2 Replies Latest reply on Aug 30, 2002 3:06 PM by 11767

    Native compilation of ctx_doc package to improve performance


      I am using Oracle text on Oracle 9i.

      I have a requirement like this -

      1. Source table has rows with company names (Say around 20,000 rows)
      2. A master table with domain index on company name column (Around 800,000 rows)
      3. I need to match rows in source table with the master table based on company name tokens

      I do it this way -

      select st.source_company_name, mt.master_rowid
      from source_table st, master_table mt
      where contains ( mt.master_table, st.source_company_name, 1 ) > 0
      -- st.source_company_name is actually broken into tokens and I seach it like this -
      --"fuzzy({oracle}) or fuzzy({corporation})"
      ctx_doc.highlight(source_string, rowid....);
      -- Here I get a pl/sql table and then I take a count of the values in pl/sql table and
      -- saw that n tokens from source string have matched in n tokens of dest string
      end loop

      But the problem is performance. Can I improve performance by using native compilation of cts_doc package?
      Is it worth the efforts?

      Is there a better way of doing this?

      Please help.

      Thanks and regards