Show sizes of FETCH queries as constants in pg_stat_statements

Prior to this patch, every FETCH call would generate a unique queryId
with a different size specified.  Depending on the workloads, this could
lead to a significant bloat in pg_stat_statements, as repeatedly calling
a specific cursor would result in a new queryId each time.  For example,
FETCH 1 c1; and FETCH 2 c1; would produce different queryIds.

This patch improves the situation by normalizing the fetch size, so as
semantically similar statements generate the same queryId.  As a result,
statements like the below, which differ syntactically but have the same
effect, will now share a single queryId:
FETCH FROM c1
FETCH NEXT c1
FETCH 1 c1

In order to do a normalization based on the keyword used in FETCH,
FetchStmt is tweaked with a new FetchDirectionKeywords.  This matters
for "howMany", which could be set to a negative value depending on the
direction, and we want to normalize the queries with enough information
about the direction keywords provided, including RELATIVE, ABSOLUTE or
all the ALL variants.

Author: Sami Imseih <samimseih@gmail.com>
Discussion: https://postgr.es/m/CAA5RZ0tA6LbHCg2qSS+KuM850BZC_+ZgHV7Ug6BXw22TNyF+MA@mail.gmail.com

contrib/pg_stat_statements/expected/cursors.out

@@ -57,8 +57,8 @@ SELECT calls, rows, query FROM pg_stat_statements ORDER BY query COLLATE "C";
      1 |    0 | COMMIT
      1 |    0 | DECLARE cursor_stats_1 CURSOR WITH HOLD FOR SELECT $1
      1 |    0 | DECLARE cursor_stats_2 CURSOR WITH HOLD FOR SELECT $1
-     1 |    1 | FETCH 1 IN cursor_stats_1
+     1 |    1 | FETCH $1 IN cursor_stats_1
-     1 |    1 | FETCH 1 IN cursor_stats_2
+     1 |    1 | FETCH $1 IN cursor_stats_2
      1 |    1 | SELECT pg_stat_statements_reset() IS NOT NULL AS t
 (9 rows)
 
@@ -68,3 +68,140 @@ SELECT pg_stat_statements_reset() IS NOT NULL AS t;
  t
 (1 row)
 
+-- Normalization of FETCH statements
+BEGIN;
+DECLARE pgss_cursor CURSOR FOR SELECT FROM generate_series(1, 10);
+-- implicit directions
+FETCH pgss_cursor;
+--
+(1 row)
+
+FETCH 1 pgss_cursor;
+--
+(1 row)
+
+FETCH 2 pgss_cursor;
+--
+(2 rows)
+
+FETCH -1 pgss_cursor;
+--
+(1 row)
+
+-- explicit NEXT
+FETCH NEXT pgss_cursor;
+--
+(1 row)
+
+-- explicit PRIOR
+FETCH PRIOR pgss_cursor;
+--
+(1 row)
+
+-- explicit FIRST
+FETCH FIRST pgss_cursor;
+--
+(1 row)
+
+-- explicit LAST
+FETCH LAST pgss_cursor;
+--
+(1 row)
+
+-- explicit ABSOLUTE
+FETCH ABSOLUTE 1 pgss_cursor;
+--
+(1 row)
+
+FETCH ABSOLUTE 2 pgss_cursor;
+--
+(1 row)
+
+FETCH ABSOLUTE -1 pgss_cursor;
+--
+(1 row)
+
+-- explicit RELATIVE
+FETCH RELATIVE 1 pgss_cursor;
+--
+(0 rows)
+
+FETCH RELATIVE 2 pgss_cursor;
+--
+(0 rows)
+
+FETCH RELATIVE -1 pgss_cursor;
+--
+(1 row)
+
+-- explicit FORWARD
+FETCH ALL pgss_cursor;
+--
+(0 rows)
+
+-- explicit FORWARD ALL
+FETCH FORWARD ALL pgss_cursor;
+--
+(0 rows)
+
+-- explicit FETCH FORWARD
+FETCH FORWARD pgss_cursor;
+--
+(0 rows)
+
+FETCH FORWARD 1 pgss_cursor;
+--
+(0 rows)
+
+FETCH FORWARD 2 pgss_cursor;
+--
+(0 rows)
+
+FETCH FORWARD -1 pgss_cursor;
+--
+(1 row)
+
+-- explicit FETCH BACKWARD
+FETCH BACKWARD pgss_cursor;
+--
+(1 row)
+
+FETCH BACKWARD 1 pgss_cursor;
+--
+(1 row)
+
+FETCH BACKWARD 2 pgss_cursor;
+--
+(2 rows)
+
+FETCH BACKWARD -1 pgss_cursor;
+--
+(1 row)
+
+-- explicit BACKWARD ALL
+FETCH BACKWARD ALL pgss_cursor;
+--
+(6 rows)
+
+COMMIT;
+SELECT calls, query FROM pg_stat_statements ORDER BY query COLLATE "C";
+ calls |                               query                                
+-------+--------------------------------------------------------------------
+     1 | BEGIN
+     1 | COMMIT
+     1 | DECLARE pgss_cursor CURSOR FOR SELECT FROM generate_series($1, $2)
+     3 | FETCH ABSOLUTE $1 pgss_cursor
+     1 | FETCH ALL pgss_cursor
+     1 | FETCH BACKWARD ALL pgss_cursor
+     4 | FETCH BACKWARD pgss_cursor
+     1 | FETCH FIRST pgss_cursor
+     1 | FETCH FORWARD ALL pgss_cursor
+     4 | FETCH FORWARD pgss_cursor
+     1 | FETCH LAST pgss_cursor
+     1 | FETCH NEXT pgss_cursor
+     1 | FETCH PRIOR pgss_cursor
+     3 | FETCH RELATIVE $1 pgss_cursor
+     4 | FETCH pgss_cursor
+     1 | SELECT pg_stat_statements_reset() IS NOT NULL AS t
+(16 rows)
+

