Return into the code the feature "QueryId based on jumbling machinery".

danolivo · danolivo · commit 2285d86719ef · 2022-10-05T16:10:23.000+05:00
diff --git a/aqo.c b/aqo.c
@@ -163,6 +163,12 @@ _PG_init(void)
 				 errmsg("AQO module could be loaded only on startup."),
 				 errdetail("Add 'aqo' into the shared_preload_libraries list.")));
 
+	/*
+	 * Inform the postmaster that we want to enable query_id calculation if
+	 * compute_query_id is set to auto.
+	 */
+	EnableQueryId();
+
 	DefineCustomEnumVariable("aqo.mode",
 							 "Mode of aqo usage.",
 							 NULL,
diff --git a/expected/aqo_fdw.out b/expected/aqo_fdw.out
@@ -54,14 +54,11 @@ SELECT x FROM frgn;
 (5 rows)
 
 -- Push down base filters. Use verbose mode to see filters.
-EXPLAIN (ANALYZE, COSTS OFF, SUMMARY OFF, TIMING OFF, VERBOSE))
-SELECT x FROM frgn WHERE x < 10;
-ERROR:  syntax error at or near ")"
-LINE 1: ...LAIN (ANALYZE, COSTS OFF, SUMMARY OFF, TIMING OFF, VERBOSE))
-                                                                      ^
-EXPLAIN (ANALYZE, COSTS OFF, SUMMARY OFF, TIMING OFF, VERBOSE)
-SELECT x FROM frgn WHERE x < 10;
-                        QUERY PLAN                         
+SELECT str FROM expln('
+  EXPLAIN (ANALYZE, COSTS OFF, SUMMARY OFF, TIMING OFF, VERBOSE)
+    SELECT x FROM frgn WHERE x < 10;
+') AS str WHERE str NOT LIKE '%Query Identifier%';
+                            str                            
 -----------------------------------------------------------
  Foreign Scan on public.frgn (actual rows=1 loops=1)
    AQO not used
@@ -72,6 +69,21 @@ SELECT x FROM frgn WHERE x < 10;
  JOINS: 0
 (7 rows)
 
+SELECT str FROM expln('
+  EXPLAIN (ANALYZE, COSTS OFF, SUMMARY OFF, TIMING OFF, VERBOSE)
+    SELECT x FROM frgn WHERE x < 10;
+') AS str WHERE str NOT LIKE '%Query Identifier%';
+                            str                            
+-----------------------------------------------------------
+ Foreign Scan on public.frgn (actual rows=1 loops=1)
+   AQO: rows=1, error=0%
+   Output: x
+   Remote SQL: SELECT x FROM public.local WHERE ((x < 10))
+ Using aqo: true
+ AQO mode: LEARN
+ JOINS: 0
+(7 rows)
+
 EXPLAIN (ANALYZE, COSTS OFF, SUMMARY OFF, TIMING OFF)
 SELECT x FROM frgn WHERE x < -10; -- AQO ignores constants
                   QUERY PLAN                  
@@ -99,9 +111,11 @@ SELECT str FROM expln('
 (6 rows)
 
 -- TODO: Should learn on postgres_fdw nodes
-EXPLAIN (ANALYZE, COSTS OFF, SUMMARY OFF, TIMING OFF, VERBOSE)
-  SELECT * FROM frgn AS a, frgn AS b WHERE a.x=b.x;
-                                               QUERY PLAN                                               
+SELECT str FROM expln('
+  EXPLAIN (ANALYZE, COSTS OFF, SUMMARY OFF, TIMING OFF, VERBOSE)
+    SELECT * FROM frgn AS a, frgn AS b WHERE a.x=b.x;
+') AS str WHERE str NOT LIKE '%Query Identifier%';
+                                                  str                                                   
 --------------------------------------------------------------------------------------------------------
  Foreign Scan (actual rows=1 loops=1)
    AQO not used
@@ -126,9 +140,11 @@ SELECT * FROM frgn AS a, frgn AS b WHERE a.x<b.x;
  JOINS: 0
 (6 rows)
 
