                  * the column is marked is_from_type, in which case marking it
                  * here would be ineffective (see MergeAttributes).  Note that
                  * this isn't effective in ALTER TABLE either, unless the
                  * column is being added in the same command.
                  */
                 if (constraint->contype == CONSTR_PRIMARY &&
                     !column->is_from_type)
                 {
                     column->is_not_null = true;
                 }
             }
             else if (SystemAttributeByName(key) != NULL)
             {
                 /*
                  * column will be a system column in the new table, so accept
                  * it. System columns can't ever be null, so no need to worry
                  * about PRIMARY/NOT NULL constraint.
                  */
                 found = true;
             }
             else if (cxt->inhRelations)
             {
                 /* try inherited tables */
                 ListCell   *inher;
  
                 foreach(inher, cxt->inhRelations)
                 {
                     RangeVar   *inh = lfirst_node(RangeVar, inher);
                     Relation    rel;
                     int         count;
  
                     rel = table_openrv(inh, AccessShareLock);
                     /* check user requested inheritance from valid relkind */
                     if (rel->rd_rel->relkind != RELKIND_RELATION &&
                         rel->rd_rel->relkind != RELKIND_FOREIGN_TABLE &&
                         rel->rd_rel->relkind != RELKIND_PARTITIONED_TABLE)
                         ereport(ERROR,
                                 (errcode(ERRCODE_WRONG_OBJECT_TYPE),
                                  errmsg("inherited relation \"%s\" is not a table or foreign table",
                                         inh->relname)));
                     for (count = 0; count < rel->rd_att->natts; count++)
                     {
                         Form_pg_attribute inhattr = TupleDescAttr(rel->rd_att,
                                                                   count);
                         char       *inhname = NameStr(inhattr->attname);
  
                         if (inhattr->attisdropped)
                             continue;
                         if (strcmp(key, inhname) == 0)
                         {
                             found = true;
                             break;
                         }
                     }
                     table_close(rel, NoLock);
                     if (found)
                         break;
                 }
             }
  
             /*
              * In the ALTER TABLE case, don't complain about index keys not
              * created in the command; they may well exist already.
              * DefineIndex will complain about them if not.
              */
             if (!found && !cxt->isalter)
                 ereport(ERROR,
                         (errcode(ERRCODE_UNDEFINED_COLUMN),
                          errmsg("column \"%s\" named in key does not exist", key),
                          parser_errposition(cxt->pstate, constraint->location)));
  
             /* Check for PRIMARY KEY(foo, foo) */
             foreach(columns, index->indexParams)
             {
                 iparam = (IndexElem *) lfirst(columns);
                 if (iparam->name && strcmp(key, iparam->name) == 0)
                 {
                     if (index->primary)
                         ereport(ERROR,
                                 (errcode(ERRCODE_DUPLICATE_COLUMN),
                                  errmsg("column \"%s\" appears twice in primary key constraint",
                                         key),
                                  parser_errposition(cxt->pstate, constraint->location)));
                     else
                         ereport(ERROR,
                                 (errcode(ERRCODE_DUPLICATE_COLUMN),
                                  errmsg("column \"%s\" appears twice in unique constraint",
                                         key),
                                  parser_errposition(cxt->pstate, constraint->location)));
                 }
             }
  
             /* OK, add it to the index definition */
             iparam = makeNode(IndexElem);
             iparam->name = pstrdup(key);
             iparam->expr = NULL;
             iparam->indexcolname = NULL;
             iparam->collation = NIL;
             iparam->opclass = NIL;
             iparam->opclassopts = NIL;
             iparam->ordering = SORTBY_DEFAULT;
             iparam->nulls_ordering = SORTBY_NULLS_DEFAULT;
             index->indexParams = lappend(index->indexParams, iparam);
  
             if (constraint->contype == CONSTR_PRIMARY)
             {
                 AlterTableCmd *notnullcmd = makeNode(AlterTableCmd);
  
                 notnullcmd->subtype = AT_SetAttNotNull;
                 notnullcmd->name = pstrdup(key);
                 notnullcmds = lappend(notnullcmds, notnullcmd);
             }
         }
     }
  
     /*
      * Add included columns to index definition.  This is much like the
      * simple-column-name-list code above, except that we don't worry about
      * NOT NULL marking; included columns in a primary key should not be
      * forced NOT NULL.  We don't complain about duplicate columns, either,
      * though maybe we should?
      */
     foreach(lc, constraint->including)
     {
         char       *key = strVal(lfirst(lc));
         bool        found = false;
         ColumnDef  *column = NULL;
         ListCell   *columns;
         IndexElem  *iparam;
  
         foreach(columns, cxt->columns)
         {
             column = lfirst_node(ColumnDef, columns);
             if (strcmp(column->colname, key) == 0)
             {
                 found = true;
                 break;
             }
         }
  
         if (!found)
         {
             if (SystemAttributeByName(key) != NULL)
             {
                 /*
                  * column will be a system column in the new table, so accept
                  * it.
                  */
                 found = true;
             }
             else if (cxt->inhRelations)
             {
                 /* try inherited tables */
                 ListCell   *inher;
  
                 foreach(inher, cxt->inhRelations)
                 {
                     RangeVar   *inh = lfirst_node(RangeVar, inher);
                     Relation    rel;
                     int         count;
  
                     rel = table_openrv(inh, AccessShareLock);
                     /* check user requested inheritance from valid relkind */
                     if (rel->rd_rel->relkind != RELKIND_RELATION &&
                         rel->rd_rel->relkind != RELKIND_FOREIGN_TABLE &&
                         rel->rd_rel->relkind != RELKIND_PARTITIONED_TABLE)
                         ereport(ERROR,
                                 (errcode(ERRCODE_WRONG_OBJECT_TYPE),
                                  errmsg("inherited relation \"%s\" is not a table or foreign table",
                                         inh->relname)));
                     for (count = 0; count < rel->rd_att->natts; count++)
                     {
                         Form_pg_attribute inhattr = TupleDescAttr(rel->rd_att,
                                                                   count);
                         char       *inhname = NameStr(inhattr->attname);
  
