cpython/Lib/test/test_pyexpat.py

314 lines
10 KiB
Python
Raw Normal View History

# Very simple test - Parse a file and print what happens
# XXX TypeErrors on calling handlers, or on bad return values from a
# handler, are obscure and unhelpful.
import pyexpat
from xml.parsers import expat
from test.test_support import sortdict, TestFailed
Get rid of the superstitious "~" in dict hashing's "i = (~hash) & mask". The comment following used to say: /* We use ~hash instead of hash, as degenerate hash functions, such as for ints <sigh>, can have lots of leading zeros. It's not really a performance risk, but better safe than sorry. 12-Dec-00 tim: so ~hash produces lots of leading ones instead -- what's the gain? */ That is, there was never a good reason for doing it. And to the contrary, as explained on Python-Dev last December, it tended to make the *sum* (i + incr) & mask (which is the first table index examined in case of collison) the same "too often" across distinct hashes. Changing to the simpler "i = hash & mask" reduced the number of string-dict collisions (== # number of times we go around the lookup for-loop) from about 6 million to 5 million during a full run of the test suite (these are approximate because the test suite does some random stuff from run to run). The number of collisions in non-string dicts also decreased, but not as dramatically. Note that this may, for a given dict, change the order (wrt previous releases) of entries exposed by .keys(), .values() and .items(). A number of std tests suffered bogus failures as a result. For dicts keyed by small ints, or (less so) by characters, the order is much more likely to be in increasing order of key now; e.g., >>> d = {} >>> for i in range(10): ... d[i] = i ... >>> d {0: 0, 1: 1, 2: 2, 3: 3, 4: 4, 5: 5, 6: 6, 7: 7, 8: 8, 9: 9} >>> Unfortunately. people may latch on to that in small examples and draw a bogus conclusion. test_support.py Moved test_extcall's sortdict() into test_support, made it stronger, and imported sortdict into other std tests that needed it. test_unicode.py Excluced cp875 from the "roundtrip over range(128)" test, because cp875 doesn't have a well-defined inverse for unicode("?", "cp875"). See Python-Dev for excruciating details. Cookie.py Chaged various output functions to sort dicts before building strings from them. test_extcall Fiddled the expected-result file. This remains sensitive to native dict ordering, because, e.g., if there are multiple errors in a keyword-arg dict (and test_extcall sets up many cases like that), the specific error Python complains about first depends on native dict ordering.
2001-05-12 21:19:31 -03:00
class Outputter:
def StartElementHandler(self, name, attrs):
Get rid of the superstitious "~" in dict hashing's "i = (~hash) & mask". The comment following used to say: /* We use ~hash instead of hash, as degenerate hash functions, such as for ints <sigh>, can have lots of leading zeros. It's not really a performance risk, but better safe than sorry. 12-Dec-00 tim: so ~hash produces lots of leading ones instead -- what's the gain? */ That is, there was never a good reason for doing it. And to the contrary, as explained on Python-Dev last December, it tended to make the *sum* (i + incr) & mask (which is the first table index examined in case of collison) the same "too often" across distinct hashes. Changing to the simpler "i = hash & mask" reduced the number of string-dict collisions (== # number of times we go around the lookup for-loop) from about 6 million to 5 million during a full run of the test suite (these are approximate because the test suite does some random stuff from run to run). The number of collisions in non-string dicts also decreased, but not as dramatically. Note that this may, for a given dict, change the order (wrt previous releases) of entries exposed by .keys(), .values() and .items(). A number of std tests suffered bogus failures as a result. For dicts keyed by small ints, or (less so) by characters, the order is much more likely to be in increasing order of key now; e.g., >>> d = {} >>> for i in range(10): ... d[i] = i ... >>> d {0: 0, 1: 1, 2: 2, 3: 3, 4: 4, 5: 5, 6: 6, 7: 7, 8: 8, 9: 9} >>> Unfortunately. people may latch on to that in small examples and draw a bogus conclusion. test_support.py Moved test_extcall's sortdict() into test_support, made it stronger, and imported sortdict into other std tests that needed it. test_unicode.py Excluced cp875 from the "roundtrip over range(128)" test, because cp875 doesn't have a well-defined inverse for unicode("?", "cp875"). See Python-Dev for excruciating details. Cookie.py Chaged various output functions to sort dicts before building strings from them. test_extcall Fiddled the expected-result file. This remains sensitive to native dict ordering, because, e.g., if there are multiple errors in a keyword-arg dict (and test_extcall sets up many cases like that), the specific error Python complains about first depends on native dict ordering.
