XMLParse.py

   1 #
   2 # XMLParse.py
   3 #
   4 # Parse arbitrary XML news streams into an object type
   5 # understandable by Planet UCC.
   6 #
   7 # (c) 2004, Davyd Madeley <[email protected]>
   8 #
   9
  10 import sys, time
  11 from xml.dom.minidom import parseString
  12
  13 class Blog:
  14         def __init__(self):
  15                 self.blogTitle  = None
  16                 self.blogURL    = None
  17                 self.imageURL   = None
  18                 self.imageLink  = None
  19                 self.items      = []
  20
  21 class BlogItem:
  22         def __init__(self):
  23                 self.itemTitle  = None
  24                 self.itemDate   = None
  25                 self.itemURL    = None
  26                 self.contents   = None
  27
  28 class XMLParse:
  29         def __init__(self, XMLString):
  30                 # parse our XML file
  31                 self.dom        = parseString(XMLString)
  32                 self.bloglist   = None
  33                 # find out what sort of XML format we're dealing with
  34                 if self.dom.documentElement.tagName == 'rss':
  35                         # this is some sort of RSS feed
  36                         # find out what version
  37                         if self.dom.documentElement.attributes.has_key('version'):
  38                                 version = self.dom.documentElement.attributes['version'].value
  39                                 if version == '2.0':
  40                                         # this is an RSS2 document
  41                                         self.news       = RSS2Parse(self.dom)
  42                                 else:
  43                                         sys.stderr.write('DEBUG: XMLParse: Unknown RSS version %s\n' % version)
  44                         else:
  45                                 sys.stderr.write('DEBUG: XMLParse: RSS document has no version information\n')
  46                 else:
  47                         sys.stderr.write('DEBUG: XMLParse: Unknown XML document %s\n' % dom.documentElement.tagname)
  48
  49         def parse(self):
  50                 "Return a list of Blog objects from the XML file we parsed"
  51                 # quick cache for XML parsing
  52                 if self.bloglist:
  53                         return self.bloglist
  54                 else:
  55                         self.bloglist   = self.news.parse()
  56                         return self.bloglist
  57
  58 class RSS2Parse:
  59         def __init__(self, dom):
  60                 self.dom        = dom
  61                 self.root       = dom.documentElement
  62
  63         def __retrieve_value__(self, fromNode):
  64                 "Returns the value from between two nodes, ie <node>text</node>"
  65                 for node in fromNode.childNodes:
  66                         if node.nodeType == 3:
  67                                 # this is the information contained within our node
  68                                 return node.nodeValue
  69                         else:
  70                                 sys.stderr.write('DEBUG: RSS2Parse: Asked to retrieve value from wrong part of tree\n')
  71                                 return None
  72
  73         def __parse_item__(self, fromNode):
  74                 "Returns a BlogItem collected from fromNode"
  75                 item    = BlogItem()
  76                 for node in fromNode.childNodes:
  77                         if node.nodeType == 1 and node.tagName == 'title':
  78                                 item.itemTitle  = self.__retrieve_value__(node)
  79                         elif node.nodeType == 1 and node.tagName == 'pubDate':
  80                                 try:
  81                                         item.itemDate   = time.mktime(time.strptime(self.__retrieve_value__(node), '%a, %d %b %Y %H:%M:%S %Z'))
  82                                 except:
  83                                         try:
  84                                                 item.itemDate   = time.mktime(time.strptime(self.__retrieve_value__(node), '%a, %d %b %Y %H:%M:%S +0000'))
  85                                         except:
  86                                                 sys.stderr.write("DEBUG: RSS2Parse: time string %s unparseable\n" % (self.__retrieve_value__(node)))
  87                         elif node.nodeType == 1 and node.tagName == 'link':
  88                                 item.itemURL    = self.__retrieve_value__(node)
  89                         elif node.nodeType == 1 and node.tagName == 'description':
  90                                 item.contents   = self.__retrieve_value__(node)
  91                 return item
  92
  93         def parse(self):
  94                 "Returns a list of Blog objects for parsing into an arbitrary data format."
  95                 channellist     = []
  96                 for node in self.root.childNodes:
  97                         if node.nodeType == 1 and node.tagName == 'channel':
  98                                 channel = Blog()
  99                                 channellist.append(channel)
 100                                 # populate channel with information from the blog
 101                                 for node2 in node.childNodes:
 102                                         if node2.nodeType == 1 and node2.tagName == 'title':
 103                                                 channel.blogTitle       = self.__retrieve_value__(node2)
 104                                         elif node2.nodeType == 1 and node2.tagName == 'link':
 105                                                 channel.blogURL         = self.__retrieve_value__(node2)
 106                                         elif node2.nodeType == 1 and node2.tagName == 'image':
 107                                                 for node3 in node2.childNodes:
 108                                                         if node3.nodeType == 1 and node3.tagName == 'url':
 109                                                                 channel.imageURL                = self.__retrieve_value__(node3)
 110                                                         elif node3.nodeType == 1 and node3.tagName == 'link':
 111                                                                 channel.imageLink       = self.__retrieve_value__(node3)
 112                                         elif node2.nodeType == 1 and node2.tagName == 'item':
 113                                                 item    = self.__parse_item__(node2)
 114                                                 channel.items.append(item)
 115                 return channellist