| 1 | """HTML 2.0 parser.
|
|---|
| 2 |
|
|---|
| 3 | See the HTML 2.0 specification:
|
|---|
| 4 | http://www.w3.org/hypertext/WWW/MarkUp/html-spec/html-spec_toc.html
|
|---|
| 5 | """
|
|---|
| 6 |
|
|---|
| 7 | from warnings import warnpy3k
|
|---|
| 8 | warnpy3k("the htmllib module has been removed in Python 3.0",
|
|---|
| 9 | stacklevel=2)
|
|---|
| 10 | del warnpy3k
|
|---|
| 11 |
|
|---|
| 12 | import sgmllib
|
|---|
| 13 |
|
|---|
| 14 | from formatter import AS_IS
|
|---|
| 15 |
|
|---|
| 16 | __all__ = ["HTMLParser", "HTMLParseError"]
|
|---|
| 17 |
|
|---|
| 18 |
|
|---|
| 19 | class HTMLParseError(sgmllib.SGMLParseError):
|
|---|
| 20 | """Error raised when an HTML document can't be parsed."""
|
|---|
| 21 |
|
|---|
| 22 |
|
|---|
| 23 | class HTMLParser(sgmllib.SGMLParser):
|
|---|
| 24 | """This is the basic HTML parser class.
|
|---|
| 25 |
|
|---|
| 26 | It supports all entity names required by the XHTML 1.0 Recommendation.
|
|---|
| 27 | It also defines handlers for all HTML 2.0 and many HTML 3.0 and 3.2
|
|---|
| 28 | elements.
|
|---|
| 29 |
|
|---|
| 30 | """
|
|---|
| 31 |
|
|---|
| 32 | from htmlentitydefs import entitydefs
|
|---|
| 33 |
|
|---|
| 34 | def __init__(self, formatter, verbose=0):
|
|---|
| 35 | """Creates an instance of the HTMLParser class.
|
|---|
| 36 |
|
|---|
| 37 | The formatter parameter is the formatter instance associated with
|
|---|
| 38 | the parser.
|
|---|
| 39 |
|
|---|
| 40 | """
|
|---|
| 41 | sgmllib.SGMLParser.__init__(self, verbose)
|
|---|
| 42 | self.formatter = formatter
|
|---|
| 43 |
|
|---|
| 44 | def error(self, message):
|
|---|
| 45 | raise HTMLParseError(message)
|
|---|
| 46 |
|
|---|
| 47 | def reset(self):
|
|---|
| 48 | sgmllib.SGMLParser.reset(self)
|
|---|
| 49 | self.savedata = None
|
|---|
| 50 | self.isindex = 0
|
|---|
| 51 | self.title = None
|
|---|
| 52 | self.base = None
|
|---|
| 53 | self.anchor = None
|
|---|
| 54 | self.anchorlist = []
|
|---|
| 55 | self.nofill = 0
|
|---|
| 56 | self.list_stack = []
|
|---|
| 57 |
|
|---|
| 58 | # ------ Methods used internally; some may be overridden
|
|---|
| 59 |
|
|---|
| 60 | # --- Formatter interface, taking care of 'savedata' mode;
|
|---|
| 61 | # shouldn't need to be overridden
|
|---|
| 62 |
|
|---|
| 63 | def handle_data(self, data):
|
|---|
| 64 | if self.savedata is not None:
|
|---|
| 65 | self.savedata = self.savedata + data
|
|---|
| 66 | else:
|
|---|
| 67 | if self.nofill:
|
|---|
| 68 | self.formatter.add_literal_data(data)
|
|---|
| 69 | else:
|
|---|
| 70 | self.formatter.add_flowing_data(data)
|
|---|
| 71 |
|
|---|
| 72 | # --- Hooks to save data; shouldn't need to be overridden
|
|---|
| 73 |
|
|---|
| 74 | def save_bgn(self):
|
|---|
| 75 | """Begins saving character data in a buffer instead of sending it
|
|---|
| 76 | to the formatter object.
|
|---|
| 77 |
|
|---|
| 78 | Retrieve the stored data via the save_end() method. Use of the
|
|---|
| 79 | save_bgn() / save_end() pair may not be nested.
