Added/Updated tests\functional\tabloid\test_cc183f59.py: Checked on 6.0.0.467. See notes about MERGE JOIN selecting.

2025-02-02 10:50:42 +01:00 · 2024-09-25 00:09:52 +03:00 · 2024-09-25 00:09:52 +03:00 · a60b8c72cb
commit a60b8c72cb
parent 7f28961f1b
1 changed files with 102 additions and 0 deletions
--- a/tests/functional/tabloid/test_cc183f59.py
+++ b/tests/functional/tabloid/test_cc183f59.py
@ -0,0 +1,102 @@
+#coding:utf-8
+
+"""
+ID:          None
+ISSUE:       https://github.com/FirebirdSQL/firebird/commit/cc183f599ee09233d6da294893f651de0ab76136
+TITLE:       Add key info to the merge join plan output
+DESCRIPTION:
+NOTES:
+    MERGE JOIN will be chosen by optimizer when several conditions are met:
+    * data sources are joined by INNER join;
+    * data sources are ORDERED, BUT *NOT* via GROUP BY (because optimizer tends to think about grouped data
+      that its cardinality *much* less than cardinality of source and it causes to decide using hash join instead).
+      This was explained by dimitr privately, letter: 24-sep-2024 16:30.
+    * number of conflicts in the hash table must be above 1009 * 1000 = 1009000.
+      Experimental shows that minimal threshold for switching from HJ to MJ is 1009883 rows.
+      This value must NOT depend on machine but can have limited dependency on page_size
+      (see letter from dimitr, 24-sep-2024 19:09).
+    * No ticket has been created for this test.
+
+    Checked on 6.0.0.467.
+"""
+
+import re
+import time
+
+import pytest
+from firebird.qa import *
+
+init_sql = """
+    create table test1(id int not null);
+    create table test2(id int not null, pid int not null);
+
+    set stat on;
+    set echo on;
+    set term ^;
+    execute block as
+        declare n_cnt int =    1009883; -- OK, plan MERGE JOIN
+        -- declare n_cnt int =    1009882; -- plan HJ
+        declare i int  = 0;
+    begin
+        while (i < n_cnt) do
+        begin
+            insert into test1(id) values(:i);
+            i = i + 1;
+        end
+    end
+    ^
+    set term ;^
+    insert into test2(id, pid) select row_number()over(), id from test1;
+    commit;
+"""
+db = db_factory(init = init_sql, page_size = 8192)
+
+substitutions =  [  ('[ \t]+', ' ')
+                   ,('keys: \\d+, total key length: \\d+', 'keys, total key length')
+                   ,('record length: \\d+, key length: \\d+', 'record length, key length')
+                 ]
+
+act = python_act('db', substitutions = substitutions)
+
+
+#-----------------------------------------------------------
+
+def replace_leading(source, char="."):
+    stripped = source.lstrip()
+    return char * (len(source) - len(stripped)) + stripped
+
+#-----------------------------------------------------------
+
+@pytest.mark.version('>=6.0')
+def test_1(act: Action, capsys):
+
+    test_sql = """
+        select a.id, b.pid
+        from (
+            select id from test1 order by id
+        ) a
+        join 
+        (
+            select pid from test2 b order by pid
+        ) b
+        on a.id = b.pid
+    """
+
+    with act.db.connect() as con:
+        cur = con.cursor()
+        ps = cur.prepare(test_sql)
+        print( '\n'.join([replace_leading(s) for s in ps.detailed_plan.split('\n')]) )
+
+    act.expected_stdout = """
+        Select Expression
+        ....-> Filter
+        ........-> Merge Join (inner) (keys: 1, total key length: 8)
+        ............-> Sort (record length: 28, key length: 8)
+        ................-> Sort (record length: 28, key length: 8)
+        ....................-> Table "TEST1" as "A TEST1" Full Scan
+        ............-> Sort (record length: 28, key length: 8)
+        ................-> Sort (record length: 28, key length: 8)
+        ....................-> Table "TEST2" as "B B" Full Scan
+    """
+    act.stdout = capsys.readouterr().out
+    assert act.clean_stdout == act.clean_expected_stdout