app.py

import gradio as gr
from openai import OpenAI
from datetime import datetime, timedelta
import pandas as pd
import os, random, tempfile, json
import re
import threading
import uuid
import csv

# --- API設定 ---
API_KEY = os.getenv("API_KEY")
BASE_URL = "https://openrouter.ai/api/v1"
LOG_FILE = os.getenv("LOG_FILE", "logs.csv")

# ✅ 管理者DL用パスワード（環境変数で設定）
ADMIN_PASSWORD = os.getenv("ADMIN_PASSWORD", "0421")

client = OpenAI(base_url=BASE_URL, api_key=API_KEY)

# --- 外部CSVとして管理する passage データ ---
# passage.csv: passage_id,level,text,question,choices,answer の順
passages_df = pd.read_csv("passage.csv")

levels = [1, 2, 3, 4, 5]

# ✅ ログの同時書き込みガード（ログは全ユーザーで共有）
_log_lock = threading.Lock()

# =========================
# ✅ ログ列（指定順）
# =========================
LOG_COLUMNS = [
    "user_id",
    "question_number",
    "reading_level",
    "passage_id",
    "question",
    "choice",
    "correct",
    "time",
    "actions",
]

# =========================
# ✅ 文字化け対策（方法②：ASCII 正規化）
# =========================
def normalize_text_for_csv(text: str) -> str:
    if not text:
        return text
    return (
        text.replace("—", "-")
            .replace("–", "-")
            .replace("’", "'")
            .replace("‘", "'")
            .replace("“", '"')
            .replace("”", '"')
            .replace("\u00a0", " ")   # NBSP 対策
    )

# --- セッション state 初期化 ---
def _new_session_state():
    return {
        "session_id": str(uuid.uuid4()),
        "used_passages": set(),
        "question_count": 0,  # ✅ 採点済み問題数
        "user_id": None,
        "action_log": [],     # ✅ 1問中の選択変更履歴（submit時にまとめて保存）
        # ✅ CSVから読んだ正解（A/B/C/D）を保持
        "current_answer": None,
    }

# --- ログ追記（CSVのみ・固定カラムで統一） ---
def log_to_csv(entry: dict):
    # ✅ pandas -> csv.writer に置換（列順は LOG_COLUMNS で保証）
    with _log_lock:
        file_exists = os.path.exists(LOG_FILE)

        with open(LOG_FILE, mode="a", encoding="utf-8", newline="") as f:
            writer = csv.DictWriter(f, fieldnames=LOG_COLUMNS)
            if not file_exists:
                writer.writeheader()

            row = {col: entry.get(col, None) for col in LOG_COLUMNS}
            writer.writerow(row)

def log_event(state: dict, **kwargs):
    # ✅ Excelで#####になりにくいよう「JST」を付けてテキスト化
    now = (datetime.utcnow() + timedelta(hours=9)).strftime("%Y-%m-%d %H:%M:%S") + " JST"

    # ✅ 文字化け対策（CSVに書く直前で正規化）
    if "question" in kwargs and isinstance(kwargs["question"], str):
        kwargs["question"] = normalize_text_for_csv(kwargs["question"])
    if "actions" in kwargs and isinstance(kwargs["actions"], str):
        kwargs["actions"] = normalize_text_for_csv(kwargs["actions"])

    if "choice" in kwargs and isinstance(kwargs["choice"], str):
        kwargs["choice"] = normalize_text_for_csv(kwargs["choice"])

    entry = {"time": now, "user_id": state.get("user_id"), **kwargs}
    log_to_csv(entry)

# =========================
# ✅ choices列を柔軟にパース
#   - JSON配列: ["...", "...", "...", "..."]
#   - 改行区切り / | / || / ; などにも対応（できるだけ）
# =========================
def _parse_choices(raw):
    if raw is None:
        return []
    s = str(raw).strip()
    if s == "":
        return []
    # JSONっぽい場合
    if (s.startswith("[") and s.endswith("]")) or (s.startswith("{") and s.endswith("}")):
        try:
            obj = json.loads(s)
            if isinstance(obj, list):
                return [str(x).strip() for x in obj]
        except Exception:
            pass

    # それ以外は区切り推定
    # 優先: 改行 -> || -> | -> ; -> ,
    if "\n" in s:
        parts = [p.strip() for p in s.split("\n") if p.strip()]
        return parts
    if "||" in s:
        parts = [p.strip() for p in s.split("||") if p.strip()]
        return parts
    if "|" in s:
        parts = [p.strip() for p in s.split("|") if p.strip()]
        return parts
    if ";" in s:
        parts = [p.strip() for p in s.split(";") if p.strip()]
        return parts
    if "," in s:
        parts = [p.strip() for p in s.split(",") if p.strip()]
        return parts

    return [s]

def _format_question_block(question, choices_list):
    # choices を A-D に整形して question_display に入れる
    labels = ["A", "B", "C", "D"]
    # 4つ未満/超過でも崩れにくくする
    choices = list(choices_list)[:4]
    while len(choices) < 4:
        choices.append("")

    lines = []
    q = "" if question is None else str(question).strip()
    lines.append(f"Q: {q}")
    for lab, opt in zip(labels, choices):
        opt_s = "" if opt is None else str(opt).strip()
        lines.append(f"{lab}. {opt_s}")
    return "\n".join(lines)