2001-05-12 21:19:31 -03:00
print 'Start element:\n\t', repr(name), sortdict(attrs)
def EndElementHandler(self, name):
print 'End element:\n\t', repr(name)
def CharacterDataHandler(self, data):
data = data.strip()
if data:
print 'Character data:'
print '\t', repr(data)
def ProcessingInstructionHandler(self, target, data):
print 'PI:\n\t', repr(target), repr(data)
def StartNamespaceDeclHandler(self, prefix, uri):
print 'NS decl:\n\t', repr(prefix), repr(uri)
def EndNamespaceDeclHandler(self, prefix):
print 'End of NS decl:\n\t', repr(prefix)
def StartCdataSectionHandler(self):
print 'Start of CDATA section'
def EndCdataSectionHandler(self):
print 'End of CDATA section'
def CommentHandler(self, text):
print 'Comment:\n\t', repr(text)
def NotationDeclHandler(self, *args):
name, base, sysid, pubid = args
print 'Notation declared:', args
def UnparsedEntityDeclHandler(self, *args):
entityName, base, systemId, publicId, notationName = args
print 'Unparsed entity decl:\n\t', args
def NotStandaloneHandler(self, userData):
print 'Not standalone'
return 1
def ExternalEntityRefHandler(self, *args):
context, base, sysId, pubId = args
print 'External entity ref:', args[1:]
return 1
def DefaultHandler(self, userData):
pass
def DefaultHandlerExpand(self, userData):
pass
def confirm(ok):
if ok:
print "OK."
else:
print "Not OK."
out = Outputter()
parser = expat.ParserCreate(namespace_separator='!')
# Test getting/setting returns_unicode
parser.returns_unicode = 0; confirm(parser.returns_unicode == 0)
parser.returns_unicode = 1; confirm(parser.returns_unicode == 1)
parser.returns_unicode = 2; confirm(parser.returns_unicode == 1)
parser.returns_unicode = 0; confirm(parser.returns_unicode == 0)
# Test getting/setting ordered_attributes
parser.ordered_attributes = 0; confirm(parser.ordered_attributes == 0)
parser.ordered_attributes = 1; confirm(parser.ordered_attributes == 1)
parser.ordered_attributes = 2; confirm(parser.ordered_attributes == 1)
parser.ordered_attributes = 0; confirm(parser.ordered_attributes == 0)
# Test getting/setting specified_attributes
parser.specified_attributes = 0; confirm(parser.specified_attributes == 0)
parser.specified_attributes = 1; confirm(parser.specified_attributes == 1)
parser.specified_attributes = 2; confirm(parser.specified_attributes == 1)
parser.specified_attributes = 0; confirm(parser.specified_attributes == 0)
HANDLER_NAMES = [
'StartElementHandler', 'EndElementHandler',
'CharacterDataHandler', 'ProcessingInstructionHandler',
'UnparsedEntityDeclHandler', 'NotationDeclHandler',
'StartNamespaceDeclHandler', 'EndNamespaceDeclHandler',
'CommentHandler', 'StartCdataSectionHandler',
'EndCdataSectionHandler',
'DefaultHandler', 'DefaultHandlerExpand',
#'NotStandaloneHandler',
'ExternalEntityRefHandler'
]
for name in HANDLER_NAMES:
setattr(parser, name, getattr(out, name))
data = '''\
<?xml version="1.0" encoding="iso-8859-1" standalone="no"?>
<?xml-stylesheet href="stylesheet.css"?>
<!-- comment data -->
<!DOCTYPE quotations SYSTEM "quotations.dtd" [
<!ELEMENT root ANY>
<!NOTATION notation SYSTEM "notation.jpeg">
<!ENTITY acirc "&#226;">
<!ENTITY external_entity SYSTEM "entity.file">
<!ENTITY unparsed_entity SYSTEM "entity.file" NDATA notation>
%unparsed_entity;
]>
<root attr1="value1" attr2="value2&#8000;">
<myns:subelement xmlns:myns="http://www.python.org/namespace">
Contents of subelements
</myns:subelement>
<sub2><![CDATA[contents of CDATA section]]></sub2>
&external_entity;
</root>
'''
# Produce UTF-8 output
parser.returns_unicode = 0
try:
parser.Parse(data, 1)
except expat.error:
print '** Error', parser.ErrorCode, expat.ErrorString(parser.ErrorCode)
print '** Line', parser.ErrorLineNumber
print '** Column', parser.ErrorColumnNumber
print '** Byte', parser.ErrorByteIndex
# Try the parse again, this time producing Unicode output
parser = expat.ParserCreate(namespace_separator='!')
parser.returns_unicode = 1
for name in HANDLER_NAMES:
setattr(parser, name, getattr(out, name))
try:
parser.Parse(data, 1)
except expat.error:
print '** Error', parser.ErrorCode, expat.ErrorString(parser.ErrorCode)
print '** Line', parser.ErrorLineNumber
print '** Column', parser.ErrorColumnNumber
print '** Byte', parser.ErrorByteIndex
# Try parsing a file
parser = expat.ParserCreate(namespace_separator='!')
parser.returns_unicode = 1
for name in HANDLER_NAMES:
setattr(parser, name, getattr(out, name))
import StringIO
file = StringIO.StringIO(data)
try:
parser.ParseFile(file)
except expat.error:
print '** Error', parser.ErrorCode, expat.ErrorString(parser.ErrorCode)
print '** Line', parser.ErrorLineNumber
print '** Column', parser.ErrorColumnNumber
print '** Byte', parser.ErrorByteIndex
# Tests that make sure we get errors when the namespace_separator value
# is illegal, and that we don't for good values:
print
print "Testing constructor for proper handling of namespace_separator values:"
expat.ParserCreate()
expat.ParserCreate(namespace_separator=None)
expat.ParserCreate(namespace_separator=' ')
print "Legal values tested o.k."
try:
expat.ParserCreate(namespace_separator=42)
except TypeError, e:
print "Caught expected TypeError:"
print e
else:
print "Failed to catch expected TypeError."
try:
expat.ParserCreate(namespace_separator='too long')
except ValueError, e:
print "Caught expected ValueError:"
print e
else:
print "Failed to catch expected ValueError."