contrib/pg_stat_statements/expected/level_tracking.out

@@ -1147,7 +1147,7 @@ SELECT toplevel, calls, query FROM pg_stat_statements
  t        |     1 | COMMIT
  t        |     1 | DECLARE FOOCUR CURSOR FOR SELECT * from stats_track_tab
  f        |     1 | DECLARE FOOCUR CURSOR FOR SELECT * from stats_track_tab;
- t        |     1 | FETCH FORWARD 1 FROM foocur
+ t        |     1 | FETCH FORWARD $1 FROM foocur
  t        |     1 | SELECT pg_stat_statements_reset() IS NOT NULL AS t
 (7 rows)
 
@@ -1176,7 +1176,7 @@ SELECT toplevel, calls, query FROM pg_stat_statements
  t        |     1 | CLOSE foocur
  t        |     1 | COMMIT
  t        |     1 | DECLARE FOOCUR CURSOR FOR SELECT * FROM stats_track_tab
- t        |     1 | FETCH FORWARD 1 FROM foocur
+ t        |     1 | FETCH FORWARD $1 FROM foocur
  t        |     1 | SELECT pg_stat_statements_reset() IS NOT NULL AS t
 (6 rows)
 

contrib/pg_stat_statements/expected/utility.out

@@ -702,7 +702,7 @@ SELECT calls, rows, query FROM pg_stat_statements ORDER BY query COLLATE "C";
      1 |   13 | CREATE MATERIALIZED VIEW pgss_matv AS SELECT * FROM pgss_ctas
      1 |   10 | CREATE TABLE pgss_ctas AS SELECT a, $1 b FROM generate_series($2, $3) a
      1 |    0 | DECLARE pgss_cursor CURSOR FOR SELECT * FROM pgss_matv
-     1 |    5 | FETCH FORWARD 5 pgss_cursor
+     1 |    5 | FETCH FORWARD $1 pgss_cursor
      1 |    7 | FETCH FORWARD ALL pgss_cursor
      1 |    1 | FETCH NEXT pgss_cursor
      1 |   13 | REFRESH MATERIALIZED VIEW pgss_matv

contrib/pg_stat_statements/sql/cursors.sql

@@ -28,3 +28,46 @@ COMMIT;
 
