1   
  2  from configParser import C3Object 
  3  from baseObjects import Parser 
  4  from xml.dom.minidom import parseString as domParseString 
  5  from record import SaxRecord, MinidomRecord, FtDomRecord, SaxContentHandler, LxmlRecord 
  6   
  7  from xml.sax import ContentHandler, make_parser, parseString as saxParseString, ErrorHandler, InputSource as SaxInput 
  8  from xml.sax.saxutils import escape 
  9   
 10   
 11  from utils import flattenTexts, elementType 
 12  import re 
 13  import cStringIO, StringIO 
 14   
 15   
 16  # utility function to update data on record from document 
 17       
 18   
 19 -class BaseParser(Parser): 
 20 -    def copy_data(self, doc, rec): 
 21          rec.filename = doc.filename 
 22          rec.schema = doc.schema 
 23          rec.processHistory = doc.processHistory 
 24          rec.processHistory.append(self.id) 
 25          if doc.documentStore: 
 26              rec.parent = ('document', doc.documentStore, doc.id) 
 27          elif doc.parent: 
 28              rec.parent = doc.parent 
 29   
 30   
 31 -class MinidomParser(BaseParser): 
 32      """ Use default Python Minidom implementation to parse document """ 
 33   
 34 -    def process_document(self, session, doc): 
 35          xml = doc.get_raw() 
 36          dom = domParseString(xml) 
 37          rec = MinidomRecord(dom, xml) 
 38          self.copy_data(doc, rec) 
 39          return rec 
 40   
 41   
 42 -class SaxParser(BaseParser): 
 43      """ Default SAX based parser. Creates SaxRecord """ 
 44   
 45 -    def __init__(self, session, parent, config): 
 46          Parser.__init__(self, session, parent, config) 
 47          self.parser = make_parser() 
 48          self.errorHandler = ErrorHandler() 
 49          self.parser.setErrorHandler(self.errorHandler) 
 50          self.inputSource = SaxInput() 
 51          ch = SaxContentHandler() 
 52          self.contentHandler  = ch 
 53          self.parser.setContentHandler(ch) 
 54          self.keepError = 1 
 55   
 56          if (self.get_setting(session, 'namespaces')): 
 57              self.parser.setFeature('http://xml.org/sax/features/namespaces', 1) 
 58          p = self.get_setting(session, 'attrHash') 
 59          if (p): 
 60              l = p.split() 
 61              for i in l: 
 62                  (a,b) = i.split("@") 
 63                  try: 
 64                      ch.hashAttributesNames[a].append(b) 
 65                  except: 
 66                      ch.hashAttributesNames[a] = [b] 
 67          if self.get_setting(session, 'stripWhitespace'): 
 68              ch.stripWS = 1 
 69   
 70 -    def process_document(self, session, doc): 
 71   
 72          xml = doc.get_raw() 
 73          self.inputSource.setByteStream(cStringIO.StringIO(xml))         
 74          ch = self.contentHandler 
 75          ch.reinit() 
 76          try: 
 77              self.parser.parse(self.inputSource) 
 78          except: 
 79              # Splat.  Reset self and reraise 
 80              if self.keepError: 
 81                  # Work out path 
 82                  path = [] 
 83                  for l in ch.pathLines: 
 84                      line = ch.currentText[l] 
 85                      elemName = line[2:line.index('{')-1] 
 86                      path.append("%s[@SAXID='%s']" % (elemName, l)) 
 87                  self.errorPath = '/'.join(path) 
 88              else: 
 89                  ch.reinit() 
 90                   
 91              raise         
 92          rec = SaxRecord(ch.currentText, xml, recordSize=ch.recordSize) 
 93          rec.elementHash = ch.elementHash 
 94          self.copy_data(doc, rec) 
 95          ch.reinit() 
 96          return rec 
 97   
 98  try: 
 99      from lxml import etree 
100   
101      class LxmlParser(BaseParser): 
102          """ lxml based Parser.  Creates LxmlRecords """ 
103          def process_document(self, session, doc): 
104              # input must be stream 
105              data = doc.get_raw() 
106              et = etree.XML(data) 
107              rec = LxmlRecord(et) 
108              self.copy_data(doc, rec) 
109              return rec