|
|---|
| 80 |
|
|---|
| 81 | """
|
|---|
| 82 | self.savedata = ''
|
|---|
| 83 |
|
|---|
| 84 | def save_end(self):
|
|---|
| 85 | """Ends buffering character data and returns all data saved since
|
|---|
| 86 | the preceding call to the save_bgn() method.
|
|---|
| 87 |
|
|---|
| 88 | If the nofill flag is false, whitespace is collapsed to single
|
|---|
| 89 | spaces. A call to this method without a preceding call to the
|
|---|
| 90 | save_bgn() method will raise a TypeError exception.
|
|---|
| 91 |
|
|---|
| 92 | """
|
|---|
| 93 | data = self.savedata
|
|---|
| 94 | self.savedata = None
|
|---|
| 95 | if not self.nofill:
|
|---|
| 96 | data = ' '.join(data.split())
|
|---|
| 97 | return data
|
|---|
| 98 |
|
|---|
| 99 | # --- Hooks for anchors; should probably be overridden
|
|---|
| 100 |
|
|---|
| 101 | def anchor_bgn(self, href, name, type):
|
|---|
| 102 | """This method is called at the start of an anchor region.
|
|---|
| 103 |
|
|---|
| 104 | The arguments correspond to the attributes of the <A> tag with
|
|---|
| 105 | the same names. The default implementation maintains a list of
|
|---|
| 106 | hyperlinks (defined by the HREF attribute for <A> tags) within
|
|---|
| 107 | the document. The list of hyperlinks is available as the data
|
|---|
| 108 | attribute anchorlist.
|
|---|
| 109 |
|
|---|
| 110 | """
|
|---|
| 111 | self.anchor = href
|
|---|
| 112 | if self.anchor:
|
|---|
| 113 | self.anchorlist.append(href)
|
|---|
| 114 |
|
|---|
| 115 | def anchor_end(self):
|
|---|
| 116 | """This method is called at the end of an anchor region.
|
|---|
| 117 |
|
|---|
| 118 | The default implementation adds a textual footnote marker using an
|
|---|
| 119 | index into the list of hyperlinks created by the anchor_bgn()method.
|
|---|
| 120 |
|
|---|
| 121 | """
|
|---|
| 122 | if self.anchor:
|
|---|
| 123 | self.handle_data("[%d]" % len(self.anchorlist))
|
|---|
| 124 | self.anchor = None
|
|---|
| 125 |
|
|---|
| 126 | # --- Hook for images; should probably be overridden
|
|---|
| 127 |
|
|---|
| 128 | def handle_image(self, src, alt, *args):
|
|---|
| 129 | """This method is called to handle images.
|
|---|
| 130 |
|
|---|
| 131 | The default implementation simply passes the alt value to the
|
|---|
| 132 | handle_data() method.
|
|---|
| 133 |
|
|---|
| 134 | """
|
|---|
| 135 | self.handle_data(alt)
|
|---|
| 136 |
|
|---|
| 137 | # --------- Top level elememts
|
|---|
| 138 |
|
|---|
| 139 | def start_html(self, attrs): pass
|
|---|
| 140 | def end_html(self): pass
|
|---|
| 141 |
|
|---|
| 142 | def start_head(self, attrs): pass
|
|---|
| 143 | def end_head(self): pass
|
|---|
| 144 |
|
|---|
| 145 | def start_body(self, attrs): pass
|
|---|
| 146 | def end_body(self): pass
|
|---|
| 147 |
|
|---|
