Oracle-SQL语句性能不佳-模糊匹配逻辑

Question

I have a fuzzy matching requirement .. for ex;

Table1 - T1Col1, T1col2, T1col3, T1col4, T1col5. 
Table 2 - T2Col1, T2col2, T2col3, T2col4, T2col5. 

so my requirement is

• T1 - Its not necessary all fields ie T1col1, T1col2, T1col3, T1col4, T1col5 are not nulls but there are scenarios where T1Col2 is populated and T1Col3, T1Col4 and 5 is null. Best case scenario here is all fields are not nulls and the worst case is except T1Col1 rest of the fields are nulls.

• I came up with a fuzzy logic matching so that if atleast one field is matching then 'where' clause should pass through.

select count(*) from T1, T2
where
  Nvl(T1COl1, nvl(T2Col1, 'x')) = nvl(T2Col1, 'x') and
  Nvl(T1COl2, nvl(T2Col2, 'x') ) = nvl(T2Col2, 'x') and
  Nvl(T1COl3, nvl(T2Col3, 'x'))  = nvl(T2Col3, 'x') and
  Nvl(T1COl4, nvl(T2Col4, 'x')) = nvl(T2Col4 'x') and
  and substr(T1COl5, 1,1) = T2Col5
  ;

Record count in T1 and T2 are 243000 and 55000 records respectively When I run the above statement, it takes 1426.809 seconds and gave me 11349 records. Looks like it is performance poor. Is that because of usage of substr or usage of too many NVLs in where clause?

Can you help me here how can I improve my query performance or is there a better way of doing that matching?

Answer 1

Maybe just a personal preference but I would write this as:

select count(*)
  from t1
      ,t2
 where (t1col1 is null or t1col1 = t2col1)
   and (t1col2 is null or t1col2 = t2col2)
   and (t1col3 is null or t1col3 = t2col3)
   and (t1col4 is null or t1col4 = t2col4)
   and substr(t1col5, 1, 1) = t2col5;

• Clearer logic that null value in t1 column is ok.
• No nvl calculations necessary
• Make use of indexes on columns possible.

The logic in this query is slightly different. This query does not match the value 'x' in t1col1 to an empty value in t2col1.