SQL [Oracle]: Count number of previous records with same client ID

Question

Had a brief search online to see if an answer to this already existed. Couldn't find anything that tackled this particular situation (which was a surprise, as I would have assumed it was fairly common).

Basically I have a table that records registrations with a service, along with the ID of the client registering, and the start and end dates of the registration. See below for an example of the table structure:

Reg ID | Client ID | Reg Start  | Reg End

R6       C1          01-06-2016   Null
R5       C2          20-05-2016   02-06-2016
R4       C2          14-03-2016   11-05-2016
R3       C1          10-03-2016   05-05-2016
R2       C2          28-01-2016   02-02-2016
R1       C2          10-11-2015   23-12-2015   

What I want to do, is pull the records from the table but also add a column that returns a count of how many previous registrations are in the table with the same client ID for each record. For the sake of clarity, I only want to count registrations for each record that have the same client ID and have a start date less than that of the record.

As such, a successful query would return the following output:

Reg ID | Client ID | Reg Start  | Reg End    | # previous reg

R6       C1          01-06-2016   Null         1
R5       C2          20-05-2016   02-06-2016   3
R4       C2          14-03-2016   11-05-2016   2
R3       C1          10-03-2016   05-05-2016   0
R2       C2          28-01-2016   02-02-2016   1
R1       C2          10-11-2015   23-12-2015   0

Has anyone had previous experience of doing this who could offer a viable solution?

My initial thought process was to create a sub query that would count registrations grouped by client ID, and then join this to my main query using client ID. For example:

SELECT
t1.reg_id
t1.client_id
t1.reg_start
t1.reg_end
FROM registrations t1
JOIN
  (SELECT
   client_id
   count(reg_id)
   FROM registrations
   GROUP BY client_id) t2
ON t1.client_id = t2.client_id

However, this would return a count of the total registrations for a given client per record, where as I specifically require the previous registrations.

Any input that the hive mind could provide would be much appreciated. I think I've hit a bit of a wall with my currently quite limited SQL knowledge here :/

Also I should probably mention that I'm using an Oracle database!

Answer 1

You can do this with row_number() :

select r.*,
       (row_number() over (partition by client_id order by reg_start) - 1) as cnt
from registrations r;

Answer 2

This can be done with a window function and a running sum:

select reg_id, client_id, reg_start, reg_end,
       count(*) over (partition by client_id order by reg_start) - 1 as previous_reg
from registrations
order by reg_start desc;

Answer 3

In SQL Server you can use the following query:

SELECT [Reg ID], [Client ID], Reg_Start, Reg_End,
       COUNT(*) OVER (PARTITION BY [Client ID] ORDER BY Reg_Start) - 1
FROM mytable

COUNT , when applied with an ORDER BY clause, returns the running total of records count. If this is available in Oracle, then it should return the value you want.

Answer 4

The easiest way, in my opinion, is to build a subquery in the SELECT statement of main query, as follow:

SELECT 
    t1.*, 
    (
        SELECT COUNT(*) 
        FROM registrations t2 
        WHERE t2.client_id = t1.client_id
          AND t2.reg_start < t1.reg_start
    ) AS previous_reg
FROM registrations t1
ORDER BY reg_id DESC