Tersus: Re: Modeling relation data

Re: Modeling relation dataDate: May 15, 2013Helge,Thank you for you kind words on Tersus.What you write on the lack of declarative definitions of database relations in Tersus is correct, and I understand your concerns.However, from our experience, this is less problematic than it seems.Reflecting on very large Tersus-based projects we have been involved with in the last few years (with hundreds of tables):

It is possible, as you mention, to define the referential integrity constraints directly in the RDBMS.

I personally view foreign key constraints as safety measures, and avoid cascading operations:

When a record Ra is added to table A, and then a record Rb referring to it is added to table B, the natural logic of the system requires that Ra is created first, and only then Rb.
When the above record Ra is deleted, I am not sure I want record Rb to be deleted as well. Many times such a situation is an indication of some bug. Proper flow requires that Ra is deleted only after verifying Rb no longer exists (i.e. was previously deleted), and this means the referential integrity is preserved.
The above does not imply, of course, that foreign keys in the database are useful to block the creation of inconsistent data.

As for readability - following strict naming conventions is very useful to clarify the semantics of columns in general, and foreign keys specifically. Without naming conventions, checking technically to which column in which table a certain foreign key refers is possible but labor intensive, and with foreign keys it usually becomes redundant.

Yes, as I said, I think the criticism is in place. Unfortunately, at this stage we do not plan any enhancement in this respect.Regards,OferSee Also:

To use the full functionality of this web site, JavaScript needs to be turned on.

For best results, use the Firefox browser..

Loading ...