app.py

import gradio as gr
import pandas as pd
from atproto import Client
from huggingface_hub import HfApi
import os
import traceback
import re
from collections import Counter

# --- 設定 ---
DATASET_ID = "Nyanpre/kaibunnko_data" 
DATA_FILE = "data.csv"

HF_TOKEN = os.getenv("HF_TOKEN")
BSKY_HANDLE = os.getenv("BSKY_HANDLE", "raira0626.bsky.social")
BSKY_PASSWORD = os.getenv("BSKY_PASSWORD", "Raira0626")

api = HfApi()

# --- カスタムCSS ---
custom_css = """
.list-wrapper { display: flex; flex-direction: column; gap: 0; }
.list-item { padding: 20px 12px !important; border-bottom: 2px solid #f0f2f5 !important; background-color: white !important; }
.row-meta { display: flex; justify-content: space-between; align-items: center; font-size: 12px; color: #6b7280; margin-bottom: 12px; }
.meta-left { font-weight: bold; color: #1f2937; }
.meta-right { display: flex; gap: 15px; }
.row-content { font-size: 15px; line-height: 1.7; color: #111827; word-break: break-all; white-space: pre-wrap !important; }
"""

def generate_html_list(df):
    if df is None or df.empty:
        return "<p style='text-align:center; padding:20px;'>条件に一致するデータがありません</p>"
    html_output = '<div class="list-wrapper">'
    for _, row in df.iterrows():
        date_str = str(row['日付']).split('T')[0]
        html_output += f"""
        <div class="list-item">
            <div class="row-meta">
                <span class="meta-left">👤 {row['投稿者']}</span>
                <div class="meta-right">
                    <span>📅 {date_str}</span>
                    <span>❤️ {row['♡数']}</span>
                    <a href="{row['URL']}" target="_blank" style="color: #008DFF; text-decoration: underline;">URL</a>
                </div>
            </div>
            <div class="row-content">{row['本文']}</div>
        </div>"""
    return html_output + '</div>'

def get_updated_dropdowns(df):
    """表示用ラベルのリストを作成する（例: '名前 (10)'）"""
    if df is None or df.empty:
        return gr.update(choices=["すべて"]), gr.update(choices=["すべて"])
    
    # 投稿者
    author_counts = df["投稿者"].value_counts()
    author_choices = ["すべて"] + [f"{name} ({count})" for name, count in author_counts.items()]
    
    # タグ
    all_tags = []
    for text in df["本文"].astype(str):
        all_tags.extend(re.findall(r'#\w+', text))
    tag_counts = Counter(all_tags)
    sorted_tags = sorted(tag_counts.items(), key=lambda x: x[1], reverse=True)
    tag_choices = ["すべて"] + [f"{tag} ({count})" for tag, count in sorted_tags]
    
    return gr.update(choices=author_choices), gr.update(choices=tag_choices)

def fetch_and_save(target_hashtag="#青空怪文庫", max_total=300):
    try:
        client = Client()
        client.login(BSKY_HANDLE, BSKY_PASSWORD)
        all_posts = []
        cursor = None
        query = target_hashtag if target_hashtag.startswith("#") else f"#{target_hashtag}"
        
        while len(all_posts) < max_total:
            response = client.app.bsky.feed.search_posts(params={'q': query, 'limit': 100, 'cursor': cursor})
            if not response.posts: break
            all_posts.extend(response.posts)
            cursor = response.cursor
            if not cursor: break

        new_data = []
        for post in all_posts[:max_total]:
            # デフォルト（引用がない場合）の情報
            display_author = post.author.display_name or post.author.handle
            display_text = getattr(post.record, 'text', "")
            display_url = f"https://bsky.app/profile/{post.author.did}/post/{post.uri.split('/')[-1]}"
            likes = getattr(post, 'like_count', 0)
            
            # --- 引用ポスト（Quote Post）の判定と差し替えロジック ---
            if hasattr(post, 'embed') and post.embed is not None:
                # 埋め込みが「レコード（引用ポスト）」であるか確認
                # オブジェクト型が app.bsky.embed.record#view である場合が多い
                embed = post.embed
                if hasattr(embed, 'record'):
                    # 引用先のデータ（recordの中身）を取得
                    quoted_record = embed.record
                    
                    # 引用先の「投稿者」や「本文」が存在するかチェック（削除済みでない場合）
                    if hasattr(quoted_record, 'author') and (hasattr(quoted_record, 'value') or hasattr(quoted_record, 'record')):
                        tagger_name = post.author.display_name or post.author.handle
                        
                        # 引用元の投稿者名
                        display_author = quoted_record.author.display_name or quoted_record.author.handle
                        
                        # 引用元の本文取得（構造が複雑なため安全に取得）
                        original_text = ""
                        if hasattr(quoted_record, 'value'): # 通常のポスト
                            original_text = getattr(quoted_record.value, 'text', "")
                        elif hasattr(quoted_record, 'record') and hasattr(quoted_record.record, 'text'):
                            original_text = quoted_record.record.text
                            
