| 1 | #!/usr/bin/env python2
|
| 2 | """
|
| 3 | lazylex/html.py - Low-Level HTML Processing.
|
| 4 |
|
| 5 | See lazylex/README.md for details.
|
| 6 |
|
| 7 | TODO: This should be an Oils library eventually. It's a "lazily-parsed data
|
| 8 | structure" like TSV8
|
| 9 | """
|
| 10 | from __future__ import print_function
|
| 11 |
|
| 12 | try:
|
| 13 | from cStringIO import StringIO
|
| 14 | except ImportError:
|
| 15 | from io import StringIO # python3
|
| 16 | import re
|
| 17 | import sys
|
| 18 |
|
| 19 | if sys.version_info.major == 2:
|
| 20 | from typing import List, Tuple, Optional
|
| 21 |
|
| 22 |
|
| 23 | def log(msg, *args):
|
| 24 | msg = msg % args
|
| 25 | print(msg, file=sys.stderr)
|
| 26 |
|
| 27 |
|
| 28 | class LexError(Exception):
|
| 29 | """
|
| 30 | Examples of lex errors:
|
| 31 |
|
| 32 | - Tok.Invalid, like <> or &&
|
| 33 | - Unclosed <!-- <? <![CDATA[ <script> <style>
|
| 34 | """
|
| 35 |
|
| 36 | def __init__(self, s, start_pos):
|
| 37 | self.s = s
|
| 38 | self.start_pos = start_pos
|
| 39 |
|
| 40 | def __str__(self):
|
| 41 | return '(LexError %r)' % (self.s[self.start_pos:self.start_pos + 20])
|
| 42 |
|
| 43 |
|
| 44 | class ParseError(Exception):
|
| 45 | """
|
| 46 | Examples of parse errors
|
| 47 |
|
| 48 | - unbalanced tag structure
|
| 49 | - ul_table.py errors
|
| 50 | """
|
| 51 |
|
| 52 | def __init__(self, msg, s=None, start_pos=-1):
|
| 53 | self.msg = msg
|
| 54 | self.s = s
|
| 55 | self.start_pos = start_pos
|
| 56 |
|
| 57 | def __str__(self):
|
| 58 | if self.s is not None:
|
| 59 | assert self.start_pos != -1, self.start_pos
|
| 60 | snippet = (self.s[self.start_pos:self.start_pos + 20])
|
| 61 | else:
|
| 62 | snippet = ''
|
| 63 | return '(ParseError %r %r)' % (self.msg, snippet)
|
| 64 |
|
| 65 |
|
| 66 | class Output(object):
|
| 67 | """Takes an underlying input buffer and an output file. Maintains a
|
| 68 | position in the input buffer.
|
| 69 |
|
| 70 | Print FROM the input or print new text to the output.
|
| 71 | """
|
| 72 |
|
| 73 | def __init__(self, s, f, left_pos=0, right_pos=-1):
|
| 74 | self.s = s
|
| 75 | self.f = f
|
| 76 | self.pos = left_pos
|
| 77 | self.right_pos = len(s) if right_pos == -1 else right_pos
|
| 78 |
|
| 79 | def SkipTo(self, pos):
|
| 80 | """Skip to a position."""
|
| 81 | self.pos = pos
|
| 82 |
|
| 83 | def PrintUntil(self, pos):
|
| 84 | """Print until a position."""
|
| 85 | piece = self.s[self.pos:pos]
|
| 86 | self.f.write(piece)
|
| 87 | self.pos = pos
|
| 88 |
|
| 89 | def PrintTheRest(self):
|
| 90 | """Print until the end of the string."""
|
| 91 | self.PrintUntil(self.right_pos)
|
| 92 |
|
| 93 | def Print(self, s):
|
| 94 | """Print text to the underlying buffer."""