-EXPLAIN (ANALYZE, COSTS OFF, SUMMARY OFF, TIMING OFF, VERBOSE)
-SELECT * FROM frgn AS a, frgn AS b WHERE a.x<b.x;
-                                               QUERY PLAN                                               
+SELECT str FROM expln('
+  EXPLAIN (ANALYZE, COSTS OFF, SUMMARY OFF, TIMING OFF, VERBOSE)
+    SELECT * FROM frgn AS a, frgn AS b WHERE a.x<b.x;
+') AS str WHERE str NOT LIKE '%Query Identifier%';
+                                                  str                                                   
 --------------------------------------------------------------------------------------------------------
  Foreign Scan (actual rows=0 loops=1)
    AQO not used
diff --git a/expected/gucs.out b/expected/gucs.out
@@ -1,4 +1,12 @@
 CREATE EXTENSION aqo;
+-- Utility tool. Allow to filter system-dependent strings from an explain output.
+CREATE OR REPLACE FUNCTION expln(query_string text) RETURNS SETOF text AS $$
+BEGIN
+    RETURN QUERY
+        EXECUTE format('%s', query_string);
+    RETURN;
+END;
+$$ LANGUAGE PLPGSQL;
 SET aqo.join_threshold = 0;
 SET aqo.mode = 'learn';
 SET aqo.show_details = true;
@@ -12,9 +20,11 @@ SELECT true FROM aqo_reset(); -- Remember! DROP EXTENSION doesn't remove any AQO
 (1 row)
 
 -- Check AQO addons to explain (the only stable data)
-EXPLAIN (ANALYZE, VERBOSE, COSTS OFF, TIMING OFF, SUMMARY OFF)
-	SELECT x FROM t;
-                   QUERY PLAN                   
+SELECT str FROM expln('
+  EXPLAIN (ANALYZE, VERBOSE, COSTS OFF, TIMING OFF, SUMMARY OFF)
+    SELECT x FROM t;
+') AS str WHERE str NOT LIKE '%Query Identifier%';
+                      str                       
 ------------------------------------------------
  Seq Scan on public.t (actual rows=100 loops=1)
    AQO not used
@@ -24,9 +34,11 @@ EXPLAIN (ANALYZE, VERBOSE, COSTS OFF, TIMING OFF, SUMMARY OFF)
  JOINS: 0
 (6 rows)
 
-EXPLAIN (ANALYZE, VERBOSE, COSTS OFF, TIMING OFF, SUMMARY OFF)
-	SELECT x FROM t;
-                   QUERY PLAN                   
+SELECT str FROM expln('
+  EXPLAIN (ANALYZE, VERBOSE, COSTS OFF, TIMING OFF, SUMMARY OFF)
+    SELECT x FROM t;
+') AS str WHERE str NOT LIKE '%Query Identifier%';
+                      str                       
 ------------------------------------------------
  Seq Scan on public.t (actual rows=100 loops=1)
    AQO: rows=100, error=0%
diff --git a/expected/unsupported.out b/expected/unsupported.out
@@ -1,4 +1,12 @@
 CREATE EXTENSION aqo;
+-- Utility tool. Allow to filter system-dependent strings from an explain output.
+CREATE OR REPLACE FUNCTION expln(query_string text) RETURNS SETOF text AS $$
+BEGIN
+  RETURN QUERY
+    EXECUTE format('%s', query_string);
+  RETURN;
+END;
+$$ LANGUAGE PLPGSQL;
 SET aqo.join_threshold = 0;
 SET aqo.mode = 'learn';
 SET aqo.show_details = 'on';
@@ -35,8 +43,8 @@ EXPLAIN (COSTS OFF)
 (11 rows)
 
