clearHtml.go 4.7 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165
  1. package util
  2. import (
  3. "github.com/PuerkitoBio/goquery"
  4. "regexp"
  5. "strings"
  6. "unicode/utf8"
  7. )
  8. //
  9. type Cut struct {
  10. annotate *regexp.Regexp
  11. tag *regexp.Regexp
  12. scripttag *regexp.Regexp
  13. inputag *regexp.Regexp
  14. isborder *regexp.Regexp
  15. hiddentag *regexp.Regexp
  16. styletag *regexp.Regexp
  17. colstag *regexp.Regexp
  18. rowstag *regexp.Regexp
  19. display *regexp.Regexp
  20. multiCR *regexp.Regexp
  21. replBlankLine *regexp.Regexp
  22. replStartWrap *regexp.Regexp
  23. replTags2CR []string
  24. retainTags2CR []string
  25. }
  26. //
  27. func NewCut() *Cut {
  28. t, _ := regexp.Compile("<[^>]+>")
  29. m, _ := regexp.Compile("([\r\n][\u3000\u2003\u00a0\\s]*)+|[\r\n]+")
  30. //sc, _ := regexp.Compile("\\<script[^\\>]*\\>*[^\\>]+\\</script\\>")
  31. //ss, _ := regexp.Compile("\\<style[^\\>]*\\>*[^\\>]+\\</style\\>")
  32. scs := regexp.MustCompile("(?s)<(script|style)[^>]*>.+?</(script|style)>")
  33. at := regexp.MustCompile("(?s)<(!%-%-|!--).*?(%-%-|--)>") //注释 css
  34. hiddentag := regexp.MustCompile(`<\s*input[^<]*type=("|')hidden("|')[^<]*>`)
  35. input := regexp.MustCompile(`<\s*input[^<]*value=("|')([^>"']*)[^<]*>`)
  36. cols, _ := regexp.Compile(`colspan="\d+"`)
  37. rows, _ := regexp.Compile(`rowspan="\d+"`)
  38. border, _ := regexp.Compile(`(border="(\d+)")|(cellpadding="(\d+)")|(cellspacing="(\d+)")`)
  39. dis, _ := regexp.Compile(`display:none`)
  40. return &Cut{
  41. annotate: at,
  42. tag: t,
  43. scripttag: scs,
  44. hiddentag: hiddentag,
  45. inputag: input,
  46. colstag: cols,
  47. isborder: border,
  48. rowstag: rows,
  49. display: dis,
  50. multiCR: m,
  51. replBlankLine: regexp.MustCompile("\\s+[\r\n]"),
  52. replStartWrap: regexp.MustCompile("^[\u3000\u2003\u00a0\\s]+|[\u3000\u2003\u00a0\\s]+$"),
  53. replTags2CR: []string{"div", "p", "br", "h1", "h2", "h3", "h4", "h5"},
  54. retainTags2CR: []string{"table", "thead", "tfoot", "tbody", "th", "td", "tr"},
  55. }
  56. }
  57. //清理HTML标签
  58. func (c *Cut) ClearHtml(src string) string {
  59. src = c.replBlankLine.ReplaceAllString(src, "")
  60. src = strings.Replace(src, ">\n", ">", -1)
  61. src = strings.Replace(src, " ", "", -1)
  62. //标签全转小写
  63. src = c.tag.ReplaceAllStringFunc(src, strings.ToLower)
  64. //清script,style
  65. src = c.scripttag.ReplaceAllString(src, "")
  66. //清理注释文本
  67. src = c.annotate.ReplaceAllString(src, "")
  68. //清理input
  69. src = c.hiddentag.ReplaceAllString(src, "")
  70. src = c.inputag.ReplaceAllString(src, "$2")
  71. document, err := goquery.NewDocumentFromReader(strings.NewReader(src))
  72. if err == nil {
  73. if tmpstr, err := document.Each(func(i int, sel *goquery.Selection) {
  74. sel.Find("td").Each(func(i int, selection *goquery.Selection) {
  75. val, b := selection.Attr("title")
  76. if b && strings.Trim(val, " ") != "" {
  77. tmpstr := strings.TrimFunc(selection.Text(), func(r rune) bool {
  78. return r == 9 || r == 32
  79. })
  80. if utf8.RuneCountInString(strings.Trim(tmpstr, " ")) < utf8.RuneCountInString(strings.Trim(val, " ")) {
  81. selection.SetText(strings.Trim(val, " "))
  82. }
  83. }
  84. })
  85. }).Html(); err == nil {
  86. src = tmpstr
  87. }
  88. }
  89. //换结束标签
  90. src = c.tag.ReplaceAllStringFunc(src, func(tmp string) string {
  91. tmp = strings.Replace(tmp, " ", "", -1)
  92. //保留这些标签
  93. for _, v := range c.retainTags2CR {
  94. if "<"+v+">" == tmp || "</"+v+">" == tmp {
  95. if tmp == "</table>" {
  96. return tmp + "\n"
  97. }
  98. return tmp
  99. }
  100. if strings.HasPrefix(tmp, "<"+v) {
  101. dispstrs := c.display.FindAllString(tmp, -1)
  102. rowstrs := c.rowstag.FindAllString(tmp, -1)
  103. colstrs := c.colstag.FindAllString(tmp, -1)
  104. con := "<" + v
  105. if con == "<table" {
  106. if isHasBoder(tmp, c.isborder) {
  107. con = con + ` border="1"`
  108. }
  109. }
  110. if len(colstrs) > 0 { //处理多列合并
  111. con += " " + colstrs[0]
  112. }
  113. if len(rowstrs) > 0 { //处理多行合并
  114. con += " " + rowstrs[0]
  115. }
  116. if len(dispstrs) > 0 {
  117. con += " style=\"" + dispstrs[0] + "\""
  118. }
  119. return con + ">"
  120. }
  121. }
  122. if tmp == "<br>" || tmp == "<br/>" || tmp == "<center>" || tmp == "</center>" || tmp == "<ul>" || tmp == "</ul>" {
  123. return "\n"
  124. }
  125. if tmp[1] != 47 { //开始标签
  126. for _, v := range c.replTags2CR {
  127. if v == tmp[1:len(tmp)-1] {
  128. return "\n"
  129. }
  130. }
  131. return ""
  132. }
  133. for _, v := range c.replTags2CR {
  134. if v == tmp[2:len(tmp)-1] {
  135. return "\n"
  136. }
  137. }
  138. return ""
  139. })
  140. src = c.replStartWrap.ReplaceAllString(src, "")
  141. src = c.replBlankLine.ReplaceAllString(src, "\n")
  142. //清除多余换行
  143. c.multiCR.ReplaceAllString(src, "\n")
  144. return strings.Replace(src, "\n", "<br/>", -1)
  145. }
  146. //判断table是否加表格线
  147. func isHasBoder(con string, reg *regexp.Regexp) bool {
  148. res := reg.FindAllStringSubmatch(con, -1)
  149. hasBorder := false
  150. for _, v := range res {
  151. for k, val := range v {
  152. if k > 0 && k%2 == 0 && IntAll(val) > 0 {
  153. hasBorder = true
  154. break
  155. }
  156. }
  157. }
  158. return hasBorder
  159. }