SQL to return the last record (SQL Server)

Friday 7 December 2018 @ 10:23 pm

A few years ago I wrote a post that showed a more efficient way to find a ‘last’ record in a SQL command, while pulling several values from that record.

The example I used then was written in the SQL Syntax for MS Access, but the same method can be used in other flavors of SQL. The main difference is the syntax for converting the date to a string. Since I just had to look up that syntax for MS SQL Server I decided to post this here so I can find it easily next time.  Here is a brief recap of what it does:

In SQL it is very easy to add a GROUP BY clause to a query and then calculate the Min() or Max() date for that group. For instance the last order date for a customer. What is not so simple is to return other fields from each last order record, like the Order Number and Amount. The typical solution is to find the maximum in initial query, and then link that to another query that matches that Max() value back to the matching record as an inner join. But this requires at least to queries, and can get complicated if the GROUP BY field and the Max() field are in separate tables.

I find that it is simpler and more efficient to do the Max() on a combined expression that includes all the fields you want. This works as long as you start the expression with the date converted to a string in “yyyymmdd” format. This format ensures that the Max() of the combined expression will be the date and values from the last record.

Take a simple example. We need to write SQL to return each customer’s last order. And from that order you also want to return the customer name, order date, order number and amount. The SQL would look like this:

SELECT
Customer.ID,
max(
CONVERT(varchar(16), Order.OrdDate, 121) + '*' +
Orders.OrderNo + '*' +
Orders.Amount + '*' +
Customer.Name
) as MaxSTring
FROM Orders Orders INNER JOIN Customer Customer ON Orders.CustomerID=Customer.ID
WHERE (Orders.OrdDate>={ts '2018-11-01 00:00:00'} AND Orders.OrdDate<{ts '2018-12-01 00:00:00'})
GROUP BY Customer.ID

This returns one row per customer and the row might look something like this:

20180630*12345*250.87*Joe Smith

You can then write a formula in Crystal to split the string back into pieces. For instance the Customer Name formula would look like this:

Split ({Command.MaxString}, '*') [4]

In English this says, split the MaxString into a list (array) based on the asterisks. Take the 4th element.

Some of you might be wondering why we don’t just do this in the report.  You have a point since this is simple in Crystal:

Group by customer, sort by date, hide the Details and Group Header. Show all the fields on the Group Footer.

This works great, but is not very efficient if you have a large volume of data, because it requires Crystal to read all the details and display the last one. The SQL version above would only return one record per customer to the report.









Leave a Reply

Jeff-Net
Recrystallize Pro

The Expert Series