 SELECT str FROM expln('
-EXPLAIN (ANALYZE, COSTS OFF, SUMMARY OFF, TIMING OFF)
-	SELECT * FROM t GROUP BY (x) HAVING x > 3;
+  EXPLAIN (ANALYZE, COSTS OFF, SUMMARY OFF, TIMING OFF)
+    SELECT * FROM t GROUP BY (x) HAVING x > 3;
 ') AS str WHERE str NOT LIKE '%Memory Usage%';
                       str                      
 -----------------------------------------------
@@ -485,17 +493,6 @@ SELECT * FROM
 -- any prediction on number of fetched tuples.
 -- So, if selectivity was wrong we could make bad choice of Scan operation.
 -- For example, we could choose suboptimal index.
---
--- Returns string-by-string explain of a query. Made for removing some strings
--- from the explain output.
---
-CREATE OR REPLACE FUNCTION expln(query_string text) RETURNS SETOF text AS $$
-BEGIN
-    RETURN QUERY
-        EXECUTE format('EXPLAIN (ANALYZE, VERBOSE, COSTS OFF, TIMING OFF, SUMMARY OFF) %s', query_string);
-    RETURN;
-END;
-$$ LANGUAGE PLPGSQL;
 -- Turn off statistics gathering for simple demonstration of filtering problem.
 ALTER TABLE t SET (autovacuum_enabled = 'false');
 CREATE INDEX ind1 ON t(x);
@@ -531,10 +528,11 @@ SELECT count(*) FROM t WHERE x < 3 AND mod(x,3) = 1;
     50
 (1 row)
 
-SELECT str AS result
-FROM expln('SELECT count(*) FROM t WHERE x < 3 AND mod(x,3) = 1') AS str
-WHERE str NOT LIKE '%Heap Blocks%';
-                             result                              
+SELECT str FROM expln('
+  EXPLAIN (ANALYZE, VERBOSE, COSTS OFF, SUMMARY OFF, TIMING OFF)
+    SELECT count(*) FROM t WHERE x < 3 AND mod(x,3) = 1') AS str
+WHERE str NOT LIKE '%Heap Blocks%' AND str NOT LIKE '%Query Identifier%';
+                               str                               
 -----------------------------------------------------------------
  Aggregate (actual rows=1 loops=1)
    AQO not used
@@ -590,6 +588,10 @@ ORDER BY (md5(query_text),error) DESC;
        |         SELECT count(*) FROM t WHERE                                                          +
        |                 x = (SELECT avg(x) FROM t t0 WHERE t0.x = t.x + 21) OR                        +
        |                 x IN (SELECT avg(x) FROM t t0 WHERE t0.x = t.x + 21);
+ 0.106 |                                                                                               +
+       |   EXPLAIN (ANALYZE, COSTS OFF, SUMMARY OFF, TIMING OFF)                                       +
+       |     SELECT * FROM t GROUP BY (x) HAVING x > 3;                                                +
+       | 
  0.454 | SELECT count(*) FROM (SELECT x, y FROM t1 GROUP BY GROUPING SETS ((x,y), (x), (y), ())) AS q1;
  0.000 | SELECT count(*) FROM (                                                                        +
        |         SELECT count(*) AS x FROM (                                                           +
@@ -602,10 +604,6 @@ ORDER BY (md5(query_text),error) DESC;
  0.000 | EXPLAIN (ANALYZE, COSTS OFF, SUMMARY OFF, TIMING OFF)                                         +
        |         SELECT count(*) FROM t WHERE (SELECT avg(x) FROM t t0 WHERE t0.x = t.x) =             +
        |         (SELECT avg(x) FROM t t0 WHERE t0.x = t.x);
- 0.106 |                                                                                               +
-       | EXPLAIN (ANALYZE, COSTS OFF, SUMMARY OFF, TIMING OFF)                                         +
-       |         SELECT * FROM t GROUP BY (x) HAVING x > 3;                                            +
-       | 
  0.000 | SELECT count(*) FROM                                                                          +
        |         (SELECT * FROM t WHERE x % 3 < (SELECT avg(x) FROM t t0 WHERE t0.x = t.x)) AS q1      +
        |                 JOIN                                                                          +
diff --git a/hash.c b/hash.c
@@ -56,24 +56,6 @@ static bool has_consts(List *lst);
 static List **get_clause_args_ptr(Expr *clause);
 static bool clause_is_eq_clause(Expr *clause);
 
