types.go 9.1 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254
  1. package backend
  2. import (
  3. "container/list"
  4. "encoding/json"
  5. "regexp"
  6. )
  7. const (
  8. JOB_RUNNING_EVENT_PROGRESS = 1
  9. JOB_RUNNING_EVENT_DEBUG = 0
  10. )
  11. var (
  12. DataResults = map[string]*list.List{} //调试运行结果
  13. VerifyResults = map[string]*SpiderConfigVerifyResult{} //调试验证结果
  14. RegSpace = regexp.MustCompile("[\\s\u3000\u2003\u00a0]+")
  15. )
  16. type (
  17. //爬虫配置信息
  18. SpiderConfig struct {
  19. Site string `json:"site"`
  20. Channel string `json:"channel"`
  21. ModifyUser string `json:"modifyuser"`
  22. Href string `json:"href"`
  23. Code string `json:"code"`
  24. ListBodyCss string `json:"listBodyCss"` //用于判断是否翻页成功
  25. ListItemCss string `json:"listItemCss"`
  26. ListLinkCss string `json:"listLinkCss"`
  27. ListPubtimeCss string `json:"listPublishTimeCss"`
  28. ListNextPageCss string `json:"listNextPageCss"`
  29. TitleCss string `json:"titleCss"`
  30. PublishUnitCss string `json:"publishUnitCss"`
  31. PublishTimeCss string `json:"publishTimeCss"`
  32. ContentCss string `json:"contentCss"`
  33. AttachCss string `json:"attachCss"`
  34. ListJSCode string `json:"listJs"`
  35. ContentJSCode string `json:"contentJs"`
  36. AttachJSCode string `json:"attachJs"` //无效
  37. ListTurnPageJSCode string `json:"listTurnPageJs"`
  38. MaxPages int64 `json:"maxPages"`
  39. //延时
  40. ListDelayTime int64 `json:"listDelayTime"`
  41. ListTurnDelayTime int64 `json:"listTurnDelayTime"`
  42. ContentDelayTime int64 `json:"contentDelayTime"`
  43. }
  44. //附件链接
  45. AttachLink struct {
  46. Title string `json:"title"`
  47. Href string `json:"href"`
  48. FileName string `json:"fileName"`
  49. FileType string `json:"fileType"`
  50. FileSize string `json:"fileSize"`
  51. FilePath string `json:"filePath"`
  52. }
  53. //爬取结果信息
  54. ResultItem struct {
  55. No int `json:"no"` //编号
  56. Site string `json:"site"`
  57. Channel string `json:"channel"`
  58. Href string `json:"href"`
  59. ListTitle string `json:"listTitle"`
  60. ListPubTime string `json:"listPublishTime"`
  61. Title string `json:"title"`
  62. PublishUnit string `json:"publishUnit"`
  63. PublishTime string `json:"publishTime"`
  64. Content string `json:"content"`
  65. ContentHtml string `json:"contentHtml"`
  66. AttachLinks []*AttachLink `json:"attachLinks"` //存放附件的标题,链接
  67. AttachJson string `json:"attachJson"` //存放附件的OSS元信息
  68. }
  69. //爬取结果信息清单
  70. ResultItems []*ResultItem
  71. SpiderConfiges []*SpiderConfig
  72. //
  73. JobItem struct {
  74. //code/site与爬虫配置一致
  75. SpiderCode string `json:"code"`
  76. SpiderSite string `json:"site"`
  77. //以下是运行参数(不配置时,默认使用通用配置)
  78. Channel string `json:"channel"`
  79. Url string `json:"url"`
  80. ProxyServe string `json:"proxyServe"`
  81. MaxPages int `json:"maxPages"`
  82. Threads int `json:"threads"`
  83. ListDelay int64 `json:"listDelay"`
  84. TrunPageDelay int64 `json:"trunPageDelay"`
  85. ContentDelay int64 `json:"contentDelay"`
  86. NeedDownloadAttaches bool `json:"needDownloadAttaches"`
  87. }
  88. //作业定义
  89. Job struct {
  90. Code string `json:"code"`
