Database Table Design for Group Values with Changing Status over Time - sql

Question

I have the following groups that have a particular designation depending on the date:
Group 1: 3/30/2017 to present: status 'on'
Group 2: 3/30/2017 to present: status 'on'
Group 3: 3/30/2017 to present: status 'on'
Group 4: 3/30/2017 to 6/1/2017: status 'off'; 6/2/2017 to present: status: 'on'
Group 5: 3/30/2017 to present: status 'off'
Group 6: 3/30/2017 to 7/10/2017: status 'off'; 7/11/2017 to present: status 'on'
I'm trying to translate this information into an effective database table so I can designate a change in status on a particular date.
I have a process that runs daily in near real time that checks the status of each group and undertakes various processes based on the status.
I have come up with the following though I think it is not sufficient:
Group Effective Date Termination Date Status
Group 1 '2017-03-30' NULL On
Group 2 '2017-03-30' NULL On
Group 3 '2017-03-30' NULL On
Group 4 '2017-03-30' '2017-06-01' On
Group 4 '2017-06-02' NULL Off
Group 5 '2017-03-30' NULL Off
Group 6 '2017-03-30' '2017-07-10' Off
Group 6 '2017-07-11' NULL On
So if I run my daily process historically, I want it to be able to consult the table and determine the status of the group. If I am running my process in real time, I want to be able to consult the table and determine the status. If I want to change the status at a particular point in time, I enter a termination date for the Group and status and start a new line.
I can't imagine this is a good way to do this.
Looking for insights.
Thanks in advance.

Answer 1

Here is one method using what I call Version Normal Form (vnf). It works for entities that have a smooth, unbroken chain of state changes. That is, there are no gaps (one state ends only upon another state taking effect) or overlaps (only one state is in effect at any time).
Design the Group table with all the group info except for status.
create table Group(
ID int auto generated primary key,
... ... -- all other Group data
);
Now create a Status table with a State field and one date field -- the date the status takes effect.
create table GroupStatus(
ID int references Group( ID ),
EffDate date not null default Now(),
State char( 1 ) check (State in ('Y', 'N')),
constraint PK_GroupStatus primary key( ID, EffDate )
);
There are two important points about the GroupStatus table to consider:
the PK definition means no two entries for the same Group can be defined for the same time. Thus, it is not possible to have overlapping status values.
there is no "end" date. A status takes effect at the designated date and time and continues in effect until replaced by another state change. Thus, it is not possible to have gaps between the status changes of any Group.
I used a single character 'Y' (for On) and 'N' (for Off) but you can define the status state any way you want. This is for illustration only.
The EffDate field may have to be Date, Datetime or Timestamp type, depending on your specific DBMS. Now() just means "current date and time" using any method available in your DBMS.
The GroupStatus data would look like this:
ID EffDate State
1 '2017-03-30' Y
2 '2017-03-30' Y
3 '2017-03-30' Y
4 '2017-03-30' Y
4 '2017-06-02' N
5 '2017-03-30' N
6 '2017-03-30' N
6 '2017-07-11' Y
For the level of data integrity enforced, the design is very simple. The queries will be a little more complicated.
To see the current status of Group 1:
select g.ID as 'Group', s.EffDate as 'Effective Date',
case s.State when 'Y' then 'On' else 'Off' end as Status
from Group g
join GroupStatus s
on s.ID = g.ID
and s.EffDate =(
select Max( s1.EffDate )
from GroupStatus s1
where s1.ID = g.ID
and s1.EffDate <= Now()
)
where g.ID = 1;
To see the current status of all groups, just omit the where clause. To see the status of group 1 that was in effect on a certain date, just change the Now() in the subquery to a variable loaded with the date and time of interest.
In fact, set the query for current status of all groups as a view. Then your daily process can simply query:
select ID, Status from CurrentGroupStatus;
Since you know there can be no gaps or overlaps, you know there will be one and only one row for each group.
Suppose upon inserting the group 6 entry on March 30, you already know the date it will be turned on. You can go ahead and insert the GroupStatus entry with the future date (July 11) and the "current" queries and view will continue to show the correct status (Off) until that date arrives, at which point the ON status will start appearing.
Create "instead of" triggers on the view(s) to correctly work with the underlying tables and your apps don't even have to know the details of how the data is stored.
This gives you rock solid data integrity and a lot of flexibility in how you view and manipulate the data.