htmlclear.go 4.8 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166
  1. package util
  2. import (
  3. mc "app.yhyue.com/moapp/jybase/common"
  4. "github.com/PuerkitoBio/goquery"
  5. "regexp"
  6. "strings"
  7. "unicode/utf8"
  8. )
  9. //
  10. type Cut struct {
  11. annotate *regexp.Regexp
  12. tag *regexp.Regexp
  13. scripttag *regexp.Regexp
  14. inputag *regexp.Regexp
  15. isborder *regexp.Regexp
  16. hiddentag *regexp.Regexp
  17. styletag *regexp.Regexp
  18. colstag *regexp.Regexp
  19. rowstag *regexp.Regexp
  20. display *regexp.Regexp
  21. multiCR *regexp.Regexp
  22. replBlankLine *regexp.Regexp
  23. replStartWrap *regexp.Regexp
  24. replTags2CR []string
  25. retainTags2CR []string
  26. }
  27. //
  28. func NewCut() *Cut {
  29. t, _ := regexp.Compile("<[^>]+>")
  30. m, _ := regexp.Compile("([\r\n][\u3000\u2003\u00a0\\s]*)+|[\r\n]+")
  31. //sc, _ := regexp.Compile("\\<script[^\\>]*\\>*[^\\>]+\\</script\\>")
  32. //ss, _ := regexp.Compile("\\<style[^\\>]*\\>*[^\\>]+\\</style\\>")
  33. scs := regexp.MustCompile("(?s)<(script|style)[^>]*>.+?</(script|style)>")
  34. at := regexp.MustCompile("(?s)<(!%-%-|!--).*?(%-%-|--)>") //注释 css
  35. hiddentag := regexp.MustCompile(`<\s*input[^<]*type=("|')hidden("|')[^<]*>`)
  36. input := regexp.MustCompile(`<\s*input[^<]*value=("|')([^>"']*)[^<]*>`)
  37. cols, _ := regexp.Compile(`colspan="\d+"`)
  38. rows, _ := regexp.Compile(`rowspan="\d+"`)
  39. border, _ := regexp.Compile(`(border="(\d+)")|(cellpadding="(\d+)")|(cellspacing="(\d+)")`)
  40. dis, _ := regexp.Compile(`display:none`)
  41. return &Cut{
  42. annotate: at,
  43. tag: t,
  44. scripttag: scs,
  45. hiddentag: hiddentag,
  46. inputag: input,
  47. colstag: cols,
  48. isborder: border,
  49. rowstag: rows,
  50. display: dis,
  51. multiCR: m,
  52. replBlankLine: regexp.MustCompile("\\s+[\r\n]"),
  53. replStartWrap: regexp.MustCompile("^[\u3000\u2003\u00a0\\s]+|[\u3000\u2003\u00a0\\s]+$"),
  54. replTags2CR: []string{"div", "p", "br", "h1", "h2", "h3", "h4", "h5"},
  55. retainTags2CR: []string{"table", "thead", "tfoot", "tbody", "th", "td", "tr"},
  56. }
  57. }
  58. //清理HTML标签
  59. func (c *Cut) ClearHtml(src string) string {
  60. src = c.replBlankLine.ReplaceAllString(src, "")
  61. src = strings.Replace(src, ">\n", ">", -1)
  62. src = strings.Replace(src, " ", "", -1)
  63. //标签全转小写
  64. src = c.tag.ReplaceAllStringFunc(src, strings.ToLower)
  65. //清script,style
  66. src = c.scripttag.ReplaceAllString(src, "")
  67. //清理注释文本
  68. src = c.annotate.ReplaceAllString(src, "")
  69. //清理input
  70. src = c.hiddentag.ReplaceAllString(src, "")
  71. src = c.inputag.ReplaceAllString(src, "$2")
  72. document, err := goquery.NewDocumentFromReader(strings.NewReader(src))
  73. if err == nil {
  74. if tmpstr, err := document.Each(func(i int, sel *goquery.Selection) {
  75. sel.Find("td").Each(func(i int, selection *goquery.Selection) {
  76. val, b := selection.Attr("title")
  77. if b && strings.Trim(val, " ") != "" {
  78. tmpstr := strings.TrimFunc(selection.Text(), func(r rune) bool {
  79. return r == 9 || r == 32
  80. })
  81. if utf8.RuneCountInString(strings.Trim(tmpstr, " ")) < utf8.RuneCountInString(strings.Trim(val, " ")) {
  82. selection.SetText(strings.Trim(val, " "))
  83. }
  84. }
  85. })
  86. }).Html(); err == nil {
  87. src = tmpstr
  88. }
  89. }
  90. //换结束标签
  91. src = c.tag.ReplaceAllStringFunc(src, func(tmp string) string {
  92. tmp = strings.Replace(tmp, " ", "", -1)
  93. //保留这些标签
  94. for _, v := range c.retainTags2CR {
  95. if "<"+v+">" == tmp || "</"+v+">" == tmp {
  96. if tmp == "</table>" {
  97. return tmp + "\n"
  98. }
  99. return tmp
  100. }
  101. if strings.HasPrefix(tmp, "<"+v) {
  102. dispstrs := c.display.FindAllString(tmp, -1)
  103. rowstrs := c.rowstag.FindAllString(tmp, -1)
  104. colstrs := c.colstag.FindAllString(tmp, -1)
  105. con := "<" + v
  106. if con == "<table" {
  107. if isHasBoder(tmp, c.isborder) {
  108. con = con + ` border="1"`
  109. }
  110. }
  111. if len(colstrs) > 0 { //处理多列合并
  112. con += " " + colstrs[0]
  113. }
  114. if len(rowstrs) > 0 { //处理多行合并
  115. con += " " + rowstrs[0]
  116. }
  117. if len(dispstrs) > 0 {
  118. con += " style=\"" + dispstrs[0] + "\""
  119. }
  120. return con + ">"
  121. }
  122. }
  123. if tmp == "<br>" || tmp == "<br/>" || tmp == "<center>" || tmp == "</center>" || tmp == "<ul>" || tmp == "</ul>" {
  124. return "\n"
  125. }
  126. if tmp[1] != 47 { //开始标签
  127. for _, v := range c.replTags2CR {
  128. if v == tmp[1:len(tmp)-1] {
  129. return "\n"
  130. }
  131. }
  132. return ""
  133. }
  134. for _, v := range c.replTags2CR {
  135. if v == tmp[2:len(tmp)-1] {
  136. return "\n"
  137. }
  138. }
  139. return ""
  140. })
  141. src = c.replStartWrap.ReplaceAllString(src, "")
  142. src = c.replBlankLine.ReplaceAllString(src, "\n")
  143. //清除多余换行
  144. c.multiCR.ReplaceAllString(src, "\n")
  145. return strings.Replace(src, "\n", "<br/>", -1)
  146. }
  147. //判断table是否加表格线
  148. func isHasBoder(con string, reg *regexp.Regexp) bool {
  149. res := reg.FindAllStringSubmatch(con, -1)
  150. hasBorder := false
  151. for _, v := range res {
  152. for k, val := range v {
  153. if k > 0 && k%2 == 0 && mc.IntAll(val) > 0 {
  154. hasBorder = true
  155. break
  156. }
  157. }
  158. }
  159. return hasBorder
  160. }