-/*
- * Computes hash for given query.Query Identifier: =
- * Hash is supposed to be constant-insensitive.
- * XXX: Hashing depend on Oids of database objects. It is restrict usability of
- * the AQO knowledge base by current database at current Postgres instance.
- */
-uint64
-get_query_hash(Query *parse, const char *query_text)
-{
-	char	   *str_repr;
-	uint64			hash;
-
-	/* XXX: remove_locations and remove_consts are heavy routines. */
-	str_repr = remove_locations(remove_consts(nodeToString(parse)));
-	hash = DatumGetUInt64(hash_any_extended((void *) str_repr, strlen(str_repr),0));
-
-	return hash;
-}
 
 /*********************************************************************************
  *
diff --git a/hash.h b/hash.h
@@ -3,7 +3,6 @@
 
 #include "nodes/pg_list.h"
 
-extern uint64 get_query_hash(Query *parse, const char *query_text);
 extern bool list_member_uint64(const List *list, uint64 datum);
 extern List *list_copy_uint64(List *list);
 extern List *lappend_uint64(List *list, uint64 datum);
diff --git a/preprocessing.c b/preprocessing.c
@@ -198,7 +198,13 @@ aqo_planner(Query *parse,
 	MemoryContextSwitchTo(oldctx);
 
 	oldctx = MemoryContextSwitchTo(AQOUtilityMemCtx);
-	query_context.query_hash = get_query_hash(parse, query_string);
+	/* Check unlucky case (get a hash of zero) */
+	if (parse->queryId == UINT64CONST(0))
+		JumbleQuery(parse, query_string);
+
+	Assert(parse->utilityStmt == NULL);
+	Assert(parse->queryId != UINT64CONST(0));
+	query_context.query_hash = parse->queryId;
 	MemoryContextSwitchTo(oldctx);
 
 	MemoryContextReset(AQOUtilityMemCtx);
diff --git a/sql/aqo_fdw.sql b/sql/aqo_fdw.sql
@@ -6,6 +6,7 @@
 
 CREATE EXTENSION aqo;
 CREATE EXTENSION postgres_fdw;
+
 SET aqo.mode = 'learn';
 SET aqo.show_details = 'true'; -- show AQO info for each node and entire query.
 SET aqo.show_hash = 'false'; -- a hash value is system-depended. Ignore it.
@@ -43,10 +44,14 @@ EXPLAIN (ANALYZE, COSTS OFF, SUMMARY OFF, TIMING OFF)
 SELECT x FROM frgn;
 
 -- Push down base filters. Use verbose mode to see filters.
-EXPLAIN (ANALYZE, COSTS OFF, SUMMARY OFF, TIMING OFF, VERBOSE))
-SELECT x FROM frgn WHERE x < 10;
-EXPLAIN (ANALYZE, COSTS OFF, SUMMARY OFF, TIMING OFF, VERBOSE)
-SELECT x FROM frgn WHERE x < 10;
+SELECT str FROM expln('
+  EXPLAIN (ANALYZE, COSTS OFF, SUMMARY OFF, TIMING OFF, VERBOSE)
+    SELECT x FROM frgn WHERE x < 10;
+') AS str WHERE str NOT LIKE '%Query Identifier%';
+SELECT str FROM expln('
+  EXPLAIN (ANALYZE, COSTS OFF, SUMMARY OFF, TIMING OFF, VERBOSE)
+    SELECT x FROM frgn WHERE x < 10;
+') AS str WHERE str NOT LIKE '%Query Identifier%';
 EXPLAIN (ANALYZE, COSTS OFF, SUMMARY OFF, TIMING OFF)
 SELECT x FROM frgn WHERE x < -10; -- AQO ignores constants
 