                        # 本文の書き換え
                        display_text = f"{original_text}\n\n({tagger_name} によるタグ付け)"
                        
                        # 引用元のURLを作成
                        # uriは 'at://did:plc:xxx/app.bsky.feed.post/rkey' の形式
                        uri_parts = quoted_record.uri.split('/')
                        display_url = f"https://bsky.app/profile/{quoted_record.author.did}/post/{uri_parts[-1]}"
                        
                        # いいね数は引用元のものがあればそちらを、なければ0
                        # ※検索APIのembed内にはlike_countが含まれないことが多いため、
                        #   厳密に取得するには別途 getPosts 等が必要ですが、ここでは簡易的に0または元の値を保持します
                        likes = getattr(quoted_record, 'like_count', 0)

            new_data.append({
                "日付": post.record.created_at[:10],
                "投稿者": display_author,
                "本文": display_text,
                "URL": display_url,
                "♡数": likes
            })
        
        df = pd.DataFrame(new_data)
        df.to_csv(DATA_FILE, index=False, encoding='utf-8-sig')
        if HF_TOKEN:
            try: api.upload_file(path_or_fileobj=DATA_FILE, path_in_repo=DATA_FILE, repo_id=DATASET_ID, repo_type="dataset", token=HF_TOKEN)
            except: pass
        
        a_up, t_up = get_updated_dropdowns(df)
        return f"✅ {len(df)}件更新完了", generate_html_list(df.sort_values(by="日付", ascending=False)), a_up, t_up
    except Exception:
        return f"❌ エラー:\n{traceback.format_exc()}", "<p>エラー</p>", gr.update(), gr.update()

def search_data(author_val, tag_val, keyword_val, sort_order):
    if not os.path.exists(DATA_FILE): return "<p>データがありません</p>"
    df = pd.read_csv(DATA_FILE)
    
    # 【重要】「名前 (10)」から「名前」だけを取り出す処理
    def clean_val(val):
        if val == "すべて" or not val: return val
        return re.sub(r'\s\(\d+\)$', '', val)

    author_clean = clean_val(author_val)
    tag_clean = clean_val(tag_val)

    if author_clean != "すべて":
        df = df[df["投稿者"] == author_clean]
    if tag_clean != "すべて":
        df = df[df["本文"].str.contains(tag_clean, na=False)]
    if keyword_search:
        df = df[df["本文"].str.contains(keyword_val, na=False, case=False)]
    
    if sort_order == "新しい順": df = df.sort_values(by="日付", ascending=False)
    elif sort_order == "古い順": df = df.sort_values(by="日付", ascending=True)
    elif sort_order == "いいねが多い順": df = df.sort_values(by=["♡数", "日付"], ascending=[False, False])
    
    return generate_html_list(df)

def auto_init():
    if os.path.exists(DATA_FILE):
        df = pd.read_csv(DATA_FILE)
        a_up, t_up = get_updated_dropdowns(df)
        return "読み込み完了", generate_html_list(df.sort_values(by="日付", ascending=False)), a_up, t_up
    return fetch_and_save()

with gr.Blocks() as demo:
    gr.Markdown("# 👻 青空怪文庫 DB")
    with gr.Tabs():
        with gr.TabItem("🔍 検索"):
            with gr.Row():
                author_dropdown = gr.Dropdown(label="投稿者", choices=["すべて"], value="すべて")
                tag_dropdown = gr.Dropdown(label="タグ", choices=["すべて"], value="すべて")
            with gr.Row():
                keyword_search = gr.Textbox(label="キーワード", placeholder="内容で検索...", scale=3)
                sort_order = gr.Radio(label="並び替え", choices=["新しい順", "古い順", "いいねが多い順"], value="新しい順", scale=2)
            with gr.Row():
                search_btn = gr.Button("検索実行", variant="primary")
                clear_btn = gr.Button("リセット")
            db_html = gr.HTML()
        with gr.TabItem("🔄 更新"):
            tag_input = gr.Textbox(label="タグ", value="青空怪文庫")
            update_btn = gr.Button("更新", variant="stop")
            update_log = gr.Textbox(label="ログ", interactive=False)

    search_btn.click(search_data, [author_dropdown, tag_dropdown, keyword_search, sort_order], db_html)
    update_btn.click(fetch_and_save, [tag_input], [update_log, db_html, author_dropdown, tag_dropdown])
    demo.load(auto_init, None, [update_log, db_html, author_dropdown, tag_dropdown])

if __name__ == "__main__":
    demo.launch(theme=gr.themes.Soft(), css=custom_css)