package recruitment import ( "context" "crypto/md5" "encoding/json" "fmt" "net/url" "os/exec" "regexp" "strings" "time" "github.com/gogf/gf/v2/errors/gerror" "github.com/gogf/gf/v2/frame/g" "github.com/gogf/gf/v2/os/gtime" "service.xpcool.com/internal/dao" "service.xpcool.com/internal/model/do" "service.xpcool.com/internal/model/dto" "service.xpcool.com/internal/model/entity" ) // crawlerRun 按数据源类型分发抓取,返回运行结果(供写日志/统计)。 func crawlerRun(ctx context.Context, src entity.CrawlSource, force bool) dto.CrawlRunResult { res := dto.CrawlRunResult{SourceId: src.Id} defer func() { // 防止单源解析panic拖垮整体调度。 if r := recover(); r != nil { res.Err = gerror.Newf("crawler panic: %v", r) } }() var ( infos []dto.RecruitmentInput err error ) switch src.SourceType { case 1: // 静态列表页 infos, err = genericStaticCrawl(ctx, src, force) case 2: // SPA / JS 渲染(需底层 JSON 接口,首批 POC) infos, err = spaCrawl(ctx, src, force) case 3: // 需登录/验证码,首批已排除 err = gerror.New("该源需登录/验证码,首批已排除") case 4: // 附件型,先按静态抓取,附件解析为后续增强 infos, err = genericStaticCrawl(ctx, src, force) default: err = gerror.Newf("未知源类型 %d", src.SourceType) } res.Err = err res.Fetched = len(infos) if err != nil { return res } newCount, updated := 0, 0 for i := range infos { isNew, e := upsertInfo(ctx, infos[i]) if e != nil { res.Err = e continue } if isNew { newCount++ } else { updated++ } } res.NewCount = newCount res.UpdatedCount = updated return res } // genericStaticCrawl 通用静态列表抓取:取列表页锚点 → 逐条抓详情 → 抽取字段。 func genericStaticCrawl(ctx context.Context, src entity.CrawlSource, force bool) ([]dto.RecruitmentInput, error) { listURL := joinURL(src.BaseUrl, src.ListPath) if listURL == "" { listURL = src.BaseUrl } html, err := fetchHTML(ctx, listURL) if err != nil { return nil, err } g.Log().Infof(ctx, "[recruit-debug] src=%d listURL=%s htmlLen=%d", src.Id, listURL, len(html)) raw := extractAnchors(html, listURL) anchors := filterArticleAnchors(raw) g.Log().Infof(ctx, "[recruit-debug] src=%d rawAnchors=%d filtered=%d", src.Id, len(raw), len(anchors)) var out []dto.RecruitmentInput limit := 60 if force { limit = 300 // 全量回溯放宽上限 } for _, a := range anchors { if len(out) >= limit { break } dHtml, e := fetchHTML(ctx, a.URL) if e != nil { continue } title, content, pub, dl, ex := extractDetail(dHtml) if title == "" { title = a.Text } pubDate := normalizeDate(pub) // 增量模式:跳过 30 天前的公告,避免无效回扫。 if !force && pubDate != "" { if t, e2 := time.Parse("2006-01-02", pubDate); e2 == nil { if time.Since(t) > 30*24*time.Hour { continue } } } out = append(out, dto.RecruitmentInput{ Title: title, SourceId: src.Id, OrgName: src.Name, Category: src.Category, Region: src.Region, PublishDate: pubDate, Deadline: normalizeDate(dl), ExamDate: normalizeDate(ex), Url: a.URL, Content: content, Status: 0, }) } return out, nil } // spaCrawl SPA 源抓取(最佳实践:逆向底层 JSON 接口)。当前为占位实现, // 若服务端渲染无锚点则返回明确错误,便于在阶段0 POC 中补齐接口。 func spaCrawl(ctx context.Context, src entity.CrawlSource, force bool) ([]dto.RecruitmentInput, error) { listURL := joinURL(src.BaseUrl, src.ListPath) if listURL == "" { listURL = src.BaseUrl } html, err := fetchHTML(ctx, listURL) if err != nil { return nil, err } // 部分 SPA 会在 HTML 内联 __NEXT_DATA__ / 初始 state,可从中抽取。 