tiktok-scripts/_probe/merge_replies.js

/**
 * 把 /replies/*.json 的子回复合并进顶级评论 JSON
 * 用法: node merge_replies.js <comments.json> <repliesDir> <outDir> [awemeId]
 */
const fs = require('fs');
const path = require('path');

const inFile = process.argv[2];
const repliesDir = process.argv[3];
const outDir = process.argv[4] || path.dirname(inFile);
const awemeId = process.argv[5] || '7552202593888603410';

const comments = JSON.parse(fs.readFileSync(inFile, 'utf8'));
const byId = {};
comments.forEach(c => { byId[c.cid] = c; c.replies = c.replies || []; });

const stats = { files: 0, badFiles: 0, rawReplies: 0, uniqueReplies: 0, commentsMatched: 0, orphan: 0 };
let files = [];
if (fs.existsSync(repliesDir)) {
    files = fs.readdirSync(repliesDir).filter(f => f.endsWith('.json'));
}
console.log('回复页面文件:', files.length, '个');

// 按 cid 分组，保证按 cursor 顺序处理
const byComment = {};
files.forEach(f => {
    const m = /^(\d+)_(\d+)\.json$/.exec(f);
    if (!m) return;
    const cid = m[1], cursor = parseInt(m[2], 10);
    (byComment[cid] = byComment[cid] || []).push({ f, cursor });
});
console.log('涉及评论数:', Object.keys(byComment).length);

const seenReply = {};
for (const cid of Object.keys(byComment)) {
    byComment[cid].sort((a, b) => a.cursor - b.cursor);
    const parent = byId[cid];
    if (!parent) { stats.orphan++; continue; }
    let n = 0;
    for (const { f } of byComment[cid]) {
        stats.files++;
        let j;
        try { j = JSON.parse(fs.readFileSync(path.join(repliesDir, f), 'utf8')); }
        catch (e) { stats.badFiles++; continue; }
        if (j.code !== 0 || !j.data) continue;
        for (const r of (j.data.comments || [])) {
            stats.rawReplies++;
            if (!r || !r.id || seenReply[r.id]) continue;
            seenReply[r.id] = 1;
            stats.uniqueReplies++;
            const u = r.user || {};
            parent.replies.push({
                cid: r.id,
                text: r.text || '',
                user: u.unique_id || u.nickname || '匿名',
                nickname: u.nickname || '',
                userId: u.id || '',
                region: u.region || '',
                likes: parseInt(r.digg_count, 10) || 0,
                createTime: r.create_time || 0,
                replyCount: parseInt(r.reply_total, 10) || 0
            });
            n++;
        }
    }
    if (n > 0) stats.commentsMatched++;
}
console.log('合并统计:', JSON.stringify(stats, null, 2));

// 排序：每个父评论的回复按点赞降序
comments.forEach(c => c.replies.sort((a, b) => b.likes - a.likes));

const withR = comments.filter(c => (c.replies || []).length > 0);
const totalReplies = comments.reduce((s, c) => s + (c.replies || []).length, 0);

fs.mkdirSync(outDir, { recursive: true });
const base = path.join(outDir, 'tiktok_' + awemeId);
fs.writeFileSync(base + '_comments.json', JSON.stringify(comments, null, 2), 'utf8');

const L = [];
L.push('视频ID: ' + awemeId);
L.push('来源: tikwm镜像(www.tikwm.com) 顶级评论 ' + comments.length + ' 条');
L.push('子回复: ' + totalReplies + ' 条（覆盖 ' + withR.length + ' 条评论）');
L.push('抓取时间: ' + new Date().toISOString());
L.push('');
comments.forEach((c, i) => {
    L.push((i + 1) + '. [' + c.user + ' | 赞' + c.likes + ' | 回复' + c.replyCount + ' | 地区' + c.region + '] ' + c.text);
    (c.replies || []).forEach(r => {
        L.push('      └─ [' + r.user + ' | 赞' + r.likes + '] ' + r.text);
    });
});
fs.writeFileSync(base + '_comments.txt', L.join('\n'), 'utf8');

const summary = JSON.parse(fs.readFileSync(base + '_stats.json', 'utf8'));
summary.repliesFetched = totalReplies;
summary.commentsWithRepliesFetched = withR.length;
summary.replyFiles = stats.files;
summary.mergedAt = new Date().toISOString();
summary.note = 'tikwm 顶级评论上限约 10369；子回复按热度抓取，可在云机断点续传';
fs.writeFileSync(base + '_stats.json', JSON.stringify(summary, null, 2), 'utf8');

console.log('\n已写出:');
['_comments.json', '_comments.txt', '_stats.json'].forEach(s => {
    const p = base + s;
    console.log('  ' + p + '  (' + fs.statSync(p).size + ' bytes)');
});