|
| 95 | self.f.write(s)
|
| 96 |
|
| 97 |
|
| 98 | # HTML Tokens
|
| 99 | # CommentBegin, ProcessingBegin, CDataBegin are "pseudo-tokens", not visible
|
| 100 | TOKENS = 'Decl Comment CommentBegin Processing ProcessingBegin CData CDataBegin StartTag StartEndTag EndTag DecChar HexChar CharEntity RawData HtmlCData Invalid EndOfStream'.split(
|
| 101 | )
|
| 102 |
|
| 103 |
|
| 104 | class Tok(object):
|
| 105 | """
|
| 106 | Avoid lint errors by using these aliases
|
| 107 | """
|
| 108 | pass
|
| 109 |
|
| 110 |
|
| 111 | TOKEN_NAMES = [None] * len(TOKENS) # type: List[str]
|
| 112 |
|
| 113 | this_module = sys.modules[__name__]
|
| 114 | for i, tok_str in enumerate(TOKENS):
|
| 115 | setattr(this_module, tok_str, i)
|
| 116 | setattr(Tok, tok_str, i)
|
| 117 | TOKEN_NAMES[i] = tok_str
|
| 118 |
|
| 119 |
|
| 120 | def TokenName(tok_id):
|
| 121 | return TOKEN_NAMES[tok_id]
|
| 122 |
|
| 123 |
|
| 124 | def MakeLexer(rules):
|
| 125 | return [(re.compile(pat, re.VERBOSE), i) for (pat, i) in rules]
|
| 126 |
|
| 127 |
|
| 128 | #
|
| 129 | # Eggex
|
| 130 | #
|
| 131 | # Tag = / ~['>']+ /
|
| 132 |
|
| 133 | # Is this valid? A single character?
|
| 134 | # Tag = / ~'>'* /
|
| 135 |
|
| 136 | # Maybe better: / [NOT '>']+/
|
| 137 | # capital letters not allowed there?
|
| 138 | #
|
| 139 | # But then this is confusing:
|
| 140 | # / [NOT ~digit]+/
|
| 141 | #
|
| 142 | # / [NOT digit] / is [^\d]
|
| 143 | # / ~digit / is \D
|
| 144 | #
|
| 145 | # Or maybe:
|
| 146 | #
|
| 147 | # / [~ digit]+ /
|
| 148 | # / [~ '>']+ /
|
| 149 | # / [NOT '>']+ /
|
| 150 |
|
| 151 | # End = / '</' Tag '>' /
|
| 152 | # StartEnd = / '<' Tag '/>' /
|
| 153 | # Start = / '<' Tag '>' /
|
| 154 | #
|
| 155 | # EntityRef = / '&' dot{* N} ';' /
|
| 156 |
|
| 157 | # Tag name, or attribute name
|
| 158 | # colon is used in XML
|
| 159 |
|
| 160 | # https://www.w3.org/TR/xml/#NT-Name
|
| 161 | # Hm there is a lot of unicode stuff. We are simplifying parsing
|
| 162 |
|
| 163 | _NAME = r'[a-zA-Z][a-zA-Z0-9:_\-]*' # must start with letter
|
| 164 |
|
| 165 | LEXER = [
|
| 166 | (r'<!--', Tok.CommentBegin),
|
| 167 |
|
| 168 | # Processing instruction are used for the XML header:
|
| 169 | # <?xml version="1.0" encoding="UTF-8"?>
|
| 170 | # They are technically XML-only, but in HTML5, they are another kind of
|
| 171 | # comment:
|
| 172 | #
|
| 173 | # https://developer.mozilla.org/en-US/docs/Web/API/ProcessingInstruction
|
| 174 | #
|
| 175 | (r'<\?', Tok.ProcessingBegin),
|
| 176 | # Not necessary in HTML5, but occurs in XML
|
| 177 | (r'<!\[CDATA\[', Tok.CDataBegin), # <![CDATA[
|
| 178 |
|
| 179 | # Markup declarations
|
| 180 | # - In HTML5, there is only <!DOCTYPE html>
|
| 181 | # - XML has 4 more declarations: <!ELEMENT ...> ATTLIST ENTITY NOTATION
|
| 182 | # - these seem to be part of DTD
|
| 183 | # - it's useful to skip these, and be able to parse the rest of the document
|
| 184 | # - Note: < is allowed?
|
| 185 | (r'<! [^>]+ >', Tok.Decl),
|
| 186 |
|
| 187 | # Tags
|
| 188 | # Notes:
|
| 189 | # - We look for a valid tag name, but we don't validate attributes.
|
| 190 | # That's done in the tag lexer.
|
| 191 | # - We don't allow leading whitespace
|
| 192 | (r'</ (%s) >' % _NAME, Tok.EndTag),
|
| 193 | # self-closing <br/> comes before StarttTag
|
| 194 | (r'< (%s) [^>]* />' % _NAME, Tok.StartEndTag), # end </a>
|
| 195 | (r'< (%s) [^>]* >' % _NAME, Tok.StartTag), # start <a>
|
| 196 |
|
| 197 | # Characters
|
| 198 | # https://www.w3.org/TR/xml/#sec-references
|
| 199 | (r'&\# [0-9]+ ;', Tok.DecChar),
|
| 200 | (r'&\# x[0-9a-fA-F]+ ;', Tok.HexChar),
|
| 201 | (r'& %s ;' % _NAME, Tok.CharEntity),
|
| 202 |
|
| 203 | # HTML5 allows unescaped > in raw data, but < is not allowed.
