如何将不带过滤器的子查询连接到带过滤器的查询？

Question

I have a query similar to the simplified one below.

SELECT
  Users.ID,
  Registrations.Timestamp
FROM
  Users,
  Registrations
WHERE
  Users.ID = Registrations.UserID AND
  Registrations.Date >= '2020-02-27'
ORDER BY
  Users.ID,
  Registrations.Timestamp;

I want to calculate a third column named Type with the values entry and departure like the following desired result:

ID  | Timestamp           | Type
----+---------------------|----------
1   | 2020-02-27 05:43:24 | entry
1   | 2020-02-27 13:48:47 | departure
1   | 2020-02-28 05:44:38 | entry
1   | 2020-02-28 13:50:11 | departure
2   | 2020-02-27 13:44:41 | entry
2   | 2020-02-27 21:47:54 | departure
2   | 2020-02-28 13:40:16 | entry
2   | 2020-02-28 21:52:57 | departure
3   | 2020-02-27 05:46:20 | departure
3   | 2020-02-28 21:44:05 | entry
3   | 2020-02-28 05:47:18 | departure

The way it calculates it is by counting the amount of registrations up to that one (without the Date filter). If the resulting count number is odd then consider it an entry else if it's even a departure .

In the example the user number 3 starts as departure because even though it is the first registration (odd number), without the Date filter it's an even number.

I have tried using a subquery but can't figure out how to count only up to the Timestamp , not all of them.

SELECT
  Users.ID,
  Registrations.Timestamp,
  CASE WHEN
    (
      SELECT
        COUNT(Registrations.Timestamp)
      FROM
        Registrations
      WHERE
        Registrations.UserID = OuterUsers.ID AND
        Registrations.Timestamp <= OuterRegistrations.Timestamp
    ) % 2 = 0 THEN
    'departure'
  ELSE
    'entry'
  END AS Type
FROM
  Users,
  Registrations
WHERE
  Users.ID = Registrations.Timestamp.UserID AND
  Registrations.Date >= '2020-02-27'
ORDER BY
  Users.ID,
  Registrations.Timestamp;

This sort of problem is likely not unique but I can't figure out even one solution to it from searching through some similar questions.

Answer 1

Use row_number() first and then filter. Also use to learn proper,explicit, standard , readable JOIN syntax:

Select u.*, r.*,
       (case when seqnum % 2 = 1  then 'entry' else 'departure' end) as type
FROM users u JOIN
     (SELECT r.*,
             ROW_NUMBER() OVER (PARTITION BY UserId ORDER BY Date) as seqnum
      FROM Registrations r
     ) r
     ON u.ID = r.Timestamp.UserID
WHERE r.Date >= '2020-02-27'
ORDER BY u.ID, r.Timestamp;

Answer 2

Why don't you add a rank/row number ordered by timestamp and use these values to determine departure and entry instead

Select *, case when row_num % 2 = 0  then 'departure' else 'entry' end as type
FROM
(SELECT
    Users.ID,
     row_number() over (partition by ID order by timestamp) as row_num
        FROM
          Users,
          Registrations
        WHERE
          Users.ID = Registrations.Timestamp.UserID AND
          Registrations.Date >= '2020-02-27'
        ORDER BY
          Users.ID,
          Registrations.Timestamp) a;