                         if (inhattr->attisdropped)
                             continue;
                         if (strcmp(key, inhname) == 0)
                         {
                             found = true;
                             break;
                         }
                     }
                     table_close(rel, NoLock);
                     if (found)
                         break;
                 }
             }
         }
  
         /*
          * In the ALTER TABLE case, don't complain about index keys not
          * created in the command; they may well exist already. DefineIndex
          * will complain about them if not.
          */
         if (!found && !cxt->isalter)
             ereport(ERROR,
                     (errcode(ERRCODE_UNDEFINED_COLUMN),
                      errmsg("column \"%s\" named in key does not exist", key),
                      parser_errposition(cxt->pstate, constraint->location)));
  
         /* OK, add it to the index definition */
         iparam = makeNode(IndexElem);
         iparam->name = pstrdup(key);
         iparam->expr = NULL;
         iparam->indexcolname = NULL;
         iparam->collation = NIL;
         iparam->opclass = NIL;
         iparam->opclassopts = NIL;
         index->indexIncludingParams = lappend(index->indexIncludingParams, iparam);
     }
  
     /*
      * If we found anything that requires run-time SET NOT NULL, build a full
      * ALTER TABLE command for that and add it to cxt->alist.
      */
     if (notnullcmds)
     {
         AlterTableStmt *alterstmt = makeNode(AlterTableStmt);
  
         alterstmt->relation = copyObject(cxt->relation);
         alterstmt->cmds = notnullcmds;
         alterstmt->objtype = OBJECT_TABLE;
         alterstmt->missing_ok = false;
  
         cxt->alist = lappend(cxt->alist, alterstmt);
     }
  
     return index;
 }

References Constraint::access_method, AccessShareLock, CreateStmtContext::alist, Assert(), AT_SetAttNotNull, attname, attnum, AlterTableStmt::cmds, IndexElem::collation, ColumnDef::colname, CreateStmtContext::columns, Constraint::conname, CONSTR_EXCLUSION, CONSTR_PRIMARY, CONSTR_UNIQUE, Constraint::contype, copyObject, DatumGetPointer(), DEFAULT_INDEX_TYPE, Constraint::deferrable, ereport, errcode(), errdetail(), errmsg(), ERROR, Constraint::exclusions, IndexElem::expr, FormData_pg_attribute, get_attoptions(), get_index_am_oid(), get_index_constraint(), get_relname_relid(), GetDefaultOpClass(), i, Constraint::including, index_open(), IndexElem::indexcolname, Constraint::indexname, INDEXRELID, Constraint::indexspace, CreateStmtContext::inhRelations, Constraint::initdeferred, InvalidOid, InvalidRelFileNumber, InvalidSubTransactionId, ColumnDef::is_from_type, ColumnDef::is_not_null, CreateStmtContext::isalter, sort-test::key, Constraint::keys, lappend(), lfirst, lfirst_node, linitial_node, list_length(), Constraint::location, lsecond_node, makeNode, makeString(), AlterTableStmt::missing_ok, IndexElem::name, AlterTableCmd::name, NameStr, TupleDescData::natts, NIL, NoLock, Constraint::nulls_not_distinct, IndexElem::nulls_ordering, OBJECT_TABLE, AlterTableStmt::objtype, OidIsValid, IndexElem::opclass, IndexElem::opclassopts, Constraint::options, IndexElem::ordering, parser_errposition(), CreateStmtContext::pkey, CreateStmtContext::pstate, pstrdup(), RelationData::rd_att, RelationData::rd_indcollation, RelationData::rd_index, RelationData::rd_indextuple, RelationData::rd_indoption, RelationData::rd_rel, CreateStmtContext::rel, CreateStmtContext::relation, AlterTableStmt::relation, relation_close(), RelationGetIndexExpressions(), RelationGetIndexPredicate(), RelationGetNamespace, RelationGetRelationName, RelationGetRelid, RangeVar::relname, Constraint::reset_default_tblspc, SORTBY_DEFAULT, SORTBY_NULLS_DEFAULT, strVal, AlterTableCmd::subtype, SysCacheGetAttrNotNull(), SystemAttributeByName(), SystemAttributeDefinition(), table_close(), table_openrv(), TupleDescAttr, oidvector::values, and Constraint::where_clause.

Referenced by transformIndexConstraints().

◆ transformIndexConstraints()

static void transformIndexConstraints ( CreateStmtContext * cxt )

static

Definition at line 2180 of file parse_utilcmd.c.

 {
     IndexStmt  *index;
     List       *indexlist = NIL;
     List       *finalindexlist = NIL;
     ListCell   *lc;
  
     /*
      * Run through the constraints that need to generate an index, and do so.
      *
      * For PRIMARY KEY, in addition we set each column's attnotnull flag true.
      * We do not create a separate not-null constraint, as that would be
      * redundant: the PRIMARY KEY constraint itself fulfills that role.  Other
      * constraint types don't need any not-null markings.
      */
     foreach(lc, cxt->ixconstraints)
     {
         Constraint *constraint = lfirst_node(Constraint, lc);
  
         Assert(constraint->contype == CONSTR_PRIMARY ||
                constraint->contype == CONSTR_UNIQUE ||
                constraint->contype == CONSTR_EXCLUSION);
  
         index = transformIndexConstraint(constraint, cxt);
  
         indexlist = lappend(indexlist, index);
     }
  
     /*
      * Scan the index list and remove any redundant index specifications. This
      * can happen if, for instance, the user writes UNIQUE PRIMARY KEY. A
      * strict reading of SQL would suggest raising an error instead, but that
      * strikes me as too anal-retentive. - tgl 2001-02-14
      *
      * XXX in ALTER TABLE case, it'd be nice to look for duplicate
      * pre-existing indexes, too.
      */
     if (cxt->pkey != NULL)
     {
         /* Make sure we keep the PKEY index in preference to others... */
         finalindexlist = list_make1(cxt->pkey);
     }
  
     foreach(lc, indexlist)
     {
         bool        keep = true;
         ListCell   *k;
  
         index = lfirst(lc);
  
