eyeprolog

Aliases and namespaces

Which names are built in, and which are just your vocabulary?

Arithmetic, comparison, text matching, and list decomposition are each written one native way. Everything else, including names that look like vocabulary from another system, is an ordinary user predicate with no special meaning.

Run it in the playground · Source

The question

Rule languages often grow many aliases for the same operation, such as math:sum, add/3, or string:matches. EyeProlog keeps one spelling per built-in and lets every other name belong to the program.

This example shows both sides in six short rules.

Native built-ins

value(nativeMath, X) :- (X is 0.125 + 0.875).
ok(nativeCompare, true) :- (2 < 3).
ok(nativeString, true) :- matches('scoped retail insight', 'retail|medical').
tail(nativeList, Tail) :- [_head|Tail] = [a, b, c].

Arithmetic is ISO is/2, comparison is </2.
matches/2 comes from library(strings) and accepts |-separated alternatives.
A list’s tail comes from unification with [Head|Tail]; no rest/2 helper is needed.

Vocabulary is data

example_label(vocabularyExample, "vocabulary names are ordinary predicate names").
label(vocabularyExample, Text) :- example_label(vocabularyExample, Text).

label/2 might look like a reserved word from some vocabulary. Here it is just a predicate defined by the program.

The answer

eyeprolog examples/aliases-and-namespaces.pl

value(nativeMath, 1.0).
ok(nativeCompare, true).
ok(nativeString, true).
tail(nativeList, "bc").
label(vocabularyExample, "vocabulary names are ordinary predicate names").

0.125 + 0.875 is the float 1.0. The tail [b, c] is printed as "bc", because a list of one-character atoms is written in double-quote notation.

The reason

eyeprolog --proof examples/aliases-and-namespaces.pl

step(ok(nativeString, true), rule(3), [], [matches('scoped retail insight', 'retail|medical')]).
step(matches('scoped retail insight', 'retail|medical'), builtin, [], []).
step(tail(nativeList, "bc"), rule(4), ['Tail' = "bc"], ["abc" = "abc"]).

Each conclusion rests on one rule and, below it, a builtin step for the native operation. The text match succeeds because retail occurs in the phrase.

The check

condition('C1', resolution, ok, 6).
condition('C5', re_decision, ok, 4).
recomputed(4).
trusted(0).
verdict(checked).

The 6 rule and fact steps are verified against the program, and the checker recomputes all 4 built-in steps itself.

Try it

Change the pattern to 'medical|legal' and predict whether ok(nativeString, true) is still derived.
Write tail/2 for the list [x]. What is printed for an empty tail?
Define your own sum/3 with is/2 and use it in a rule. Does it behave any differently from a native built-in in the proof?

Learn more in The Art of EyeProlog, Chapter 39.

This site is open source. Improve this page.