| 148 | # ------ Head elements
|
|---|
| 149 |
|
|---|
| 150 | def start_title(self, attrs):
|
|---|
| 151 | self.save_bgn()
|
|---|
| 152 |
|
|---|
| 153 | def end_title(self):
|
|---|
| 154 | self.title = self.save_end()
|
|---|
| 155 |
|
|---|
| 156 | def do_base(self, attrs):
|
|---|
| 157 | for a, v in attrs:
|
|---|
| 158 | if a == 'href':
|
|---|
| 159 | self.base = v
|
|---|
| 160 |
|
|---|
| 161 | def do_isindex(self, attrs):
|
|---|
| 162 | self.isindex = 1
|
|---|
| 163 |
|
|---|
| 164 | def do_link(self, attrs):
|
|---|
| 165 | pass
|
|---|
| 166 |
|
|---|
| 167 | def do_meta(self, attrs):
|
|---|
| 168 | pass
|
|---|
| 169 |
|
|---|
| 170 | def do_nextid(self, attrs): # Deprecated
|
|---|
| 171 | pass
|
|---|
| 172 |
|
|---|
| 173 | # ------ Body elements
|
|---|
| 174 |
|
|---|
| 175 | # --- Headings
|
|---|
| 176 |
|
|---|
| 177 | def start_h1(self, attrs):
|
|---|
| 178 | self.formatter.end_paragraph(1)
|
|---|
| 179 | self.formatter.push_font(('h1', 0, 1, 0))
|
|---|
| 180 |
|
|---|
| 181 | def end_h1(self):
|
|---|
| 182 | self.formatter.end_paragraph(1)
|
|---|
| 183 | self.formatter.pop_font()
|
|---|
| 184 |
|
|---|
| 185 | def start_h2(self, attrs):
|
|---|
| 186 | self.formatter.end_paragraph(1)
|
|---|
| 187 | self.formatter.push_font(('h2', 0, 1, 0))
|
|---|
| 188 |
|
|---|
| 189 | def end_h2(self):
|
|---|
| 190 | self.formatter.end_paragraph(1)
|
|---|
| 191 | self.formatter.pop_font()
|
|---|
| 192 |
|
|---|
| 193 | def start_h3(self, attrs):
|
|---|
| 194 | self.formatter.end_paragraph(1)
|
|---|
| 195 | self.formatter.push_font(('h3', 0, 1, 0))
|
|---|
| 196 |
|
|---|
| 197 | def end_h3(self):
|
|---|
| 198 | self.formatter.end_paragraph(1)
|
|---|
| 199 | self.formatter.pop_font()
|
|---|
| 200 |
|
|---|
| 201 | def start_h4(self, attrs):
|
|---|
| 202 | self.formatter.end_paragraph(1)
|
|---|
| 203 | self.formatter.push_font(('h4', 0, 1, 0))
|
|---|
| 204 |
|
|---|
| 205 | def end_h4(self):
|
|---|
| 206 | self.formatter.end_paragraph(1)
|
|---|
| 207 | self.formatter.pop_font()
|
|---|
| 208 |
|
|---|
| 209 | def start_h5(self, attrs):
|
|---|
| 210 | self.formatter.end_paragraph(1)
|
|---|
| 211 | self.formatter.push_font(('h5', 0, 1, 0))
|
|---|
| 212 |
|
|---|
| 213 | def end_h5(self):
|
|---|
| 214 | self.formatter.end_paragraph(1)
|
|---|
| 215 | self.formatter.pop_font()
|
|---|
| 216 |
|
|---|
| 217 | def start_h6(self, attrs):
|
|---|
| 218 | self.formatter.end_paragraph(1)
|
|---|
| 219 | self.formatter.push_font(('h6', 0, 1, 0))
|
|---|
| 220 |
|
|---|
| 221 | def end_h6(self):
|
|---|
| 222 | self.formatter.end_paragraph(1)
|
|---|
| 223 | self.formatter.pop_font()
|
|---|
| 224 |
|
|---|
| 225 | # --- Block Structuring Elements
|
|---|
| 226 |
|
|---|
| 227 | def do_p(self, attrs):
|
|---|