         /* if it's pkey, it's already in finalindexlist */
         if (index == cxt->pkey)
             continue;
  
         foreach(k, finalindexlist)
         {
             IndexStmt  *priorindex = lfirst(k);
  
             if (equal(index->indexParams, priorindex->indexParams) &&
                 equal(index->indexIncludingParams, priorindex->indexIncludingParams) &&
                 equal(index->whereClause, priorindex->whereClause) &&
                 equal(index->excludeOpNames, priorindex->excludeOpNames) &&
                 strcmp(index->accessMethod, priorindex->accessMethod) == 0 &&
                 index->nulls_not_distinct == priorindex->nulls_not_distinct &&
                 index->deferrable == priorindex->deferrable &&
                 index->initdeferred == priorindex->initdeferred)
             {
                 priorindex->unique |= index->unique;
  
                 /*
                  * If the prior index is as yet unnamed, and this one is
                  * named, then transfer the name to the prior index. This
                  * ensures that if we have named and unnamed constraints,
                  * we'll use (at least one of) the names for the index.
                  */
                 if (priorindex->idxname == NULL)
                     priorindex->idxname = index->idxname;
                 keep = false;
                 break;
             }
         }
  
         if (keep)
             finalindexlist = lappend(finalindexlist, index);
     }
  
     /*
      * Now append all the IndexStmts to cxt->alist.
      */
     cxt->alist = list_concat(cxt->alist, finalindexlist);
 }

References IndexStmt::accessMethod, CreateStmtContext::alist, Assert(), CONSTR_EXCLUSION, CONSTR_PRIMARY, CONSTR_UNIQUE, Constraint::contype, IndexStmt::deferrable, equal(), IndexStmt::excludeOpNames, IndexStmt::idxname, IndexStmt::indexIncludingParams, IndexStmt::indexParams, IndexStmt::initdeferred, CreateStmtContext::ixconstraints, lappend(), lfirst, lfirst_node, list_concat(), list_make1, NIL, IndexStmt::nulls_not_distinct, CreateStmtContext::pkey, transformIndexConstraint(), IndexStmt::unique, and IndexStmt::whereClause.

Referenced by transformAlterTableStmt(), and transformCreateStmt().

◆ transformIndexStmt()

IndexStmt* transformIndexStmt	(	Oid	relid,
		IndexStmt *	stmt,
		const char *	queryString
	)

Definition at line 2918 of file parse_utilcmd.c.

 {
     ParseState *pstate;
     ParseNamespaceItem *nsitem;
     ListCell   *l;
     Relation    rel;
  
     /* Nothing to do if statement already transformed. */
     if (stmt->transformed)
         return stmt;
  
     /* Set up pstate */
     pstate = make_parsestate(NULL);
     pstate->p_sourcetext = queryString;
  
     /*
      * Put the parent table into the rtable so that the expressions can refer
      * to its fields without qualification.  Caller is responsible for locking
      * relation, but we still need to open it.
      */
     rel = relation_open(relid, NoLock);
     nsitem = addRangeTableEntryForRelation(pstate, rel,
                                            AccessShareLock,
                                            NULL, false, true);
  
     /* no to join list, yes to namespaces */
     addNSItemToQuery(pstate, nsitem, false, true, true);
  
     /* take care of the where clause */
     if (stmt->whereClause)
     {
         stmt->whereClause = transformWhereClause(pstate,
                                                  stmt->whereClause,
                                                  EXPR_KIND_INDEX_PREDICATE,
                                                  "WHERE");
         /* we have to fix its collations too */
         assign_expr_collations(pstate, stmt->whereClause);
     }
  
     /* take care of any index expressions */
     foreach(l, stmt->indexParams)
     {
         IndexElem  *ielem = (IndexElem *) lfirst(l);
  
         if (ielem->expr)
         {
             /* Extract preliminary index col name before transforming expr */
             if (ielem->indexcolname == NULL)
                 ielem->indexcolname = FigureIndexColname(ielem->expr);
  
             /* Now do parse transformation of the expression */
             ielem->expr = transformExpr(pstate, ielem->expr,
                                         EXPR_KIND_INDEX_EXPRESSION);
  
             /* We have to fix its collations too */
             assign_expr_collations(pstate, ielem->expr);
  
             /*
              * transformExpr() should have already rejected subqueries,
              * aggregates, window functions, and SRFs, based on the EXPR_KIND_
              * for an index expression.
              *
              * DefineIndex() will make more checks.
              */
         }
     }
  
     /*
      * Check that only the base rel is mentioned.  (This should be dead code
      * now that add_missing_from is history.)
      */
     if (list_length(pstate->p_rtable) != 1)
         ereport(ERROR,
                 (errcode(ERRCODE_INVALID_COLUMN_REFERENCE),
                  errmsg("index expressions and predicates can refer only to the table being indexed")));
  
     free_parsestate(pstate);
  
     /* Close relation */
     table_close(rel, NoLock);
  
     /* Mark statement as successfully transformed */
     stmt->transformed = true;
  
     return stmt;
 }

References AccessShareLock, addNSItemToQuery(), addRangeTableEntryForRelation(), assign_expr_collations(), ereport, errcode(), errmsg(), ERROR, IndexElem::expr, EXPR_KIND_INDEX_EXPRESSION, EXPR_KIND_INDEX_PREDICATE, FigureIndexColname(), free_parsestate(), IndexElem::indexcolname, lfirst, list_length(), make_parsestate(), NoLock, ParseState::p_rtable, ParseState::p_sourcetext, relation_open(), stmt, table_close(), transformExpr(), and transformWhereClause().

Referenced by ATPostAlterTypeParse(), ProcessUtilitySlow(), and transformAlterTableStmt().

◆ transformOfType()

static void transformOfType	(	CreateStmtContext *	cxt,
		TypeName *	ofTypename
	)

static

Definition at line 1594 of file parse_utilcmd.c.