|
| 204 | # https://stackoverflow.com/questions/10462348/right-angle-bracket-in-html
|
| 205 | #
|
| 206 | # - My early blog has THREE errors when disallowing >
|
| 207 | # - So do some .wwz files
|
| 208 | (r'[^&<]+', Tok.RawData),
|
| 209 | (r'.', Tok.Invalid), # error!
|
| 210 | ]
|
| 211 |
|
| 212 | # Old notes:
|
| 213 | #
|
| 214 | # Non-greedy matches are regular and can be matched in linear time
|
| 215 | # with RE2.
|
| 216 | #
|
| 217 | # https://news.ycombinator.com/item?id=27099798
|
| 218 | #
|
| 219 | # Maybe try combining all of these for speed.
|
| 220 |
|
| 221 | # . is any char except newline
|
| 222 | # https://re2c.org/manual/manual_c.html
|
| 223 |
|
| 224 | # Discarded options
|
| 225 | #(r'<!-- .*? -->', Tok.Comment),
|
| 226 |
|
| 227 | # Hack from Claude: \s\S instead of re.DOTALL. I don't like this
|
| 228 | #(r'<!-- [\s\S]*? -->', Tok.Comment),
|
| 229 | #(r'<!-- (?:.|[\n])*? -->', Tok.Comment),
|
| 230 |
|
| 231 | LEXER = MakeLexer(LEXER)
|
| 232 |
|
| 233 |
|
| 234 | class Lexer(object):
|
| 235 |
|
| 236 | def __init__(self, s, left_pos=0, right_pos=-1):
|
| 237 | self.s = s
|
| 238 | self.pos = left_pos
|
| 239 | self.right_pos = len(s) if right_pos == -1 else right_pos
|
| 240 | self.cache = {} # string -> compiled regex pattern object
|
| 241 |
|
| 242 | # either </script> or </style> - we search until we see that
|
| 243 | self.search_state = None # type: Optional[str]
|
| 244 |
|
| 245 | # Position of tag name, if applicable
|
| 246 | # - Set after you get a StartTag, EndTag, or StartEndTag
|
| 247 | # - Unset on other tags
|
| 248 | self.tag_pos_left = -1
|
| 249 | self.tag_pos_right = -1
|
| 250 |
|
| 251 | def _Peek(self):
|
| 252 | # type: () -> Tuple[int, int]
|
| 253 | """
|
| 254 | Note: not using _Peek() now
|
| 255 | """
|
| 256 | if self.pos == self.right_pos:
|
| 257 | return Tok.EndOfStream, self.pos
|
| 258 |
|
| 259 | assert self.pos < self.right_pos, self.pos
|
| 260 |
|
| 261 | if self.search_state is not None:
|
| 262 | pos = self.s.find(self.search_state, self.pos)
|
| 263 | if pos == -1:
|
| 264 | # unterminated <script> or <style>
|
| 265 | raise LexError(self.s, self.pos)
|
| 266 | self.search_state = None
|
| 267 | # beginning
|
| 268 | return Tok.HtmlCData, pos
|
| 269 |
|
| 270 | # Find the first match.
|
| 271 | # Note: frontend/match.py uses _LongestMatch(), which is different!
|
| 272 | # TODO: reconcile them. This lexer should be expressible in re2c.
|
| 273 |
|
| 274 | for pat, tok_id in LEXER:
|
| 275 | m = pat.match(self.s, self.pos)
|
| 276 | if m:
|
| 277 | if tok_id in (Tok.StartTag, Tok.EndTag, Tok.StartEndTag):
|
| 278 | self.tag_pos_left = m.start(1)
|
| 279 | self.tag_pos_right = m.end(1)
|
| 280 | else:
|
| 281 | # Reset state
|
| 282 | self.tag_pos_left = -1
|
| 283 | self.tag_pos_right = -1
|
| 284 |
|
| 285 | if tok_id == Tok.CommentBegin:
|
| 286 | pos = self.s.find('-->', self.pos)
|
| 287 | if pos == -1:
|
| 288 | # unterminated <!--
|
| 289 | raise LexError(self.s, self.pos)
|
| 290 | return Tok.Comment, pos + 3 # -->
|
| 291 |
|
| 292 | if tok_id == Tok.ProcessingBegin:
|
| 293 | pos = self.s.find('?>', self.pos)
|
| 294 | if pos == -1:
|
| 295 | # unterminated <?
