Uploaded image for project: 'MariaDB Server'
  1. MariaDB Server
  2. MDEV-4369

column_stats.histogram contents doesnt make sense

Details

    • Bug
    • Status: Closed (View Workflow)
    • Major
    • Resolution: Fixed
    • None
    • None
    • None

    Description

      I used DBT-3 data, scale=10, InnoDB engine. I create statistical tables like
      specified in scripts/mysql_system_tables.sql.

      Then I use the following to generate the histogram:

      MariaDB [dbt3sf10]> set histogram_size=200;
      		 
      Query OK, 0 rows affected (0.00 sec)
      		 
       
      		 
      MariaDB [dbt3sf10]> set histogram_type='single_prec_hb';
      		 
      Query OK, 0 rows affected (0.00 sec)
      		 
      MariaDB [dbt3sf10]> analyze table customer persistent for all;

      Then I want to take a look at the histogram:

      MariaDB [dbt3sf10]> select *,hex(histogram),length(histogram) from mysql.column_stats where column_name='c_acctbal' and table_name='customer'\G
      		 
      *************************** 1. row ***************************
      		 
                db_name: dbt3sf10
      		 
             table_name: customer
      		 
            column_name: c_acctbal
      		 
              min_value: -999.99
      		 
              max_value: 9999.99
      		 
            nulls_ratio: 0.0000
      		 
             avg_length: 8.0000
      		 
          avg_frequency: 1.8319
      		 
              hist_size: 200
      		 
              hist_type: 
              histogram: ���	
      ������� "%'),.0357:<>ACEHJLOQSVXZ\_acfhjmoqtvx{}������������������������������������������������������   �       �           
                                                                                                                                       �P      �             �?                �w���          �   mmen
      		 
         hex(histogram): 020406090B0D10121417191B1E20222527292C2E303335373A3C3E414345484A4C4F515356585A5C5F616366686A6D6F717476787B7D7F828486898B8D90929497999B9EA0A2A5A7A9ACAEB0B2B5B7B9BCBEC0C3C5C7CACCCED1D3D5D8DADCDFE1E3E6E8EAEDEFF1F4F6F8FBFD00000001000000000000000300000000000000000000000B00000003500000000000000100000000000000000000000000F03F00000000000000000000000000000000F0770298C67F00000000000000000000010000006D6D656E

      Look at the hex(histogram) data. The first half of it has growing values:
      02,04,06,09,0B,0D, ..., F1,F4,F6,F8,FB,FD.. Then it goes back to zero, and
      continues at zero, with a few spikes.

      As far as I understand the meaning of values in a histogram, they should always
      increase. That is, histogram[i+1] >= histogram[i]. This is not the case for
      this histogram.

      Attachments

        Issue Links

          relates to

          Task - A task that needs to be done. MDEV-4145 Take into account the selectivity of single-table range predicates on non-indexed columns when searching for the best execution plan

          • Major - Major loss of function.
          • Closed

          Activity

            People

              igor Igor Babaev (Inactive)
              psergei Sergei Petrunia
              Votes:
              0 Vote for this issue
              Watchers:
              2 Start watching this issue

              Dates

                Created:
                Updated:
                Resolved:

                Git Integration

                  Error rendering 'com.xiplink.jira.git.jira_git_plugin:git-issue-webpanel'. Please contact your Jira administrators.