It is difficult to find a clean book on Database Relational model.
Typical books on database are of two sorts:
 - Very huge theoretical model books
 
 - Very plain vanilla SQL manual for a specific vendor
 
C.J.Date books doesn't fall in ether of these categories: let's see why!
Database in Depth
C.J.Date is a veteran about this topic: he worked with
 E.F.Codd from almost the beginning of the relational era.
E.F. Codd introduced the E-R theory in the late 1968.
So Date's point of view is valuable.
The Book is clean, and at the end of every chapter you find a summary 
and a good number of exercises:  this book was thought for self-studying 
people too.
"Database in depth" is focused on IT professionals which already knows 
the field, but doesn't have a strong theoretical background.
Date guides us long the E-R model, stressing a lot about the difference 
between SQL and E-R model.
I haven't follow a course on Database at the University (too annoying me 
:), so I was surprised by the big difference between the theoretical model 
(ER) and the most known  (or only one known?) SQL implementation.
After some definition, very carefully explained, the Chapter 3 takes us 
two very interesting issues: "Why duplicate tuples are Prohibited" and 
"Why Nulls are Prohibited".
In The Author's opinion, the SQL implementation lacks on these two 
fronts, a regrettable thing, because these two principles can lead us to 
a polite and terse implementation model.
I read these paragraph with a lot of surprise in my heart, because I 
didn't even think these two "limitations" of SQL-implementation can 
carry us so much problems.
Also, in these paragraphs Date express a very brave idea: always use 
Select DISTINCT in all your queries, to avoid a small subset of the 
weird things caused from repeated tuples.
These statements are well analyzed, and in my own opinion, they bring a 
new light on the E-R model.
The only mistake I can move to Date is the excessive stress about 
distinction of theory and model, to isolate the latter from  reader 
misunderstanding.
But this is not a big issue.
So this book can give us a new vision on the old E-R theory.
Date can teach us old and new theory concepts in a new way.
A true clean exposition makes this book a must for every IT Technical 
Guru or Geek.
I cannot avoid to suggest you this book, even if DataBase aren't your 
primary focus: because you can learn a lot from its pages!
Global Rating: 8 of 10, a strong buy!