|
| 296 | raise LexError(self.s, self.pos)
|
| 297 | return Tok.Processing, pos + 2 # ?>
|
| 298 |
|
| 299 | if tok_id == Tok.CDataBegin:
|
| 300 | pos = self.s.find(']]>', self.pos)
|
| 301 | if pos == -1:
|
| 302 | # unterminated <![CDATA[
|
| 303 | raise LexError(self.s, self.pos)
|
| 304 | return Tok.CData, pos + 3 # ]]>
|
| 305 |
|
| 306 | if tok_id == Tok.StartTag:
|
| 307 | if self.TagNameEquals('script'):
|
| 308 | self.search_state = '</script>'
|
| 309 | elif self.TagNameEquals('style'):
|
| 310 | self.search_state = '</style>'
|
| 311 |
|
| 312 | return tok_id, m.end()
|
| 313 | else:
|
| 314 | raise AssertionError('Tok.Invalid rule should have matched')
|
| 315 |
|
| 316 | def TagNameEquals(self, expected):
|
| 317 | # type: (str) -> bool
|
| 318 | assert self.tag_pos_left != -1, self.tag_pos_left
|
| 319 | assert self.tag_pos_right != -1, self.tag_pos_right
|
| 320 |
|
| 321 | # TODO: In C++, this does not need an allocation
|
| 322 | return expected == self.s[self.tag_pos_left:self.tag_pos_right]
|
| 323 |
|
| 324 | def TagName(self):
|
| 325 | # type: () -> None
|
| 326 | assert self.tag_pos_left != -1, self.tag_pos_left
|
| 327 | assert self.tag_pos_right != -1, self.tag_pos_right
|
| 328 |
|
| 329 | return self.s[self.tag_pos_left:self.tag_pos_right]
|
| 330 |
|
| 331 | def Read(self):
|
| 332 | # type: () -> Tuple[int, int]
|
| 333 | tok_id, end_pos = self._Peek()
|
| 334 | self.pos = end_pos # advance
|
| 335 | return tok_id, end_pos
|
| 336 |
|
| 337 | def LookAhead(self, regex):
|
| 338 | # Cache the regex compilation. This could also be LookAheadFor(THEAD)
|
| 339 | # or something.
|
| 340 | pat = self.cache.get(regex)
|
| 341 | if pat is None:
|
| 342 | pat = re.compile(regex)
|
| 343 | self.cache[regex] = pat
|
| 344 |
|
| 345 | m = pat.match(self.s, self.pos)
|
| 346 | return m is not None
|
| 347 |
|
| 348 |
|
| 349 | def _Tokens(s, left_pos, right_pos):
|
| 350 | """
|
| 351 | Args:
|
| 352 | s: string to parse
|
| 353 | left_pos, right_pos: Optional span boundaries.
|
| 354 | """
|
| 355 | lx = Lexer(s, left_pos, right_pos)
|
| 356 | while True:
|
| 357 | tok_id, pos = lx.Read()
|
| 358 | yield tok_id, pos
|
| 359 | if tok_id == Tok.EndOfStream:
|
| 360 | break
|
| 361 |
|
| 362 |
|
| 363 | def ValidTokens(s, left_pos=0, right_pos=-1):
|
| 364 | """Wrapper around _Tokens to prevent callers from having to handle Invalid.
|
| 365 |
|
| 366 | I'm not combining the two functions because I might want to do a
|
| 367 | 'yield' transformation on Tokens()? Exceptions might complicate the
|
| 368 | issue?
|
| 369 | """
|
| 370 | pos = left_pos
|
| 371 | for tok_id, end_pos in _Tokens(s, left_pos, right_pos):
|
| 372 | if tok_id == Tok.Invalid:
|
| 373 | raise LexError(s, pos)
|
| 374 | yield tok_id, end_pos
|
| 375 | pos = end_pos
|
| 376 |
|
| 377 |
|
| 378 | # Tag names:
|
| 379 | # Match <a or </a
|
| 380 | # Match <h2, but not <2h
|
| 381 | #
|
| 382 | # HTML 5 doesn't restrict tag names at all
|
| 383 | # https://html.spec.whatwg.org/#toc-syntax
|
| 384 | #
|
| 385 | # XML allows : - .