 SELECT calls, rows, query FROM pg_stat_statements ORDER BY query COLLATE "C";
 SELECT pg_stat_statements_reset() IS NOT NULL AS t;
+
+-- Normalization of FETCH statements
+BEGIN;
+DECLARE pgss_cursor CURSOR FOR SELECT FROM generate_series(1, 10);
+-- implicit directions
+FETCH pgss_cursor;
+FETCH 1 pgss_cursor;
+FETCH 2 pgss_cursor;
+FETCH -1 pgss_cursor;
+-- explicit NEXT
+FETCH NEXT pgss_cursor;
+-- explicit PRIOR
+FETCH PRIOR pgss_cursor;
+-- explicit FIRST
+FETCH FIRST pgss_cursor;
+-- explicit LAST
+FETCH LAST pgss_cursor;
+-- explicit ABSOLUTE
+FETCH ABSOLUTE 1 pgss_cursor;
+FETCH ABSOLUTE 2 pgss_cursor;
+FETCH ABSOLUTE -1 pgss_cursor;
+-- explicit RELATIVE
+FETCH RELATIVE 1 pgss_cursor;
+FETCH RELATIVE 2 pgss_cursor;
+FETCH RELATIVE -1 pgss_cursor;
+-- explicit FORWARD
+FETCH ALL pgss_cursor;
+-- explicit FORWARD ALL
+FETCH FORWARD ALL pgss_cursor;
+-- explicit FETCH FORWARD
+FETCH FORWARD pgss_cursor;
+FETCH FORWARD 1 pgss_cursor;
+FETCH FORWARD 2 pgss_cursor;
+FETCH FORWARD -1 pgss_cursor;
+-- explicit FETCH BACKWARD
+FETCH BACKWARD pgss_cursor;
+FETCH BACKWARD 1 pgss_cursor;
+FETCH BACKWARD 2 pgss_cursor;
+FETCH BACKWARD -1 pgss_cursor;
+-- explicit BACKWARD ALL
+FETCH BACKWARD ALL pgss_cursor;
+COMMIT;
+SELECT calls, query FROM pg_stat_statements ORDER BY query COLLATE "C";