# --- 自動難易度調整 ---
def adaptive_test(prev_level, prev_correct):
    idx = levels.index(prev_level)
    if prev_correct and idx < len(levels) - 1:
        return levels[idx + 1]
    elif not prev_correct and idx > 0:
        return levels[idx - 1]
    return prev_level

# --- passage取得（✅ CSVから question/choices/answer も返す） ---
def get_passage(level, used_passages):
    subset = passages_df[passages_df["level"] == level]
    available = [pid for pid in subset["passage_id"] if pid not in used_passages]
    if not available:
        available = list(subset["passage_id"])

    passage_id = random.choice(available)
    row = subset[subset["passage_id"] == passage_id].iloc[0]

    text = row["text"]
    question = row["question"]
    choices_raw = row["choices"]
    answer = row["answer"]

    choices_list = _parse_choices(choices_raw)
    question_block = _format_question_block(question, choices_list)

    # answer は A/B/C/D を想定（念のため整形）
    ans = "" if answer is None else str(answer).strip().upper()
    ans = ans[:1]  # "A" など1文字に寄せる

    return passage_id, text, question_block, ans

# --- 開始ボタン動作（✅ start時だけログ） ---
def start_test(student_id, state):
    state = _new_session_state()

    if not student_id or student_id.strip() == "":
        return (
            state,
            "", "", 0, "", None,
            "⚠️ 学生番号を入力してからテストを開始してください",
            False, "", "", "",
            gr.update(interactive=True),   # Student ID 入力：有効のまま
            gr.update(interactive=True),   # Start Test ボタン：有効のまま
        )

    state["user_id"] = student_id.strip()

    level = 3
    passage_id, text, question_block, ans = get_passage(level, state["used_passages"])
    state["used_passages"].add(passage_id)

    # ✅ この問題の正解を state に保持
    state["current_answer"] = ans

    displayed_time = datetime.utcnow() + timedelta(hours=9)

    # ✅ startログ：actions列に "start test pushed" を入れる
    log_event(
        state,
        question_number=None,
        reading_level=None,
        passage_id=None,
        question=None,
        choice=None,
        correct=None,
        actions="start test pushed",
    )

    return (
        state,
        text, question_block, level, passage_id, None,
        "", True, displayed_time.isoformat(), 1, state["user_id"],
        gr.update(interactive=False),  # ✅ Student ID 入力：無効化（リロードまで）
        gr.update(interactive=False),  # ✅ Start Test ボタン：無効化（リロードまで）
    )

# --- 選択肢変更（✅ CSVには書かない。actionsにだけ溜める） ---
def log_choice_change(choice, question_number, user_id, state):
    if not isinstance(state, dict):
        state = _new_session_state()
    if choice:
        t_iso = (datetime.utcnow() + timedelta(hours=9)).isoformat()
        state["action_log"].append({"action": "choice", "choice": choice, "time": t_iso})
    return state

# --- 回答送信（✅ submit時だけログ） ---
def next_step(prev_level, user_answer, question_text, passage_text,
              displayed_time, question_number, user_id, passage_id, state):

    if not isinstance(state, dict):
        state = _new_session_state()

    if not user_answer:
        return (
            state,
            "⚠️ Please select an answer!", passage_text, question_text, prev_level,
            None, "", True, displayed_time, question_number, user_id, passage_id
        )

    submit_time = datetime.utcnow() + timedelta(hours=9)
    state["question_count"] += 1

    # ✅ 変更点①：「わからない」は例外なく不正解
    if str(user_answer).strip() == "わからない":
        correct = False
    else:
        # ✅ CSVから読んだ正解（A/B/C/D）と比較（AI採点はしない）
        correct_answer = (state.get("current_answer") or "").strip().upper()
        correct = (str(user_answer).strip().upper() == correct_answer)

    new_level = adaptive_test(prev_level, correct)

    # ✅ submitログ（ここだけ記録）
    log_event(
        state,
        question_number=state["question_count"],
        reading_level=prev_level,
        passage_id=passage_id,
        question=question_text,
        choice=user_answer,
        correct=correct,
        actions=json.dumps(state["action_log"], ensure_ascii=False),
    )

    # ✅ 最終問題なら結果だけ大きく表示（仕様そのまま）
    if state["question_count"] >= 5:
        final_level = prev_level if correct else new_level
        return (
            state,
            f"<h1>🎯 Your Reading level: <strong>Level {final_level}</strong></h1>",
            "", "", final_level,
            None, "", False, "", "", user_id, passage_id
        )

    # --- 次の問題へ ---
    next_passage_id, next_text, next_question_block, next_ans = get_passage(new_level, state["used_passages"])
    state["used_passages"].add(next_passage_id)

    # ✅ 次の問題の正解を state に保持
    state["current_answer"] = next_ans

    next_display_time = datetime.utcnow() + timedelta(hours=9)