if data := extractFromInlineJSON(html); len(data) > 0 { return data, nil } // 纯客户端渲染(hash 路由)无服务端内容,需 POC 接入接口。 if len(extractAnchors(html, listURL)) == 0 { return nil, gerror.New("SPA 无服务端渲染内容,需接入底层 JSON 接口(阶段0 POC)") } return genericStaticCrawl(ctx, src, force) } // extractFromInlineJSON 从 SPA 内联 JSON(__NEXT_DATA__ / window.__INITIAL_STATE__)抽取标题与链接。 // 适配 Vue/Next 等 SSR/CSR 混合页面,作为 SPA 源的兜底解析。 func extractFromInlineJSON(html string) []dto.RecruitmentInput { var out []dto.RecruitmentInput // 匹配常见内联 JSON 块,逐块扫描其中的标题/链接字段。 blocks := inlineJSONRe.FindAllStringSubmatch(html, -1) for _, b := range blocks { raw := b[1] var generic map[string]any if err := json.Unmarshal([]byte(raw), &generic); err != nil { continue } walkJSON(generic, &out) } return out } // walkJSON 递归遍历内联 JSON,提取含标题与链接的条目(最佳实践,避免硬规则)。 func walkJSON(node any, out *[]dto.RecruitmentInput) { switch v := node.(type) { case map[string]any: title, _ := v["title"].(string) link, _ := v["url"].(string) if link == "" { link, _ = v["href"].(string) } if title != "" && link != "" { *out = append(*out, dto.RecruitmentInput{ Title: title, Url: link, Status: 0, }) } for _, val := range v { walkJSON(val, out) } case []any: for _, item := range v { walkJSON(item, out) } } } // fetchHTML 带超时与浏览器 UA 的请求,降低被反爬概率。 // 部分政府站对 Go HTTP 客户端做 TLS 指纹级拦截(返回无锚点页面), // 此时回退系统 curl 重试(容器镜像内置 curl,已验证可拿到完整列表页)。 func fetchHTML(ctx context.Context, u string) (string, error) { client := g.Client() client.SetTimeout(15 * time.Second) client.SetHeader("User-Agent", "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/124.0 Safari/537.36") client.SetHeader("Accept", "text/html,application/xhtml+xml,*/*") client.SetHeader("Accept-Language", "zh-CN,zh;q=0.9") resp, err := client.Get(ctx, u) if err != nil { // 网络层失败:尝试 curl 兜底,避免单点失败。 if body, e := fetchWithCurl(u); e == nil { return body, nil } return "", gerror.Wrapf(err, "fetch %s", u) } defer resp.Close() if resp.StatusCode != 200 { // 非 200(如 302/403):回退 curl 重试(curl 自动跟随重定向)。 if body, e := fetchWithCurl(u); e == nil { return body, nil } return "", gerror.Newf("HTTP %d for %s", resp.StatusCode, u) } body := resp.ReadAllString() // 内容可疑(疑似 WAF/JS 挑战页:无 html 特征或锚点过少)时回退 curl。 if !looksLikeHtml(body) || len(body) < 3000 { if cb, e := fetchWithCurl(u); e == nil { return cb, nil } } return body, nil } // fetchWithCurl 调用系统 curl 抓取(自动跟随重定向,带浏览器 UA)。 func fetchWithCurl(u string) (string, error) { args := []string{ "-s", "-L", "--max-time", "25", "-A", "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/124.0 Safari/537.36", "-H", "Accept: text/html,application/xhtml+xml,*/*", "-H", "Accept-Language: zh-CN,zh;q=0.9", u, } out, err := exec.Command("curl", args...).Output() if err != nil { return "", gerror.Wrapf(err, "curl %s", u) } return string(out), nil } // looksLikeHtml 粗略判断响应是否为正常 HTML 文档(含 ... 块;hrefRe/titleAttrRe 从块内分别抽取链接与 title 属性。 anchorBlockRe = regexp.MustCompile(`(?is)]*>([\s\S]*?)`) hrefRe = regexp.MustCompile(`(?i)\bhref\s*=\s*["']([^"']+)["']`) titleAttrRe = regexp.MustCompile(`(?i)\btitle\s*=\s*["']([^"']+)["']`) tagRe = regexp.MustCompile(`<[^>]+>`) scriptRe = regexp.MustCompile(`(?is)`) styleRe = regexp.MustCompile(`(?is)`) titleRe = regexp.MustCompile(`(?is)]*>([\s\S]*?)