# ParserCreate() needs to accept a namespace_separator of zero length
# to satisfy the requirements of RDF applications that are required
# to simply glue together the namespace URI and the localname. Though
# considered a wart of the RDF specifications, it needs to be supported.
#
# See XML-SIG mailing list thread starting with
# http://mail.python.org/pipermail/xml-sig/2001-April/005202.html
#
expat.ParserCreate(namespace_separator='') # too short
# Test the interning machinery.
p = expat.ParserCreate()
L = []
def collector(name, *args):
L.append(name)
p.StartElementHandler = collector
p.EndElementHandler = collector
p.Parse("<e> <e/> <e></e> </e>", 1)
tag = L[0]
if len(L) != 6:
print "L should only contain 6 entries; found", len(L)
for entry in L:
if tag is not entry:
print "expected L to contain many references to the same string",
print "(it didn't)"
print "L =", `L`
break
# Tests of the buffer_text attribute.
import sys
class TextCollector:
def __init__(self, parser):
self.stuff = []
def check(self, expected, label):
require(self.stuff == expected,
"%s\nstuff = %s\nexpected = %s"
% (label, `self.stuff`, `map(unicode, expected)`))
def CharacterDataHandler(self, text):
self.stuff.append(text)
def StartElementHandler(self, name, attrs):
self.stuff.append("<%s>" % name)
bt = attrs.get("buffer-text")
if bt == "yes":
parser.buffer_text = 1
elif bt == "no":
parser.buffer_text = 0
def EndElementHandler(self, name):
self.stuff.append("</%s>" % name)
def CommentHandler(self, data):
self.stuff.append("<!--%s-->" % data)
def require(cond, label):
# similar to confirm(), but no extraneous output
if not cond:
raise TestFailed(label)
def setup(handlers=[]):
parser = expat.ParserCreate()
require(not parser.buffer_text,
"buffer_text not disabled by default")
parser.buffer_text = 1
handler = TextCollector(parser)
parser.CharacterDataHandler = handler.CharacterDataHandler
for name in handlers:
setattr(parser, name, getattr(handler, name))
return parser, handler
parser, handler = setup()
require(parser.buffer_text,
"text buffering either not acknowledged or not enabled")
parser.Parse("<a>1<b/>2<c/>3</a>", 1)
handler.check(["123"],
"buffered text not properly collapsed")
# XXX This test exposes more detail of Expat's text chunking than we
# XXX like, but it tests what we need to concisely.
parser, handler = setup(["StartElementHandler"])
parser.Parse("<a>1<b buffer-text='no'/>2\n3<c buffer-text='yes'/>4\n5</a>", 1)
handler.check(["<a>", "1", "<b>", "2", "\n", "3", "<c>", "4\n5"],
"buffering control not reacting as expected")
parser, handler = setup()
parser.Parse("<a>1<b/>&lt;2&gt;<c/>&#32;\n&#x20;3</a>", 1)
handler.check(["1<2> \n 3"],
"buffered text not properly collapsed")
parser, handler = setup(["StartElementHandler"])
parser.Parse("<a>1<b/>2<c/>3</a>", 1)
handler.check(["<a>", "1", "<b>", "2", "<c>", "3"],
"buffered text not properly split")
parser, handler = setup(["StartElementHandler", "EndElementHandler"])
parser.CharacterDataHandler = None
parser.Parse("<a>1<b/>2<c/>3</a>", 1)
handler.check(["<a>", "<b>", "</b>", "<c>", "</c>", "</a>"],
"huh?")
parser, handler = setup(["StartElementHandler", "EndElementHandler"])
parser.Parse("<a>1<b></b>2<c/>3</a>", 1)
handler.check(["<a>", "1", "<b>", "</b>", "2", "<c>", "</c>", "3", "</a>"],
"huh?")
parser, handler = setup(["CommentHandler", "EndElementHandler",
"StartElementHandler"])
parser.Parse("<a>1<b/>2<c></c>345</a> ", 1)
handler.check(["<a>", "1", "<b>", "</b>", "2", "<c>", "</c>", "345", "</a>"],
"buffered text not properly split")
parser, handler = setup(["CommentHandler", "EndElementHandler",
"StartElementHandler"])
parser.Parse("<a>1<b/>2<c></c>3<!--abc-->4<!--def-->5</a> ", 1)
handler.check(["<a>", "1", "<b>", "</b>", "2", "<c>", "</c>", "3",
"<!--abc-->", "4", "<!--def-->", "5", "</a>"],
"buffered text not properly split")