TE DAS PES TEA ETIL

We consider splitting entities with optional arrows
and merging those with similar relationships.

indicate solutions that are likely, but not
absolutely certain. They warn us to investi-
gate further.

Consider the following database de-
sign for a pipe-tobacco wholesaler. Cut to-
bacco is stored in warehouses and
distributed in trucks.

The location/blend record tells how
many pounds of one tobacco blend there
are at a specific warehouse. If there are 20
warehouses and 50 blends in all, we could
have up to 1,000 occurrences of this record
(perhaps not every blend will be stored at
each location). Any blend may be shipped
or received many times. Similarly, each
truck may be involved in many movements.
But, by definition, each movement repre-
sents an occasion where a single location's
inventory of a blend was either increment-
ed because tobacco arrived or decremented
because some was sent off. Also, each
movement involves only one vehicle.

User
informatnn

Now there's a data
design methodolo-
g9y for structured
analysis that takes
advantage of the best of
the data-oriented sys-
tems design technigues.

Logical Data Design for Structured s?"
Analysis, a new five-day seminar from
Ken Orr 8 Associates, Inc., will teach
you how to logically develop a fully normalized data
base without the frustrations of classic normalization.

The situation seems straightfor-
ward. The fields in the location/blend rec-
ord include location number, blend
number, inventory balance, and the like.
Truck record holds vehicle number, cargo
capacity, miles-since-maintenance, etc.
And movement contains date, guantity
shipped, and an in-or-out flag. But doesn't
the truck-to-movement arrow mean each
movement must involve a truck?

Think about transferring tobacco
within a location. Many pipe tobaccos are
produced by blending others—that's why
they're called blends. The process is one of
simply shoveling a measured amount from
one hopper to another and, though it cer-
tainly affects the inventory balances, no ve-
hicle is involved. The truck-movement
relationship is then optional in some sense.
Only one truck is involved in any ship-
ments that go by truck, but some shipments
do not involve trucks at all.

The optional arrow is a warning
that the movement box, in an information
modeling sense, could represent two funda-
mentally different entities. Redrawing it as
two boxes, we have:

"The Missing Link"

System Processing
Model

4 a Logical
t Logical  ", Dala Model
Dala

-—..

Working from your system data flows, you will use steps from Ken Orr's
Data Structured Systems Development (DSSD") methodology to build
a logical data model that can easily be translated to any physical DBMS
environment. DSSD logical data design makes explicit the implicit link
between structured analysis and detailed design.

L ogical Data Design for Structured Analysis — The Missing Link.

Jan 13-17 Atlanta Aug. 23-29 Chicago
Mar. 24-28 Portland Nov, 10-14 Kansas City
May 5-9 Washington

Call tolktree for more informalion: 800/255-2459

1725 Gage Blvd.,
Topeka, Ks. 66604

Ken Orr 8% Associates, Inc.

CIRACLE 64 ON READER CARD

430 DATAMATION

Shipment and transfer tell about dif-
ferent real-world events. As we refine the
design further, their record layouts con-
tinue to diverge until, eventually, we find
the only data common to both are date and
guantity.

Our first indication of their duality
was that optional arrow.

Now look at the arrow between lo-
cation/blend and transfer. We can't say
each transfer is associated with only one lo-
cation/blend because the transfer event af-
fects two different balances. It decrements
the inventory balances and increments the
location/blend balance it goes to.

The diagram does not mean that
each transfer is related twice to the same
location/blend, by the way. On the con-
trary, each transfer has two different rela-
tionships with two different location/blend
record occurrences—a "from" relationship
with one and a "to" with the other.

Merging records with similar rela-
tionships is basically the reverse of the pro-
cess just described. For example:

Both intersection records, inbound
shipment and outbound shipment, hold the
same data elements and are related to ev-
erything around them in precisely the same
way. This situation warns us to look at
them more closely and see if they're not
really modeling the same class of real-
world event.

Next time, we'li look at hard sets,
soft sets, and summary fields. e

Frank Sweet is corporate manager of
data administation for the Charter Co.,
Jacksonville, Fla.

Z oma,