Improve performance of "multi-step" queries (an on-going process).

We already had code that detected when we can reduce the entire query
to a single step, including joins. We also had general code that
allowed for handling arbitrary queries that could not be reduced
into one step, with the catch that joins will be done at the
coordinator level.

This commit allows for some joins to be pushed down when possible.
For example, in a three-way join, if two tables are co-located,
they may be joined in a single step down at the data nodes, and
those results will be joined with the third table (on the
coordinator, however).

In addition, the previous code was based on a SQL/MED patch,
which meant when accessing remote tables, it selected all of
the columns. With this commit, we project only the needed columns.

Written primarily by Pavan Deolasee (join push-dwon and column selection)
and Mason Sharp (safe push-down detection).

Squashed commit of the following:

commit eb50a76cb929fbe4a31d093b43e1589382c892a0
Author: Pavan Deolasee <pavan.deolasee@gmail.com>
Date:   Wed Oct 27 16:09:28 2010 +0530

    Set remote relation stats (pages, rows etc) to a lower value so that
    NestLoop joins are preferred over other join types. This is necessary
    until we can handle other join types for remote join reduction

commit 69bb66c62f71b9be918475ea65931adb3bbfba20
Author: Pavan Deolasee <pavan.deolasee@gmail.com>
Date:   Tue Oct 19 12:20:44 2010 +0530

    Set aliases properly for join reduction

commit 2a313446f3e714ba36c9ccc5c5167309b7c89a95
Author: Mason Sharp <mason_s@users.sourceforge.net>
Date:   Mon Oct 18 16:15:16 2010 -0400

    Added IsJoinReducible to determine if the two plan nodes can be joined.

    See comments for this function for more details. Basically, we use
    examine_conditions_walker to check if it is safe to join the two.

    Partitioned-partitioned joins are safe to collapse, and
    partitioned-replicated are safe iff one of the nodes does not already
    contain such a collapsed node.

commit f275fa535e9673af0964ecc7ca93ab1b49df2317
Author: Pavan Deolasee <pavan.deolasee@gmail.com>
Date:   Mon Oct 18 11:53:54 2010 +0530

    Fix a bug where rte/alias were not getting set up properly

commit 6af07721357944af801a384ed1eb54e363839403
Author: Pavan Deolasee <pavan.deolasee@gmail.com>
Date:   Mon Oct 18 11:35:43 2010 +0530

    Update some missing copy/out/read functions

commit 7bcb490dc50eeb1ad1569d90cc5eb759b766aa91
Author: Pavan Deolasee <pavan.deolasee@gmail.com>
Date:   Mon Oct 18 11:33:54 2010 +0530

    Initial implementation of remote join reduction. We still don't have
    the logic to determine whether its safe to reduce two join trees or not

commit aefc06e7bd90c657fb093a923f7b66177687561d
Author: Pavan Deolasee <pavan.deolasee@gmail.com>
Date:   Mon Oct 18 11:29:21 2010 +0530

    First step to SQL-med integration. Moving query generation to planning stage

7 files changed, 46 insertions, 0 deletions

diff --git a/src/include/nodes/parsenodes.h b/src/include/nodes/parsenodes.h
index 6250de3774..355c44847e 100644
--- a/src/include/nodes/parsenodes.h
+++ b/src/include/nodes/parsenodes.h
@@ -24,6 +24,9 @@
 #include "nodes/bitmapset.h"
 #include "nodes/primnodes.h"
 #include "nodes/value.h"
+#ifdef PGXC
+#include "access/tupdesc.h"
+#endif
 
 /* Possible sources of a Query */
 typedef enum QuerySource
@@ -660,6 +663,11 @@ typedef struct RangeTblEntry
 	 * code that is being actively worked on.  FIXME someday.
 	 */
 
+#ifdef PGXC
+	char		*relname;
+	TupleDesc	reltupdesc;
+#endif
+
 	/*
 	 * Fields valid for a plain relation RTE (else zero):
 	 */
diff --git a/src/include/nodes/relation.h b/src/include/nodes/relation.h
index 7eb15dbeec..4acd48bbcb 100644
--- a/src/include/nodes/relation.h
+++ b/src/include/nodes/relation.h
@@ -189,6 +189,11 @@ typedef struct PlannerInfo
 										 * pseudoconstant = true */
 	bool		hasRecursion;	/* true if planning a recursive WITH item */
 
+#ifdef PGXC
+	/* This field is used only when RemoteScan nodes are involved */
+	int         rs_alias_index; /* used to build the alias reference */
+#endif
+
 	/* These fields are used only when hasRecursion is true: */
 	int			wt_param_id;	/* PARAM_EXEC ID for the work table */
 	struct Plan *non_recursive_plan;	/* plan for non-recursive term */
@@ -377,6 +382,10 @@ typedef struct RelOptInfo
 										 * clauses */
 	List	   *index_inner_paths;		/* InnerIndexscanInfo nodes */
 
