parse_clause.c

来自「PostgreSQL7.4.6 for Linux」· C语言 代码 · 共 1,591 行 · 第 1/3 页

	 * Handle two special cases as mandated by the SQL92 spec:
	 *
	 * 1. Bare ColumnName (no qualifier or subscripts)
	 *	  For a bare identifier, we search for a matching column name
	 *	  in the existing target list.	Multiple matches are an error
	 *	  unless they refer to identical values; for example,
	 *	  we allow	SELECT a, a FROM table ORDER BY a
	 *	  but not	SELECT a AS b, b FROM table ORDER BY b
	 *	  If no match is found, we fall through and treat the identifier
	 *	  as an expression.
	 *	  For GROUP BY, it is incorrect to match the grouping item against
	 *	  targetlist entries: according to SQL92, an identifier in GROUP BY
	 *	  is a reference to a column name exposed by FROM, not to a target
	 *	  list column.	However, many implementations (including pre-7.0
	 *	  PostgreSQL) accept this anyway.  So for GROUP BY, we look first
	 *	  to see if the identifier matches any FROM column name, and only
	 *	  try for a targetlist name if it doesn't.  This ensures that we
	 *	  adhere to the spec in the case where the name could be both.
	 *	  DISTINCT ON isn't in the standard, so we can do what we like there;
	 *	  we choose to make it work like ORDER BY, on the rather flimsy
	 *	  grounds that ordinary DISTINCT works on targetlist entries.
	 *
	 * 2. IntegerConstant
	 *	  This means to use the n'th item in the existing target list.
	 *	  Note that it would make no sense to order/group/distinct by an
	 *	  actual constant, so this does not create a conflict with our
	 *	  extension to order/group by an expression.
	 *	  GROUP BY column-number is not allowed by SQL92, but since
	 *	  the standard has no other behavior defined for this syntax,
	 *	  we may as well accept this common extension.
	 *
	 * Note that pre-existing resjunk targets must not be used in either case,
	 * since the user didn't write them in his SELECT list.
	 *
	 * If neither special case applies, fall through to treat the item as
	 * an expression.
	 *----------
	 */
	if (IsA(node, ColumnRef) &&
		length(((ColumnRef *) node)->fields) == 1 &&
		((ColumnRef *) node)->indirection == NIL)
	{
		char	   *name = strVal(lfirst(((ColumnRef *) node)->fields));

		if (clause == GROUP_CLAUSE)
		{
			/*
			 * In GROUP BY, we must prefer a match against a FROM-clause
			 * column to one against the targetlist.  Look to see if there
			 * is a matching column.  If so, fall through to let
			 * transformExpr() do the rest.  NOTE: if name could refer
			 * ambiguously to more than one column name exposed by FROM,
			 * colNameToVar will ereport(ERROR).  That's just what we want
			 * here.
			 *
			 * Small tweak for 7.4.3: ignore matches in upper query levels.
			 * This effectively changes the search order for bare names to
			 * (1) local FROM variables, (2) local targetlist aliases,
			 * (3) outer FROM variables, whereas before it was (1) (3) (2).
			 * SQL92 and SQL99 do not allow GROUPing BY an outer reference,
			 * so this breaks no cases that are legal per spec, and it
			 * seems a more self-consistent behavior.
			 */
			if (colNameToVar(pstate, name, true) != NULL)
				name = NULL;
		}

		if (name != NULL)
		{
			foreach(tl, tlist)
			{
				TargetEntry *tle = (TargetEntry *) lfirst(tl);
				Resdom	   *resnode = tle->resdom;

				if (!resnode->resjunk &&
					strcmp(resnode->resname, name) == 0)
				{
					if (target_result != NULL)
					{
						if (!equal(target_result->expr, tle->expr))
							ereport(ERROR,
									(errcode(ERRCODE_AMBIGUOUS_COLUMN),
									 /* translator: first %s is name of a SQL construct, eg ORDER BY */
									 errmsg("%s \"%s\" is ambiguous",
											clauseText[clause], name)));
					}
					else
						target_result = tle;
					/* Stay in loop to check for ambiguity */
				}
			}
			if (target_result != NULL)
				return target_result;	/* return the first match */
		}
	}
	if (IsA(node, A_Const))
	{
		Value	   *val = &((A_Const *) node)->val;
		int			targetlist_pos = 0;
		int			target_pos;