 {
     HeapTuple   tuple;
     TupleDesc   tupdesc;
     int         i;
     Oid         ofTypeId;
  
     Assert(ofTypename);
  
     tuple = typenameType(NULL, ofTypename, NULL);
     check_of_type(tuple);
     ofTypeId = ((Form_pg_type) GETSTRUCT(tuple))->oid;
     ofTypename->typeOid = ofTypeId; /* cached for later */
  
     tupdesc = lookup_rowtype_tupdesc(ofTypeId, -1);
     for (i = 0; i < tupdesc->natts; i++)
     {
         Form_pg_attribute attr = TupleDescAttr(tupdesc, i);
         ColumnDef  *n;
  
         if (attr->attisdropped)
             continue;
  
         n = makeColumnDef(NameStr(attr->attname), attr->atttypid,
                           attr->atttypmod, attr->attcollation);
         n->is_from_type = true;
  
         cxt->columns = lappend(cxt->columns, n);
     }
     ReleaseTupleDesc(tupdesc);
  
     ReleaseSysCache(tuple);
 }

References Assert(), check_of_type(), CreateStmtContext::columns, GETSTRUCT, i, ColumnDef::is_from_type, lappend(), lookup_rowtype_tupdesc(), makeColumnDef(), NameStr, TupleDescData::natts, ReleaseSysCache(), ReleaseTupleDesc, TupleDescAttr, typenameType(), and TypeName::typeOid.

Referenced by transformCreateStmt().

◆ transformPartitionBound()

PartitionBoundSpec* transformPartitionBound	(	ParseState *	pstate,
		Relation	parent,
		PartitionBoundSpec *	spec
	)

Definition at line 4110 of file parse_utilcmd.c.

 {
     PartitionBoundSpec *result_spec;
     PartitionKey key = RelationGetPartitionKey(parent);
     char        strategy = get_partition_strategy(key);
     int         partnatts = get_partition_natts(key);
     List       *partexprs = get_partition_exprs(key);
  
     /* Avoid scribbling on input */
     result_spec = copyObject(spec);
  
     if (spec->is_default)
     {
         /*
          * Hash partitioning does not support a default partition; there's no
          * use case for it (since the set of partitions to create is perfectly
          * defined), and if users do get into it accidentally, it's hard to
          * back out from it afterwards.
          */
         if (strategy == PARTITION_STRATEGY_HASH)
             ereport(ERROR,
                     (errcode(ERRCODE_INVALID_TABLE_DEFINITION),
                      errmsg("a hash-partitioned table may not have a default partition")));
  
         /*
          * In case of the default partition, parser had no way to identify the
          * partition strategy. Assign the parent's strategy to the default
          * partition bound spec.
          */
         result_spec->strategy = strategy;
  
         return result_spec;
     }
  
     if (strategy == PARTITION_STRATEGY_HASH)
     {
         if (spec->strategy != PARTITION_STRATEGY_HASH)
             ereport(ERROR,
                     (errcode(ERRCODE_INVALID_TABLE_DEFINITION),
                      errmsg("invalid bound specification for a hash partition"),
                      parser_errposition(pstate, exprLocation((Node *) spec))));
  
         if (spec->modulus <= 0)
             ereport(ERROR,
                     (errcode(ERRCODE_INVALID_TABLE_DEFINITION),
                      errmsg("modulus for hash partition must be an integer value greater than zero")));
  
         Assert(spec->remainder >= 0);
  
         if (spec->remainder >= spec->modulus)
             ereport(ERROR,
                     (errcode(ERRCODE_INVALID_TABLE_DEFINITION),
                      errmsg("remainder for hash partition must be less than modulus")));
     }
     else if (strategy == PARTITION_STRATEGY_LIST)
     {
         ListCell   *cell;
         char       *colname;
         Oid         coltype;
         int32       coltypmod;
         Oid         partcollation;
  
         if (spec->strategy != PARTITION_STRATEGY_LIST)
             ereport(ERROR,
                     (errcode(ERRCODE_INVALID_TABLE_DEFINITION),
                      errmsg("invalid bound specification for a list partition"),
                      parser_errposition(pstate, exprLocation((Node *) spec))));
  
         /* Get the only column's name in case we need to output an error */
         if (key->partattrs[0] != 0)
             colname = get_attname(RelationGetRelid(parent),
                                   key->partattrs[0], false);
         else
             colname = deparse_expression((Node *) linitial(partexprs),
                                          deparse_context_for(RelationGetRelationName(parent),
                                                              RelationGetRelid(parent)),
                                          false, false);
         /* Need its type data too */
         coltype = get_partition_col_typid(key, 0);
         coltypmod = get_partition_col_typmod(key, 0);
         partcollation = get_partition_col_collation(key, 0);
  
         result_spec->listdatums = NIL;
         foreach(cell, spec->listdatums)
         {
             Node       *expr = lfirst(cell);
             Const      *value;
             ListCell   *cell2;
             bool        duplicate;
  
             value = transformPartitionBoundValue(pstate, expr,
                                                  colname, coltype, coltypmod,
                                                  partcollation);
  
             /* Don't add to the result if the value is a duplicate */
             duplicate = false;
             foreach(cell2, result_spec->listdatums)
             {
                 Const      *value2 = lfirst_node(Const, cell2);
  
                 if (equal(value, value2))
                 {
                     duplicate = true;
                     break;
                 }
             }
             if (duplicate)
                 continue;
  
             result_spec->listdatums = lappend(result_spec->listdatums,
                                               value);
         }
     }
     else if (strategy == PARTITION_STRATEGY_RANGE)
     {
         if (spec->strategy != PARTITION_STRATEGY_RANGE)
             ereport(ERROR,
                     (errcode(ERRCODE_INVALID_TABLE_DEFINITION),
                      errmsg("invalid bound specification for a range partition"),
                      parser_errposition(pstate, exprLocation((Node *) spec))));
  
         if (list_length(spec->lowerdatums) != partnatts)
             ereport(ERROR,
                     (errcode(ERRCODE_INVALID_TABLE_DEFINITION),
                      errmsg("FROM must specify exactly one value per partitioning column")));
         if (list_length(spec->upperdatums) != partnatts)
             ereport(ERROR,
                     (errcode(ERRCODE_INVALID_TABLE_DEFINITION),
                      errmsg("TO must specify exactly one value per partitioning column")));
  