+#ifdef PGXC
+	TupleDesc	reltupdesc;
+#endif
+
 	/*
 	 * Inner indexscans are not in the main pathlist because they are not
 	 * usable except in specific join contexts.  We use the index_inner_paths
diff --git a/src/include/optimizer/cost.h b/src/include/optimizer/cost.h
index 839270f355..15cdf599a2 100644
--- a/src/include/optimizer/cost.h
+++ b/src/include/optimizer/cost.h
@@ -59,6 +59,9 @@ extern bool enable_hashagg;
 extern bool enable_nestloop;
 extern bool enable_mergejoin;
 extern bool enable_hashjoin;
+#ifdef PGXC
+extern bool enable_remotejoin;
+#endif
 extern int	constraint_exclusion;
 
 extern double clamp_row_est(double nrows);
diff --git a/src/include/optimizer/planmain.h b/src/include/optimizer/planmain.h
index 3ffd80003a..2c3dcd4dbd 100644
--- a/src/include/optimizer/planmain.h
+++ b/src/include/optimizer/planmain.h
@@ -119,4 +119,7 @@ extern void extract_query_dependencies(List *queries,
 						   List **relationOids,
 						   List **invalItems);
 
+#ifdef PGXC
+extern Var *search_tlist_for_var(Var *var, List *jtlist);
+#endif
 #endif   /* PLANMAIN_H */
diff --git a/src/include/optimizer/var.h b/src/include/optimizer/var.h
index 5d19572e0d..0e7ac5f212 100644
--- a/src/include/optimizer/var.h
+++ b/src/include/optimizer/var.h
@@ -25,6 +25,9 @@ typedef enum
 
 extern Relids pull_varnos(Node *node);
 extern void pull_varattnos(Node *node, Bitmapset **varattnos);
+#ifdef PGXC
+extern Bitmapset * pull_varattnos_varno(Node *node, Index varno, Bitmapset *varattnos);
+#endif
 extern bool contain_var_clause(Node *node);
 extern bool contain_vars_of_level(Node *node, int levelsup);
 extern int	locate_var_of_level(Node *node, int levelsup);
diff --git a/src/include/pgxc/planner.h b/src/include/pgxc/planner.h
index 1e31fa38c1..8aae356fd2 100644
--- a/src/include/pgxc/planner.h
+++ b/src/include/pgxc/planner.h
@@ -23,6 +23,7 @@
 #include "nodes/primnodes.h"
 #include "pgxc/locator.h"
 #include "tcop/dest.h"
+#include "nodes/relation.h"
 
 
 typedef enum
@@ -85,6 +86,18 @@ typedef struct
 	bool		read_only;          /* do not use 2PC when committing read only steps */
 	bool		force_autocommit;	/* some commands like VACUUM require autocommit mode */
 	RemoteQueryExecType		exec_type;
+
+	char	  *relname;
+	bool	   remotejoin;			/* True if this is a reduced remote join  */
+	bool	   partitioned_replicated;	/* True if reduced and contains replicated-partitioned join */
+	int		   reduce_level;		/* in case of reduced JOIN, it's level    */
+	List	  *base_tlist;			/* in case of isReduced, the base tlist   */
+	char	  *outer_alias;
+	char	  *inner_alias;
+	int		   outer_reduce_level;
+	int		   inner_reduce_level;
+	Relids	   outer_relids;
+	Relids	   inner_relids;
 } RemoteQuery;
 
 
@@ -165,4 +178,8 @@ extern PlannedStmt *pgxc_planner(Query *query, int cursorOptions,
 extern bool IsHashDistributable(Oid col_type);
 
 extern bool is_immutable_func(Oid funcid);
+
+extern bool IsJoinReducible(RemoteQuery *innernode, RemoteQuery *outernode,
+					List *rtable_list, JoinPath *join_path, bool *partitioned_replicated);
+
 #endif   /* PGXCPLANNER_H */
diff --git a/src/include/utils/builtins.h b/src/include/utils/builtins.h
index 90955b73f9..4b1463ff74 100644
--- a/src/include/utils/builtins.h
+++ b/src/include/utils/builtins.h
@@ -595,7 +595,10 @@ extern Datum pg_get_function_identity_arguments(PG_FUNCTION_ARGS);
 extern Datum pg_get_function_result(PG_FUNCTION_ARGS);
 extern char *deparse_expression(Node *expr, List *dpcontext,
 				   bool forceprefix, bool showimplicit);
+extern List *deparse_context_for_remotequery(const char *aliasname, Oid relid);
+#ifdef PGXC
 extern List *deparse_context_for(const char *aliasname, Oid relid);
+#endif
 extern List *deparse_context_for_plan(Node *plan, Node *outer_plan,
 						 List *rtable, List *subplans);
 extern const char *quote_identifier(const char *ident);