clearHtml.go 4.7 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164
  1. package spider_com
  2. import (
  3. "regexp"
  4. "strings"
  5. "unicode/utf8"
  6. "github.com/PuerkitoBio/goquery"
  7. )
  8. type Cut struct {
  9. annotate *regexp.Regexp
  10. tag *regexp.Regexp
  11. scripttag *regexp.Regexp
  12. inputag *regexp.Regexp
  13. isborder *regexp.Regexp
  14. hiddentag *regexp.Regexp
  15. styletag *regexp.Regexp
  16. colstag *regexp.Regexp
  17. rowstag *regexp.Regexp
  18. display *regexp.Regexp
  19. multiCR *regexp.Regexp
  20. replBlankLine *regexp.Regexp
  21. replStartWrap *regexp.Regexp
  22. replTags2CR []string
  23. retainTags2CR []string
  24. }
  25. func NewCut() *Cut {
  26. t, _ := regexp.Compile("<[^>]+>")
  27. m, _ := regexp.Compile("([\r\n][\u3000\u2003\u00a0\\s]*)+|[\r\n]+")
  28. //sc, _ := regexp.Compile("\\<script[^\\>]*\\>*[^\\>]+\\</script\\>")
  29. //ss, _ := regexp.Compile("\\<style[^\\>]*\\>*[^\\>]+\\</style\\>")
  30. scs := regexp.MustCompile("(?s)<(script|style)[^>]*>.+?</(script|style)>")
  31. at := regexp.MustCompile("(?s)<(!%-%-|!--).*?(%-%-|--)>") //注释 css
  32. hiddentag := regexp.MustCompile(`<\s*input[^<]*type=("|')hidden("|')[^<]*>`)
  33. input := regexp.MustCompile(`<\s*input[^<]*value=("|')([^>"']*)[^<]*>`)
  34. cols, _ := regexp.Compile(`colspan="\d+"`)
  35. rows, _ := regexp.Compile(`rowspan="\d+"`)
  36. border, _ := regexp.Compile(`(border="(\d+)")|(cellpadding="(\d+)")|(cellspacing="(\d+)")`)
  37. dis, _ := regexp.Compile(`display:none`)
  38. return &Cut{
  39. annotate: at,
  40. tag: t,
  41. scripttag: scs,
  42. hiddentag: hiddentag,
  43. inputag: input,
  44. colstag: cols,
  45. isborder: border,
  46. rowstag: rows,
  47. display: dis,
  48. multiCR: m,
  49. replBlankLine: regexp.MustCompile("\\s+[\r\n]"),
  50. replStartWrap: regexp.MustCompile("^[\u3000\u2003\u00a0\\s]+|[\u3000\u2003\u00a0\\s]+$"),
  51. replTags2CR: []string{"div", "p", "br", "h1", "h2", "h3", "h4", "h5"},
  52. retainTags2CR: []string{"table", "thead", "tfoot", "tbody", "th", "td", "tr"},
  53. }
  54. }
  55. // 清理HTML标签
  56. func (c *Cut) ClearHtml(src string) string {
  57. src = c.replBlankLine.ReplaceAllString(src, "")
  58. src = strings.Replace(src, ">\n", ">", -1)
  59. src = strings.Replace(src, " ", "", -1)
  60. //标签全转小写
  61. src = c.tag.ReplaceAllStringFunc(src, strings.ToLower)
  62. //清script,style
  63. src = c.scripttag.ReplaceAllString(src, "")
  64. //清理注释文本
  65. src = c.annotate.ReplaceAllString(src, "")
  66. //清理input
  67. src = c.hiddentag.ReplaceAllString(src, "")
  68. src = c.inputag.ReplaceAllString(src, "$2")
  69. document, err := goquery.NewDocumentFromReader(strings.NewReader(src))
  70. if err == nil {
  71. if tmpstr, err := document.Each(func(i int, sel *goquery.Selection) {
  72. sel.Find("td").Each(func(i int, selection *goquery.Selection) {
  73. val, b := selection.Attr("title")
  74. if b && strings.Trim(val, " ") != "" {
  75. tmpstr := strings.TrimFunc(selection.Text(), func(r rune) bool {
  76. return r == 9 || r == 32
  77. })
  78. if utf8.RuneCountInString(strings.Trim(tmpstr, " ")) < utf8.RuneCountInString(strings.Trim(val, " ")) {
  79. selection.SetText(strings.Trim(val, " "))
  80. }
  81. }
  82. })
  83. }).Html(); err == nil {
  84. src = tmpstr
  85. }
  86. }
  87. //换结束标签
  88. src = c.tag.ReplaceAllStringFunc(src, func(tmp string) string {
  89. tmp = strings.Replace(tmp, " ", "", -1)
  90. //保留这些标签
  91. for _, v := range c.retainTags2CR {
  92. if "<"+v+">" == tmp || "</"+v+">" == tmp {
  93. if tmp == "</table>" {
  94. return tmp + "\n"
  95. }
  96. return tmp
  97. }
  98. if strings.HasPrefix(tmp, "<"+v) {
  99. dispstrs := c.display.FindAllString(tmp, -1)
  100. rowstrs := c.rowstag.FindAllString(tmp, -1)
  101. colstrs := c.colstag.FindAllString(tmp, -1)
  102. con := "<" + v
  103. if con == "<table" {
  104. if isHasBoder(tmp, c.isborder) {
  105. con = con + ` border="1"`
  106. }
  107. }
  108. if len(colstrs) > 0 { //处理多列合并
  109. con += " " + colstrs[0]
  110. }
  111. if len(rowstrs) > 0 { //处理多行合并
  112. con += " " + rowstrs[0]
  113. }
  114. if len(dispstrs) > 0 {
  115. con += " style=\"" + dispstrs[0] + "\""
  116. }
  117. return con + ">"
  118. }
  119. }
  120. if tmp == "<br>" || tmp == "<br/>" || tmp == "<center>" || tmp == "</center>" || tmp == "<ul>" || tmp == "</ul>" {
  121. return "\n"
  122. }
  123. if tmp[1] != 47 { //开始标签
  124. for _, v := range c.replTags2CR {
  125. if v == tmp[1:len(tmp)-1] {
  126. return "\n"
  127. }
  128. }
  129. return ""
  130. }
  131. for _, v := range c.replTags2CR {
  132. if v == tmp[2:len(tmp)-1] {
  133. return "\n"
  134. }
  135. }
  136. return ""
  137. })
  138. src = c.replStartWrap.ReplaceAllString(src, "")
  139. src = c.replBlankLine.ReplaceAllString(src, "\n")
  140. //清除多余换行
  141. c.multiCR.ReplaceAllString(src, "\n")
  142. return strings.Replace(src, "\n", "<br/>", -1)
  143. }
  144. // 判断table是否加表格线
  145. func isHasBoder(con string, reg *regexp.Regexp) bool {
  146. res := reg.FindAllStringSubmatch(con, -1)
  147. hasBorder := false
  148. for _, v := range res {
  149. for k, val := range v {
  150. if k > 0 && k%2 == 0 && IntAll(val) > 0 {
  151. hasBorder = true
  152. break
  153. }
  154. }
  155. }
  156. return hasBorder
  157. }