         /*
          * Convert raw parse nodes into PartitionRangeDatum nodes and perform
          * any necessary validation.
          */
         result_spec->lowerdatums =
             transformPartitionRangeBounds(pstate, spec->lowerdatums,
                                           parent);
         result_spec->upperdatums =
             transformPartitionRangeBounds(pstate, spec->upperdatums,
                                           parent);
     }
     else
         elog(ERROR, "unexpected partition strategy: %d", (int) strategy);
  
     return result_spec;
 }

References Assert(), copyObject, deparse_context_for(), deparse_expression(), elog(), equal(), ereport, errcode(), errmsg(), ERROR, exprLocation(), get_attname(), get_partition_col_collation(), get_partition_col_typid(), get_partition_col_typmod(), get_partition_exprs(), get_partition_natts(), get_partition_strategy(), PartitionBoundSpec::is_default, sort-test::key, lappend(), lfirst, lfirst_node, linitial, list_length(), PartitionBoundSpec::listdatums, PartitionBoundSpec::lowerdatums, NIL, parser_errposition(), PARTITION_STRATEGY_HASH, PARTITION_STRATEGY_LIST, PARTITION_STRATEGY_RANGE, RelationGetPartitionKey(), RelationGetRelationName, RelationGetRelid, PartitionBoundSpec::strategy, transformPartitionBoundValue(), transformPartitionRangeBounds(), PartitionBoundSpec::upperdatums, and value.

Referenced by DefineRelation(), and transformPartitionCmd().

◆ transformPartitionBoundValue()

static Const * transformPartitionBoundValue	(	ParseState *	pstate,
		Node *	val,
		const char *	colName,
		Oid	colType,
		int32	colTypmod,
		Oid	partCollation
	)

static

Definition at line 4420 of file parse_utilcmd.c.

 {
     Node       *value;
  
     /* Transform raw parsetree */
     value = transformExpr(pstate, val, EXPR_KIND_PARTITION_BOUND);
  
     /*
      * transformExpr() should have already rejected column references,
      * subqueries, aggregates, window functions, and SRFs, based on the
      * EXPR_KIND_ of a partition bound expression.
      */
     Assert(!contain_var_clause(value));
  
     /*
      * Coerce to the correct type.  This might cause an explicit coercion step
      * to be added on top of the expression, which must be evaluated before
      * returning the result to the caller.
      */
     value = coerce_to_target_type(pstate,
                                   value, exprType(value),
                                   colType,
                                   colTypmod,
                                   COERCION_ASSIGNMENT,
                                   COERCE_IMPLICIT_CAST,
                                   -1);
  
     if (value == NULL)
         ereport(ERROR,
                 (errcode(ERRCODE_DATATYPE_MISMATCH),
                  errmsg("specified value cannot be cast to type %s for column \"%s\"",
                         format_type_be(colType), colName),
                  parser_errposition(pstate, exprLocation(val))));
  
     /*
      * Evaluate the expression, if needed, assigning the partition key's data
      * type and collation to the resulting Const node.
      */
     if (!IsA(value, Const))
     {
         assign_expr_collations(pstate, value);
         value = (Node *) expression_planner((Expr *) value);
         value = (Node *) evaluate_expr((Expr *) value, colType, colTypmod,
                                        partCollation);
         if (!IsA(value, Const))
             elog(ERROR, "could not evaluate partition bound expression");
     }
     else
     {
         /*
          * If the expression is already a Const, as is often the case, we can
          * skip the rather expensive steps above.  But we still have to insert
          * the right collation, since coerce_to_target_type doesn't handle
          * that.
          */
         ((Const *) value)->constcollid = partCollation;
     }
  
     /*
      * Attach original expression's parse location to the Const, so that
      * that's what will be reported for any later errors related to this
      * partition bound.
      */
     ((Const *) value)->location = exprLocation(val);
  
     return (Const *) value;
 }

References Assert(), assign_expr_collations(), COERCE_IMPLICIT_CAST, coerce_to_target_type(), COERCION_ASSIGNMENT, contain_var_clause(), elog(), ereport, errcode(), errmsg(), ERROR, evaluate_expr(), EXPR_KIND_PARTITION_BOUND, expression_planner(), exprLocation(), exprType(), format_type_be(), IsA, parser_errposition(), transformExpr(), val, and value.

Referenced by transformPartitionBound(), and transformPartitionRangeBounds().

◆ transformPartitionCmd()

static void transformPartitionCmd	(	CreateStmtContext *	cxt,
		PartitionCmd *	cmd
	)

static

Definition at line 4057 of file parse_utilcmd.c.

 {
     Relation    parentRel = cxt->rel;
  
     switch (parentRel->rd_rel->relkind)
     {
         case RELKIND_PARTITIONED_TABLE:
             /* transform the partition bound, if any */
             Assert(RelationGetPartitionKey(parentRel) != NULL);
             if (cmd->bound != NULL)
                 cxt->partbound = transformPartitionBound(cxt->pstate, parentRel,
                                                          cmd->bound);
             break;
         case RELKIND_PARTITIONED_INDEX:
  
             /*
              * A partitioned index cannot have a partition bound set.  ALTER
              * INDEX prevents that with its grammar, but not ALTER TABLE.
              */
             if (cmd->bound != NULL)
                 ereport(ERROR,
                         (errcode(ERRCODE_INVALID_OBJECT_DEFINITION),
                          errmsg("\"%s\" is not a partitioned table",
                                 RelationGetRelationName(parentRel))));
             break;
         case RELKIND_RELATION:
             /* the table must be partitioned */
             ereport(ERROR,
                     (errcode(ERRCODE_INVALID_OBJECT_DEFINITION),
                      errmsg("table \"%s\" is not partitioned",
                             RelationGetRelationName(parentRel))));
             break;
         case RELKIND_INDEX:
             /* the index must be partitioned */
             ereport(ERROR,
                     (errcode(ERRCODE_INVALID_OBJECT_DEFINITION),
                      errmsg("index \"%s\" is not partitioned",
                             RelationGetRelationName(parentRel))));
             break;
         default:
             /* parser shouldn't let this case through */
             elog(ERROR, "\"%s\" is not a partitioned table or index",
                  RelationGetRelationName(parentRel));
             break;
     }
 }

References Assert(), PartitionCmd::bound, elog(), ereport, errcode(), errmsg(), ERROR, CreateStmtContext::partbound, CreateStmtContext::pstate, RelationData::rd_rel, CreateStmtContext::rel, RelationGetPartitionKey(), RelationGetRelationName, and transformPartitionBound().

