lib-python.2.7.xml.sax.expatreader.py Maven / Gradle / Ivy
Go to download
Show more of this group Show more artifacts with this name
Show all versions of jython-installer Show documentation
Show all versions of jython-installer Show documentation
Jython is an implementation of the high-level, dynamic, object-oriented
language Python written in 100% Pure Java, and seamlessly integrated with
the Java platform. It thus allows you to run Python on any Java platform.
"""
SAX driver for the pyexpat C module. This driver works with
pyexpat.__version__ == '2.22'.
"""
version = "0.20"
from xml.sax._exceptions import *
from xml.sax.handler import feature_validation, feature_namespaces
from xml.sax.handler import feature_namespace_prefixes
from xml.sax.handler import feature_external_ges, feature_external_pes
from xml.sax.handler import feature_string_interning
from xml.sax.handler import property_xml_string, property_interning_dict
# xml.parsers.expat does not raise ImportError in Jython
import sys
if sys.platform[:4] == "java":
raise SAXReaderNotAvailable("expat not available in Java", None)
del sys
try:
from xml.parsers import expat
except ImportError:
raise SAXReaderNotAvailable("expat not supported", None)
else:
if not hasattr(expat, "ParserCreate"):
raise SAXReaderNotAvailable("expat not supported", None)
from xml.sax import xmlreader, saxutils, handler
AttributesImpl = xmlreader.AttributesImpl
AttributesNSImpl = xmlreader.AttributesNSImpl
# If we're using a sufficiently recent version of Python, we can use
# weak references to avoid cycles between the parser and content
# handler, otherwise we'll just have to pretend.
try:
import _weakref
except ImportError:
def _mkproxy(o):
return o
else:
import weakref
_mkproxy = weakref.proxy
del weakref, _weakref
class _ClosedParser:
pass
# --- ExpatLocator
class ExpatLocator(xmlreader.Locator):
"""Locator for use with the ExpatParser class.
This uses a weak reference to the parser object to avoid creating
a circular reference between the parser and the content handler.
"""
def __init__(self, parser):
self._ref = _mkproxy(parser)
def getColumnNumber(self):
parser = self._ref
if parser._parser is None:
return None
return parser._parser.ErrorColumnNumber
def getLineNumber(self):
parser = self._ref
if parser._parser is None:
return 1
return parser._parser.ErrorLineNumber
def getPublicId(self):
parser = self._ref
if parser is None:
return None
return parser._source.getPublicId()
def getSystemId(self):
parser = self._ref
if parser is None:
return None
return parser._source.getSystemId()
# --- ExpatParser
class ExpatParser(xmlreader.IncrementalParser, xmlreader.Locator):
"""SAX driver for the pyexpat C module."""
def __init__(self, namespaceHandling=0, bufsize=2**16-20):
xmlreader.IncrementalParser.__init__(self, bufsize)
self._source = xmlreader.InputSource()
self._parser = None
self._namespaces = namespaceHandling
self._lex_handler_prop = None
self._parsing = 0
self._entity_stack = []
self._external_ges = 1
self._interning = None
# XMLReader methods
def parse(self, source):
"Parse an XML document from a URL or an InputSource."