`) h1Re = regexp.MustCompile(`(?is)]*>([\s\S]*?)`) dateRe = regexp.MustCompile(`(\d{4})[-/年.](\d{1,2})[-/月.](\d{1,2})`) inlineJSONRe = regexp.MustCompile(`(?is)(?:__NEXT_DATA__|__INITIAL_STATE__|window\.__[A-Z_]+)\s*=\s*(\{[\s\S]*?\})\s*;?`) spaceRe = regexp.MustCompile(`\s+`) // recruitKw 招聘类公告常见语义词,用于从政府站链接中识别公告条目。 recruitKw = regexp.MustCompile(`招聘|招考|公招|选聘|引进|人才|公告|公示|录用|录取|面试|笔试|报名|选调|遴选|拟聘|聘用|招募|招录|考试|考录|体检|考察|资格复审|递补`) ) // extractAnchors 从 HTML 抽取绝对化后的锚点(URL + 文本)。 // 政府站常把真实标题放在 title 属性、inner text 仅"详细/更多",故 inner 过短时回退 title。 func extractAnchors(html, base string) []anchor { out := make([]anchor, 0) for _, block := range anchorBlockRe.FindAllStringSubmatch(html, -1) { full := block[0] hm := hrefRe.FindStringSubmatch(full) if len(hm) < 2 { continue } href := strings.TrimSpace(hm[1]) if href == "" || strings.HasPrefix(href, "javascript:") || strings.HasPrefix(href, "#") || strings.HasPrefix(href, "mailto:") { continue } inner := strings.TrimSpace(stripTags(block[1])) text := inner if tm := titleAttrRe.FindStringSubmatch(full); len(tm) >= 2 { t := strings.TrimSpace(tm[1]) if len([]rune(text)) < 4 && t != "" { text = t } } if text == "" { continue } abs := resolveURL(base, href) if abs == "" { continue } out = append(out, anchor{URL: abs, Text: text}) } return out } // filterArticleAnchors 过滤导航/无用链接,保留疑似招聘公告条目并去重。 func filterArticleAnchors(as []anchor) []anchor { out := make([]anchor, 0, len(as)) for _, a := range as { text := strings.TrimSpace(a.Text) if len([]rune(text)) < 4 { continue } u := strings.ToLower(a.URL) // URL 形态线索:静态详情页、政务公开/人事招考栏目、含年份。 urlHint := strings.Contains(u, ".html") || strings.Contains(u, ".shtml") || strings.Contains(u, ".php") || strings.Contains(u, "/tzgg") || strings.Contains(u, "/rszk") || strings.Contains(u, "/rsxx") || strings.Contains(u, "/zfxx") || dateRe.MatchString(a.URL) // 文本语义线索:政府站公告标题常含这些词(必须命中,过滤"市长信箱/重点领域"等非招聘页)。 textHint := recruitKw.MatchString(text) || recruitKw.MatchString(a.URL) // 同时满足"内容页形态"与"招聘语义",避免误抓栏目/互动页。 if urlHint && textHint { out = append(out, a) } } seen := map[string]bool{} uniq := out[:0] for _, a := range out { if seen[a.URL] { continue } seen[a.URL] = true uniq = append(uniq, a) } return uniq } // extractDetail 从详情页抽取标题/正文/发布日期/报名截止/笔试日期(启发式,可按源调优)。 func extractDetail(html string) (title, content, publishDate, deadline, examDate string) { if m := titleRe.FindStringSubmatch(html); len(m) > 1 { title = stripTags(m[1]) } if title == "" { if m := h1Re.FindStringSubmatch(html); len(m) > 1 { title = stripTags(m[1]) } } title = strings.TrimSpace(title) body := scriptRe.ReplaceAllString(html, " ") body = styleRe.ReplaceAllString(body, " ") text := stripTags(body) text = collapseSpace(text) if len([]rune(text)) > 4000 { text = string([]rune(text)[:4000]) } content = text publishDate = normalizeDate(firstDate(html)) deadline = normalizeDate(findDateAfter(text, "报名")) examDate = normalizeDate(findDateAfter(text, "笔试")) return } // firstDate 返回文本中首个日期(归一化后)。 func firstDate(s string) string { m := dateRe.FindStringSubmatch(s) if len(m) == 0 { return "" } return normalizeDate(m[0]) } // findDateAfter 在 keyword 之后查找下一个日期(用于报名/笔试时间)。 