|
| 386 | # https://www.w3.org/TR/xml/#NT-NameChar
|
| 387 |
|
| 388 | # Namespaces for MathML, SVG
|
| 389 | # XLink, XML, XMLNS
|
| 390 | #
|
| 391 | # https://infra.spec.whatwg.org/#namespaces
|
| 392 | #
|
| 393 | # Allow - for td-attrs
|
| 394 |
|
| 395 | _ATTR_VALUE = r'[a-zA-Z0-9_\-]+' # allow hyphens
|
| 396 |
|
| 397 | # TODO: we don't need to capture the tag name here? That's done at the top
|
| 398 | # level
|
| 399 | _TAG_RE = re.compile(r'/? \s* (%s)' % _NAME, re.VERBOSE)
|
| 400 |
|
| 401 | # To match href="foo"
|
| 402 |
|
| 403 | _ATTR_RE = re.compile(
|
| 404 | r'''
|
| 405 | \s+ # Leading whitespace is required
|
| 406 | (%s) # Attribute name
|
| 407 | (?: # Optional attribute value
|
| 408 | \s* = \s*
|
| 409 | (?:
|
| 410 | " ([^>"]*) " # double quoted value
|
| 411 | | (%s) # Attribute value
|
| 412 | # TODO: relax this? for href=$foo
|
| 413 | )
|
| 414 | )?
|
| 415 | ''' % (_NAME, _ATTR_VALUE), re.VERBOSE)
|
| 416 |
|
| 417 | TagName, AttrName, UnquotedValue, QuotedValue = range(4)
|
| 418 |
|
| 419 |
|
| 420 | class TagLexer(object):
|
| 421 | """
|
| 422 | Given a tag like <a href="..."> or <link type="..." />, the TagLexer
|
| 423 | provides a few operations:
|
| 424 |
|
| 425 | - What is the tag?
|
| 426 | - Iterate through the attributes, giving (name, value_start_pos, value_end_pos)
|
| 427 | """
|
| 428 |
|
| 429 | def __init__(self, s):
|
| 430 | self.s = s
|
| 431 | self.start_pos = -1 # Invalid
|
| 432 | self.end_pos = -1
|
| 433 |
|
| 434 | def Reset(self, start_pos, end_pos):
|
| 435 | """Reuse instances of this object."""
|
| 436 | self.start_pos = start_pos
|
| 437 | self.end_pos = end_pos
|
| 438 |
|
| 439 | def TagString(self):
|
| 440 | return self.s[self.start_pos:self.end_pos]
|
| 441 |
|
| 442 | def TagName(self):
|
| 443 | # First event
|
| 444 | tok_id, start, end = next(self.Tokens())
|
| 445 | return self.s[start:end]
|
| 446 |
|
| 447 | def GetSpanForAttrValue(self, attr_name):
|
| 448 | # Algorithm: search for QuotedValue or UnquotedValue after AttrName
|
| 449 | # TODO: Could also cache these
|
| 450 |
|
| 451 | events = self.Tokens()
|
| 452 | val = (-1, -1)
|
| 453 | try:
|
| 454 | while True:
|
| 455 | tok_id, start, end = next(events)
|
| 456 | if tok_id == AttrName:
|
| 457 | name = self.s[start:end]
|
| 458 | if name == attr_name:
|
| 459 | # The value should come next
|
| 460 | tok_id, start, end = next(events)
|
| 461 | if tok_id in (QuotedValue, UnquotedValue):
|
| 462 | # Note: quoted values may have &
|
| 463 | # We would need ANOTHER lexer to unescape them.
|
| 464 | # Right now help_gen.py and oils_doc.py
|
| 465 | val = start, end
|
| 466 | break
|
| 467 |
|
| 468 | except StopIteration:
|
| 469 | pass
|
| 470 | return val
|
| 471 |
|
| 472 | def GetAttrRaw(self, attr_name):
|
| 473 | """
|
| 474 | Return the value, which may be UNESCAPED.
|
| 475 | """
|
| 476 | # Algorithm: search for QuotedValue or UnquotedValue after AttrName
|
| 477 | # TODO: Could also cache these
|
| 478 | start, end = self.GetSpanForAttrValue(attr_name)
|
| 479 | if start == -1:
|
| 480 | return None
|
| 481 | return self.s[start:end]
|
| 482 |
|
| 483 | def AllAttrsRaw(self):
|
| 484 | """
|
| 485 | Get a list of pairs [('class', 'foo'), ('href', '?foo=1&bar=2')]
|
| 486 |
|
| 487 | The quoted values may be escaped. We would need another lexer to
|
| 488 | unescape them.