source = saxutils.prepare_input_source(source)
self._source = source
self.reset()
self._cont_handler.setDocumentLocator(ExpatLocator(self))
xmlreader.IncrementalParser.parse(self, source)
def prepareParser(self, source):
if source.getSystemId() is not None:
base = source.getSystemId()
if isinstance(base, unicode):
base = base.encode('utf-8')
self._parser.SetBase(base)
# Redefined setContentHandler to allow changing handlers during parsing
def setContentHandler(self, handler):
xmlreader.IncrementalParser.setContentHandler(self, handler)
if self._parsing:
self._reset_cont_handler()
def getFeature(self, name):
if name == feature_namespaces:
return self._namespaces
elif name == feature_string_interning:
return self._interning is not None
elif name in (feature_validation, feature_external_pes,
feature_namespace_prefixes):
return 0
elif name == feature_external_ges:
return self._external_ges
raise SAXNotRecognizedException("Feature '%s' not recognized" % name)
def setFeature(self, name, state):
if self._parsing:
raise SAXNotSupportedException("Cannot set features while parsing")
if name == feature_namespaces:
self._namespaces = state
elif name == feature_external_ges:
self._external_ges = state
elif name == feature_string_interning:
if state:
if self._interning is None:
self._interning = {}
else:
self._interning = None
elif name == feature_validation:
if state:
raise SAXNotSupportedException(
"expat does not support validation")
elif name == feature_external_pes:
if state:
raise SAXNotSupportedException(
"expat does not read external parameter entities")
elif name == feature_namespace_prefixes:
if state:
raise SAXNotSupportedException(
"expat does not report namespace prefixes")
else:
raise SAXNotRecognizedException(
"Feature '%s' not recognized" % name)
def getProperty(self, name):
if name == handler.property_lexical_handler:
return self._lex_handler_prop
elif name == property_interning_dict:
return self._interning
elif name == property_xml_string:
if self._parser:
if hasattr(self._parser, "GetInputContext"):
return self._parser.GetInputContext()
else:
raise SAXNotRecognizedException(
"This version of expat does not support getting"
" the XML string")
else:
raise SAXNotSupportedException(
"XML string cannot be returned when not parsing")
raise SAXNotRecognizedException("Property '%s' not recognized" % name)
def setProperty(self, name, value):
if name == handler.property_lexical_handler:
self._lex_handler_prop = value
if self._parsing:
self._reset_lex_handler_prop()
elif name == property_interning_dict:
self._interning = value
elif name == property_xml_string:
raise SAXNotSupportedException("Property '%s' cannot be set" %
name)
else:
raise SAXNotRecognizedException("Property '%s' not recognized" %
name)
# IncrementalParser methods
def feed(self, data, isFinal = 0):
if not self._parsing:
self.reset()
self._parsing = 1
self._cont_handler.startDocument()
try:
# The isFinal parameter is internal to the expat reader.
# If it is set to true, expat will check validity of the entire
# document. When feeding chunks, they are not normally final -
# except when invoked from close.
self._parser.Parse(data, isFinal)
except expat.error, e:
exc = SAXParseException(expat.ErrorString(e.code), e, self)
# FIXME: when to invoke error()?
self._err_handler.fatalError(exc)
def close(self):
if (self._entity_stack or self._parser is None or
isinstance(self._parser, _ClosedParser)):
# If we are completing an external entity, do nothing here
return
try:
self.feed("", isFinal = 1)
self._cont_handler.endDocument()
self._parsing = 0
# break cycle created by expat handlers pointing to our methods
self._parser = None
finally:
self._parsing = 0
if self._parser is not None:
# Keep ErrorColumnNumber and ErrorLineNumber after closing.
parser = _ClosedParser()
parser.ErrorColumnNumber = self._parser.ErrorColumnNumber
parser.ErrorLineNumber = self._parser.ErrorLineNumber
self._parser = parser
def _reset_cont_handler(self):
self._parser.ProcessingInstructionHandler = \
self._cont_handler.processingInstruction
self._parser.CharacterDataHandler = self._cont_handler.characters
def _reset_lex_handler_prop(self):
lex = self._lex_handler_prop
parser = self._parser
if lex is None:
parser.CommentHandler = None
parser.StartCdataSectionHandler = None
parser.EndCdataSectionHandler = None
parser.StartDoctypeDeclHandler = None
parser.EndDoctypeDeclHandler = None
else:
parser.CommentHandler = lex.comment
parser.StartCdataSectionHandler = lex.startCDATA
parser.EndCdataSectionHandler = lex.endCDATA
parser.StartDoctypeDeclHandler = self.start_doctype_decl