		if (!IsA(val, Integer))
			ereport(ERROR,
					(errcode(ERRCODE_SYNTAX_ERROR),
			/* translator: %s is name of a SQL construct, eg ORDER BY */
					 errmsg("non-integer constant in %s",
							clauseText[clause])));
		target_pos = intVal(val);
		foreach(tl, tlist)
		{
			TargetEntry *tle = (TargetEntry *) lfirst(tl);
			Resdom	   *resnode = tle->resdom;

			if (!resnode->resjunk)
			{
				if (++targetlist_pos == target_pos)
					return tle; /* return the unique match */
			}
		}
		ereport(ERROR,
				(errcode(ERRCODE_INVALID_COLUMN_REFERENCE),
		/* translator: %s is name of a SQL construct, eg ORDER BY */
				 errmsg("%s position %d is not in select list",
						clauseText[clause], target_pos)));
	}

	/*
	 * Otherwise, we have an expression (this is a Postgres extension not
	 * found in SQL92).  Convert the untransformed node to a transformed
	 * expression, and search for a match in the tlist. NOTE: it doesn't
	 * really matter whether there is more than one match.	Also, we are
	 * willing to match a resjunk target here, though the above cases must
	 * ignore resjunk targets.
	 */
	expr = transformExpr(pstate, node);

	foreach(tl, tlist)
	{
		TargetEntry *tle = (TargetEntry *) lfirst(tl);

		if (equal(expr, tle->expr))
			return tle;
	}

	/*
	 * If no matches, construct a new target entry which is appended to
	 * the end of the target list.	This target is given resjunk = TRUE so
	 * that it will not be projected into the final tuple.
	 */
	target_result = transformTargetEntry(pstate, node, expr, NULL, true);
	lappend(tlist, target_result);

	return target_result;
}


/*
 * transformGroupClause -
 *	  transform a GROUP BY clause
 */
List *
transformGroupClause(ParseState *pstate, List *grouplist,
					 List *targetlist, List *sortClause)
{
	List	   *glist = NIL,
			   *gl;

	foreach(gl, grouplist)
	{
		TargetEntry *tle;
		Oid			restype;
		Oid			ordering_op;
		GroupClause *grpcl;

		tle = findTargetlistEntry(pstate, lfirst(gl),
								  targetlist, GROUP_CLAUSE);

		/* avoid making duplicate grouplist entries */
		if (targetIsInSortList(tle, glist))
			continue;

		/* if tlist item is an UNKNOWN literal, change it to TEXT */
		restype = tle->resdom->restype;

		if (restype == UNKNOWNOID)
		{
			tle->expr = (Expr *) coerce_type(pstate, (Node *) tle->expr,
											 restype, TEXTOID,
											 COERCION_IMPLICIT,
											 COERCE_IMPLICIT_CAST);
			restype = tle->resdom->restype = TEXTOID;
			tle->resdom->restypmod = -1;
		}

		/*
		 * If the GROUP BY clause matches the ORDER BY clause, we want to
		 * adopt the ordering operators from the latter rather than using
		 * the default ops.  This allows "GROUP BY foo ORDER BY foo DESC"
		 * to be done with only one sort step.	Note we are assuming that
		 * any user-supplied ordering operator will bring equal values
		 * together, which is all that GROUP BY needs.
		 */
		if (sortClause &&
			((SortClause *) lfirst(sortClause))->tleSortGroupRef ==
			tle->resdom->ressortgroupref)
		{
			ordering_op = ((SortClause *) lfirst(sortClause))->sortop;
			sortClause = lnext(sortClause);
		}
		else
		{
			ordering_op = ordering_oper_opid(restype);
			sortClause = NIL;	/* disregard ORDER BY once match fails */
		}

		grpcl = makeNode(GroupClause);
		grpcl->tleSortGroupRef = assignSortGroupRef(tle, targetlist);
		grpcl->sortop = ordering_op;
		glist = lappend(glist, grpcl);
	}

	return glist;
}

/*
 * transformSortClause -
 *	  transform an ORDER BY clause
 */
List *
transformSortClause(ParseState *pstate,
					List *orderlist,
					List *targetlist,
					bool resolveUnknown)
{
	List	   *sortlist = NIL;
	List	   *olitem;

	foreach(olitem, orderlist)
	{
		SortBy	   *sortby = lfirst(olitem);
		TargetEntry *tle;

		tle = findTargetlistEntry(pstate, sortby->node,
								  targetlist, ORDER_CLAUSE);

		sortlist = addTargetToSortList(pstate, tle,
									   sortlist, targetlist,
									   sortby->sortby_kind,
									   sortby->useOp,
									   resolveUnknown);
	}

	return sortlist;
}

/*
 * transformDistinctClause -
 *	  transform a DISTINCT or DISTINCT ON clause
 *
 * Since we may need to add items to the query's sortClause list, that list
 * is passed by reference.	We might also need to add items to the query's
 * targetlist, but we assume that cannot be empty initially, so we can
 * lappend to it even though the pointer is passed by value.
 */
List *
transformDistinctClause(ParseState *pstate, List *distinctlist,
						List *targetlist, List **sortClause)
{
	List	   *result = NIL;
	List	   *slitem;
	List	   *dlitem;

