Home/ Questions/Q 8547453
In Process

The Archive Base Latest Questions

Editorial Team
  • 0
Asked: 2026-06-11T13:16:10+00:00

I have to compare three COLUMNS for each row and get the MAX data

  • 0

I have to compare three COLUMNS for each row and get the MAX data from these columns. My sample data is:

create table sept18 ( t1 number, t2 number, t3 number );

insert all
  into sept18 values (1,2,3)
  into sept18 values (-1,-2,3)
  into sept18 values (1,null,13)
  into sept18 values (-1,-2,-3)
  into sept18 values (-1,0,0)
select * from dual;

My operation should show the three columns and a field with the maximum value from each row. For instance, for the first and second rows it is 3, for the third it is 13, -1 for the fourth and 0 for the fifth.

How can I achieve this? Do I have to use partition?

Share

Leave an answer

You must login to add an answer.

Need An Account,

1 Answer

  1. Editorial Team

    No, you don’t have to use an analytic query (I assume that’s what you mean by partition?). An analytic function is used to compute an aggregate amount, of some description. You’re not looking to aggregate anything across a column; you want to find the maximum in a row.

    For this you need to use the greatest() function, the opposite of least().

    select a.*, greatest(t1, t2, t3)
  from sept18 a

    Be warned; both of these functions will return null if a single column is null. How you deal with this is your decision. My normal tactic is to use nvl() with a completely ridiculous value. This can be dangerous, as the “ridiculous” value you choose might become correct so ensure you understand your data.

    select a.*, greatest(nvl(t1,-9999), nvl(t2, -9999), nvl(t3, -9999))
  from sept18 a

    Here’s a little SQL Fiddle to demonstrate both queries.

    If all columns are null then you can descend even further into ridiculousness. Either you have to nullif() the greatest(), or, more simply, just use WHERE to exclude rows where all columns are null.

    One little extra point. The existence of a table sept18 seems a little suspicious. If you actually have one table per day then you should consider normalising your database and creating, for instance, a temporal table, which has a created_date column.

    • 0