extension/packages/find-link.coffee

   1 { interfaces: Ci } = Components
   2
   3 HTMLDocument = Ci.nsIDOMHTMLDocument
   4 XPathResult  = Ci.nsIDOMXPathResult
   5
   6 # All the following elements qualify as a link
   7 LINK_ELEMENTS = [
   8   "a"
   9   "area[@href]"
  10   "button"
  11 ]
  12
  13 # All elements that have one or more of the following properties
  14 # qualify as a link
  15 LINK_ELEMENT_PROPERTIES = [
  16   "@onclick"
  17   "@onmousedown"
  18   "@onmouseup"
  19   "@oncommand"
  20   "@role='link'"
  21   "@role='button'"
  22   "contains(@class, 'button')"
  23 ]
  24
  25 # Find a link with ref equals value
  26 findLinkRef = (document, value) ->
  27   relTags = ["link", "a", "area"]
  28   for tag in relTags
  29     elements = document.getElementsByTagName(tag)
  30     for element in elements
  31       if (element.hasAttribute("rel") && element.rel == value)
  32         return element
  33   null
  34
  35
  36 # Find a link with match the patterns
  37 findLinkPattern = (document, patterns) ->
  38   links = getLinkElements(document)
  39   candidateLinks = []
  40
  41   # at the end of this loop, candidateLinks will contain all visible links that match our patterns
  42   # links lower in the page are more likely to be the ones we want, so we loop through the snapshot backwards
  43   for i in [0...links.snapshotLength] by 1
  44     link = links.snapshotItem(i)
  45
  46     if (isVisibleElement(link) &&
  47         isElementMatchPattern(link, patterns))
  48       candidateLinks.push(link)
  49
  50   return if (candidateLinks.length == 0)
  51
  52   for link in candidateLinks
  53     link.wordCount = link.textContent.trim().split(/\s+/).length
  54
  55   # We can use this trick to ensure that Array.sort is stable. We need this property to retain the reverse
  56   # in-page order of the links.
  57   candidateLinks.forEach((a,i) -> a.originalIndex = i)
  58
  59   # favor shorter links, and ignore those that are more than one word longer than the shortest link
  60   candidateLinks =
  61     candidateLinks.sort((a, b) ->
  62         if (a.wordCount == b.wordCount)
  63           a.originalIndex - b.originalIndex
  64         else
  65           a.wordCount - b.wordCount
  66       ).filter((a) ->
  67         a.wordCount <= candidateLinks[0].wordCount + 1)
  68
  69   for pattern in patterns
  70     exactWordRegex =
  71       if /\b/.test(pattern[0]) or /\b/.test(pattern[pattern.length - 1])
  72         new RegExp "\\b" + pattern + "\\b", "i"
  73       else
  74         new RegExp pattern, "i"
  75
  76     for candidateLink in candidateLinks
  77       if (exactWordRegex.test(candidateLink.textContent))
  78         return candidateLink
  79   null
  80
  81
  82 # Find a followable link match ref or patterns
  83 find = (document, ref, patterns) ->
  84   findLinkRef(document, ref) || findLinkPattern(document, patterns)
  85
  86
  87 # Returns elements that qualify as links
  88 # Generates and memoizes an XPath query internally
  89 getLinkElements = do ->
  90   # Some preparations done on startup
  91   elements = [
  92     LINK_ELEMENTS...
  93     "*[#{ LINK_ELEMENT_PROPERTIES.join(' or ') }]"
  94   ]
  95
  96   reduce = (m, rule) -> m.concat(["//#{ rule }", "//xhtml:#{ rule }"])
  97   xpath = elements.reduce(reduce, []).join(' | ')
  98
  99   namespaceResolver = (namespace) ->
 100     if namespace == 'xhtml' then 'http://www.w3.org/1999/xhtml' else null
 101
 102   # The actual function that will return the desired elements
 103   return (document, resultType = XPathResult.ORDERED_NODE_SNAPSHOT_TYPE) ->
 104     return document.evaluate(xpath, document.documentElement, namespaceResolver, resultType, null)
 105
 106
 107 # Determine the link is visible
 108 isVisibleElement = (element) ->
 109   document = element.ownerDocument
 110   window   = document.defaultView
 111
 112   # element that isn't visible on the page
 113   computedStyle = window.getComputedStyle(element, null)
 114   if (computedStyle.getPropertyValue('visibility') != 'visible' ||
 115         computedStyle.getPropertyValue('display') == 'none' ||
 116         computedStyle.getPropertyValue('opacity') == '0')
 117       return false
 118
 119   # element that has 0 dimension
 120   clientRect = element.getBoundingClientRect()
 121   if (clientRect.width == 0 || clientRect.height == 0)
 122     return false
 123
 124   true
 125
 126
 127 # Determine the link has a pattern matched
 128 isElementMatchPattern = (element, patterns) ->
 129     for pattern in patterns
 130       if (element.textContent.toLowerCase().indexOf(pattern) != -1)
 131         return true
 132     false
 133
 134
 135 exports.find = find