	/* No work if there was no DISTINCT clause */
	if (distinctlist == NIL)
		return NIL;

	if (lfirst(distinctlist) == NIL)
	{
		/* We had SELECT DISTINCT */

		/*
		 * All non-resjunk elements from target list that are not already
		 * in the sort list should be added to it.	(We don't really care
		 * what order the DISTINCT fields are checked in, so we can leave
		 * the user's ORDER BY spec alone, and just add additional sort
		 * keys to it to ensure that all targetlist items get sorted.)
		 */
		*sortClause = addAllTargetsToSortList(pstate,
											  *sortClause,
											  targetlist,
											  true);

		/*
		 * Now, DISTINCT list consists of all non-resjunk sortlist items.
		 * Actually, all the sortlist items had better be non-resjunk!
		 * Otherwise, user wrote SELECT DISTINCT with an ORDER BY item
		 * that does not appear anywhere in the SELECT targetlist, and we
		 * can't implement that with only one sorting pass...
		 */
		foreach(slitem, *sortClause)
		{
			SortClause *scl = (SortClause *) lfirst(slitem);
			TargetEntry *tle = get_sortgroupclause_tle(scl, targetlist);

			if (tle->resdom->resjunk)
				ereport(ERROR,
						(errcode(ERRCODE_INVALID_COLUMN_REFERENCE),
						 errmsg("for SELECT DISTINCT, ORDER BY expressions must appear in select list")));
			else
				result = lappend(result, copyObject(scl));
		}
	}
	else
	{
		/* We had SELECT DISTINCT ON (expr, ...) */

		/*
		 * If the user writes both DISTINCT ON and ORDER BY, then the two
		 * expression lists must match (until one or the other runs out).
		 * Otherwise the ORDER BY requires a different sort order than the
		 * DISTINCT does, and we can't implement that with only one sort
		 * pass (and if we do two passes, the results will be rather
		 * unpredictable). However, it's OK to have more DISTINCT ON
		 * expressions than ORDER BY expressions; we can just add the
		 * extra DISTINCT values to the sort list, much as we did above
		 * for ordinary DISTINCT fields.
		 *
		 * Actually, it'd be OK for the common prefixes of the two lists to
		 * match in any order, but implementing that check seems like more
		 * trouble than it's worth.
		 */
		List	   *nextsortlist = *sortClause;

		foreach(dlitem, distinctlist)
		{
			TargetEntry *tle;

			tle = findTargetlistEntry(pstate, lfirst(dlitem),
									  targetlist, DISTINCT_ON_CLAUSE);

			if (nextsortlist != NIL)
			{
				SortClause *scl = (SortClause *) lfirst(nextsortlist);

				if (tle->resdom->ressortgroupref != scl->tleSortGroupRef)
					ereport(ERROR,
							(errcode(ERRCODE_INVALID_COLUMN_REFERENCE),
							 errmsg("SELECT DISTINCT ON expressions must match initial ORDER BY expressions")));
				result = lappend(result, copyObject(scl));
				nextsortlist = lnext(nextsortlist);
			}
			else
			{
				*sortClause = addTargetToSortList(pstate, tle,
												  *sortClause, targetlist,
												  SORTBY_ASC, NIL, true);

				/*
				 * Probably, the tle should always have been added at the
				 * end of the sort list ... but search to be safe.
				 */
				foreach(slitem, *sortClause)
				{
					SortClause *scl = (SortClause *) lfirst(slitem);

					if (tle->resdom->ressortgroupref == scl->tleSortGroupRef)
					{
						result = lappend(result, copyObject(scl));
						break;
					}
				}
				if (slitem == NIL)		/* should not happen */
					elog(ERROR, "failed to add DISTINCT ON clause to target list");
			}
		}
	}

