annotate roundup/cgi/TAL/HTMLParser.py @ 1202:01a143f9382e

*** empty log message ***
author Richard Jones <richard@users.sourceforge.net>
date Tue, 24 Sep 2002 05:50:26 +0000
parents b9988e118055
children fc52d57c6c3e
Ignore whitespace changes - Everywhere: Within whitespace: At end of lines:
rev   line source
1049
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
1 """A parser for HTML and XHTML."""
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
2
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
3 # This file is based on sgmllib.py, but the API is slightly different.
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
4
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
5 # XXX There should be a way to distinguish between PCDATA (parsed
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
6 # character data -- the normal case), RCDATA (replaceable character
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
7 # data -- only char and entity references and end tags are special)
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
8 # and CDATA (character data -- only end tags are special).
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
9
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
10
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
11 import markupbase
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
12 import re
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
13 import string
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
14
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
15 # Regular expressions used for parsing
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
16
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
17 interesting_normal = re.compile('[&<]')
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
18 interesting_cdata = re.compile(r'<(/|\Z)')
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
19 incomplete = re.compile('&[a-zA-Z#]')
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
20
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
21 entityref = re.compile('&([a-zA-Z][-.a-zA-Z0-9]*)[^a-zA-Z0-9]')
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
22 charref = re.compile('&#(?:[0-9]+|[xX][0-9a-fA-F]+)[^0-9a-fA-F]')
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
23
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
24 starttagopen = re.compile('<[a-zA-Z]')
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
25 piclose = re.compile('>')
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
26 endtagopen = re.compile('</')
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
27 commentclose = re.compile(r'--\s*>')
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
28 tagfind = re.compile('[a-zA-Z][-.a-zA-Z0-9:_]*')
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
29 attrfind = re.compile(
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
30 r'\s*([a-zA-Z_][-.:a-zA-Z_0-9]*)(\s*=\s*'
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
31 r'(\'[^\']*\'|"[^"]*"|[-a-zA-Z0-9./:;+*%?!&$\(\)_#=~]*))?')
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
32
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
33 locatestarttagend = re.compile(r"""
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
34 <[a-zA-Z][-.a-zA-Z0-9:_]* # tag name
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
35 (?:\s+ # whitespace before attribute name
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
36 (?:[a-zA-Z_][-.:a-zA-Z0-9_]* # attribute name
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
37 (?:\s*=\s* # value indicator
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
38 (?:'[^']*' # LITA-enclosed value
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
39 |\"[^\"]*\" # LIT-enclosed value
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
40 |[^'\">\s]+ # bare value
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
41 )
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
42 )?
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
43 )
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
44 )*
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
45 \s* # trailing whitespace
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
46 """, re.VERBOSE)
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
47 endendtag = re.compile('>')
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
48 endtagfind = re.compile('</\s*([a-zA-Z][-.a-zA-Z0-9:_]*)\s*>')
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
49
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
50
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
51 class HTMLParseError(Exception):
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
52 """Exception raised for all parse errors."""
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
53
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
54 def __init__(self, msg, position=(None, None)):
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
55 assert msg
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
56 self.msg = msg
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
57 self.lineno = position[0]
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
58 self.offset = position[1]
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
59
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
60 def __str__(self):
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
61 result = self.msg
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
62 if self.lineno is not None:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
63 result = result + ", at line %d" % self.lineno
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
64 if self.offset is not None:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
65 result = result + ", column %d" % (self.offset + 1)
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
66 return result
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
67
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
68
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
69 def _contains_at(s, sub, pos):
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
70 return s[pos:pos+len(sub)] == sub
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
71
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
72
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
73 class HTMLParser(markupbase.ParserBase):
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
74 """Find tags and other markup and call handler functions.
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
75
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
76 Usage:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
77 p = HTMLParser()
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
78 p.feed(data)
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
79 ...
