I have a project that I am working on and I can’t figure out which would be a “better” table relationship schema.
The scope of the area in question are:
-
User uploads documents (become the owner/author)
-
User can share document with other users (set share rights)
-
Any user who has access to document can checkout the document (exclusive lock)
My original schema looks like this:
Benefits are:
-
only one user can be author. (authorid)
-
Rights table contain only “sharerights” (Read, Write)
-
User has easy distinction as to which files they “own” (authorid) vs sharedfiles (sharedfiles table). (this one is a weak benefit, I know)
After thinking things through I thought that this might be a better schema:
Benefits are:
-
All document associations are located in one location (UserFiles)
-
Future ability to allow multiple authors/owner of a single document
Rights table would now have Read, Write, and Owner. As soon as a document was uploaded by a user, an automatic association would be made to a new document and the user would be given “owner” rights.
This led me to the final schema:
Benefits are:
- If a user file association is removed (removed share) and that user had a lock on the file (Checked out), then that exclusive lock would be automatically removed.
The only problem with this last model is that I plan to add “special” user for each department so that a user could share a document to an entire department. So I am not sure if I want to associate the share association to the checkoutID (if that makes sense). The query for the users files would look like “select all files where userfiles.userid = me.userid || (userfiles.id == SpecialDepID && me.depid == SpecialDepID)” (major pseudo code)
It has been a long time since I have done database schemas and this one design decision is really racking my brain. It is really bugging me as to which design would be “better” and by better I mean better design principles, better decision based off of previous experience, allows for easier “growth” in the design, etc. Please let me know your thoughts!
FINAL SOLUTION
With the help from Michael Madsen the final solution looks like:
There will be a trigger on UserFiles for delete which will determine if a lock should be deleted when a relationship is removed.
If it were only those three options, I’d go with the second one and place a trigger on deletion from UserFiles to handle the issue you’re trying to handle with the third design.
You’ve already provided good reasons for choosing that one over the first one, so I’m not going to repeat that.
The third design isn’t good, though – it’s not straightforward to see if a file is locked; you have to see if a sharedFileID exists where the fileID matches the one you’re after, meaning multiple records per table. It’s also not nice that you’re missing a primary key on CheckedOutFiles, so that also counts against that one.
However, we can of course fix those problems. If you were to use FileID as the primary key in CheckedOutFiles, you would be able to avoid those two issues – you have a meaningful primary key, and you can easily check if a given file is locked.
Of course, even if you do that, you still have the problem of the “special” user. A simple way you could use to handle that is by storing the actual user as part of your checkout table – the sharedFileID references the department user, while you still have a reference to the actual user to verify that you’re dealing with the right user.
With those changes, the third design seems best – you only reserve space for lock information for the files that are actually locked.
TL;DR: Third design, but with fileID as the PK in CheckedOutFiles, and a specific UserID as part of CheckedOutFiles to handle “meta”-users.