Referenced by transformAlterTableStmt().

◆ transformPartitionRangeBounds()

static List * transformPartitionRangeBounds	(	ParseState *	pstate,
		List *	blist,
		Relation	parent
	)

static

Definition at line 4264 of file parse_utilcmd.c.

 {
     List       *result = NIL;
     PartitionKey key = RelationGetPartitionKey(parent);
     List       *partexprs = get_partition_exprs(key);
     ListCell   *lc;
     int         i,
                 j;
  
     i = j = 0;
     foreach(lc, blist)
     {
         Node       *expr = lfirst(lc);
         PartitionRangeDatum *prd = NULL;
  
         /*
          * Infinite range bounds -- "minvalue" and "maxvalue" -- get passed in
          * as ColumnRefs.
          */
         if (IsA(expr, ColumnRef))
         {
             ColumnRef  *cref = (ColumnRef *) expr;
             char       *cname = NULL;
  
             /*
              * There should be a single field named either "minvalue" or
              * "maxvalue".
              */
             if (list_length(cref->fields) == 1 &&
                 IsA(linitial(cref->fields), String))
                 cname = strVal(linitial(cref->fields));
  
             if (cname == NULL)
             {
                 /*
                  * ColumnRef is not in the desired single-field-name form. For
                  * consistency between all partition strategies, let the
                  * expression transformation report any errors rather than
                  * doing it ourselves.
                  */
             }
             else if (strcmp("minvalue", cname) == 0)
             {
                 prd = makeNode(PartitionRangeDatum);
                 prd->kind = PARTITION_RANGE_DATUM_MINVALUE;
                 prd->value = NULL;
             }
             else if (strcmp("maxvalue", cname) == 0)
             {
                 prd = makeNode(PartitionRangeDatum);
                 prd->kind = PARTITION_RANGE_DATUM_MAXVALUE;
                 prd->value = NULL;
             }
         }
  
         if (prd == NULL)
         {
             char       *colname;
             Oid         coltype;
             int32       coltypmod;
             Oid         partcollation;
             Const      *value;
  
             /* Get the column's name in case we need to output an error */
             if (key->partattrs[i] != 0)
                 colname = get_attname(RelationGetRelid(parent),
                                       key->partattrs[i], false);
             else
             {
                 colname = deparse_expression((Node *) list_nth(partexprs, j),
                                              deparse_context_for(RelationGetRelationName(parent),
                                                                  RelationGetRelid(parent)),
                                              false, false);
                 ++j;
             }
  
             /* Need its type data too */
             coltype = get_partition_col_typid(key, i);
             coltypmod = get_partition_col_typmod(key, i);
             partcollation = get_partition_col_collation(key, i);
  
             value = transformPartitionBoundValue(pstate, expr,
                                                  colname,
                                                  coltype, coltypmod,
                                                  partcollation);
             if (value->constisnull)
                 ereport(ERROR,
                         (errcode(ERRCODE_INVALID_OBJECT_DEFINITION),
                          errmsg("cannot specify NULL in range bound")));
             prd = makeNode(PartitionRangeDatum);
             prd->kind = PARTITION_RANGE_DATUM_VALUE;
             prd->value = (Node *) value;
             ++i;
         }
  
         prd->location = exprLocation(expr);
  
         result = lappend(result, prd);
     }
  
     /*
      * Once we see MINVALUE or MAXVALUE for one column, the remaining columns
      * must be the same.
      */
     validateInfiniteBounds(pstate, result);
  
     return result;
 }

References deparse_context_for(), deparse_expression(), ereport, errcode(), errmsg(), ERROR, exprLocation(), ColumnRef::fields, get_attname(), get_partition_col_collation(), get_partition_col_typid(), get_partition_col_typmod(), get_partition_exprs(), i, IsA, j, sort-test::key, PartitionRangeDatum::kind, lappend(), lfirst, linitial, list_length(), list_nth(), PartitionRangeDatum::location, makeNode, NIL, PARTITION_RANGE_DATUM_MAXVALUE, PARTITION_RANGE_DATUM_MINVALUE, PARTITION_RANGE_DATUM_VALUE, RelationGetPartitionKey(), RelationGetRelationName, RelationGetRelid, strVal, transformPartitionBoundValue(), validateInfiniteBounds(), PartitionRangeDatum::value, and value.

Referenced by transformPartitionBound().

◆ transformRuleStmt()

void transformRuleStmt	(	RuleStmt *	stmt,
		const char *	queryString,
		List **	actions,
		Node **	whereClause
	)

Definition at line 3088 of file parse_utilcmd.c.

 {
     Relation    rel;
     ParseState *pstate;
     ParseNamespaceItem *oldnsitem;
     ParseNamespaceItem *newnsitem;
  
     /*
      * To avoid deadlock, make sure the first thing we do is grab
      * AccessExclusiveLock on the target relation.  This will be needed by
      * DefineQueryRewrite(), and we don't want to grab a lesser lock
      * beforehand.
      */
     rel = table_openrv(stmt->relation, AccessExclusiveLock);
  
     if (rel->rd_rel->relkind == RELKIND_MATVIEW)
         ereport(ERROR,
                 (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
                  errmsg("rules on materialized views are not supported")));
  
     /* Set up pstate */
     pstate = make_parsestate(NULL);
     pstate->p_sourcetext = queryString;
  
     /*
      * NOTE: 'OLD' must always have a varno equal to 1 and 'NEW' equal to 2.
      * Set up their ParseNamespaceItems in the main pstate for use in parsing
      * the rule qualification.
      */
     oldnsitem = addRangeTableEntryForRelation(pstate, rel,
                                               AccessShareLock,
                                               makeAlias("old", NIL),
                                               false, false);
     newnsitem = addRangeTableEntryForRelation(pstate, rel,
                                               AccessShareLock,
                                               makeAlias("new", NIL),
                                               false, false);
  
