v0.4.0 Beta — Initial Public Beta

Added

• Unified exception hierarchy — all library errors now inherit from ConformlyError, providing consistent interface for error handling and logging.
• Specialized exception classes:
  • SchemaError — invalid or malformed schema definitions.
  • ResolutionError — failures during schema reference resolution or type mapping.
  • PlanningError — invalid task configuration or impossible violation combinations.
  • GenerationError — runtime failures during synthetic data generation.
• Structured error context — every exception includes:
  • message: str — human-readable description.
  • context: dict[str, Any] — machine-readable diagnostic data (e.g., code, field, constraint, strategy).
• Collection-level constraints support:
  • MinItems(n) — minimum number of elements
  • MaxItems(n) — maximum number of elements
  • UniqueItems(bool) — enforce uniqueness of elements (for dict will ingnore)
• Path DSL for explicit field targeting: path("field.name").violate(V.TOO_SHORT) explicit violation targeting via DSL
• Violation type alias: V — shorthand for ViolationType enum used in DSL expressions
• Basic dict support — Added generation for dict[K, V], where K must be str or Enum, and V accepts Any.

Changed

• Public API error handling — functions case(), cases(), and low-level pipeline stages now raise GenerationError or PlanningError instead of built-in exceptions. This enables precise except clauses and structured logging.
• set[T] and frozenset[T] types are now normalized to list-based generation with UniqueItems(True) semantics, ensuring deterministic output format and consistent constraint handling across collection types
• Parser module refactored:
  • Unified IR building logic under parser/core.
  • Introduced a centralized type analysis pipeline.
  • Added ElementSpec for standardized type representation.