Release 1.28.0

What's New

Features

• compareRoots — Root-AST structural similarity API — A new
  deterministic, model-free API for comparing two RIDDL Root ASTs and
  getting back a similarity score with a full breakdown. Surfaced on the
  cross-platform RiddlLib facade (JVM, JS, and Native):

  def compareRoots(a: Root, b: Root)(using PlatformContext): RootSimilarity
  def similarityMarkdown(a: Root, b: Root)(using PlatformContext): String

  RootSimilarity carries per-kind counts (Map[String, (Int, Int)]),
  fuzzy-matched and unmatched name pairs per definition kind, structural
  metrics (max depth and top-level breadth), and an overall weighted score
  in [0.0, 1.0] — exactly 1.0 for identical models, low for disjoint
  ones. similarityMarkdown renders the same data as a human/AI-readable
  report.

  Matching is on definition kind + fuzzy name (lowercase-alphanumeric
  normalization with greedy best Levenshtein-ratio pairing), making it
  location- and case-independent — a renamed-but-structurally-equal
  model still scores high on structure, and identically named definitions
  at different source locations still match. It never uses
  Definition.equals/hashCode (which include source location). Scoring
  weights DDD-structural kinds (contexts, entities, message types) above
  incidental ones, and blends a count-vector cosine term so that structure
  is rewarded independently of naming.

  This was built for consumers that generate RIDDL models and need a
  regression guard on how closely a generated model matches a
  hand-authored reference — but it is a general-purpose AST diff/compare
  utility usable by any tool.

Notes

• Additive only — new methods on the RiddlLib trait and object; no
  existing signatures changed. Fully backward compatible with the 1.x line.
• The comparison core (RootComparison, RootSimilarity,
  KindComparison) lives in com.ossuminc.riddl.passes.analysis and is
  Native-safe (no regex or platform formatting).
• Verified across JVM, JS, and Native builds.