prize18-5  Exercise 1: Canonical and Compatibility Equivalence
==============================================================

Predictions: e-acute pair: both True (canonical implies compatibility). A-ring and angstrom sign: both True. The ffi ligature and "ffi": canonical False,
compatibility True. Two mark orders that share a combining class (both class 230 here): the order is kept, so both False. Fullwidth A and A: canonical False,
compatibility True. a and A: both False (normalization does not change case).

ex1.py:

  import unicodedata as ud

  def canon_eq(a, b):
      return ud.normalize("NFD", a) == ud.normalize("NFD", b)

  def compat_eq(a, b):
      return ud.normalize("NFKD", a) == ud.normalize("NFKD", b)

  pairs = [
      ("e acute", "\u00e9", "e\u0301"),
      ("angstrom", "\u00c5", "\u212b"),
      ("ffi ligature", "\ufb03", "ffi"),
      ("marks swapped", "e\u0302\u0301", "e\u0301\u0302"),
      ("fullwidth", "\uff21", "A"),
      ("case", "a", "A"),
  ]
  for label, a, b in pairs:
      print(f"{label:14} canonical={canon_eq(a, b)!s:5} compatibility={compat_eq(a, b)}")

Real output (Python 3.14):

  e acute        canonical=True  compatibility=True
  angstrom       canonical=True  compatibility=True
  ffi ligature   canonical=False compatibility=True
  marks swapped  canonical=False compatibility=False
  fullwidth      canonical=False compatibility=True
  case           canonical=False compatibility=False

All six match the predictions. The "marks swapped" row is the subtle one: the circumflex and acute both have combining class 230, so normalization must not reorder them.