Errata

On p. 91 of Date's book (the last four sentences on the page), he says

"the semidifference between r and s is the difference between r and r MATCHING s; that is, r SEMIMINUS s is equivalent to r MINUS (r MATCHING s)."

He then shows the SQL equivalent of S SEMIMINUS SP as

SELECT S.*
FROM S
EXCEPT
SELECT S.*
FROM S,SP
WHERE S.SNO=SP.PNO;

But on page 89 (at the top) he shows S SEMIJOIN SP to be equivalent to this SQL

SELECT DISTINCT S.*
FROM S,SP
WHERE S.SNO=SP.PNO;

The DISTINCT keyword is need here; otherwise, the result is not be a valid relation. Without it there would be duplicate tuples. So why did Date omit DISTINCT from the SQL equivalent of S SEMIMINUS SP, at the bottom of p. 91? Shouldn't the correct SQL analogue of S SEMIMUS SP be

SELECT S.*
FROM S
EXCEPT
SELECT DISTINCT S.*
FROM S,SP
WHERE S.SNO=SP.PNO;

even though the result is the same without DISTINCT?

regards,
Kurt Krueckeberg

First, project distributes over intersect and union, though not difference.
MUST BE:
First, project distributes over union, though not intersect and difference.

On p.146 of Database in Depth, the last sentence of the 4th reads: "The only nontrivial join dependency satisfied by this relvar is the fuctional dependency*:

{ SNO, PNO, JNO } -> { QTY }

which is an arrow out of a superkey."

This should read: "The only nontrivial functional dependency satisfied by this relvar is the fuctional dependency:

{ SNO, PNO, JNO } -> { QTY }"

which is an "arrow out of a superkey."

-----------
It should read this way because the functional dependency of { SNO, PNO, JNO } -> { QTY } implies that the only JD is the trivial JD consisting of all the attributes of the reval: * { {SNO, PNO, JNO, QTY } }.

The asterick at the bottom of page 146 makes this point. It reads: "Equivalently, using the name SPJQ to refer to the set of all attributes of SPJ, the revlar satisfies the trivial JD * {SPJQ}.(Recall from chapter 5 that the join of a single relations r is just r itself.)"