	return result;
}

/*
 * addAllTargetsToSortList
 *		Make sure all non-resjunk targets in the targetlist are in the
 *		ORDER BY list, adding the not-yet-sorted ones to the end of the list.
 *		This is typically used to help implement SELECT DISTINCT.
 *
 * See addTargetToSortList for info about pstate and resolveUnknown inputs.
 *
 * Returns the updated ORDER BY list.
 */
List *
addAllTargetsToSortList(ParseState *pstate, List *sortlist,
						List *targetlist, bool resolveUnknown)
{
	List	   *i;

	foreach(i, targetlist)
	{
		TargetEntry *tle = (TargetEntry *) lfirst(i);

		if (!tle->resdom->resjunk)
			sortlist = addTargetToSortList(pstate, tle,
										   sortlist, targetlist,
										   SORTBY_ASC, NIL,
										   resolveUnknown);
	}
	return sortlist;
}

/*
 * addTargetToSortList
 *		If the given targetlist entry isn't already in the ORDER BY list,
 *		add it to the end of the list, using the sortop with given name
 *		or the default sort operator if opname == NIL.
 *
 * If resolveUnknown is TRUE, convert TLEs of type UNKNOWN to TEXT.  If not,
 * do nothing (which implies the search for a sort operator will fail).
 * pstate should be provided if resolveUnknown is TRUE, but can be NULL
 * otherwise.
 *
 * Returns the updated ORDER BY list.
 */
List *
addTargetToSortList(ParseState *pstate, TargetEntry *tle,
					List *sortlist, List *targetlist,
					int sortby_kind, List *sortby_opname,
					bool resolveUnknown)
{
	/* avoid making duplicate sortlist entries */
	if (!targetIsInSortList(tle, sortlist))
	{
		SortClause *sortcl = makeNode(SortClause);
		Oid			restype = tle->resdom->restype;

		/* if tlist item is an UNKNOWN literal, change it to TEXT */
		if (restype == UNKNOWNOID && resolveUnknown)
		{
			tle->expr = (Expr *) coerce_type(pstate, (Node *) tle->expr,
											 restype, TEXTOID,
											 COERCION_IMPLICIT,
											 COERCE_IMPLICIT_CAST);
			restype = tle->resdom->restype = TEXTOID;
			tle->resdom->restypmod = -1;
		}

		sortcl->tleSortGroupRef = assignSortGroupRef(tle, targetlist);

		switch (sortby_kind)
		{
			case SORTBY_ASC:
				sortcl->sortop = ordering_oper_opid(restype);
				break;
			case SORTBY_DESC:
				sortcl->sortop = reverse_ordering_oper_opid(restype);
				break;
			case SORTBY_USING:
				Assert(sortby_opname != NIL);
				sortcl->sortop = compatible_oper_opid(sortby_opname,
													  restype,
													  restype,
													  false);
				break;
			default:
				elog(ERROR, "unrecognized sortby_kind: %d", sortby_kind);
				break;
		}

		sortlist = lappend(sortlist, sortcl);
	}
	return sortlist;
}

/*
 * assignSortGroupRef
 *	  Assign the targetentry an unused ressortgroupref, if it doesn't
 *	  already have one.  Return the assigned or pre-existing refnumber.
 *
 * 'tlist' is the targetlist containing (or to contain) the given targetentry.
 */
Index
assignSortGroupRef(TargetEntry *tle, List *tlist)
{
	Index		maxRef;
	List	   *l;

	if (tle->resdom->ressortgroupref)	/* already has one? */
		return tle->resdom->ressortgroupref;

	/* easiest way to pick an unused refnumber: max used + 1 */
	maxRef = 0;
	foreach(l, tlist)
	{
		Index		ref = ((TargetEntry *) lfirst(l))->resdom->ressortgroupref;

		if (ref > maxRef)
			maxRef = ref;
	}
	tle->resdom->ressortgroupref = maxRef + 1;
	return tle->resdom->ressortgroupref;
}

/*
 * targetIsInSortList
 *		Is the given target item already in the sortlist?
 *
 * Works for both SortClause and GroupClause lists.  Note that the main
 * reason we need this routine (and not just a quick test for nonzeroness
 * of ressortgroupref) is that a TLE might be in only one of the lists.
 */
bool
targetIsInSortList(TargetEntry *tle, List *sortList)
{
	Index		ref = tle->resdom->ressortgroupref;
	List	   *i;

	/* no need to scan list if tle has no marker */
	if (ref == 0)
		return false;

	foreach(i, sortList)
	{
		SortClause *scl = (SortClause *) lfirst(i);

		if (scl->tleSortGroupRef == ref)
			return true;
	}
	return false;
}