    # ✅ 次の問題のために選択変更履歴をリセット
    state["action_log"] = []

    feedback = "✅ Correct!" if correct else "❌ Incorrect."

    return (
        state,
        feedback + "\n➡️ Loading next question…",
        next_text, next_question_block, new_level,
        None, "", True, next_display_time.isoformat(), state["question_count"] + 1, user_id, next_passage_id
    )

# --- 管理者DL ---
def download_logs(admin_password):
    if not ADMIN_PASSWORD:
        return None, "⚠️ ADMIN_PASSWORD が設定されていません（環境変数で設定してください）"

    if (admin_password or "") != ADMIN_PASSWORD:
        return None, "❌ Password incorrect."

    with _log_lock:
        if not os.path.exists(LOG_FILE):
            return None, "⚠️ logs.csv がまだ存在しません（ログが1件もありません）"

        tmp_dir = tempfile.mkdtemp()

        # ✅ 変更点：DL用ファイル名を日付・時間付きにする（JST）
        ts = (datetime.utcnow() + timedelta(hours=9)).strftime("%Y%m%d_%H%M")
        out_path = os.path.join(tmp_dir, f"test_logs_{ts}.csv")

        with open(LOG_FILE, "rb") as fsrc, open(out_path, "wb") as fdst:
            fdst.write(fsrc.read())

    return out_path, "✅ logs.csv is ready."

# =========================
# ✅ 追加：Student ID の入力に応じて Start Test の有効/無効を切替
# =========================
def toggle_start_button(student_id):
    ok = bool((student_id or "").strip())
    return gr.update(interactive=ok)

# --- Gradio UI ---
with gr.Blocks() as demo:
    gr.Markdown("# 📘 Reading Level Test")

    session_state = gr.State(_new_session_state())

    student_id_input = gr.Textbox(label="Student ID", placeholder="例: B123456")

    # ✅ 初期状態は押せない（Student ID が空だから）
    start_btn = gr.Button("▶️ Start Test", interactive=False)

    # ✅ Student ID が入力されたら Start Test を押せるようにする
    student_id_input.change(
        fn=toggle_start_button,
        inputs=[student_id_input],
        outputs=[start_btn]
    )

    text_display = gr.Textbox(label="Reading Passage", lines=15, interactive=False)
    question_display = gr.Textbox(label="Question", lines=6, interactive=False)

    # ✅ 変更点②：共通選択肢「わからない」を追加（ログ列は触らない）
    user_answer = gr.Radio(["A", "B", "C", "D", "わからない"], label="Your Answer")

    submit_btn = gr.Button("Submit Answer")

    feedback_display = gr.Markdown()

    hidden_level = gr.Number(visible=False)
    hidden_passage = gr.Textbox(visible=False)  # 互換用（未使用）
    hidden_display_time = gr.Textbox(visible=False)
    hidden_question_number = gr.Number(visible=False)
    hidden_user_id = gr.Textbox(visible=False)
    hidden_passage_id = gr.Textbox(visible=False)
    test_visible = gr.State(False)

    start_btn.click(
        fn=start_test,
        inputs=[student_id_input, session_state],
        outputs=[
            session_state,
            text_display, question_display, hidden_level, hidden_passage_id, user_answer,
            feedback_display, test_visible, hidden_display_time,
            hidden_question_number, hidden_user_id,
            student_id_input,   # ✅ 追加：Start後に入力欄を無効化
            start_btn,          # ✅ 追加：Start後にボタンを無効化
        ]
    )

    # ✅ 選択肢変更はCSVに書かず、actionsにだけ溜める
    user_answer.change(
        fn=log_choice_change,
        inputs=[user_answer, hidden_question_number, hidden_user_id, session_state],
        outputs=[session_state]
    )

    submit_btn.click(
        fn=next_step,
        inputs=[hidden_level, user_answer, question_display, text_display,
                hidden_display_time, hidden_question_number, hidden_user_id, hidden_passage_id, session_state],
        outputs=[
            session_state,
            feedback_display, text_display, question_display, hidden_level,
            user_answer, hidden_passage, test_visible, hidden_display_time,
            hidden_question_number, hidden_user_id, hidden_passage_id
        ]
    )

    def toggle_visibility(show):
        v = bool(show)
        return (
            gr.update(visible=v), gr.update(visible=v),
            gr.update(visible=v), gr.update(visible=v)
        )

    test_visible.change(
        fn=toggle_visibility,
        inputs=test_visible,
        outputs=[text_display, question_display, user_answer, submit_btn]
    )

    gr.Markdown("---")
    gr.Markdown("## 🔒 Admin: Download logs (since last restart)")

    admin_pw = gr.Textbox(label="Admin Password", type="password", placeholder="Enter admin password")
    dl_btn = gr.Button("⬇️ Download logs.csv")
    dl_file = gr.File(label="logs.csv (download)")
    dl_msg = gr.Markdown()

    dl_btn.click(
        fn=download_logs,
        inputs=[admin_pw],
        outputs=[dl_file, dl_msg]
    )

# ✅ 追加：queue（同時アクセス耐性UP）
demo.queue(max_size=64).launch()