从非规范化表中选择最新记录

Question

I have a table set up with columns:

• fname
• lname
• address
• city
• state
• zip
• Customer ID
• date_modified

the data is basically denormalized so looks like

Ben -- Smith--***123first*** st -- NY -- NY -- 12101 -- 123 --  1-1-2011
Ben -- Smith--***123 1st st*** -- NY -- NY -- 12101 -- 123 -- 1-1-2011
Sara -- Smith -- BLAH BLAH BLAh

I am trying to copy these records to a new table but I only want one record per Customer ID.

I tried doing something like

 insert into new_table(fname,lname,address,city,state,zip,Customer_ID,
                      date_modified)

 select fname,lname,address,city,state,zip,Customer_ID,date_modified
 group by (fname,lname,address,city,state,zip,Customer_ID,date_modified)

the issue is there are too many addresses and other columns that have the same meaning but different text (first vs 1st). so the group by leaves both of these records in the new table. how do i esentialy get one record for each customer id by choosing the max(date modified). Basically I would want to group by just the customer_id and not the rest of the columns but that isn't allowed in oracle.

Answer 1

The following is one of several ways to get what you seem to want. Keep in mind though that this does not normalize your database. You still have a customer_id in with a first and last name and an address. I'd probably turn this into several inserts, one to get all of the unique customer IDs along with the latest name information for the Customers table, then another insert for the addresses. If you want historical information of changes then you would need to adjust appropriately.

Also, the code below will not work properly if you have two rows that have the same exact customer ID and the same exact date_modified . If you run into that case you'll need to come up with the appropriate business logic to handle that.

INSERT INTO New_Table (
    fname,
    lname,
    address,
    city,
    state,
    zip,
    Customer_ID,
    date_modified )
SELECT
    OT1.fname,
    OT1.lname,
    OT1.address,
    OT1.city,
    OT1.state,
    OT1.zip,
    OT1.customer_id,
    OT1.date_modified
FROM
    (
    SELECT
        customer_id,
        MAX(date_modified) AS latest_date_modified
    FROM
        Old_Table
    GROUP BY customer_id) SQ
INNER JOIN Old_Table OT1 ON
    OT1.customer_id = OT1.customer_id AND
    OT1.date_modified = SQ.latest_date_modified

Answer 2

This is quite easy by using analytical (aka windowing) functions to select the first row for each customer. In case two rows have the same date_modified, it is not defined which one is taken.

INSERT INTO new_table (fname,lname,address,city,state,zip,Customer_ID,date_modified)
SELECT fname,
       lname,
       address,
       city,
       state,
       zip,
       Customer_ID,
       date_modified
FROM (
   SELECT fname,
          lname,
          address,
          city,
          state,
          zip,
          Customer_ID,
          date_modified,
          row_number() over (partition by customer_id order by date_modified desc) as rn
) 
WHERE rn = 1