nVision Performance Tuning: 10. Maintaining Statistics on Non-Partitioned Tree Selector Tables

This blog post is part of a series that discusses how to get optimal performance from PeopleSoft nVision reporting as used in General Ledger.

If you do not partition your selector tables, and if you predominantly use static selectors then you can maintain the statistics on the selector table as the new static selectors are created or updated.  PeopleTools table PSTREESELCTL is used to track static selectors.  It is keyed by selector number.  A row is inserted when a new selector is created, and the version number is updated when an existing selector is re-extracted.  Therefore, a trigger on this table can be used to submit a job to refresh the statistics on the tree selector.

CREATE OR REPLACE TRIGGER sysadm.pstreeselector_stats
BEFORE INSERT OR UPDATE ON sysadm.pstreeselctl
FOR EACH ROW
DECLARE
  l_jobno      NUMBER;
  l_cmd        VARCHAR2(1000);
  l_table_name VARCHAR2(18);
  l_suffix     VARCHAR2(2);
BEGIN
  l_table_name := 'PSTREESELECT'||LTRIM(TO_CHAR(:new.length,'00'));
  l_suffix     := SUBSTR(l_table_name,-2);
  l_cmd := 'dbms_stats.gather_table_stats(ownname=>user,tabname=>'''||l_table_name||''',force=>TRUE);'
       ||'dbms_stats.set_column_stats(ownname=>user,tabname=>'''||l_table_name||''',colname=>''RANGE_FROM_'||l_suffix||''',density=>1,force=>TRUE);'
       ||'dbms_stats.set_column_stats(ownname=>user,tabname=>'''||l_table_name||''',colname=>''RANGE_TO_'||l_suffix||''',density=>1,force=>TRUE);'
  dbms_output.put_line(l_cmd);
  dbms_job.submit(l_jobno,l_cmd);
EXCEPTION WHEN OTHERS THEN NULL;
END;
/

When all the selectors exist in a single segment, the optimizer will not correctly calculate the cardinality of the expressions in the nVision queries on the tree selector tables.  Therefore, I have found it necessary to manually set the density on the range columns to 1 to get the effective execution plans at least some of the time.
N.B. This trigger should not be used if the selector tables are partitioned.