package space.anyi.service;

import space.anyi.client.GzCmcClient;
import space.anyi.cleaner.DetailCleaner;
import space.anyi.config.Config;
import space.anyi.dto.ChannelAllContentsResponse;
import space.anyi.dto.NewsItem;
import space.anyi.dto.NewsItemData;
import space.anyi.entity.News;
import space.anyi.mapper.NewsMapper;

import java.time.LocalDateTime;
import java.util.ArrayList;
import java.util.HashSet;
import java.util.List;
import java.util.Set;

/**
 * 阶段一：列表采集服务。
 *
 * <p>对每个配置关键词全量分页采集列表，按发布时间窗口过滤，通过 id 去重后
 * 仅插入缺失记录（content 置空，作为阶段二待补队列）。可重复运行，幂等。</p>
 */
public class ListCollectService {

    private final GzCmcClient client;
    private final NewsMapper mapper;
    private final Config config;

    /**
     * @param client 接口客户端
     * @param mapper NewsMapper
     * @param config 运行时配置
     */
    public ListCollectService(GzCmcClient client, NewsMapper mapper, Config config) {
        this.client = client;
        this.mapper = mapper;
        this.config = config;
    }

    /**
     * 执行列表采集主流程。
     *
     * <p>外层遍历配置的全部 channel、内层遍历全部关键词（双层循环，组合数为
     * O(频道数 × 关键词数) = O(n²)），每个频道×关键词组合独立分页采集。某页重试后
     * 仍失败（通常是服务端 offset≥10000 的检索上限）时优雅停止该组合的采集，
     * 不中断整体运行。完成后打印各组合与合计新增数。</p>
     *
     * @throws Exception 网络或序列化等未预期异常
     */
    public void run() throws Exception {
        int totalInserted = 0;
        for (String channelId : config.channelIds()) {
            int channelInserted = 0;
            for (String keyword : config.keywords()) {
                int pageNum = 1;
                int keywordInserted = 0;
                while (true) {
                    ChannelAllContentsResponse resp;
                    try {
                        resp = client.search(keyword, channelId, pageNum, config.pageSize());
                    } catch (Exception e) {
                        System.err.printf("[列表] 频道=%s 关键词=%s pageNum=%d 请求失败(重试后)，疑似达到服务端检索上限，停止本组合采集: %s%n",
                                channelId, keyword, pageNum, e.getMessage());
                        break;
                    }
                    List<NewsItem> items = resp.getList();
                    if (items == null || items.isEmpty()) {
                        break;
                    }
                    List<News> batch = new ArrayList<>();
                    for (NewsItem item : items) {
                        News news = toNews(item);
                        if (news != null) {
                            batch.add(news);
                        }
                    }
                    keywordInserted += insertMissing(batch);
                    int pages = resp.getPages() == null ? 1 : resp.getPages();
                    System.out.printf("[列表] 频道=%s 关键词=%s pageNum=%d/%d 命中窗口=%d%n",
                            channelId, keyword, pageNum, pages, batch.size());
                    if (pageNum >= pages) {
                        break;
                    }
                    pageNum++;
                }
                channelInserted += keywordInserted;
                System.out.printf("[列表] 频道=%s 关键词=%s 新增入库=%d%n", channelId, keyword, keywordInserted);
            }
            totalInserted += channelInserted;
            System.out.printf("[列表] 频道=%s 新增入库=%d%n", channelId, channelInserted);
        }
        System.out.printf("[列表] 全部完成, 新增入库合计=%d%n", totalInserted);
    }

    /**
     * 将列表元素映射为实体（含窗口过滤）。
     *
     * @param item 列表元素
     * @return News 实体；数据缺失、时间解析失败或不在发布窗口内时返回 null
     */
    private News toNews(NewsItem item) {
        NewsItemData data = item.getData();
        if (data == null || data.getId() == null) {
            return null;
        }
        LocalDateTime publishTime;
        try {
            publishTime = LocalDateTime.parse(data.getPublishTime(), Config.TIME_FORMATTER);
        } catch (Exception e) {
            return null;
        }
        if (publishTime.isBefore(config.publishStart()) || publishTime.isAfter(config.publishEnd())) {
            return null;
        }
        News news = new News();
        news.setId(data.getId());
        news.setTitle(DetailCleaner.titleClean(data.getTitle()));
        news.setUrl(data.getUrl());
        news.setPublishTime(publishTime);
        news.setChannelId(data.getChannelId());
        news.setChannelName(data.getChannelName());
        news.setEditor(data.getUserName());
        return news;
    }

    /**
     * 批量去重并插入缺失记录。
     *
     * <p>先用本批 id 查询库中已存在 id，仅插入不存在的记录，保证可重复运行。</p>
     *
     * @param batch 本页映射出的实体（已通过窗口过滤）
     * @return 实际新增行数
     */
    private int insertMissing(List<News> batch) {
        if (batch.isEmpty()) {
            return 0;
        }
        Set<String> ids = new HashSet<>();
        for (News n : batch) {
            ids.add(n.getId());
        }
        List<News> existing = mapper.selectBatchIds(ids);
        Set<String> existingIds = new HashSet<>();
        for (News n : existing) {
            existingIds.add(n.getId());
        }
        int inserted = 0;
        for (News n : batch) {
            if (existingIds.add(n.getId())) {
                mapper.insert(n);
                inserted++;
            }
        }
        return inserted;
    }
}