src/backend/parser/gram.y

@@ -7477,6 +7477,8 @@ fetch_args:	cursor_name
 					n->portalname = $1;
 					n->direction = FETCH_FORWARD;
 					n->howMany = 1;
+					n->location = -1;
+					n->direction_keyword = FETCH_KEYWORD_NONE;
 					$$ = (Node *) n;
 				}
 			| from_in cursor_name
@@ -7486,60 +7488,8 @@ fetch_args:	cursor_name
 					n->portalname = $2;
 					n->direction = FETCH_FORWARD;
 					n->howMany = 1;
-					$$ = (Node *) n;
+					n->location = -1;
-				}
+					n->direction_keyword = FETCH_KEYWORD_NONE;
-			| NEXT opt_from_in cursor_name
-				{
-					FetchStmt *n = makeNode(FetchStmt);
-
-					n->portalname = $3;
-					n->direction = FETCH_FORWARD;
-					n->howMany = 1;
-					$$ = (Node *) n;
-				}
-			| PRIOR opt_from_in cursor_name
-				{
-					FetchStmt *n = makeNode(FetchStmt);
-
-					n->portalname = $3;
-					n->direction = FETCH_BACKWARD;
-					n->howMany = 1;
-					$$ = (Node *) n;
-				}
-			| FIRST_P opt_from_in cursor_name
-				{
-					FetchStmt *n = makeNode(FetchStmt);
-
-					n->portalname = $3;
-					n->direction = FETCH_ABSOLUTE;
-					n->howMany = 1;
-					$$ = (Node *) n;
-				}
-			| LAST_P opt_from_in cursor_name
-				{
-					FetchStmt *n = makeNode(FetchStmt);
-
-					n->portalname = $3;
-					n->direction = FETCH_ABSOLUTE;
-					n->howMany = -1;
-					$$ = (Node *) n;
-				}
-			| ABSOLUTE_P SignedIconst opt_from_in cursor_name
-				{
-					FetchStmt *n = makeNode(FetchStmt);
-
-					n->portalname = $4;
-					n->direction = FETCH_ABSOLUTE;
-					n->howMany = $2;
-					$$ = (Node *) n;
-				}
-			| RELATIVE_P SignedIconst opt_from_in cursor_name
-				{
-					FetchStmt *n = makeNode(FetchStmt);
-
-					n->portalname = $4;
-					n->direction = FETCH_RELATIVE;
-					n->howMany = $2;
 					$$ = (Node *) n;
 				}
 			| SignedIconst opt_from_in cursor_name
@@ -7549,6 +7499,74 @@ fetch_args:	cursor_name
 					n->portalname = $3;
 					n->direction = FETCH_FORWARD;
 					n->howMany = $1;
+					n->location = @1;
+					n->direction_keyword = FETCH_KEYWORD_NONE;
+					$$ = (Node *) n;
+				}
+			| NEXT opt_from_in cursor_name
+				{
+					FetchStmt *n = makeNode(FetchStmt);
+
+					n->portalname = $3;
+					n->direction = FETCH_FORWARD;
+					n->howMany = 1;
+					n->location = -1;
+					n->direction_keyword = FETCH_KEYWORD_NEXT;
+					$$ = (Node *) n;
+				}
+			| PRIOR opt_from_in cursor_name
+				{
+					FetchStmt *n = makeNode(FetchStmt);
+
+					n->portalname = $3;
+					n->direction = FETCH_BACKWARD;
+					n->howMany = 1;
+					n->location = -1;
+					n->direction_keyword = FETCH_KEYWORD_PRIOR;
+					$$ = (Node *) n;
+				}
+			| FIRST_P opt_from_in cursor_name
+				{
+					FetchStmt *n = makeNode(FetchStmt);
+
+					n->portalname = $3;
+					n->direction = FETCH_ABSOLUTE;
+					n->howMany = 1;
+					n->location = -1;
+					n->direction_keyword = FETCH_KEYWORD_FIRST;
+					$$ = (Node *) n;
+				}
+			| LAST_P opt_from_in cursor_name
+				{
+					FetchStmt *n = makeNode(FetchStmt);
+
+					n->portalname = $3;
+					n->direction = FETCH_ABSOLUTE;
+					n->howMany = -1;
+					n->location = -1;
+					n->direction_keyword = FETCH_KEYWORD_LAST;
+					$$ = (Node *) n;
+				}
+			| ABSOLUTE_P SignedIconst opt_from_in cursor_name
+				{
+					FetchStmt *n = makeNode(FetchStmt);
+
+					n->portalname = $4;
+					n->direction = FETCH_ABSOLUTE;
+					n->howMany = $2;
+					n->location = @2;
+					n->direction_keyword = FETCH_KEYWORD_ABSOLUTE;
+					$$ = (Node *) n;
+				}
+			| RELATIVE_P SignedIconst opt_from_in cursor_name
+				{
+					FetchStmt *n = makeNode(FetchStmt);
+
+					n->portalname = $4;
+					n->direction = FETCH_RELATIVE;
+					n->howMany = $2;
+					n->location = @2;
+					n->direction_keyword = FETCH_KEYWORD_RELATIVE;
 					$$ = (Node *) n;
 				}
 			| ALL opt_from_in cursor_name
@@ -7558,6 +7576,8 @@ fetch_args:	cursor_name
 					n->portalname = $3;
 					n->direction = FETCH_FORWARD;
 					n->howMany = FETCH_ALL;
+					n->location = -1;
+					n->direction_keyword = FETCH_KEYWORD_ALL;
 					$$ = (Node *) n;
 				}
 			| FORWARD opt_from_in cursor_name
@@ -7567,6 +7587,8 @@ fetch_args:	cursor_name
 					n->portalname = $3;
 					n->direction = FETCH_FORWARD;
 					n->howMany = 1;
+					n->location = -1;
+					n->direction_keyword = FETCH_KEYWORD_FORWARD;
 					$$ = (Node *) n;
 				}
 			| FORWARD SignedIconst opt_from_in cursor_name
@@ -7576,6 +7598,8 @@ fetch_args:	cursor_name
 					n->portalname = $4;
 					n->direction = FETCH_FORWARD;
 					n->howMany = $2;
+					n->location = @2;
+					n->direction_keyword = FETCH_KEYWORD_FORWARD;
 					$$ = (Node *) n;
 				}
 			| FORWARD ALL opt_from_in cursor_name
@@ -7585,6 +7609,8 @@ fetch_args:	cursor_name
 					n->portalname = $4;
 					n->direction = FETCH_FORWARD;
 					n->howMany = FETCH_ALL;
+					n->location = -1;
+					n->direction_keyword = FETCH_KEYWORD_FORWARD_ALL;
 					$$ = (Node *) n;
 				}
 			| BACKWARD opt_from_in cursor_name
@@ -7594,6 +7620,8 @@ fetch_args:	cursor_name
 					n->portalname = $3;
 					n->direction = FETCH_BACKWARD;
 					n->howMany = 1;
+					n->location = -1;
+					n->direction_keyword = FETCH_KEYWORD_BACKWARD;
 					$$ = (Node *) n;
 				}
 			| BACKWARD SignedIconst opt_from_in cursor_name
@@ -7603,6 +7631,8 @@ fetch_args:	cursor_name
 					n->portalname = $4;
 					n->direction = FETCH_BACKWARD;
 					n->howMany = $2;
+					n->location = @2;
+					n->direction_keyword = FETCH_KEYWORD_BACKWARD;
 					$$ = (Node *) n;
 				}
 			| BACKWARD ALL opt_from_in cursor_name
@@ -7612,6 +7642,8 @@ fetch_args:	cursor_name
 					n->portalname = $4;
 					n->direction = FETCH_BACKWARD;
 					n->howMany = FETCH_ALL;
+					n->location = -1;
+					n->direction_keyword = FETCH_KEYWORD_BACKWARD_ALL;
 					$$ = (Node *) n;
 				}
 		;

