Result shapes#

parcel_runtime::shape applies suppression and aggregate noise to DataFusion plans. Active rules are selected with active_shapes.

`apply`#

pub fn apply(
    plan: LogicalPlan,
    optimized: &LogicalPlan,
    active: &[&ShapeRule],
) -> parcel_runtime::plan::Result<Shaped>;

Parameter	Meaning
`plan`	Caller plan to rewrite. Consumed by this function.
`optimized`	The same plan after optimization, used to identify columns actually read.
`active`	Shape rules selected for this caller, potentially from multiple contracts.

Returns the rewritten plan, privacy-budget charges, and suppression metadata. Unsupported plan shapes or invalid rewrites return RuntimeError.

Field	Type	Meaning
`plan`	`LogicalPlan`	Rewritten plan to execute.
`charges`	`Vec<Charge>`	Budget charges to apply before execution; the maximum epsilon for each budget.
`suppress_k`	`Option<u64>`	Suppression threshold, if any.
`has_aggregate`	`bool`	Whether the plan contains an aggregate.

Field	Type	Meaning
`budget`	`String`	Name of the privacy budget.
`epsilon`	`f64`	Amount charged to that budget.

apply calculates charges; the application must check and record them before executing the plan.

Signature	Parameters and result
`suppress_ungrouped(batches: Vec<RecordBatch>, k: u64) -> Vec<RecordBatch>`	Treats all result batches as one group; removes the result when its total row count is below `k`. Consumes the batches.
`suppress_groups(plan: LogicalPlan, k: u64) -> plan::Result<LogicalPlan>`	Rewrites an aggregate plan to suppress groups below `k`.
`has_aggregate(plan: &LogicalPlan) -> bool`	Reports whether a plan contains an aggregate.
`columns_read(optimized: &LogicalPlan) -> Vec<String>`	Returns column names read by optimized scans.

After executing Shaped.plan, use suppress_ungrouped when has_aggregate is false and suppress_k is set.