Question

I’m building a database table representing a ‘note’ that has a parent-child relationship in two of it’s columns, like so:

ID           INT(10) PK NOT NULL UNSIGNED ZEROFILL AUTOINCREMENT  -- Primary key
parent_ID    INT(10) UNSIGNED ZEROFILL                 -- References Primary key
username     VARCHAR(30) NOT NULL
information  VARCHAR(256) NOT NULL
comment      VARCHAR(256) NOT NULL

where username is a foreign key to a user table and is the username of the person who wrote the note, information is the information contained in the note and comment is what username added. comment will always have something in it, and is not necessarily the same as the previous user.

The concept is that someone can “copy” a note and have their own comments, and say “I got this note from xxx”, hence the parent-child relationship. A bit like sharing in facebook, perhaps.

How do I correctly form the parent_ID column? Should I use an identifying relationship; which sides are mandatory?
I figure this has to be a 1-to-Many relationship since five people can copy the same note.

I expect there to be more copying than creating new notes, so there will be a relatively small number of NULL‘s in the records, but to get rid of nulls entirely should I apply a NOT NULL constraint and simply make the default parent 0 with an essentially meaningless record at ID 0 and take note of this in the software?

Is this even the correct way? Should I employ a two-table system with:

ID          INT(10)      PK
information VARCHAR(256)
orig_user   VARHCAR(30)  FK  -- Potentially

and

ID          INT(10)      PK FK
username    VARCHAR(30)  PK FK
comment     VARCHAR(256)

Which eliminates any possible NULL‘s by definiton?

Thanks

Answer 1

Should I use an identifying relationship

No, only ID should be in PK since it alone is unique. Also, a root (parent-less) note will have a NULL parent_ID and NULLs cannot be in PK.

I figure this has to be a 1-to-Many relationship since five people can copy the same note.

Correct. It’s also a tree, since there can be multiple levels of copying.

… should I apply a NOT NULL constraint and simply make the default parent 0 with an essentially meaningless record at ID 0 and take note of this in the software?

No need, FK would still be enforced on 0 and you’d need a “dummy” row just to satisfy it, as you already noted. FK ignores NULLs, so just put a NULL in the root note’s parent_ID.

Should I employ a two-table system…

This doesn’t model the same thing. It just allows you to connect to the original user, but not original note.

A different two-table design would be viable if a single note could have multiple parents, but that doesn’t seem to be the case here.

Which eliminates any possible NULL’s by definiton?

You seem very intent on eliminating NULLs. Any reason for that?

---

All in all, you should store notes in just one table, like this:

CREATE TABLE note (
    ID           INT(10) UNSIGNED AUTO_INCREMENT PRIMARY KEY,
    parent_ID    INT(10) UNSIGNED,
    username     VARCHAR(30) NOT NULL,
    information  VARCHAR(256) NOT NULL,
    comment      VARCHAR(256) NOT NULL,
    FOREIGN KEY (parent_ID) REFERENCES note (ID),
    FOREIGN KEY (username) REFERENCES `user` (username)
);