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") and 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) and isElementMatchPattern(link, patterns)
  47       candidateLinks.push(link)
  48
  49   return if candidateLinks.length == 0
  50
  51   for link in candidateLinks
  52     link.wordCount = link.textContent.trim().split(/\s+/).length
  53
  54   # We can use this trick to ensure that Array.sort is stable. We need this property to retain the reverse
  55   # in-page order of the links.
  56   candidateLinks.forEach((a,i) -> a.originalIndex = i)
  57
  58   # favor shorter links, and ignore those that are more than one word longer than the shortest link
  59   candidateLinks =
  60     candidateLinks.sort((a, b) ->
  61       if a.wordCount == b.wordCount
  62         a.originalIndex - b.originalIndex
  63       else
  64         a.wordCount - b.wordCount
  65     ).filter((a) -> a.wordCount <= candidateLinks[0].wordCount + 1)
  66
  67   for pattern in patterns
  68     exactWordRegex =
  69       if /\b/.test(pattern[0]) or /\b/.test(pattern[pattern.length - 1])
  70         new RegExp '\\b' + pattern + '\\b', 'i'
  71       else
  72         new RegExp pattern, 'i'
  73
  74     for candidateLink in candidateLinks
  75       if exactWordRegex.test(candidateLink.textContent)
  76         return candidateLink
  77   null
  78
  79
  80 # Find a followable link match ref or patterns
  81 find = (document, ref, patterns) ->
  82   findLinkRef(document, ref) or findLinkPattern(document, patterns)
  83
  84
  85 # Returns elements that qualify as links
  86 # Generates and memoizes an XPath query internally
  87 getLinkElements = do ->
  88   # Some preparations done on startup
  89   elements = [
  90     LINK_ELEMENTS...
  91     "*[#{ LINK_ELEMENT_PROPERTIES.join(' or ') }]"
  92   ]
  93
  94   reduce = (m, rule) -> m.concat(["//#{ rule }", "//xhtml:#{ rule }"])
  95   xpath = elements.reduce(reduce, []).join(' | ')
  96
  97   namespaceResolver = (namespace) ->
  98     if namespace == 'xhtml' then 'http://www.w3.org/1999/xhtml' else null
  99
 100   # The actual function that will return the desired elements
 101   return (document, resultType = XPathResult.ORDERED_NODE_SNAPSHOT_TYPE) ->
 102     return document.evaluate(xpath, document.documentElement, namespaceResolver, resultType, null)
 103
 104
 105 # Determine if the link is visible
 106 isVisibleElement = (element) ->
 107   document = element.ownerDocument
 108   window   = document.defaultView
 109
 110   # element that isn't visible on the page
 111   computedStyle = window.getComputedStyle(element, null)
 112   if computedStyle.getPropertyValue('visibility') != 'visible' or
 113       computedStyle.getPropertyValue('display') == 'none' or
 114       computedStyle.getPropertyValue('opacity') == '0'
 115     return false
 116
 117   # element that has 0 dimension
 118   clientRect = element.getBoundingClientRect()
 119   if clientRect.width == 0 or clientRect.height == 0
 120     return false
 121
 122   true
 123
 124
 125 # Determine if the link has a pattern matched
 126 isElementMatchPattern = (element, patterns) ->
 127   for pattern in patterns
 128     if element.textContent.toLowerCase().indexOf(pattern) != -1
 129       return true
 130   false
 131
 132
 133 exports.find = find