| 228 | self.formatter.end_paragraph(1)
|
|---|
| 229 |
|
|---|
| 230 | def start_pre(self, attrs):
|
|---|
| 231 | self.formatter.end_paragraph(1)
|
|---|
| 232 | self.formatter.push_font((AS_IS, AS_IS, AS_IS, 1))
|
|---|
| 233 | self.nofill = self.nofill + 1
|
|---|
| 234 |
|
|---|
| 235 | def end_pre(self):
|
|---|
| 236 | self.formatter.end_paragraph(1)
|
|---|
| 237 | self.formatter.pop_font()
|
|---|
| 238 | self.nofill = max(0, self.nofill - 1)
|
|---|
| 239 |
|
|---|
| 240 | def start_xmp(self, attrs):
|
|---|
| 241 | self.start_pre(attrs)
|
|---|
| 242 | self.setliteral('xmp') # Tell SGML parser
|
|---|
| 243 |
|
|---|
| 244 | def end_xmp(self):
|
|---|
| 245 | self.end_pre()
|
|---|
| 246 |
|
|---|
| 247 | def start_listing(self, attrs):
|
|---|
| 248 | self.start_pre(attrs)
|
|---|
| 249 | self.setliteral('listing') # Tell SGML parser
|
|---|
| 250 |
|
|---|
| 251 | def end_listing(self):
|
|---|
| 252 | self.end_pre()
|
|---|
| 253 |
|
|---|
| 254 | def start_address(self, attrs):
|
|---|
| 255 | self.formatter.end_paragraph(0)
|
|---|
| 256 | self.formatter.push_font((AS_IS, 1, AS_IS, AS_IS))
|
|---|
| 257 |
|
|---|
| 258 | def end_address(self):
|
|---|
| 259 | self.formatter.end_paragraph(0)
|
|---|
| 260 | self.formatter.pop_font()
|
|---|
| 261 |
|
|---|
| 262 | def start_blockquote(self, attrs):
|
|---|
| 263 | self.formatter.end_paragraph(1)
|
|---|
| 264 | self.formatter.push_margin('blockquote')
|
|---|
| 265 |
|
|---|
| 266 | def end_blockquote(self):
|
|---|
| 267 | self.formatter.end_paragraph(1)
|
|---|
| 268 | self.formatter.pop_margin()
|
|---|
| 269 |
|
|---|
| 270 | # --- List Elements
|
|---|
| 271 |
|
|---|
| 272 | def start_ul(self, attrs):
|
|---|
| 273 | self.formatter.end_paragraph(not self.list_stack)
|
|---|
| 274 | self.formatter.push_margin('ul')
|
|---|
| 275 | self.list_stack.append(['ul', '*', 0])
|
|---|
| 276 |
|
|---|
| 277 | def end_ul(self):
|
|---|
| 278 | if self.list_stack: del self.list_stack[-1]
|
|---|
| 279 | self.formatter.end_paragraph(not self.list_stack)
|
|---|
| 280 | self.formatter.pop_margin()
|
|---|
| 281 |
|
|---|
| 282 | def do_li(self, attrs):
|
|---|
| 283 | self.formatter.end_paragraph(0)
|
|---|
| 284 | if self.list_stack:
|
|---|
| 285 | [dummy, label, counter] = top = self.list_stack[-1]
|
|---|
| 286 | top[2] = counter = counter+1
|
|---|
| 287 | else:
|
|---|
| 288 | label, counter = '*', 0
|
|---|
| 289 | self.formatter.add_label_data(label, counter)
|
|---|
| 290 |
|
|---|
| 291 | def start_ol(self, attrs):
|
|---|
| 292 | self.formatter.end_paragraph(not self.list_stack)
|
|---|
| 293 | self.formatter.push_margin('ol')
|
|---|
| 294 | label = '1.'
|
|---|
| 295 | for a, v in attrs:
|
|---|
| 296 | if a == 'type':
|
|---|
| 297 | if len(v) == 1: v = v + '.'