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
80 p.close()
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
81
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
82 Start tags are handled by calling self.handle_starttag() or
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
83 self.handle_startendtag(); end tags by self.handle_endtag(). The
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
84 data between tags is passed from the parser to the derived class
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
85 by calling self.handle_data() with the data as argument (the data
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
86 may be split up in arbitrary chunks). Entity references are
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
87 passed by calling self.handle_entityref() with the entity
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
88 reference as the argument. Numeric character references are
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
89 passed to self.handle_charref() with the string containing the
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
90 reference as the argument.
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
91 """
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
92
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
93 CDATA_CONTENT_ELEMENTS = ("script", "style")
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
94
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
95
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
96 def __init__(self):
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
97 """Initialize and reset this instance."""
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
98 self.reset()
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
99
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
100 def reset(self):
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
101 """Reset this instance. Loses all unprocessed data."""
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
102 self.rawdata = ''
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
103 self.stack = []
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
104 self.lasttag = '???'
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
105 self.interesting = interesting_normal
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
106 markupbase.ParserBase.reset(self)
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
107
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
108 def feed(self, data):
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
109 """Feed data to the parser.
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
110
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
111 Call this as often as you want, with as little or as much text
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
112 as you want (may include '\n').
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
113 """
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
114 self.rawdata = self.rawdata + data
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
115 self.goahead(0)
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
116
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
117 def close(self):
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
118 """Handle any buffered data."""
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
119 self.goahead(1)
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
120
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
121 def error(self, message):
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
122 raise HTMLParseError(message, self.getpos())
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
123
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
124 __starttag_text = None
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
125
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
126 def get_starttag_text(self):
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
127 """Return full source of start tag: '<...>'."""
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
128 return self.__starttag_text
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
129
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
130 cdata_endtag = None
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
131
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
132 def set_cdata_mode(self, endtag=None):
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
133 self.cdata_endtag = endtag
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
134 self.interesting = interesting_cdata
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
135
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
136 def clear_cdata_mode(self):
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
137 self.cdata_endtag = None
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
138 self.interesting = interesting_normal
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
139
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
140 # Internal -- handle data as far as reasonable. May leave state
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
141 # and data to be processed by a subsequent call. If 'end' is
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
142 # true, force handling all data as if followed by EOF marker.
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
143 def goahead(self, end):
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
144 rawdata = self.rawdata
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
145 i = 0
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
146 n = len(rawdata)
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
147 while i < n:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
148 match = self.interesting.search(rawdata, i) # < or &
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
149 if match:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
150 j = match.start()
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
151 else:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
152 j = n
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
153 if i < j: self.handle_data(rawdata[i:j])
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
154 i = self.updatepos(i, j)
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
155 if i == n: break
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
156 if rawdata[i] == '<':
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
157 if starttagopen.match(rawdata, i): # < + letter
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
158 k = self.parse_starttag(i)
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
159 elif endtagopen.match(rawdata, i): # </
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
160 k = self.parse_endtag(i)
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
161 elif _contains_at(rawdata, "<!--", i): # <!--
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
162 k = self.parse_comment(i)
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
163 elif _contains_at(rawdata, "<!", i): # <!
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
164 k = self.parse_declaration(i)
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
165 elif _contains_at(rawdata, "<?", i): # <?
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
166 k = self.parse_pi(i)
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
167 elif _contains_at(rawdata, "<?", i): # <!
