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