src/include/nodes/parsenodes.h

@@ -3422,15 +3422,44 @@ typedef enum FetchDirection
 	FETCH_RELATIVE,
 } FetchDirection;
 
+typedef enum FetchDirectionKeywords
+{
+	FETCH_KEYWORD_NONE = 0,
+	FETCH_KEYWORD_NEXT,
+	FETCH_KEYWORD_PRIOR,
+	FETCH_KEYWORD_FIRST,
+	FETCH_KEYWORD_LAST,
+	FETCH_KEYWORD_ABSOLUTE,
+	FETCH_KEYWORD_RELATIVE,
+	FETCH_KEYWORD_ALL,
+	FETCH_KEYWORD_FORWARD,
+	FETCH_KEYWORD_FORWARD_ALL,
+	FETCH_KEYWORD_BACKWARD,
+	FETCH_KEYWORD_BACKWARD_ALL,
+} FetchDirectionKeywords;
+
 #define FETCH_ALL	LONG_MAX
 
 typedef struct FetchStmt
 {
 	NodeTag		type;
 	FetchDirection direction;	/* see above */
-	long		howMany;		/* number of rows, or position argument */
+	/* number of rows, or position argument */
-	char	   *portalname;		/* name of portal (cursor) */
+	long		howMany pg_node_attr(query_jumble_ignore);
-	bool		ismove;			/* true if MOVE */
+	/* name of portal (cursor) */
+	char	   *portalname;
+	/* true if MOVE */
+	bool		ismove;
+
+	/*
+	 * Set when a direction_keyword (e.g., FETCH FORWARD) is used, to
+	 * distinguish it from a numeric variant (e.g., FETCH 1) for the purpose
+	 * of query jumbling.
+	 */
+	FetchDirectionKeywords direction_keyword;
+
+	/* token location, or -1 if unknown */
+	ParseLoc	location pg_node_attr(query_jumble_location);
 } FetchStmt;
 
 /* ----------------------

src/tools/pgindent/typedefs.list

@@ -805,6 +805,7 @@ FastPathStrongRelationLockData
 FdwInfo
 FdwRoutine
 FetchDirection
+FetchDirectionKeywords
 FetchStmt
 FieldSelect
 FieldStore