06 — Rule Language¶

Rules are stored as readable text and compiled into internal indices before evaluation.

Grammar¶

rule       := "IF" antecedent "THEN" clause [ "WITH" number ]
antecedent := clause { ("AND" | "OR") clause }
clause     := identifier "IS" [ "NOT" ] identifier
identifier := letter { letter | digit | "_" }

Examples¶

IF Temperature IS Cold THEN Fan IS Slow
IF Temperature IS Hot AND Humidity IS High THEN Fan IS Fast
IF Battery IS NOT Low OR Distance IS Near THEN Aggression IS Bold
IF Speed IS Extreme THEN Brake IS Hard WITH 0.8

Names and Keywords¶

IF, IS, NOT, AND, OR, THEN, WITH are case-insensitive. Spaces, tabs, and line breaks are free-form.

Variable and set names must exist in the system. They are stored as FName, so they keep the case you typed for display but are matched without regard to case: IF speed IS fast resolves against a system that declares Speed and Fast. For the same reason, two sets named Low and low on one variable are reported as a duplicate.

The seven keywords above are reserved and cannot be used as a variable or set name. A system that declares a variable called Or compiles without complaint, but every rule mentioning it fails to parse with a message about the rule — rename the variable.

letter in the grammar above resolves through the platform's character classification, so non-ASCII names (Cyrillic, accented Latin, CJK) may be accepted on one machine and rejected on another. Keep identifiers to ASCII letters, digits, and _; use the set name only as a key and put the localized wording in your UI.

Premises refer only to inputs. The consequence refers only to an output. An unknown name, using an output in the premise, or using an input in the consequence produces a diagnostic error.

AND¶

AND combines degrees via the chosen T-norm. With the default Minimum, the rule's strength equals its weakest condition:

Distance.Near = 0.7
Integrity.High = 0.4
AND / Minimum = 0.4

Product multiplies the degrees: 0.7 × 0.4 = 0.28.

OR¶

OR uses an S-norm. Maximum takes the strongest condition. Probabilistic Sum computes a + b - ab and gives mutual reinforcement without exceeding 1.

NOT¶

NOT complements a specific statement:

IF Health IS NOT Low THEN Speed IS Fast

If Health.Low = 0.25, then Health NOT Low = 0.75. NOT is also allowed in the consequence, though a separate, positively named output set is usually more readable.

WITH¶

WITH sets a rule weight from 0 to 1:

IF Visibility IS Poor THEN Caution IS High WITH 0.65

The weight multiplies the premise's strength. An absent weight equals 1.

Single-Connective Constraint¶

A single rule uses either AND or OR throughout its premise. A mixed expression is rejected so that hidden operator precedence doesn't silently change intent:

# Invalid
IF A IS High AND B IS Low OR C IS Near THEN Result IS Strong

Split it into two explicit rules, or introduce a separate variable if more complex logic is needed.

Tips¶

One rule should express one clear idea.
Use weights for fine-tuning after the set shapes are already correct.
Don't duplicate fully identical rules: under summing aggregation they can reinforce each other.
The index in RuleActivations corresponds to the rule's position in the authored array, even when the rule failed to compile.

The mathematical behavior of the operators is explained in 07 — Inference And Settings.