|
|---|
| 298 | label = v
|
|---|
| 299 | self.list_stack.append(['ol', label, 0])
|
|---|
| 300 |
|
|---|
| 301 | def end_ol(self):
|
|---|
| 302 | if self.list_stack: del self.list_stack[-1]
|
|---|
| 303 | self.formatter.end_paragraph(not self.list_stack)
|
|---|
| 304 | self.formatter.pop_margin()
|
|---|
| 305 |
|
|---|
| 306 | def start_menu(self, attrs):
|
|---|
| 307 | self.start_ul(attrs)
|
|---|
| 308 |
|
|---|
| 309 | def end_menu(self):
|
|---|
| 310 | self.end_ul()
|
|---|
| 311 |
|
|---|
| 312 | def start_dir(self, attrs):
|
|---|
| 313 | self.start_ul(attrs)
|
|---|
| 314 |
|
|---|
| 315 | def end_dir(self):
|
|---|
| 316 | self.end_ul()
|
|---|
| 317 |
|
|---|
| 318 | def start_dl(self, attrs):
|
|---|
| 319 | self.formatter.end_paragraph(1)
|
|---|
| 320 | self.list_stack.append(['dl', '', 0])
|
|---|
| 321 |
|
|---|
| 322 | def end_dl(self):
|
|---|
| 323 | self.ddpop(1)
|
|---|
| 324 | if self.list_stack: del self.list_stack[-1]
|
|---|
| 325 |
|
|---|
| 326 | def do_dt(self, attrs):
|
|---|
| 327 | self.ddpop()
|
|---|
| 328 |
|
|---|
| 329 | def do_dd(self, attrs):
|
|---|
| 330 | self.ddpop()
|
|---|
| 331 | self.formatter.push_margin('dd')
|
|---|
| 332 | self.list_stack.append(['dd', '', 0])
|
|---|
| 333 |
|
|---|
| 334 | def ddpop(self, bl=0):
|
|---|
| 335 | self.formatter.end_paragraph(bl)
|
|---|
| 336 | if self.list_stack:
|
|---|
| 337 | if self.list_stack[-1][0] == 'dd':
|
|---|
| 338 | del self.list_stack[-1]
|
|---|
| 339 | self.formatter.pop_margin()
|
|---|
| 340 |
|
|---|
| 341 | # --- Phrase Markup
|
|---|
| 342 |
|
|---|
| 343 | # Idiomatic Elements
|
|---|
| 344 |
|
|---|
| 345 | def start_cite(self, attrs): self.start_i(attrs)
|
|---|
| 346 | def end_cite(self): self.end_i()
|
|---|
| 347 |
|
|---|
| 348 | def start_code(self, attrs): self.start_tt(attrs)
|
|---|
| 349 | def end_code(self): self.end_tt()
|
|---|
| 350 |
|
|---|
| 351 | def start_em(self, attrs): self.start_i(attrs)
|
|---|
| 352 | def end_em(self): self.end_i()
|
|---|
| 353 |
|
|---|
| 354 | def start_kbd(self, attrs): self.start_tt(attrs)
|
|---|
| 355 | def end_kbd(self): self.end_tt()
|
|---|
| 356 |
|
|---|
| 357 | def start_samp(self, attrs): self.start_tt(attrs)
|
|---|
| 358 | def end_samp(self): self.end_tt()
|
|---|
| 359 |
|
|---|
| 360 | def start_strong(self, attrs): self.start_b(attrs)
|
|---|
| 361 | def end_strong(self): self.end_b()
|
|---|
| 362 |
|
|---|
| 363 | def start_var(self, attrs): self.start_i(attrs)
|
|---|
| 364 | def end_var(self): self.end_i()
|
|---|
| 365 |
|
|---|
| 366 | # Typographic Elements
|
|---|
| 367 |
|
|---|
| 368 | def start_i(self, attrs):
|
|---|
| 369 | self.formatter.push_font((AS_IS, 1, AS_IS, AS_IS))
|
|---|
| 370 | def end_i(self):
|
|---|
| 371 | self.formatter.pop_font()
|
|---|
| 372 |
|
|---|
| 373 | def start_b(self, attrs):
|
|---|
| 374 | self.formatter.push_font((AS_IS, AS_IS, 1, AS_IS))
|
|---|
| 375 | def end_b(self):
|
|---|
| 376 | self.formatter.pop_font()
|
|---|
| 377 |
|
|---|
| 378 | def start_tt(self, attrs):
|
|---|
| 379 | self.formatter.push_font((AS_IS, AS_IS, AS_IS, 1))
|
|---|
| 380 | def end_tt(self):
|
|---|
| 381 | self.formatter.pop_font()
|
|---|
| 382 |
|
|---|
| 383 | def start_a(self, attrs):
|
|---|
| 384 | href = ''
|
|---|
| 385 | name = ''
|
|---|
| 386 | type = ''
