Lib/htmllib.py

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