Selecting a distinct column that is a CLOB datatpye

The supplied PL\SQL package: dbms_lob has lots of procedures and functions that you can use on CLOB columns, but no, you cannot use most standard SQL commands on CLOB columns.

Just a thought - do what Oracle does when you enter a query - hash the description into something more manageable, and put it in another column, and select distinct on that instead.

Assuming you arent updating these descriptions continuously, it will be much more efficient than scanning your CLOBs again and again.

I havent got a decent hash algorithm at hand, but a simple convert it all to ASCII character by character and add it up isnt a bad starting point.

HTH
Bob

A perhaps not-so-elegant solution (but one that works - if slowly) would be to take the first 4000 characters of your clob, cast it to a varchar2 and then do a distinct on that!  Fewer characters would be preferable, and this would only work if the differences do occur in the first 4000 characters.

SELECT DISTINCT CAST (SUBSTR(description, 0, 4000) AS VARCHAR2(4000)) AS description, threat_id
FROM   foo

Also, be careful with the "distinct" operator in Oracle queries.  If you have SQL Server experience, you may be used to including that in most queries out of habit. ***DON'T DO THAT IN ORACLE QUERIES!***  If you only select one column, then it is OK, but if you want to select multiple columns based on a distinct value in one (or more) columns, test the results and the response times thoroughly before you deploy this.  "Distinct" will always force a sort that you may not need.   Make sure that Oracle's implementation of "distinct" matches your knowledge of the data - the results may surprise you.  It usually works better in Oracle to use a group operator (min, max, avg, count, etc.) when you want a distinct result set.