|
|---|
| 387 | for attrname, value in attrs:
|
|---|
| 388 | value = value.strip()
|
|---|
| 389 | if attrname == 'href':
|
|---|
| 390 | href = value
|
|---|
| 391 | if attrname == 'name':
|
|---|
| 392 | name = value
|
|---|
| 393 | if attrname == 'type':
|
|---|
| 394 | type = value.lower()
|
|---|
| 395 | self.anchor_bgn(href, name, type)
|
|---|
| 396 |
|
|---|
| 397 | def end_a(self):
|
|---|
| 398 | self.anchor_end()
|
|---|
| 399 |
|
|---|
| 400 | # --- Line Break
|
|---|
| 401 |
|
|---|
| 402 | def do_br(self, attrs):
|
|---|
| 403 | self.formatter.add_line_break()
|
|---|
| 404 |
|
|---|
| 405 | # --- Horizontal Rule
|
|---|
| 406 |
|
|---|
| 407 | def do_hr(self, attrs):
|
|---|
| 408 | self.formatter.add_hor_rule()
|
|---|
| 409 |
|
|---|
| 410 | # --- Image
|
|---|
| 411 |
|
|---|
| 412 | def do_img(self, attrs):
|
|---|
| 413 | align = ''
|
|---|
| 414 | alt = '(image)'
|
|---|
| 415 | ismap = ''
|
|---|
| 416 | src = ''
|
|---|
| 417 | width = 0
|
|---|
| 418 | height = 0
|
|---|
| 419 | for attrname, value in attrs:
|
|---|
| 420 | if attrname == 'align':
|
|---|
| 421 | align = value
|
|---|
| 422 | if attrname == 'alt':
|
|---|
| 423 | alt = value
|
|---|
| 424 | if attrname == 'ismap':
|
|---|
| 425 | ismap = value
|
|---|
| 426 | if attrname == 'src':
|
|---|
| 427 | src = value
|
|---|
| 428 | if attrname == 'width':
|
|---|
| 429 | try: width = int(value)
|
|---|
| 430 | except ValueError: pass
|
|---|
| 431 | if attrname == 'height':
|
|---|
| 432 | try: height = int(value)
|
|---|
| 433 | except ValueError: pass
|
|---|
| 434 | self.handle_image(src, alt, ismap, align, width, height)
|
|---|
| 435 |
|
|---|
| 436 | # --- Really Old Unofficial Deprecated Stuff
|
|---|
| 437 |
|
|---|
| 438 | def do_plaintext(self, attrs):
|
|---|
| 439 | self.start_pre(attrs)
|
|---|
| 440 | self.setnomoretags() # Tell SGML parser
|
|---|
| 441 |
|
|---|
| 442 | # --- Unhandled tags
|
|---|
| 443 |
|
|---|
| 444 | def unknown_starttag(self, tag, attrs):
|
|---|
| 445 | pass
|
|---|
| 446 |
|
|---|
| 447 | def unknown_endtag(self, tag):
|
|---|
| 448 | pass
|
|---|
| 449 |
|
|---|
| 450 |
|
|---|
| 451 | def test(args = None):
|
|---|
| 452 | import sys, formatter
|
|---|
| 453 |
|
|---|
| 454 | if not args:
|
|---|
| 455 | args = sys.argv[1:]
|
|---|
| 456 |
|
|---|
| 457 | silent = args and args[0] == '-s'
|
|---|
| 458 | if silent:
|
|---|
| 459 | del args[0]
|
|---|
| 460 |
|
|---|
| 461 | if args:
|
|---|
| 462 | file = args[0]
|
|---|
| 463 | else:
|
|---|
| 464 | file = 'test.html'
|
|---|
| 465 |
|
|---|
| 466 | if file == '-':
|
|---|
| 467 | f = sys.stdin
|
|---|
| 468 | else:
|
|---|
| 469 | try:
|
|---|
| 470 | f = open(file, 'r')
|
|---|
| 471 | except IOError, msg:
|
|---|
| 472 | print file, ":", msg
|
|---|
| 473 | sys.exit(1)
|
|---|
| 474 |
|
|---|
| 475 | data = f.read()
|
|---|
| 476 |
|
|---|
| 477 | if f is not sys.stdin:
|
|---|
| 478 | f.close()
|
|---|
| 479 |
|
|---|
| 480 | if silent:
|
|---|
| 481 | f = formatter.NullFormatter()
|
|---|
| 482 | else:
|
|---|
| 483 | f = formatter.AbstractFormatter(formatter.DumbWriter())
|
|---|
| 484 |
|
|---|
| 485 | p = HTMLParser(f)
|
|---|
| 486 | p.feed(data)
|
|---|
| 487 | p.close()
|
|---|
| 488 |
|
|---|
| 489 |
|
|---|
| 490 | if __name__ == '__main__':
|
|---|
| 491 | test()
|
|---|
