In order to see what the principle we need might look like, let’s consider a simpler example. As you know, normalization as such—in particular, normalization as used in the example of the previous section—has to do with “vertical” decomposition of relvars (meaning decomposition via projection). But “horizontal” decomposition (that is, decomposition via restriction) is clearly possible, too. Consider the design illustrated in Figure 14-2, in which the parts relvar P has been split horizontally—in fact, partitioned—into two relvars, one (“light parts,” LP) containing parts with weight less than 17.0 pounds and the other (“heavy parts,” HP) containing parts with weight greater than or equal to 17.0 pounds. (I assume for definiteness that WEIGHT values represent weights in pounds avoirdupois.)

Figure 14-2. Relvars LP and HP—sample values

The predicates are as follows:

LP:

*Part PNO is named PNAME, has color COLOR and weight WEIGHT (which is less than 17.0), and is stored in city CITY*.HP:

*Part PNO is named PNAME, has color COLOR and weight WEIGHT (which is greater than or equal to 17.0), and is stored in city CITY*.

Note that the original relvar P can be recovered by taking the (disjoint) union of relvars LP and HP.

Why might we want to perform such a horizontal decomposition? Frankly, I’m not aware of any good logical reason for doing so, though of course that’s not to say ...

Start Free Trial

No credit card required