Uploaded image for project: 'MariaDB Server'
  1. MariaDB Server
  2. MDEV-428

Select type DEPENDENT SUBQUERY in SHOW EXPLAIN vs SUBQUERY in EXPLAIN

    Details

    • Type: Bug
    • Status: Closed
    • Priority: Minor
    • Resolution: Won't Fix
    • Affects Version/s: None
    • Fix Version/s: 10.0.0
    • Component/s: None
    • Labels:
      None

      Description

      EXPLAIN SELECT * FROM t1 
      WHERE a IN ( SELECT b FROM t2 WHERE b <= a OR b IN ( SELECT a FROM t1, t3 WHERE a != b ) );
      id	select_type	table	type	possible_keys	key	key_len	ref	rows	Extra
      1	PRIMARY	t1	ALL	NULL	NULL	NULL	NULL	43	
      1	PRIMARY	t2	ALL	NULL	NULL	NULL	NULL	12	Using where; FirstMatch(t1); Using join buffer (flat, BNL join)
      3	SUBQUERY	NULL	NULL	NULL	NULL	NULL	NULL	NULL	Impossible WHERE noticed after reading const tables
      #---------------
      # SHOW EXPLAIN output:
      id	select_type	table	type	possible_keys	key	key_len	ref	rows	Extra
      1	PRIMARY	t1	ALL	NULL	NULL	NULL	NULL	43	
      1	PRIMARY	t2	ALL	NULL	NULL	NULL	NULL	12	Using where; FirstMatch(t1); Using join buffer (flat, BNL join)
      3	DEPENDENT SUBQUERY	NULL	NULL	NULL	NULL	NULL	NULL	NULL	no matching row in const table
      

      bzr version-info

      revision-id: psergey@askmonty.org-20120801115152-qgxtqd0a5vlqrrhx
      date: 2012-08-01 15:51:52 +0400
      build-date: 2012-08-02 13:13:01 +0300
      revno: 3461
      

      Default optimizer_switch:

      index_merge=on,index_merge_union=on,index_merge_sort_union=on,index_merge_intersection=on,index_merge_sort_intersection=off,engine_condition_pushdown=off,index_condition_pushdown=on,derived_merge=on,derived_with_keys=on,firstmatch=on,loosescan=on,materialization=on,in_to_exists=on,semijoin=on,partial_match_rowid_merge=on,partial_match_table_scan=on,subquery_cache=on,mrr=off,mrr_cost_based=off,mrr_sort_keys=off,outer_join_with_cache=on,semijoin_with_cache=on,join_cache_incremental=on,join_cache_hashed=on,join_cache_bka=on,optimize_join_buffer_size=off,table_elimination=on,extended_keys=off
      

      Test case:

      CREATE TABLE t1 (a INT);
      INSERT INTO t1 VALUES 
      (12),(22),(9),(45),(13),(11),(17),(33),(26),(11),(9),(33),(17),(10),(10),
      (42),(20),(28),(14),(66),(25),(65),(23),(11),(11),(10),(9),(24),(19),(59),
      (9),(14),(9),(14),(29),(10),(10),(10),(11),(9),(9),(12),(18);
      
      CREATE TABLE t2 (b INT);
      INSERT INTO t2 VALUES (1),(86),(1),(1),(1),(0),(0),(0),(0),(8),(0),(0);
      
      CREATE TABLE t3 (c INT);
      
      
      let $query =
      SELECT * FROM t1 
      WHERE a IN ( SELECT b FROM t2 WHERE b <= a OR b IN ( SELECT a FROM t1, t3 WHERE a != b ) );
      
      
      eval EXPLAIN $query;
      --echo #---------------
      --echo # SHOW EXPLAIN output:
      
      --connect (con1,localhost,root,,)
      --let $con_id = `SELECT CONNECTION_ID()`
      
      --let $trials = 50
      
      --disable_query_log
      
      while ($trials)
      {
        --dec $trials
        --let $run = 1000
      
        --send_eval $query
      
        --connection default
        while ($run)
        {
          --error 0,1932
          eval SHOW EXPLAIN FOR $con_id;
          --dec $run
          if (!$mysql_errno)
          {
            --let $run = 0
            --let $trials = 0
            --let $found = 1
          }
        }
      
        --disable_result_log
        --connection con1
        --reap
        --enable_result_log
      
      }
      
      if (!$found)
      {
        --echo ########### Could not catch the query by SHOW EXPLAIN, try again  #############
      }
      
      DROP TABLE t1, t2, t3;
      

        Gliffy Diagrams

          Attachments

            Issue Links

              Activity

              Hide
              psergey Sergei Petrunia added a comment -

              Repetable only when the tables have type=MyISAM.

              Look at the query:

              SELECT * FROM t1
              WHERE a IN ( SELECT b FROM t2 WHERE b <= a OR b IN ( SELECT a FROM t1, t3 WHERE a != b ) );

              the subquery is
              ( SELECT a FROM t1, t3 WHERE a != b )

              t1 has column a, t3 has column c. `b` is only in table t2, which means the subquery IS correlated.

