Compiled rule types

These public structs are in parcel_core::compile and are returned within Compilation. They have no public constructors or methods of their own.

CelRule

Field	Type	Meaning
id	String	Rule ID.
cel	String	Expression evaluated by the CEL interpreter.

GuaranteeRule

Field	Type	Meaning
id	String	Rule ID.
cel	String	Dataset requirement expression.
on_fail	GuaranteeOnFail	Action when false: Deny or Annotate.
reads	Vec<String>	Statistics-column names read by the expression.

ShapeRule

Field	Type	Meaning
id	String	Rule ID.
shape	ShapeOp	Typed shape operation.
unless	Option<String>	Optional caller condition that skips this shape.

Flag

Field	Type	Meaning
assert_id	String	Assertion ID.
column	String	Stored flag column name.
expr	Expr	DataFusion assertion expression.
on_fail	AssertOnFail	Action when false: Drop, Deny, or Report.

RowRule

Field	Type	Meaning
id	String	Rule ID.
kind	RowRuleKind	Admit, Assert, or Transform { column: String }.
cel	String	Reference CEL expression.
reads	Vec<String>	Columns read by value; nulls affect rule evaluation.
ty	Type	Expression return type.

Derived

Field	Type	Meaning
column	String	Stored calculation column name.
cel	String	Canonical CEL calculation.
expr	Expr	DataFusion calculation.
ty	Type	Result type.

EnrichSpec

Field	Type	Meaning
field	String	Custom row field name.
cel	String	CEL expression over raw columns.
reads	Vec<String>	Raw columns read.
expr	Expr	DataFusion calculation.
ty	Type	Result type.

StatSpec

Field	Type	Meaning
field	DatasetField	Typed dataset-field identifier.
column	String	Output name in the verdict row.
path	String	CEL field path, such as dataset.amount.min.
ty	Type	Statistic type.

ShapeOp and NoiseAt

ShapeOp is defined in parcel_core::check:

pub enum ShapeOp {
    Noise { column: String, sensitivity: f64, epsilon: f64, budget: String, at: NoiseAt },
    Suppress { k: u64 },
    Sample { fraction: f64, key: String },
}
pub enum NoiseAt { Aggregate, Row }

column identifies the noised output, sensitivity and epsilon control noise, and budget identifies the charged budget. k is the minimum group size. fraction is the sampling proportion and key determines the stable sample.