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
168 k = self.parse_declaration(i)
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
169 elif (i + 1) < n:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
170 self.handle_data("<")
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
171 k = i + 1
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
172 else:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
173 break
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
174 if k < 0:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
175 if end:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
176 self.error("EOF in middle of construct")
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
177 break
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
178 i = self.updatepos(i, k)
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
179 elif rawdata[i:i+2] == "&#":
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
180 match = charref.match(rawdata, i)
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
181 if match:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
182 name = match.group()[2:-1]
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
183 self.handle_charref(name)
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
184 k = match.end()
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
185 if rawdata[k-1] != ';':
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
186 k = k - 1
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
187 i = self.updatepos(i, k)
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
188 continue
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
189 else:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
190 break
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
191 elif rawdata[i] == '&':
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
192 match = entityref.match(rawdata, i)
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
193 if match:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
194 name = match.group(1)
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
195 self.handle_entityref(name)
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
196 k = match.end()
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
197 if rawdata[k-1] != ';':
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
198 k = k - 1
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
199 i = self.updatepos(i, k)
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
200 continue
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
201 match = incomplete.match(rawdata, i)
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
202 if match:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
203 # match.group() will contain at least 2 chars
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
204 rest = rawdata[i:]
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
205 if end and match.group() == rest:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
206 self.error("EOF in middle of entity or char ref")
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
207 # incomplete
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
208 break
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
209 elif (i + 1) < n:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
210 # not the end of the buffer, and can't be confused
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
211 # with some other construct
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
212 self.handle_data("&")
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
213 i = self.updatepos(i, i + 1)
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
214 else:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
215 break
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
216 else:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
217 assert 0, "interesting.search() lied"
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
218 # end while
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
219 if end and i < n:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
220 self.handle_data(rawdata[i:n])
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
221 i = self.updatepos(i, n)
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
222 self.rawdata = rawdata[i:]
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
223
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
224 # Internal -- parse comment, return end or -1 if not terminated
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
225 def parse_comment(self, i, report=1):
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
226 rawdata = self.rawdata
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
227 assert rawdata[i:i+4] == '<!--', 'unexpected call to parse_comment()'
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
228 match = commentclose.search(rawdata, i+4)
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
229 if not match:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
230 return -1
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
231 if report:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
232 j = match.start()
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
233 self.handle_comment(rawdata[i+4: j])
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
234 j = match.end()
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
235 return j
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
236
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
237 # Internal -- parse processing instr, return end or -1 if not terminated
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
238 def parse_pi(self, i):
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
239 rawdata = self.rawdata
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
240 assert rawdata[i:i+2] == '<?', 'unexpected call to parse_pi()'
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
241 match = piclose.search(rawdata, i+2) # >
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
242 if not match:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
243 return -1
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
244 j = match.start()
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
245 self.handle_pi(rawdata[i+2: j])
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
246 j = match.end()
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
247 return j
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
248
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
249 # Internal -- handle starttag, return end or -1 if not terminated
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
250 def parse_starttag(self, i):
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
251 self.__starttag_text = None
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
252 endpos = self.check_for_whole_start_tag(i)
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
253 if endpos < 0:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
254 return endpos
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
255 rawdata = self.rawdata
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
256 self.__starttag_text = rawdata[i:endpos]
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
257
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
258 # Now parse the data between i+1 and j into a tag and attrs
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
259 attrs = []
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
260 match = tagfind.match(rawdata, i+1)
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
261 assert match, 'unexpected call to parse_starttag()'
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
262 k = match.end()
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
263 self.lasttag = tag = string.lower(rawdata[i+1:k])
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
264
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
265 while k < endpos:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
266 m = attrfind.match(rawdata, k)
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
267 if not m:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
268 break
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
269 attrname, rest, attrvalue = m.group(1, 2, 3)
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
270 if not rest:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
271 attrvalue = None
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
272 elif attrvalue[:1] == '\'' == attrvalue[-1:] or \
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
273 attrvalue[:1] == '"' == attrvalue[-1:]:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
274 attrvalue = attrvalue[1:-1]
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
275 attrvalue = self.unescape(attrvalue)
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
276 attrs.append((string.lower(attrname), attrvalue))
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
277 k = m.end()
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
278
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
279 end = string.strip(rawdata[k:endpos])
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
280 if end not in (">", "/>"):
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
281 lineno, offset = self.getpos()
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
282 if "\n" in self.__starttag_text:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
283 lineno = lineno + string.count(self.__starttag_text, "\n")
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
284 offset = len(self.__starttag_text) \
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
285 - string.rfind(self.__starttag_text, "\n")
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
286 else:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
287 offset = offset + len(self.__starttag_text)
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
288 self.error("junk characters in start tag: %s"
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
289 % `rawdata[k:endpos][:20]`)
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
290 if end[-2:] == '/>':
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
291 # XHTML-style empty tag: <span attr="value" />
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
292 self.handle_startendtag(tag, attrs)
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
293 else:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
294 self.handle_starttag(tag, attrs)
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
295 if tag in self.CDATA_CONTENT_ELEMENTS:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
296 self.set_cdata_mode(tag)
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
297 return endpos
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
298
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
299 # Internal -- check to see if we have a complete starttag; return end
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
300 # or -1 if incomplete.
