More complex forms of test

Oh, and there's just one more thing . . .

The decision to take a particular action may sometimes be based on a simple test. You may cut a pack of cards to see who deals. You may toss a coin if you can't decide to stay in or go out. One test decides the outcome. But very often, a decision cannot be reached so easily. If you remember the case of the Spanish, you will know that it is not enough to determine the ending of a verb word to ask if its grammatical staus is first.person or not. You also need to know if it is singular or not and moreover if it is past or not. It is only when all of these tests have proved positive that you can start to feel safe about the form of the word. In fact, even then, you are only half safe, since there are some other features which come into play. But let's labour the point. Let us just agree that sometimes a decision will depend on a number of conditions, all of which must be met.

Or do you want an alternative . . .

Adding one new requirement upon another is not the only way to combine tests. If you want to select the appropriate sounds to add to the end of a (regular) English verb stem in order to derive the past tense form, then you need to check if it the stem ends with a /t/ or with a /d/. If either one or the other is true - and clearly both together would be an impossibility - then you add /id/ as a final syllable. (bat becomes batted = /batid/)

Let's be logical

Logo copes with the piling on of conditions and the provision of alternatives within conditional instructions by the use of the primitive predicates, and and or. Despite appearances, these are indeed procedures. Both normally expect two inputs, which must evaluate to "true or "false, and as predicates themselves they are restricted to outputting "true or "false. The following are, therefore, acceptable even if complex Logo expressions and, since each one reduces in the end to a single "true or "false, either could be used to provide the test (i.e. the first input to if) in a conditional instruction.:

and :first.name="donald :last.name="duck

or :first.name="micky :first.name="minnie

In the first case, and outputs "true if, and only if, both of its inputs individually output "true, i.e. if :first.name="donald outputs "true and also :last.name="duck outputs "true . Otherwise it outputs "false . As for or, it outputs "true if either one or the other or even both of its inputs output "true. If both output "false then it too outputs :false.

If you have had any dealings with logic, you will recognise here not only the names but also the behaviour of two of the connectives of propositional logic. Logicians are accustomed to set out the inferences which it is possible to draw when propositions are combined with these connectives by means of tables called truth tables. The truth table for logical and which conjoins Proposition 1 (P1) and Proposition 2 (P2) would look like this:

The corresponding table for logical or would be:

In fact you can just as easily view these tables as specifying the output from Logo's and and or if you take columns P1 and P2 to refer to the inputs to the procedure and the last column to identify its output. In other words, exactly the same relationships apply. For that reason, and and or are sometimes referred to as logical predicates or Boolean predicates.

How to be negative

The connectives of propositional logic include not only and and or but also not. On the face of it, connective is an odd choice of terminology in this case, since not is constucted with only one proposition rather than two. But if that is hard to swallow, its truth table by way of compensation is inevitably simpler:

Logo, not surprisingly, matches this connective with a primitive predicate called not which behaves in the same way. It expects one input, which, like those of and and or, must evaluate to "true or "false. (Now you can see why it belongs in the same family.) If its input is evaluates to "true then not outputs "false and vice-versa.

The advantage of such a predicate shows up clearly whenever it is necessary to specify something to be done in all circumstances except one, i.e. where the negative conditions are easier to specify than the positive ones. Suppose, to take a trivial example, you are dealing with a language which distinguishes three persons - first (I, we), second (you) and third (he, she, it) - and suppose that first and second person forms of verbs behave similarly in some respect. It is clearly less long-winded to check that thing "person is "first or "second by using not in an expression such as:

not :person="third

than it is to use or:

or :person="first :person="second

And this and that and the other

You may have noticed that the correct choice of ending for a Spanish verb is dependent on more than two tests proving positive. The solution to this potential problem - given that and and or are limitied to two inputs - is essentially the same as the one we used to concatenate more than one object with se in Lonely Hearts. (Look here for a reminder.) Either you use more than one and (with every and except the last receiving one of its inputs indirectly from another and):

and member? :features "1st and member? :features "singular member? :features "past

Or you take advantage of the fact that Logo is happy to accept more (or less) than two inputs to and as long as you signal your intentions clearly with a left parenthesis before the procedure name and a right parenthesis after the last of the inputs:

(and member? :features "1st member? :features "singular member? :features "past)

Whichever method you adopt, the expression will only reduce to "true when every one of the sub-tests outputs "true.

Or, you will be pleased to hear, is treated in exactly the same way so that more than two alternatives can be easily accommodated, as in this instruction:

if (or :month="september :month="april :month="june :month="november) [op 30]

Masculine and dative or plural

Sometimes, of course, you might need to combine and with or and indeed both or either of these with not. For example, to handle some particular aspect of word structure you might need to formulate a conditional instruction corresponding to something like like this:

If the word is masculine and dative or plural then suffix something or other

But notice that this English instruction is dangerously ambiguous. Should a suffix is be added when the word is either on the one hand (a) both masculine and dative or on the other hand (b) simply plural? Or does it mean instead that a suffix is to be added only when the word is (a) of necessity masculine and additionally (b) either dative or plural? And the problem is aggravated with the introduction of not:

If the word is not masculine and dative or plural then suffix something or other

Aggravation and prefixation

It may have bothered you so far to see that Logo's and and or precede both of their inputs. You must say and :late? :tired? or or :late? :tired? instead of using the usual English patterns, late and tired? and late or tired?. In other words, although Logo allows infixing of the equality predicate (=), it requires that and and or as well as not are prefixed. The awkwardness of this convention has its compensations, however. Given the way in which the Logo interpreter reads and :masc? or :dat? :plural?, there is no ambiguity. The first input to and is provided by :masc? and the output of or :dat? :plural? must provide the second. It follows that, in this case, for and itself to output true, :masc? must output "true and or :dat? :plural? must output "true. Something quite different, on the other hand, is expressed by and :masc? or :dat? :plural?

A little discipline and some first aid

From these examples, you can see clearly that when you come to formulate a test you must be very clear about what you intend and you must be careful to use an expression which says just what you intend. But you were warned that Logo expects clear thinking! (If you want a quick test of your understanding click here.)

You might also feel at this point that you could do with some kind of crutch to help you step through the construction of a possibly long and complex test structure. If so, then take advantage of a pair of parentheses. Logo does not mind you using parentheses redundantly and you at least might find that it helps your reading of a line if you say or (:masc?) (and :dat? :plural?) instead of the balder or :masc? and :dat? :plural?. They mean the same thing.



Ron Brasington
Department of Linguistic Science
The University of Reading
Reading
UK

E-mail: ron.brasington@rdg.ac.uk