  91. Name string `json:"name"`
  92. Items []*JobItem `json:"items"`
  93. //通用参数
  94. ProxyServe string `json:"proxyServe"`
  95. MaxPages int `json:"maxPages"`
  96. Threads int `json:"threads"`
  97. ListDelay int64 `json:"listDelay"`
  98. TrunPageDelay int64 `json:"trunPageDelay"`
  99. ContentDelay int64 `json:"contentDelay"`
  100. State int `json:"state"`
  101. StateType string `json:"stateType"`
  102. Progress int `json:"progress"`
  103. NeedDownloadAttaches bool `json:"needDownloadAttaches"`
  104. }
  105. Jobs []*Job
  106. //推送事件
  107. JobRunningEvent struct {
  108. Progress int `json:"progress"`
  109. Msg string `json:"msg"`
  110. Act int `json:"act"`
  111. Code string `json:"code"`
  112. }
  113. //Job运行时状态,中间结果
  114. JobRunningState struct {
  115. Code string
  116. State int
  117. Progress int
  118. ResultCache *list.List //缓存
  119. ExitCh chan bool
  120. }
  121. //爬虫配置验证结果
  122. SpiderConfigVerifyResult struct {
  123. Title bool `json:"title"`
  124. PublishUnit bool `json:"publishUnit"`
  125. PublishTime bool `json:"publishTime"`
  126. Content bool `json:"content"`
  127. Attaches bool `json:"attaches"`
  128. ListItems bool `json:"listItems"`
  129. ListTrunPage bool `json:"listTrunPage"`
  130. }
  131. )
  132. func (sc SpiderConfiges) Len() int {
  133. return len(sc)
  134. }
  135. func (sc SpiderConfiges) Swap(i, j int) {
  136. sc[i], sc[j] = sc[j], sc[i]
  137. }
  138. func (sc SpiderConfiges) Less(i, j int) bool {
  139. return sc[i].Code > sc[j].Code
  140. }
  141. func (js Jobs) Len() int {
  142. return len(js)
  143. }
  144. func (js Jobs) Swap(i, j int) {
  145. js[i], js[j] = js[j], js[i]
  146. }
  147. func (js Jobs) Less(i, j int) bool {
  148. return js[i].Code > js[j].Code
  149. }
  150. // CopyAttribute
  151. func CopyAttribute(dst *string, value1, value2 string) {
  152. if value1 == "EMPTY" { //特殊符号,强制修改
  153. *dst = ""
  154. } else if value1 != "" {
  155. *dst = value1
  156. } else if value2 != "" {
  157. *dst = value2
  158. }
  159. }
  160. // NewSpiderConfig 生成css对象
  161. func NewSpiderConfig(cssmark map[string]interface{}) (*SpiderConfig, error) {
  162. sc := &SpiderConfig{}
  163. cssmark_byte, err := json.Marshal(cssmark)
  164. if err != nil {
  165. return nil, err
  166. }
  167. err = json.Unmarshal(cssmark_byte, &sc)
  168. return sc, err
  169. //param_common := param["param_common"].(map[string]interface{})
  170. //css_list, _ := param["css_list"].(map[string]interface{})
  171. //css_content, _ := param["css_content"].(map[string]interface{})
  172. //js_list, _ := param["js_list"].(map[string]interface{})
  173. //js_nextpage, _ := param["js_nextpage"].(map[string]interface{})
  174. //js_content, _ := param["js_content"].(map[string]interface{})
  175. //sc := &SpiderConfig{
  176. // Site: qu.ObjToString(param_common["site"]),
  177. // Channel: qu.ObjToString(param_common["channel"]),
  178. // Modifyuser: qu.ObjToString(param_common["modifyuser"]),
  179. // Href: qu.ObjToString(param_common["href"]),
  180. // Code: qu.ObjToString(param_common["code"]),
  181. // ListBodyCss: qu.ObjToString(css_list["body"]),
  182. // ListItemCss: qu.ObjToString(css_list["title"]),
