designspace.ConstraintEval

class designspace.ConstraintEval(constraint: Constraint, instance_path: str | None, applicable: bool, satisfied: bool | None, margin: float | None)

Bases: object

One constraint evaluated against one configuration.

Read violated rather than satisfied unless you mean to handle polarity yourself: a forbid names a bad state, so satisfying it is the unhealthy outcome.

constraint

The constraint evaluated. For a per-element constraint inside a .repeat(), the template rather than a per-instance copy.

Type:

Constraint

instance_path

Which element this evaluation is for, when the constraint lives inside a lift; None otherwise.

Type:

str | None

applicable

Whether the constraint could be decided at all. False when a parameter it reads is inactive, in which case it neither holds nor fails, since an inapplicable constraint is never violated.

Type:

bool

satisfied

Whether the stored predicate held. None when the outcome is unknown, which is also when applicable is False.

Type:

bool | None

margin

How far the configuration sits from the constraint boundary, signed so that a positive value means the stored predicate holds. None for a predicate with no numeric distance, such as a Boolean composition or an opaque ds.value(…, returns=bool).

Type:

float | None

property violated: bool

Whether this evaluation counts against feasibility (for a hard forbid/require) or is flagged as a violation (for a soft encourage/discourage). It is polarity-correct across all four kinds. An inapplicable (Kleene-Unknown) eval is never violated (rule 4); otherwise the stored predicate is violated when satisfied differs from the constraint’s desired polarity. This is the public, display-ready reading the reference sampler and validate use internally (eval.is_violated); see Constraint.feasible_when_satisfied.