Since the entity integrity rule ensures that a relational table cannot have
any duplicate rows, does that imply that each row in a table maps to a
distinct entity?

Mr. Scott wrote:

I am unfamiliar with an entity integrity rule. It sounds like some shit
somone just made up to market ER diagrams.

http://en.wikipedia.org/wiki/Codd's_12_rules

The guaranteed access rule is sometimes called the logical identity rule.

"Mr. Scott" <do_not_reply (AT) noone (DOT) com> wrote

It's more reasonable to say that a tuple can be understood to represent some
fact which is normally assumed to be true. Entity = Thing. I don't see any
special reason why a tuple MUST map to a "thing" (whatever that means) but
if you find that interpretation useful then use it.

The concept of "Entity Integrity" is not much more than a restatement of
what a relation is anyway.

--
David Portas

"Bob Badour" <bbadour (AT) pei (DOT) sympatico.ca> wrote

I learned the entity integrity rule back in 1984, at the same time I learned
the referential integrity rule.
It didn't seem like shit then, and it doesn't seem like shit now.

The entity integrity rule just says a table's key data may not be null, and
must be unique. In SQL, declaring a primary key is sufficient to enforce
both the uniqueness rule and the not null rule. This rule was presented
particularly for tables that represent entities. The concept is that each
entity must have a unique key. Otherwise you can't keep entities distinct.

"David Portas" <REMOVE_BEFORE_REPLYING_dportas (AT) acm (DOT) org> wrote

subject matter entities. In this context, the entity integrity rule just
says that every instance of an entity can be identified by its key data.

Some peopl, especially the regulars in c.d.t. don't use the concept of
subject matter entities in their discussion of relational data. So be it.

"Bob Badour" <bbadour (AT) pei (DOT) sympatico.ca> wrote


I thought Codd referred to the the entity integrity and referential
integrity rules as the insert-update-delete rules of the relational model.

The entity integrity rule requires that no primary key value or part of a
primary key value can be missing, which ensures that there can be no
duplicate rows in a relational table.

Does that mean that entity integrity implies entity identity? Is logical
identity the same thing as entity identity?

Mr. Scott wrote:

Can you cite a reference for that?


That sounds more like the guaranteed access rule, which is sometimes
referred to as logical identity.


I am unaware of anything called an entity in either the algebra or the
calculus. The guaranteed access rule is sometimes referred to as logical
identity.

http://en.wikipedia.org/wiki/Identity_of_indiscernibles

"Mr. Scott" <do_not_reply (AT) noone (DOT) com> wrote


Each row in a table represents either an instance of an entity or an
instance of a relationship between or among entities. I'm not sure whether
or not "represents" in my wording is equivalent to "maps to" in your
wording.

Also the way I learned it, entities and relationship among them are part of
what was called the "conceptual data model". The conceptual data model was
not a relational model as such, although it's very straight forward to start
with an ER model and transform it into a relational model that expresses the
same facts.

The conceptual model was used for data analysis resulting in a clearer
definition of the information requirements. The conceptual model was NOT a
design model.

The design model could be broken into two layers: logical model and
physical model. I could go into more detail here, but it doesn't pertain to
your question. Suffice it to say the logical model should be relational if
the goal is to design a relational database. Back in 1984, they were quite
loose about what was considered a relational DBMS. It was that looseness
that led Codd to formulate the 12 rules, in order to distinguish between soi
disant relational DBMSes and really relational DBMSes.

The entity integrity rule is generally presented as a subrule under Codd's
rule 10. I do not know if Codd used the term "entity integrity" or not.
And I also do not know if he explicitly made the rule that no part of a
primary key can be missing. I started with practice rather than theory, and
only learned enough of the theory to help me with the practice.

http://en.wikipedia.org/wiki/Entity_integrity
http://en.wikipedia.org/wiki/Integrity_constraints
http://en.wikipedia.org/wiki/Codd%27s_rules

(sorry this is reverse order).

"Bob Badour" <bbadour (AT) pei (DOT) sympatico.ca> wrote

The "RM/T" paper, Extending the Database Relational Model to Capture More
Meaning, ACM TODS, Vol. 4, No. 4, December 1979:

"Rule 1 (entity integrity): No primary key value of a base relation is
allowed to be null or to have a null component."

It therefore originates from the point at which Codd decided to "extend" the
model with E-relations, nulls and other exotic things. The rule begs more
questions than it answers (what about derived relations and relations with
multiple keys?). Myself and likeminded people are happy with the concept of
a pre-RM/T relation that needs no such rules.

--
David Portas

David Portas wrote:

Ah, yes. Codd introduced a lot of things in RM/T and RM/V2 that other
relational proponents find questionable.