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
301 def check_for_whole_start_tag(self, i):
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
302 rawdata = self.rawdata
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
303 m = locatestarttagend.match(rawdata, i)
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
304 if m:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
305 j = m.end()
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
306 next = rawdata[j:j+1]
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
307 if next == ">":
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
308 return j + 1
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
309 if next == "/":
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
310 s = rawdata[j:j+2]
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
311 if s == "/>":
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
312 return j + 2
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
313 if s == "/":
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
314 # buffer boundary
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
315 return -1
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
316 # else bogus input
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
317 self.updatepos(i, j + 1)
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
318 self.error("malformed empty start tag")
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
319 if next == "":
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
320 # end of input
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
321 return -1
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
322 if next in ("abcdefghijklmnopqrstuvwxyz=/"
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
323 "ABCDEFGHIJKLMNOPQRSTUVWXYZ"):
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
324 # end of input in or before attribute value, or we have the
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
325 # '/' from a '/>' ending
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
326 return -1
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
327 self.updatepos(i, j)
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
328 self.error("malformed start tag")
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
329 raise AssertionError("we should not get here!")
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
330
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
331 # Internal -- parse endtag, return end or -1 if incomplete
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
332 def parse_endtag(self, i):
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
333 rawdata = self.rawdata
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
334 assert rawdata[i:i+2] == "</", "unexpected call to parse_endtag"
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
335 match = endendtag.search(rawdata, i+1) # >
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
336 if not match:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
337 return -1
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
338 j = match.end()
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
339 match = endtagfind.match(rawdata, i) # </ + tag + >
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
340 if not match:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
341 self.error("bad end tag: %s" % `rawdata[i:j]`)
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
342 tag = string.lower(match.group(1))
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
343 if ( self.cdata_endtag is not None
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
344 and tag != self.cdata_endtag):
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
345 # Should be a mismatched end tag, but we'll treat it
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
346 # as text anyway, since most HTML authors aren't
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
347 # interested in the finer points of syntax.
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
348 self.handle_data(match.group(0))
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
349 else:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
350 self.handle_endtag(tag)
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
351 self.clear_cdata_mode()
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
352 return j
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
353
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
354 # Overridable -- finish processing of start+end tag: <tag.../>
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
355 def handle_startendtag(self, tag, attrs):
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
356 self.handle_starttag(tag, attrs)
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
357 self.handle_endtag(tag)
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
358
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
359 # Overridable -- handle start tag
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
360 def handle_starttag(self, tag, attrs):
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
361 pass
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
362
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
363 # Overridable -- handle end tag
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
364 def handle_endtag(self, tag):
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
365 pass
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
366
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
367 # Overridable -- handle character reference
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
368 def handle_charref(self, name):
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
369 pass
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
370
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
371 # Overridable -- handle entity reference
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
372 def handle_entityref(self, name):
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
373 pass
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
374
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
375 # Overridable -- handle data
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
376 def handle_data(self, data):
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
377 pass
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
378
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
379 # Overridable -- handle comment
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
380 def handle_comment(self, data):
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
381 pass
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
382
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
383 # Overridable -- handle declaration
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
384 def handle_decl(self, decl):
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
385 pass
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
386
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
387 # Overridable -- handle processing instruction
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
388 def handle_pi(self, data):
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
389 pass
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
390
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
391 def unknown_decl(self, data):
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
392 self.error("unknown declaration: " + `data`)
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
393
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
394 # Internal -- helper to remove special character quoting
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
395 def unescape(self, s):
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
396 if '&' not in s:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
397 return s
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
398 s = string.replace(s, "&lt;", "<")
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
399 s = string.replace(s, "&gt;", ">")
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
400 s = string.replace(s, "&apos;", "'")
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
401 s = string.replace(s, "&quot;", '"')
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
402 s = string.replace(s, "&amp;", "&") # Must be last
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
403 return s

Roundup Issue Tracker: http://roundup-tracker.org/