              Show
              psergey Sergei Petrunia added a comment - Repetable only when the tables have type=MyISAM. Look at the query: SELECT * FROM t1 WHERE a IN ( SELECT b FROM t2 WHERE b <= a OR b IN ( SELECT a FROM t1, t3 WHERE a != b ) ); the subquery is ( SELECT a FROM t1, t3 WHERE a != b ) t1 has column a, t3 has column c. `b` is only in table t2, which means the subquery IS correlated.
              Hide
              psergey Sergei Petrunia added a comment -

              I tried the testcase on MariaDB 5.2 (explicitly specified type=MYISAM for the tables to be certain). I've got:

              +id select_type table type possible_keys key key_len ref rows Extra
              +1 PRIMARY t1 ALL NULL NULL NULL NULL 43 Using where
              +2 DEPENDENT SUBQUERY t2 ALL NULL NULL NULL NULL 12 Using where
              +3 DEPENDENT SUBQUERY NULL NULL NULL NULL NULL NULL NULL Impossible WHERE noticed after reading const tables

              Show
              psergey Sergei Petrunia added a comment - I tried the testcase on MariaDB 5.2 (explicitly specified type=MYISAM for the tables to be certain). I've got: +id select_type table type possible_keys key key_len ref rows Extra +1 PRIMARY t1 ALL NULL NULL NULL NULL 43 Using where +2 DEPENDENT SUBQUERY t2 ALL NULL NULL NULL NULL 12 Using where +3 DEPENDENT SUBQUERY NULL NULL NULL NULL NULL NULL NULL Impossible WHERE noticed after reading const tables
              Hide
              psergey Sergei Petrunia added a comment -

              I've debugged both EXPLAIN and SELECT. Both conclude that the subquery can only produce zero rows, however they arrive at the conclusion through different routes.

              EXPLAIN hits these lines in JOIN::optimize:

              if (conds && const_table_map != found_const_table_map &&
              (select_options & SELECT_DESCRIBE))

              { conds=new Item_int((longlong) 0,1); // Always false }

              Note that the WHERE clause is replaced with Item_int(1) here. This is how correlated subquery becomes un-correlated.

              SELECT hits these lines in JOIN::optimize:

              if (const_table_map != found_const_table_map &&
              !(select_options & SELECT_DESCRIBE) &&
              (!conds ||
              !(conds->used_tables() & RAND_TABLE_BIT) ||
              select_lex->master_unit() == &thd->lex->unit)) // upper level SELECT

              { zero_result_cause= "no matching row in const table"; DBUG_PRINT("error",("Error: %s", zero_result_cause)); error= 0; goto setup_subq_exit; }

              The subquery is declared as "no matching row in const table", but its WHERE clause is not removed.

              Show
              psergey Sergei Petrunia added a comment - I've debugged both EXPLAIN and SELECT. Both conclude that the subquery can only produce zero rows, however they arrive at the conclusion through different routes. EXPLAIN hits these lines in JOIN::optimize: if (conds && const_table_map != found_const_table_map && (select_options & SELECT_DESCRIBE)) { conds=new Item_int((longlong) 0,1); // Always false } Note that the WHERE clause is replaced with Item_int(1) here. This is how correlated subquery becomes un-correlated. SELECT hits these lines in JOIN::optimize: if (const_table_map != found_const_table_map && !(select_options & SELECT_DESCRIBE) && (!conds || !(conds->used_tables() & RAND_TABLE_BIT) || select_lex->master_unit() == &thd->lex->unit)) // upper level SELECT { zero_result_cause= "no matching row in const table"; DBUG_PRINT("error",("Error: %s", zero_result_cause)); error= 0; goto setup_subq_exit; } The subquery is declared as "no matching row in const table", but its WHERE clause is not removed.
              Hide
              psergey Sergei Petrunia added a comment -

              Elena, please add a suppression, so that RQG doesn't consider a difference
              between

              type=SUBQUERY Extra='Impossible WHERE noticed after reading const tables'
              type=DEPENDENT SUBQUERY Extra='no matching row in const table'

              meaningful.

              Show
              psergey Sergei Petrunia added a comment - Elena, please add a suppression, so that RQG doesn't consider a difference between type=SUBQUERY Extra='Impossible WHERE noticed after reading const tables' type=DEPENDENT SUBQUERY Extra='no matching row in const table' meaningful.
              Hide
              psergey Sergei Petrunia added a comment -

              Won't fix; natural property of the code.

              Show
              psergey Sergei Petrunia added a comment - Won't fix; natural property of the code.

                People

                • Assignee:
                  psergey Sergei Petrunia
                  Reporter:
                  elenst Elena Stepanova
                • Votes:
                  0 Vote for this issue
                  Watchers:
                  2 Start watching this issue

                  Dates

                  • Created:
                    Updated:
                    Resolved:

                    Time Tracking

                    Estimated:
                    Original Estimate - Not Specified
                    Not Specified
                    Remaining:
                    Remaining Estimate - 0 minutes
                    0m
                    Logged:
                    Time Spent - 1 hour, 30 minutes
                    1h 30m