基于数据库ID的引用

Question

I have a table called branch which will store the details of a branch office like its manager, location etc. The fields are:

branch_id
branch_manager
branch_location

Data in this table would look like:

1    Mr. John Doe      Manhattan
2    Miss. Jane Doe    Utah
3    Mr. Random Guy    Somewhere

There is another table called services which stores all service details. The fields are:

service_id
service_name

Data would be like:

1   New Sales
2   After-Sales Support
3   Replacement
4   Exchange Offer

Now I want to store the services available in each branch. Normally I guess it is done by creating a new table called branch_services which will contain branch_id and service_id . So if I have following record in branch_services

branch-id  service-id
1          2
1          3
2          1

This means that the Manhattan branch does After-Sales Support and Replacement, while the Utah branch does only New Sales.

Now my question is - instead of having this branch_services table, what if I stored a comma separated list of services in the branch table itself?

1    Mr. John Doe      Manhattan    2,3
2    Miss. Jane Doe    Utah         1
3    Mr. Random Guy    Somewhere

What are the main arguments against this? Is it that queries would be more difficult/slower to execute (eg if I wanted to see which all branches did a particular service)? Is there any advantage by structuring it comma separated?

Answer 1

You can store your data in the way that you prefer (ie you can have your text fields in Cyrillic alphabet) but you have also to consider what you want to do with your data (ie you want to store English words and you and your colleagues don't know Cyrillic alphabet).

If you are working with a relational database, I suspect you are going to use SQL. Now I can write a query like:

select b.branch_location,
       s.service_name
  from branch b
inner join branch_services b_s 
        on b.branch_id = b_s.branch_id
inner join services s
        on b_s.service_id = s.service_id

To obtain:

branch_location service_name 
Manhattan       After-Sales Support 
Manhattan       Replacement 
Utah            New Sales

Because SQL is designed to query a RDMS and in our case the relationship is between the content of the columns branch_id in branch and branch_services (or service_id in branch_services and services ).

Of course, we can understand that 2,3 mean the same than 2 and 3 in two different rows, but SQL is not able to understand that. [Actually what we do is to split the 2,3 string in two values and associate each of them to a row in the services tables. Not that simple to do it in pure SQL.]

Your question is about performance, before to get there the real question is "Can I query such kind of structure with SQL?" Yes and no.

If to get a list of branches and services, like my query above, can be quite complicated, to get the branch that does a specific service is still doable, after all if need to check who does New Sales you need only to see if in your services list there is a 1, but exclude all 10, 11, 12 and so on. Uhm... not that straight forward.

What I'm trying to show you is that the three tables structure is more flexible, if you need to use it with SQL and not with a human brain. Unfortunately computers are not yet that smart.

So:

Is it possible to store data comma-separated in a column? Yes .

Is a good idea? Usually No if you want to use it in a relationship with another table.

Usually? If they need just to expose the long comma-separated string, then yes.