func findDateAfter(text, keyword string) string { idx := strings.Index(text, keyword) if idx < 0 { return "" } rest := text[idx:] m := dateRe.FindStringSubmatch(rest) if len(m) == 0 { return "" } return normalizeDate(m[0]) } // normalizeDate 将多种中文/西式日期归一化为 YYYY-MM-DD。 func normalizeDate(s string) string { s = strings.TrimSpace(s) if s == "" { return "" } m := dateRe.FindStringSubmatch(s) if len(m) == 0 { return "" } y, mo, d := m[1], atoiSafe(m[2]), atoiSafe(m[3]) return fmt.Sprintf("%s-%02d-%02d", y, mo, d) } // stripTags 去除 HTML 标签。 func stripTags(s string) string { return tagRe.ReplaceAllString(s, " ") } // collapseSpace 折叠空白字符。 func collapseSpace(s string) string { return spaceRe.ReplaceAllString(s, " ") } // resolveURL 将相对链接解析为绝对 URL。 func resolveURL(base, href string) string { if strings.HasPrefix(href, "http://") || strings.HasPrefix(href, "https://") { return href } u, err := url.Parse(base) if err != nil { return "" } ref, err := url.Parse(href) if err != nil { return "" } return u.ResolveReference(ref).String() } // joinURL 拼接基础域名与路径。 func joinURL(base, p string) string { if p == "" { return base } if strings.HasPrefix(p, "http") { return p } u, err := url.Parse(base) if err != nil { return base + p } ref, err := url.Parse(p) if err != nil { return base + p } return u.ResolveReference(ref).String() } func atoiSafe(s string) int { n := 0 for _, c := range s { if c < '0' || c > '9' { break } n = n*10 + int(c-'0') } return n } // ---------- 去重与写入 ---------- func fingerprintOf(in dto.RecruitmentInput) string { h := md5.Sum([]byte(fmt.Sprintf("%d|%s|%s|%s", in.SourceId, in.Title, in.PublishDate, in.Url))) return fmt.Sprintf("%x", h) } func groupKeyOf(in dto.RecruitmentInput) string { h := md5.Sum([]byte(fmt.Sprintf("%s|%s", in.Title, in.PublishDate))) return fmt.Sprintf("%x", h) } // upsertInfo 按指纹去重写入公告;已存在则更新可变字段(内容/状态/日期)。 func upsertInfo(ctx context.Context, in dto.RecruitmentInput) (bool, error) { if in.Fingerprint == "" { in.Fingerprint = fingerprintOf(in) } if in.GroupKey == "" { in.GroupKey = groupKeyOf(in) } var exist entity.RecruitmentInfo if err := dao.RecruitmentInfo.Ctx(ctx).Where("fingerprint", in.Fingerprint).Scan(&exist); err != nil { // gf 的 Scan 在查无记录时返回 "no rows" 错误,视为未存在,继续插入。 if !strings.Contains(err.Error(), "no rows") { return false, gerror.Wrap(err, "query exist info") } } if exist.Id > 0 { _, err := dao.RecruitmentInfo.Ctx(ctx).Where("id", exist.Id).Data(do.RecruitmentInfo{ Content: in.Content, Status: in.Status, Deadline: in.Deadline, ExamDate: in.ExamDate, UpdatedAt: gtime.Now(), }).Update() if err != nil { return false, gerror.Wrap(err, "update info") } return false, nil } orgId, err := upsertOrg(ctx, in.OrgName, in.Category, in.Region) if err != nil { return false, err } in.OrgId = orgId if _, err = dao.RecruitmentInfo.Ctx(ctx).Data(toRecruitmentDO(in)).InsertAndGetId(); err != nil { return false, gerror.Wrap(err, "insert info") } return true, nil } // upsertOrg 发布主体按名称去重,返回主键。 