parser.EndDoctypeDeclHandler = lex.endDTD
def reset(self):
if self._namespaces:
self._parser = expat.ParserCreate(self._source.getEncoding(), " ",
intern=self._interning)
self._parser.namespace_prefixes = 1
self._parser.StartElementHandler = self.start_element_ns
self._parser.EndElementHandler = self.end_element_ns
else:
self._parser = expat.ParserCreate(self._source.getEncoding(),
intern = self._interning)
self._parser.StartElementHandler = self.start_element
self._parser.EndElementHandler = self.end_element
self._reset_cont_handler()
self._parser.UnparsedEntityDeclHandler = self.unparsed_entity_decl
self._parser.NotationDeclHandler = self.notation_decl
self._parser.StartNamespaceDeclHandler = self.start_namespace_decl
self._parser.EndNamespaceDeclHandler = self.end_namespace_decl
self._decl_handler_prop = None
if self._lex_handler_prop:
self._reset_lex_handler_prop()
# self._parser.DefaultHandler =
# self._parser.DefaultHandlerExpand =
# self._parser.NotStandaloneHandler =
self._parser.ExternalEntityRefHandler = self.external_entity_ref
try:
self._parser.SkippedEntityHandler = self.skipped_entity_handler
except AttributeError:
# This pyexpat does not support SkippedEntity
pass
self._parser.SetParamEntityParsing(
expat.XML_PARAM_ENTITY_PARSING_UNLESS_STANDALONE)
self._parsing = 0
self._entity_stack = []
# Locator methods
def getColumnNumber(self):
if self._parser is None:
return None
return self._parser.ErrorColumnNumber
def getLineNumber(self):
if self._parser is None:
return 1
return self._parser.ErrorLineNumber
def getPublicId(self):
return self._source.getPublicId()
def getSystemId(self):
return self._source.getSystemId()
# event handlers
def start_element(self, name, attrs):
self._cont_handler.startElement(name, AttributesImpl(attrs))
def end_element(self, name):
self._cont_handler.endElement(name)
def start_element_ns(self, name, attrs):
pair = name.split()
if len(pair) == 1:
# no namespace
pair = (None, name)
elif len(pair) == 3:
pair = pair[0], pair[1]
else:
# default namespace
pair = tuple(pair)
newattrs = {}
qnames = {}
for (aname, value) in attrs.items():
parts = aname.split()
length = len(parts)
if length == 1:
# no namespace
qname = aname
apair = (None, aname)
elif length == 3:
qname = "%s:%s" % (parts[2], parts[1])
apair = parts[0], parts[1]
else:
# default namespace
qname = parts[1]
apair = tuple(parts)
newattrs[apair] = value
qnames[apair] = qname
self._cont_handler.startElementNS(pair, None,
AttributesNSImpl(newattrs, qnames))
def end_element_ns(self, name):
pair = name.split()
if len(pair) == 1:
pair = (None, name)
elif len(pair) == 3:
pair = pair[0], pair[1]
else:
pair = tuple(pair)
self._cont_handler.endElementNS(pair, None)
# this is not used (call directly to ContentHandler)
def processing_instruction(self, target, data):
self._cont_handler.processingInstruction(target, data)
# this is not used (call directly to ContentHandler)
def character_data(self, data):
self._cont_handler.characters(data)
def start_namespace_decl(self, prefix, uri):
self._cont_handler.startPrefixMapping(prefix, uri)
def end_namespace_decl(self, prefix):
self._cont_handler.endPrefixMapping(prefix)
def start_doctype_decl(self, name, sysid, pubid, has_internal_subset):
self._lex_handler_prop.startDTD(name, pubid, sysid)
def unparsed_entity_decl(self, name, base, sysid, pubid, notation_name):
self._dtd_handler.unparsedEntityDecl(name, pubid, sysid, notation_name)
def notation_decl(self, name, base, sysid, pubid):
self._dtd_handler.notationDecl(name, pubid, sysid)
def external_entity_ref(self, context, base, sysid, pubid):
if not self._external_ges:
return 1
source = self._ent_handler.resolveEntity(pubid, sysid)
source = saxutils.prepare_input_source(source,
self._source.getSystemId() or
"")
self._entity_stack.append((self._parser, self._source))
self._parser = self._parser.ExternalEntityParserCreate(context)
self._source = source
try:
xmlreader.IncrementalParser.parse(self, source)
except:
return 0 # FIXME: save error info here?
(self._parser, self._source) = self._entity_stack[-1]
del self._entity_stack[-1]
return 1
def skipped_entity_handler(self, name, is_pe):
if is_pe:
# The SAX spec requires to report skipped PEs with a '%'
name = '%'+name
self._cont_handler.skippedEntity(name)
# ---
def create_parser(*args, **kwargs):
return ExpatParser(*args, **kwargs)
# ---
if __name__ == "__main__":
import xml.sax.saxutils
p = create_parser()
p.setContentHandler(xml.sax.saxutils.XMLGenerator())
p.setErrorHandler(xml.sax.ErrorHandler())
p.parse("http://www.ibiblio.org/xml/examples/shakespeare/hamlet.xml")