     /*
      * They must be in the namespace too for lookup purposes, but only add the
      * one(s) that are relevant for the current kind of rule.  In an UPDATE
      * rule, quals must refer to OLD.field or NEW.field to be unambiguous, but
      * there's no need to be so picky for INSERT & DELETE.  We do not add them
      * to the joinlist.
      */
     switch (stmt->event)
     {
         case CMD_SELECT:
             addNSItemToQuery(pstate, oldnsitem, false, true, true);
             break;
         case CMD_UPDATE:
             addNSItemToQuery(pstate, oldnsitem, false, true, true);
             addNSItemToQuery(pstate, newnsitem, false, true, true);
             break;
         case CMD_INSERT:
             addNSItemToQuery(pstate, newnsitem, false, true, true);
             break;
         case CMD_DELETE:
             addNSItemToQuery(pstate, oldnsitem, false, true, true);
             break;
         default:
             elog(ERROR, "unrecognized event type: %d",
                  (int) stmt->event);
             break;
     }
  
     /* take care of the where clause */
     *whereClause = transformWhereClause(pstate,
                                         stmt->whereClause,
                                         EXPR_KIND_WHERE,
                                         "WHERE");
     /* we have to fix its collations too */
     assign_expr_collations(pstate, *whereClause);
  
     /* this is probably dead code without add_missing_from: */
     if (list_length(pstate->p_rtable) != 2) /* naughty, naughty... */
         ereport(ERROR,
                 (errcode(ERRCODE_INVALID_OBJECT_DEFINITION),
                  errmsg("rule WHERE condition cannot contain references to other relations")));
  
     /*
      * 'instead nothing' rules with a qualification need a query rangetable so
      * the rewrite handler can add the negated rule qualification to the
      * original query. We create a query with the new command type CMD_NOTHING
      * here that is treated specially by the rewrite system.
      */
     if (stmt->actions == NIL)
     {
         Query      *nothing_qry = makeNode(Query);
  
         nothing_qry->commandType = CMD_NOTHING;
         nothing_qry->rtable = pstate->p_rtable;
         nothing_qry->rteperminfos = pstate->p_rteperminfos;
         nothing_qry->jointree = makeFromExpr(NIL, NULL);    /* no join wanted */
  
         *actions = list_make1(nothing_qry);
     }
     else
     {
         ListCell   *l;
         List       *newactions = NIL;
  
         /*
          * transform each statement, like parse_sub_analyze()
          */
         foreach(l, stmt->actions)
         {
             Node       *action = (Node *) lfirst(l);
             ParseState *sub_pstate = make_parsestate(NULL);
             Query      *sub_qry,
                        *top_subqry;
             bool        has_old,
                         has_new;
  
             /*
              * Since outer ParseState isn't parent of inner, have to pass down
              * the query text by hand.
              */
             sub_pstate->p_sourcetext = queryString;
  
             /*
              * Set up OLD/NEW in the rtable for this statement.  The entries
              * are added only to relnamespace, not varnamespace, because we
              * don't want them to be referred to by unqualified field names
              * nor "*" in the rule actions.  We decide later whether to put
              * them in the joinlist.
              */
             oldnsitem = addRangeTableEntryForRelation(sub_pstate, rel,
                                                       AccessShareLock,
                                                       makeAlias("old", NIL),
                                                       false, false);
             newnsitem = addRangeTableEntryForRelation(sub_pstate, rel,
                                                       AccessShareLock,
                                                       makeAlias("new", NIL),
                                                       false, false);
             addNSItemToQuery(sub_pstate, oldnsitem, false, true, false);
             addNSItemToQuery(sub_pstate, newnsitem, false, true, false);
  
             /* Transform the rule action statement */
             top_subqry = transformStmt(sub_pstate, action);
  
             /*
              * We cannot support utility-statement actions (eg NOTIFY) with
              * nonempty rule WHERE conditions, because there's no way to make
              * the utility action execute conditionally.
              */
             if (top_subqry->commandType == CMD_UTILITY &&
                 *whereClause != NULL)
                 ereport(ERROR,
                         (errcode(ERRCODE_INVALID_OBJECT_DEFINITION),
                          errmsg("rules with WHERE conditions can only have SELECT, INSERT, UPDATE, or DELETE actions")));
  
             /*
              * If the action is INSERT...SELECT, OLD/NEW have been pushed down
              * into the SELECT, and that's what we need to look at. (Ugly
              * kluge ... try to fix this when we redesign querytrees.)
              */
             sub_qry = getInsertSelectQuery(top_subqry, NULL);
  
             /*
              * If the sub_qry is a setop, we cannot attach any qualifications
              * to it, because the planner won't notice them.  This could
              * perhaps be relaxed someday, but for now, we may as well reject
              * such a rule immediately.
              */
             if (sub_qry->setOperations != NULL && *whereClause != NULL)
                 ereport(ERROR,
                         (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
                          errmsg("conditional UNION/INTERSECT/EXCEPT statements are not implemented")));
  
             /*
              * Validate action's use of OLD/NEW, qual too
              */
             has_old =
                 rangeTableEntry_used((Node *) sub_qry, PRS2_OLD_VARNO, 0) ||
                 rangeTableEntry_used(*whereClause, PRS2_OLD_VARNO, 0);
             has_new =
                 rangeTableEntry_used((Node *) sub_qry, PRS2_NEW_VARNO, 0) ||
                 rangeTableEntry_used(*whereClause, PRS2_NEW_VARNO, 0);
  
             switch (stmt->event)
             {
                 case CMD_SELECT:
                     if (has_old)
                         ereport(ERROR,
                                 (errcode(ERRCODE_INVALID_OBJECT_DEFINITION),
                                  errmsg("ON SELECT rule cannot use OLD")));
                     if (has_new)
                         ereport(ERROR,
                                 (errcode(ERRCODE_INVALID_OBJECT_DEFINITION),
                                  errmsg("ON SELECT rule cannot use NEW")));
                     break;
                 case CMD_UPDATE:
                     /* both are OK */
                     break;
                 case CMD_INSERT:
                     if (has_old)
                         ereport(ERROR,
                                 (errcode(ERRCODE_INVALID_OBJECT_DEFINITION),
                                  errmsg("ON INSERT rule cannot use OLD")));
                     break;
                 case CMD_DELETE:
                     if (has_new)
                         ereport(ERROR,
                                 (errcode(ERRCODE_INVALID_OBJECT_DEFINITION),
                                  errmsg("ON DELETE rule cannot use NEW")));
                     break;
                 default:
                     elog(ERROR, "unrecognized event type: %d",
                          (int) stmt->event);
                     break;
             }
  
