"""
Verification for the review comments in docs/uce8-article.md.

Run from the repo root:   python3 docs/verify.py
"""

import gzip
import os
import sys

sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))

from uce import MAX_CODE_POINT, TAIL_COUNT, decode, encode


def uce(s):
    return b"".join(encode(ord(c)) for c in s)


def blen(c):
    return len(encode(ord(c)))


# --- 1. round-trip every Unicode code point ------------------------------
bad = 0
for cp in range(MAX_CODE_POINT + 1):
    if 0xD800 <= cp <= 0xDFFF:      # surrogates are not encodable characters
        continue
    if decode(encode(cp)) != cp:
        bad += 1

print("round-trip failures:", bad)

# --- 2. capacity arithmetic ----------------------------------------------
print("tier2 capacity:", 128 * TAIL_COUNT)
print("tier3 capacity:", 128 * 128 * TAIL_COUNT)
print("unicode total :", MAX_CODE_POINT + 1)
print()

# --- 3. raw vs gzipped saving on real sentences --------------------------
samples = {
    "Chinese (simp)": "我们每天使用的文本中，字符以一到四个字节的长度存储。",
    "Chinese (trad)": "我們每天使用的文本中，字符以一到四個位元組的長度儲存。",
    "Japanese": "私たちが毎日使うテキストでは、文字は1〜4バイトの長さで保存されます。",
    "Korean": "우리가 매일 사용하는 텍스트에서 문자는 1~4바이트 길이로 저장됩니다.",
    "Hindi": "हम हर दिन जिस पाठ का उपयोग करते हैं उसमें वर्ण संग्रहीत होते हैं।",
    "Thai": "ในข้อความที่เราใช้ทุกวัน อักขระจะถูกจัดเก็บด้วยความยาว",
    "Amharic": "በየቀኑ በምንጠቀምበት ጽሑፍ ውስጥ ቁምፊዎች ይቀመጣሉ።",
    "Mongolian(Cyr)": "Бидний өдөр тутмын хэрэглэдэг текст дотор 1-4 байт урттайгаар хадгалагддаг",
    "Mongolian(trad)": "ᠮᠣᠩᠭᠣᠯ ᠪᠢᠴᠢᠭ᠌",
    "Vietnamese": "Trong văn bản chúng ta sử dụng hàng ngày, các ký tự được lưu trữ với độ dài từ 1 đến 4 byte.",
    "JSON w/ Chinese": '{"name":"张三","city":"北京","note":"你好世界"}',
}

for name, text in samples.items():
    u8, u = text.encode("utf-8"), uce(text)
    g8, gu = len(gzip.compress(u8, 9)), len(gzip.compress(u, 9))
    print(
        f"{name:<17} utf8={len(u8):>4} uce8={len(u):>4} "
        f"raw={(1 - len(u) / len(u8)) * 100:>5.1f}%  "
        f"gzip={(1 - gu / g8) * 100:>5.1f}%"
    )
print()

# --- 4. coverage of the CJK/Hangul blocks --------------------------------
hangul = [chr(c) for c in range(0xAC00, 0xD7A4)]
han = [chr(c) for c in range(0x4E00, 0xA000)]
print("Hangul syllables at 2 bytes:", sum(blen(c) == 2 for c in hangul), "/", len(hangul))
print("CJK Unified   at 2 bytes:", sum(blen(c) == 2 for c in han), "/", len(han))

# --- 5. trail bytes are ASCII digits/letters -> naive-search collisions ---
doc = uce("नमस्ते") + b" ISO 8601"
print("literal 'I' bytes in doc:", doc.count(b"I"), "(only 1 is a real letter)")