  183. // ListLinkCss: qu.ObjToString(css_list["href"]),
  184. // ListPubtimeCss: qu.ObjToString(css_list["ptime"]),
  185. // ListNextPageCss: qu.ObjToString(css_list["nextpage"]),
  186. // TitleCss: qu.ObjToString(css_content["title"]),
  187. // PublishUnitCss: qu.ObjToString(css_content["source"]),
  188. // PublishTimeCss: qu.ObjToString(css_content["ptime"]),
  189. // ContentCss: qu.ObjToString(css_content["content"]),
  190. // AttachCss: qu.ObjToString(css_content["file"]),
  191. // ListJSCode: qu.ObjToString(js_list["js"]),
  192. // ContentJSCode: qu.ObjToString(js_content["js"]),
  193. // ListTrunPageJSCode: qu.ObjToString(js_nextpage["js"]),
  194. // //AttachJSCode : string `json:"attachJs"` //无效
  195. //}
  196. }
  197. // MergeSpiderConfig 合并
  198. func MergeSpiderConfig(src1, src2 *SpiderConfig) *SpiderConfig {
  199. nsc := new(SpiderConfig)
  200. CopyAttribute(&nsc.Code, src2.Code, src1.Code)
  201. CopyAttribute(&nsc.Site, src2.Site, src1.Site)
  202. CopyAttribute(&nsc.Channel, src2.Channel, src1.Channel)
  203. CopyAttribute(&nsc.Href, src2.Href, src1.Href)
  204. CopyAttribute(&nsc.ModifyUser, src2.ModifyUser, src1.ModifyUser)
  205. CopyAttribute(&nsc.ListBodyCss, src2.ListBodyCss, src1.ListBodyCss)
  206. CopyAttribute(&nsc.ListItemCss, src2.ListItemCss, src1.ListItemCss)
  207. CopyAttribute(&nsc.ListLinkCss, src2.ListLinkCss, src1.ListLinkCss)
  208. CopyAttribute(&nsc.ListPubtimeCss, src2.ListPubtimeCss, src1.ListPubtimeCss)
  209. CopyAttribute(&nsc.ListNextPageCss, src2.ListNextPageCss, src1.ListNextPageCss)
  210. CopyAttribute(&nsc.TitleCss, src2.TitleCss, src1.TitleCss)
  211. CopyAttribute(&nsc.PublishTimeCss, src2.PublishTimeCss, src1.PublishTimeCss)
  212. CopyAttribute(&nsc.PublishUnitCss, src2.PublishUnitCss, src1.PublishUnitCss)
  213. CopyAttribute(&nsc.ContentCss, src2.ContentCss, src1.ContentCss)
  214. CopyAttribute(&nsc.AttachCss, src2.AttachCss, src1.AttachCss)
  215. CopyAttribute(&nsc.ListJSCode, src2.ListJSCode, src1.ListJSCode)
  216. CopyAttribute(&nsc.ContentJSCode, src2.ContentJSCode, src1.ContentJSCode)
  217. CopyAttribute(&nsc.AttachJSCode, src2.AttachJSCode, src1.AttachJSCode)
  218. CopyAttribute(&nsc.ListTurnPageJSCode, src2.ListTurnPageJSCode, src1.ListTurnPageJSCode)
  219. return nsc
  220. }
  221. // CopySpiderConfig 复制
  222. func CopySpiderConfig(src1, src2 *SpiderConfig) {
  223. src1.Code = src2.Code
  224. src1.Site = src2.Site
  225. src1.ModifyUser = src2.ModifyUser
  226. src1.Channel = src2.Channel
  227. src1.Href = src2.Href
  228. src1.ListBodyCss = src2.ListBodyCss
  229. src1.ListItemCss = src2.ListItemCss
  230. src1.ListPubtimeCss = src2.ListPubtimeCss
  231. src1.ListNextPageCss = src2.ListNextPageCss
  232. src1.ListLinkCss = src2.ListLinkCss
  233. src1.TitleCss = src2.TitleCss
  234. src1.PublishTimeCss = src2.PublishTimeCss
  235. src1.PublishUnitCss = src2.PublishUnitCss
  236. src1.ContentCss = src2.ContentCss
  237. src1.AttachCss = src2.AttachCss
  238. src1.ListJSCode = src2.ListJSCode
  239. src1.ListTurnPageJSCode = src2.ListTurnPageJSCode
  240. src1.ContentJSCode = src2.ContentJSCode
  241. src1.AttachJSCode = src2.AttachJSCode
  242. }