             /*
              * OLD/NEW are not allowed in WITH queries, because they would
              * amount to outer references for the WITH, which we disallow.
              * However, they were already in the outer rangetable when we
              * analyzed the query, so we have to check.
              *
              * Note that in the INSERT...SELECT case, we need to examine the
              * CTE lists of both top_subqry and sub_qry.
              *
              * Note that we aren't digging into the body of the query looking
              * for WITHs in nested sub-SELECTs.  A WITH down there can
              * legitimately refer to OLD/NEW, because it'd be an
              * indirect-correlated outer reference.
              */
             if (rangeTableEntry_used((Node *) top_subqry->cteList,
                                      PRS2_OLD_VARNO, 0) ||
                 rangeTableEntry_used((Node *) sub_qry->cteList,
                                      PRS2_OLD_VARNO, 0))
                 ereport(ERROR,
                         (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
                          errmsg("cannot refer to OLD within WITH query")));
             if (rangeTableEntry_used((Node *) top_subqry->cteList,
                                      PRS2_NEW_VARNO, 0) ||
                 rangeTableEntry_used((Node *) sub_qry->cteList,
                                      PRS2_NEW_VARNO, 0))
                 ereport(ERROR,
                         (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
                          errmsg("cannot refer to NEW within WITH query")));
  
             /*
              * For efficiency's sake, add OLD to the rule action's jointree
              * only if it was actually referenced in the statement or qual.
              *
              * For INSERT, NEW is not really a relation (only a reference to
              * the to-be-inserted tuple) and should never be added to the
              * jointree.
              *
              * For UPDATE, we treat NEW as being another kind of reference to
              * OLD, because it represents references to *transformed* tuples
              * of the existing relation.  It would be wrong to enter NEW
              * separately in the jointree, since that would cause a double
              * join of the updated relation.  It's also wrong to fail to make
              * a jointree entry if only NEW and not OLD is mentioned.
              */
             if (has_old || (has_new && stmt->event == CMD_UPDATE))
             {
                 RangeTblRef *rtr;
  
                 /*
                  * If sub_qry is a setop, manipulating its jointree will do no
                  * good at all, because the jointree is dummy. (This should be
                  * a can't-happen case because of prior tests.)
                  */
                 if (sub_qry->setOperations != NULL)
                     ereport(ERROR,
                             (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
                              errmsg("conditional UNION/INTERSECT/EXCEPT statements are not implemented")));
                 /* hackishly add OLD to the already-built FROM clause */
                 rtr = makeNode(RangeTblRef);
                 rtr->rtindex = oldnsitem->p_rtindex;
                 sub_qry->jointree->fromlist =
                     lappend(sub_qry->jointree->fromlist, rtr);
             }
  
             newactions = lappend(newactions, top_subqry);
  
             free_parsestate(sub_pstate);
         }
  
         *actions = newactions;
     }
  
     free_parsestate(pstate);
  
     /* Close relation, but keep the exclusive lock */
     table_close(rel, NoLock);
 }

References AccessExclusiveLock, AccessShareLock, generate_unaccent_rules::action, addNSItemToQuery(), addRangeTableEntryForRelation(), assign_expr_collations(), CMD_DELETE, CMD_INSERT, CMD_NOTHING, CMD_SELECT, CMD_UPDATE, CMD_UTILITY, Query::commandType, Query::cteList, elog(), ereport, errcode(), errmsg(), ERROR, EXPR_KIND_WHERE, free_parsestate(), FromExpr::fromlist, getInsertSelectQuery(), Query::jointree, lappend(), lfirst, list_length(), list_make1, make_parsestate(), makeAlias(), makeFromExpr(), makeNode, NIL, NoLock, ParseState::p_rtable, ParseState::p_rteperminfos, ParseNamespaceItem::p_rtindex, ParseState::p_sourcetext, PRS2_NEW_VARNO, PRS2_OLD_VARNO, rangeTableEntry_used(), RelationData::rd_rel, Query::rtable, RangeTblRef::rtindex, Query::setOperations, stmt, table_close(), table_openrv(), transformStmt(), and transformWhereClause().

Referenced by DefineRule().

◆ transformStatsStmt()

CreateStatsStmt* transformStatsStmt	(	Oid	relid,
		CreateStatsStmt *	stmt,
		const char *	queryString
	)

Definition at line 3013 of file parse_utilcmd.c.

 {
     ParseState *pstate;
     ParseNamespaceItem *nsitem;
     ListCell   *l;
     Relation    rel;
  
     /* Nothing to do if statement already transformed. */
     if (stmt->transformed)
         return stmt;
  
     /* Set up pstate */
     pstate = make_parsestate(NULL);
     pstate->p_sourcetext = queryString;
  
     /*
      * Put the parent table into the rtable so that the expressions can refer
 

Data Structures

Macros

Functions

Macro Definition Documentation

◆ SUPPORTS_ATTRS

Function Documentation

◆ expandTableLikeClause()

◆ generateClonedExtStatsStmt()

◆ generateClonedIndexStmt()

◆ generateSerialExtraStmts()

◆ get_collation()

◆ get_opclass()

◆ setSchemaName()

◆ transformAlterTableStmt()

◆ transformCheckConstraints()

◆ transformColumnDefinition()

◆ transformColumnType()

◆ transformConstraintAttrs()

◆ transformCreateSchemaStmtElements()

◆ transformCreateStmt()

◆ transformExtendedStatistics()

◆ transformFKConstraints()

◆ transformIndexConstraint()

◆ transformIndexConstraints()

◆ transformIndexStmt()

◆ transformOfType()

◆ transformPartitionBound()

◆ transformPartitionBoundValue()

◆ transformPartitionCmd()

◆ transformPartitionRangeBounds()

◆ transformRuleStmt()

◆ transformStatsStmt()