"""
Input normalization, run BEFORE matching. Defuses the unicode/homoglyph/whitespace
evasion class both reviewers flagged: full-width text, soft hyphens, zero-width
chars, Cyrillic/Greek look-alike letters, diacritics, trailing punctuation/quotes.
"""

import re
import unicodedata

# Cyrillic / Greek letters that look like Latin -> fold to Latin (homoglyph attack)
_CONFUSABLES = {
    # Cyrillic
    "а": "a", "е": "e", "о": "o", "р": "p", "с": "c",
    "х": "x", "у": "y", "і": "i", "ѕ": "s", "к": "k",
    "м": "m", "н": "h", "т": "t", "в": "b",
    # Greek
    "α": "a", "ο": "o", "ρ": "p", "ε": "e", "υ": "u",
    "κ": "k", "ν": "v", "τ": "t", "χ": "x",
}
_ZERO_WIDTH = {"​", "‌", "‍", "", "­", "⁠", "‎", "‏"}
_STRIP_SYMBOLS = {"®", "™", "©", "℠", "°"}


def normalize_input(text: str) -> str:
    if not text:
        return ""
    # 1. NFKC: fold full-width (ＴＹＬＥＮＯＬ -> TYLENOL), ligatures, etc.
    t = unicodedata.normalize("NFKC", text)
    # 2. map confusable scripts to Latin
    t = "".join(_CONFUSABLES.get(ch, ch) for ch in t)
    # 3. drop zero-width / soft-hyphen / control chars (incl. null byte); keep spaces
    t = "".join(
        ch for ch in t
        if ch not in _ZERO_WIDTH and ch not in _STRIP_SYMBOLS
        and (ch == " " or unicodedata.category(ch)[0] != "C")
    )
    # 4. strip diacritics (Depakote <- Dépakote)
    t = "".join(c for c in unicodedata.normalize("NFKD", t) if not unicodedata.combining(c))
    # 5. collapse whitespace, strip surrounding quotes + trailing punctuation
    t = re.sub(r"\s+", " ", t).strip()
    t = t.strip("'\"“”‘’`")
    t = re.sub(r"[?!.,;:]+$", "", t).strip()
    return t


if __name__ == "__main__":
    tests = ["ＴＹＬＥＮＯＬ", "ti­zanidine",
             "хanax", "аspirin", "lipitor®", "Dépakote",
             "  metformin  ", "'lipitor'", "metformin?", "Omeprazole\x00"]
    for t in tests:
        print(f"{t!r:30s} -> {normalize_input(t)!r}")