| 1 |
1708
|
aaronmk
|
# XML parsing
|
| 2 |
|
|
|
| 3 |
1712
|
aaronmk
|
import re
|
| 4 |
1708
|
aaronmk
|
import xml.dom.minidom as minidom
|
| 5 |
1712
|
aaronmk
|
import xml.parsers.expat as expat
|
| 6 |
1708
|
aaronmk
|
|
| 7 |
1757
|
aaronmk
|
import exc
|
| 8 |
1712
|
aaronmk
|
import streams
|
| 9 |
|
|
|
| 10 |
1708
|
aaronmk
|
def parse_str(str_): return minidom.parseString(str_).documentElement
|
| 11 |
1712
|
aaronmk
|
|
| 12 |
|
|
class ConsecXmlInputStream(streams.TracedStream):
|
| 13 |
|
|
'''Wraps an input stream, inserting EOF after each consecutive XML document
|
| 14 |
|
|
'''
|
| 15 |
|
|
def __init__(self, stream):
|
| 16 |
|
|
self.eof = False
|
| 17 |
|
|
self.root_close_tag = None
|
| 18 |
|
|
|
| 19 |
|
|
def trace(line):
|
| 20 |
|
|
if self.root_close_tag == None: # before XML document
|
| 21 |
|
|
match = re.match(r'^<(?!\?)([^\s>]+)', line)
|
| 22 |
|
|
if match: self.root_close_tag = '</'+match.group(1)+'>'
|
| 23 |
|
|
else: # inside XML document
|
| 24 |
|
|
if line.rstrip().endswith(self.root_close_tag):
|
| 25 |
|
|
self.root_close_tag = None # now outside XML document
|
| 26 |
|
|
self.eof = True # next read will produce EOF
|
| 27 |
|
|
|
| 28 |
|
|
streams.TracedStream.__init__(self, trace, stream)
|
| 29 |
|
|
|
| 30 |
|
|
def readline(self):
|
| 31 |
|
|
if self.eof:
|
| 32 |
|
|
self.eof = False
|
| 33 |
|
|
return '' # don't read from underlying stream
|
| 34 |
|
|
else: return streams.TracedStream.readline(self)
|
| 35 |
|
|
|
| 36 |
|
|
def read(self, n): return self.readline() # forward all reads to readline()
|
| 37 |
|
|
|
| 38 |
1757
|
aaronmk
|
def parse_next(stream, on_error=exc.raise_):
|
| 39 |
1712
|
aaronmk
|
'''Parses the next document in a consecutive sequence of XML documents.
|
| 40 |
|
|
@return The next document's root, or None if no more documents available
|
| 41 |
|
|
'''
|
| 42 |
|
|
try: return minidom.parse(ConsecXmlInputStream(stream)).documentElement
|
| 43 |
|
|
except expat.ExpatError, e:
|
| 44 |
1757
|
aaronmk
|
# Save in case another exception raised, overwriting sys.exc_info()
|
| 45 |
|
|
exc.add_traceback(e)
|
| 46 |
1712
|
aaronmk
|
if str(e).startswith('no element found:'): return None # no more docs
|
| 47 |
1757
|
aaronmk
|
else: on_error(e)
|
| 48 |
1712
|
aaronmk
|
|
| 49 |
1757
|
aaronmk
|
def docs_iter(stream, *args, **kw_args):
|
| 50 |
|
|
return iter(lambda: parse_next(stream, *args, **kw_args), None)
|