func upsertOrg(ctx context.Context, name string, category int, region string) (uint64, error) { if name == "" { return 0, nil } var org entity.Organization if err := dao.Organization.Ctx(ctx).Where("name", name).Scan(&org); err != nil { if !strings.Contains(err.Error(), "no rows") { return 0, gerror.Wrap(err, "query org") } } if org.Id > 0 { return org.Id, nil } id, err := dao.Organization.Ctx(ctx).Data(do.Organization{ Name: name, Type: mapCategoryToOrgType(category), Region: region, CreatedAt: gtime.Now(), UpdatedAt: gtime.Now(), }).InsertAndGetId() if err != nil { return 0, gerror.Wrap(err, "insert org") } return uint64(id), nil } func mapCategoryToOrgType(c int) int { switch c { case 2: return 2 // 事业单位 case 3: return 3 // 国企 case 4: return 4 // 央企 case 5: return 5 // 私企 default: return 6 } } func toRecruitmentDO(in dto.RecruitmentInput) do.RecruitmentInfo { now := gtime.Now() return do.RecruitmentInfo{ Title: in.Title, SourceId: in.SourceId, OrgId: in.OrgId, OrgName: in.OrgName, Category: in.Category, Region: in.Region, PublishDate: nullIfEmpty(in.PublishDate), Deadline: nullIfEmpty(in.Deadline), ExamDate: nullIfEmpty(in.ExamDate), Url: in.Url, Content: in.Content, Attachments: in.Attachments, Status: in.Status, Fingerprint: in.Fingerprint, GroupKey: in.GroupKey, CreatedAt: now, UpdatedAt: now, } } // nullIfEmpty 将空字符串转为 nil,便于插入 NULL(DATE 等可空列不接受空串)。 func nullIfEmpty(s string) any { if strings.TrimSpace(s) == "" { return nil } return s } // recordCrawlLog 写抓取日志并更新数据源运行状态(成功清失败计数;失败累加)。 func recordCrawlLog(ctx context.Context, res dto.CrawlRunResult) { errMsg := "" if res.Err != nil { // 错误可能含整页 SQL,截断避免超过 error 列长度导致日志也写不进。 errMsg = res.Err.Error() const maxErr = 900 if len(errMsg) > maxErr { errMsg = errMsg[:maxErr] + "...(truncated)" } } if _, err := dao.CrawlLog.Ctx(ctx).Data(do.CrawlLog{ SourceId: res.SourceId, RunAt: gtime.Now(), Fetched: res.Fetched, NewCount: res.NewCount, UpdatedCount: res.UpdatedCount, Error: errMsg, }).Insert(); err != nil { g.Log().Errorf(ctx, "write crawl log failed: %v", err) } if res.Err != nil { var src entity.CrawlSource _ = dao.CrawlSource.Ctx(ctx).Where("id", res.SourceId).Scan(&src) _, _ = dao.CrawlSource.Ctx(ctx).Where("id", res.SourceId). Data(do.CrawlSource{FailCount: src.FailCount + 1}).Update() return } _, _ = dao.CrawlSource.Ctx(ctx).Where("id", res.SourceId). Data(do.CrawlSource{LastSuccessAt: gtime.Now().String(), FailCount: 0}).Update() } // ---------- 订阅辅助 ---------- // loadSubscription 加载推送订阅:id>0 按 id;否则取首个启用订阅。 func loadSubscription(ctx context.Context, id uint64) (*entity.PushSubscription, error) { var sub entity.PushSubscription m := dao.PushSubscription.Ctx(ctx) if id > 0 { m = m.Where("id", id) } else { m = m.Where("enabled", 1) } if err := m.OrderDesc("id").Scan(&sub); err != nil { return nil, gerror.Wrap(err, "query subscription") } if sub.Id == 0 { return nil, nil } return &sub, nil } func doPushSubscription(in dto.SubscriptionInput) do.PushSubscription { regions, _ := json.Marshal(in.Regions) cats, _ := json.Marshal(in.Categories) return do.PushSubscription{ Name: in.Name, DeviceKey: in.DeviceKey, Regions: string(regions), Categories: string(cats), OnlyNew: in.OnlyNew, PushTime: in.PushTime, Enabled: in.Enabled, UpdatedAt: gtime.Now(), } } func parseJSONStrings(s string) []string { if s == "" { return nil } var out []string _ = json.Unmarshal([]byte(s), &out) return out } func parseJSONInts(s string) []int { if s == "" { return nil } var out []int _ = json.Unmarshal([]byte(s), &out) return out }