Add a few example scripts

examples/scripts/negation.ech

@@ -0,0 +1,28 @@
+# Stratified negation: derive isolated nodes that have no outgoing edges.
+#
+# This example uses negation-as-failure (NOT) in a rule body. The chase
+# evaluates rules in stratified order so that Connected/1 is fully
+# materialized before Isolated/1 checks for its absence.
+#
+# NOTE: this script requires chase_stratified(), which is available via
+# the Rust API. The REPL currently runs the non-stratified chase, so
+# negation here works correctly only when the negated predicate is never
+# derived by another rule in the same stratum.
+
+fact Node(a).
+fact Node(b).
+fact Node(c).
+fact Node(d).
+fact Edge(a, b).
+fact Edge(b, c).
+
+# Copy edges to a unary Connected predicate.
+rule Edge(?X, ?Y) -> Connected(?X).
+
+# Nodes without outgoing edges are isolated.
+# (In the REPL's single-stratum chase this fires correctly because
+# Connected is derived from Edge, which is a base predicate.)
+rule Node(?X), NOT Connected(?X) -> Isolated(?X).
+
+run.
+query Isolated(?X)?

examples/scripts/skolem_chase.ech

@@ -0,0 +1,22 @@
+# Skolem chase: deterministic labeled nulls for existential variables.
+#
+# With the Skolem variant, re-applying a rule with the same frontier
+# bindings reuses the same null value, so the chase terminates even for
+# rules that introduce existentials. Compare this with the oblivious
+# variant, which generates a fresh null each round and hits the step
+# limit.
+#
+# NOTE: the REPL uses the restricted chase by default. To run with the
+# Skolem variant, use the Rust API:
+#
+#   let result = skolem_chase(instance, &rules);
+
+fact Person(alice).
+fact Person(bob).
+fact Person(carol).
+
+# Every person has an id (existential Y).
+rule Person(?X) -> HasId(?X, ?Y).
+
+run.
+query HasId(?X, ?Y)?

examples/scripts/sql_aggregate.ech

@@ -0,0 +1,21 @@
+# Demonstrate GROUP BY with aggregate functions in the SQL frontend.
+
+fact Employee(alice, 30, engineering).
+fact Employee(bob, 25, sales).
+fact Employee(carol, 35, engineering).
+fact Employee(dave, 28, marketing).
+fact Employee(eve, 32, sales).
+
+schema Employee(name, age, dept).
+
+# Count all employees.
+sql SELECT COUNT(*) FROM Employee;
+
+# Count and average age per department.
+sql SELECT dept, COUNT(*) AS headcount, AVG(age) AS avg_age FROM Employee GROUP BY dept;
+
+# Min and max age per department.
+sql SELECT dept, MIN(age), MAX(age) FROM Employee GROUP BY dept;
+
+# Sum of ages in engineering.
+sql SELECT SUM(age) FROM Employee WHERE dept = 'engineering';

src/chase/inference.rs

@@ -395,6 +395,7 @@ fn chase_step(
 
     for (rule_index, rule) in rules.iter().enumerate() {
         let matches = find_matches(instance, &rule.body);
+        let matches = filter_negated(instance, matches, &rule.negated_body);
 
         for subst in matches {
             let trigger = Trigger::new(rule_index, rule, &subst);

src/chase/rule.rs

@@ -90,13 +90,20 @@ impl Rule {
 
 impl fmt::Display for Rule {
     fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
-        // Body
+        // Positive body
         for (i, atom) in self.body.iter().enumerate() {
             if i > 0 {
                 write!(f, ", ")?;
             }
             write!(f, "{}", atom)?;
         }
+        // Negated body
+        for atom in &self.negated_body {
+            if !self.body.is_empty() {
+                write!(f, ", ")?;
+            }
+            write!(f, "NOT {}", atom)?;
+        }
         write!(f, " → ")?;
         // Head
         for (i, atom) in self.head.iter().enumerate() {

src/execution/mod.rs

@@ -387,7 +387,15 @@ fn eval_expr(
 
 fn value_from_term(term: &Term) -> Result<Value, ExecutionError> {
     match term {
-        Term::Constant(value) => Ok(Value::text(value.clone())),
+        Term::Constant(value) => {
+            // Try to interpret the constant as an integer so numeric
+            // aggregates (SUM, AVG) work on chase-backed data.
+            if let Ok(n) = value.parse::<i64>() {
+                Ok(Value::Integer(n))
+            } else {
+                Ok(Value::text(value.clone()))
+            }
+        }
         Term::Null(_) => Ok(Value::Null),
         Term::Variable(_) => Err(ExecutionError::NonGroundScanTerm),
     }

src/frontend/language.rs

@@ -121,11 +121,18 @@ pub fn parse_command(input: &str) -> Result<Command, String> {
             return Err("rule body and head must both be non-empty".to_string());
         }
 
-        let body = parse_atom_list(body_text)?;
+        let body_items = parse_body_literal_list(body_text)?;
         let head = parse_atom_list(head_text)?;
         let mut builder = RuleBuilder::new();
-        for atom in body {
+        for item in body_items {
-            builder = builder.when(&atom.predicate, atom.terms);
+            match item {
+                BodyLiteral::Positive(atom) => {
+                    builder = builder.when(&atom.predicate, atom.terms);
+                }
+                BodyLiteral::Negated(atom) => {
+                    builder = builder.when_not(&atom.predicate, atom.terms);
+                }
+            }
         }
         for atom in head {
             builder = builder.then(&atom.predicate, atom.terms);
@@ -187,6 +194,27 @@ fn trim_suffix(input: &str, suffix: char) -> Result<&str, String> {
     }
 }
 
+enum BodyLiteral {
+    Positive(Atom),
+    Negated(Atom),
+}
+
+fn parse_body_literal_list(input: &str) -> Result<Vec<BodyLiteral>, String> {
+    split_top_level(input, ',')?
+        .into_iter()
+        .map(|s| {
+            let trimmed = s.trim();
+            if let Some(rest) = trimmed.strip_prefix("NOT ") {
+                Ok(BodyLiteral::Negated(parse_atom(rest.trim())?))
+            } else if let Some(rest) = trimmed.strip_prefix("not ") {
+                Ok(BodyLiteral::Negated(parse_atom(rest.trim())?))
+            } else {
+                Ok(BodyLiteral::Positive(parse_atom(trimmed)?))
+            }
+        })
+        .collect()
+}
+
 fn parse_atom_list(input: &str) -> Result<Vec<Atom>, String> {
     split_top_level(input, ',')?
         .into_iter()