@@ -57,14 +62,18 @@ SELECT str FROM expln('
 ') AS str WHERE str NOT LIKE '%Sort Method%';
 
 -- TODO: Should learn on postgres_fdw nodes
-EXPLAIN (ANALYZE, COSTS OFF, SUMMARY OFF, TIMING OFF, VERBOSE)
-  SELECT * FROM frgn AS a, frgn AS b WHERE a.x=b.x;
+SELECT str FROM expln('
+  EXPLAIN (ANALYZE, COSTS OFF, SUMMARY OFF, TIMING OFF, VERBOSE)
+    SELECT * FROM frgn AS a, frgn AS b WHERE a.x=b.x;
+') AS str WHERE str NOT LIKE '%Query Identifier%';
 
 -- TODO: Non-mergejoinable join condition.
 EXPLAIN (ANALYZE, COSTS OFF, SUMMARY OFF, TIMING OFF)
 SELECT * FROM frgn AS a, frgn AS b WHERE a.x<b.x;
-EXPLAIN (ANALYZE, COSTS OFF, SUMMARY OFF, TIMING OFF, VERBOSE)
-SELECT * FROM frgn AS a, frgn AS b WHERE a.x<b.x;
+SELECT str FROM expln('
+  EXPLAIN (ANALYZE, COSTS OFF, SUMMARY OFF, TIMING OFF, VERBOSE)
+    SELECT * FROM frgn AS a, frgn AS b WHERE a.x<b.x;
+') AS str WHERE str NOT LIKE '%Query Identifier%';
 
 DROP EXTENSION aqo CASCADE;
 DROP EXTENSION postgres_fdw CASCADE;
diff --git a/sql/gucs.sql b/sql/gucs.sql
@@ -1,6 +1,15 @@
 CREATE EXTENSION aqo;
-SET aqo.join_threshold = 0;
 
+-- Utility tool. Allow to filter system-dependent strings from an explain output.
+CREATE OR REPLACE FUNCTION expln(query_string text) RETURNS SETOF text AS $$
+BEGIN
+    RETURN QUERY
+        EXECUTE format('%s', query_string);
+    RETURN;
+END;
+$$ LANGUAGE PLPGSQL;
+
+SET aqo.join_threshold = 0;
 SET aqo.mode = 'learn';
 SET aqo.show_details = true;
 
@@ -10,10 +19,14 @@ ANALYZE t;
 
 SELECT true FROM aqo_reset(); -- Remember! DROP EXTENSION doesn't remove any AQO data gathered.
 -- Check AQO addons to explain (the only stable data)
-EXPLAIN (ANALYZE, VERBOSE, COSTS OFF, TIMING OFF, SUMMARY OFF)
-	SELECT x FROM t;
-EXPLAIN (ANALYZE, VERBOSE, COSTS OFF, TIMING OFF, SUMMARY OFF)
-	SELECT x FROM t;
+SELECT str FROM expln('
+  EXPLAIN (ANALYZE, VERBOSE, COSTS OFF, TIMING OFF, SUMMARY OFF)
+    SELECT x FROM t;
+') AS str WHERE str NOT LIKE '%Query Identifier%';
+SELECT str FROM expln('
+  EXPLAIN (ANALYZE, VERBOSE, COSTS OFF, TIMING OFF, SUMMARY OFF)
+    SELECT x FROM t;
+') AS str WHERE str NOT LIKE '%Query Identifier%';
 SET aqo.mode = 'disabled';
 
 -- Check existence of the interface functions.
diff --git a/sql/unsupported.sql b/sql/unsupported.sql
