Add GUC gp_use_streaming_hashagg

Swtich GUC to check results and avoid a lot of plan diffs
in pax test cases.

Authored-by: Zhang Mingli avamingli@gmail.com

Author: avamingli

contrib/pax_storage/src/test/regress/sql/agg_pushdown.sql

@@ -1,5 +1,8 @@
 -- disable ORCA
 SET optimizer TO off;
+-- start_ignore
+set gp_use_streaming_hashagg = off;
+-- end_ignore
 
 -- Test case group 1: basic functions
 CREATE TABLE agg_pushdown_parent (

contrib/pax_storage/src/test/regress/sql/aggregates.sql

@@ -1,6 +1,9 @@
 --
 -- AGGREGATES
 --
+-- start_ignore
+set gp_use_streaming_hashagg = off;
+-- end_ignore
 
 -- start_ignore
 -- end_ignore

contrib/pax_storage/src/test/regress/sql/bfv_aggregate.sql

@@ -1,5 +1,8 @@
 create schema bfv_aggregate;
 set search_path=bfv_aggregate;
+-- start_ignore
+set gp_use_streaming_hashagg = off;
+-- end_ignore
 
 --
 -- Window function with outer references in PARTITION BY/ORDER BY clause

contrib/pax_storage/src/test/regress/sql/direct_dispatch.sql

@@ -1,5 +1,8 @@
 -- turn off autostats so we don't have to worry about the logging of the autostat queries
 set gp_autostats_mode = None;
+-- start_ignore
+set gp_use_streaming_hashagg = off;
+-- end_ignore
 
 -- create needed tables (in a transaction, for speed)
 begin;

contrib/pax_storage/src/test/regress/sql/eagerfree.sql

@@ -1,5 +1,8 @@
 create schema eagerfree;
 set search_path=eagerfree;
+-- start_ignore
+set gp_use_streaming_hashagg = off;
+-- end_ignore
 
 create table smallt (i int, t text, d date) distributed by (i);
 insert into smallt select i%10, 'text ' || (i%15), '2011-01-01'::date + ((i%20) || ' days')::interval

contrib/pax_storage/src/test/regress/sql/gp_aggregates_costs.sql

@@ -5,6 +5,9 @@ insert into cost_agg_t2 select i, random() * 99999, i % 300000 from generate_ser
 analyze cost_agg_t1;
 analyze cost_agg_t2;
 
+-- start_ignore
+set gp_use_streaming_hashagg = off;
+-- end_ignore
 --
 -- Test planner's decisions on aggregates when only little memory is available.
 --

contrib/pax_storage/src/test/regress/sql/gp_hashagg.sql

@@ -10,6 +10,9 @@ insert into hashagg_test values (1,1,'1/1/2006','hi',2);
 insert into hashagg_test values (1,1,'1/2/2006','hi',3);
 insert into hashagg_test values (1,1,'1/3/2006','hi',4);
 
+-- start_ignore
+set gp_use_streaming_hashagg = off;
+-- end_ignore
 -- this will get the wrong answer (right number of rows, wrong aggregates)
 set enable_seqscan=off;
 select grp,sum(v) from hashagg_test where id1 = 1 and id2 = 1 and day between '1/1/2006' and '1/31/2006' group by grp order by sum(v) desc;

contrib/pax_storage/src/test/regress/sql/gporca.sql

@@ -1,6 +1,9 @@
 --
 -- ORCA tests
 --
+-- start_ignore
+set gp_use_streaming_hashagg = off;
+-- end_ignore
 
 -- show version
 SELECT count(*) from gp_opt_version();

contrib/pax_storage/src/test/regress/sql/indexjoin.sql

@@ -4,6 +4,9 @@ CREATE TABLE my_tt_agg_small (
     trade_price numeric,
     trade_volume bigint
 ) DISTRIBUTED BY (symbol);
+-- start_ignore
+set gp_use_streaming_hashagg = off;
+-- end_ignore
 
 
 CREATE TABLE my_tq_agg_small (

contrib/pax_storage/src/test/regress/sql/limit_gp.sql

@@ -2,6 +2,9 @@
 -- Check for MPP-19310 and MPP-19857 where mksort produces wrong result
 -- on OPT build, and fails assertion on debug build if a "LIMIT" query
 -- spills to disk.
+-- start_ignore
+set gp_use_streaming_hashagg = off;
+-- end_ignore
 
 CREATE TABLE mksort_limit_test_table(dkey INT, jkey INT, rval REAL, tval TEXT default repeat('abcdefghijklmnopqrstuvwxyz', 300)) DISTRIBUTED BY (dkey);
 INSERT INTO mksort_limit_test_table VALUES(generate_series(1, 10000), generate_series(10001, 20000), sqrt(generate_series(10001, 20000)));