|
| 489 | """
|
| 490 | pairs = []
|
| 491 | events = self.Tokens()
|
| 492 | try:
|
| 493 | while True:
|
| 494 | tok_id, start, end = next(events)
|
| 495 | if tok_id == AttrName:
|
| 496 | name = self.s[start:end]
|
| 497 |
|
| 498 | # The value should come next
|
| 499 | tok_id, start, end = next(events)
|
| 500 | if tok_id in (QuotedValue, UnquotedValue):
|
| 501 | # Note: quoted values may have &
|
| 502 | # We would need ANOTHER lexer to unescape them, but we
|
| 503 | # don't need that for ul-table
|
| 504 |
|
| 505 | val = self.s[start:end]
|
| 506 | pairs.append((name, val))
|
| 507 | except StopIteration:
|
| 508 | pass
|
| 509 | return pairs
|
| 510 |
|
| 511 | def Tokens(self):
|
| 512 | """
|
| 513 | Yields a sequence of tokens: Tag (AttrName AttrValue?)*
|
| 514 |
|
| 515 | Where each Token is (Type, start_pos, end_pos)
|
| 516 |
|
| 517 | Note that start and end are NOT redundant! We skip over some unwanted
|
| 518 | characters.
|
| 519 | """
|
| 520 | m = _TAG_RE.match(self.s, self.start_pos + 1)
|
| 521 | if not m:
|
| 522 | raise RuntimeError("Couldn't find HTML tag in %r" %
|
| 523 | self.TagString())
|
| 524 | yield TagName, m.start(1), m.end(1)
|
| 525 |
|
| 526 | pos = m.end(0)
|
| 527 |
|
| 528 | while True:
|
| 529 | # don't search past the end
|
| 530 | m = _ATTR_RE.match(self.s, pos, self.end_pos)
|
| 531 | if not m:
|
| 532 | # A validating parser would check that > or /> is next -- there's no junk
|
| 533 | break
|
| 534 |
|
| 535 | yield AttrName, m.start(1), m.end(1)
|
| 536 |
|
| 537 | # Quoted is group 2, unquoted is group 3.
|
| 538 | if m.group(2) is not None:
|
| 539 | yield QuotedValue, m.start(2), m.end(2)
|
| 540 | elif m.group(3) is not None:
|
| 541 | yield UnquotedValue, m.start(3), m.end(3)
|
| 542 |
|
| 543 | # Skip past the "
|
| 544 | pos = m.end(0)
|
| 545 |
|
| 546 |
|
| 547 | def ReadUntilStartTag(it, tag_lexer, tag_name):
|
| 548 | """Find the next <foo>, returning its (start, end) positions
|
| 549 |
|
| 550 | Raise ParseError if it's not found.
|
| 551 |
|
| 552 | tag_lexer is RESET.
|
| 553 | """
|
| 554 | pos = 0
|
| 555 | while True:
|
| 556 | try:
|
| 557 | tok_id, end_pos = next(it)
|
| 558 | except StopIteration:
|
| 559 | break
|
| 560 | tag_lexer.Reset(pos, end_pos)
|
| 561 | if tok_id == Tok.StartTag and tag_lexer.TagName() == tag_name:
|
| 562 | return pos, end_pos
|
| 563 |
|
| 564 | pos = end_pos
|
| 565 |
|
| 566 | raise ParseError('No start tag %r' % tag_name)
|
| 567 |
|
| 568 |
|
| 569 | def ReadUntilEndTag(it, tag_lexer, tag_name):
|
| 570 | """Find the next </foo>, returning its (start, end) position
|
| 571 |
|
| 572 | Raise ParseError if it's not found.
|
| 573 |
|
| 574 | tag_lexer is RESET.
|
| 575 | """
|
| 576 | pos = 0
|
| 577 | while True:
|
| 578 | try:
|
| 579 | tok_id, end_pos = next(it)
|
| 580 | except StopIteration:
|
| 581 | break
|
| 582 | tag_lexer.Reset(pos, end_pos)
|
| 583 | if tok_id == Tok.EndTag and tag_lexer.TagName() == tag_name:
|
| 584 | return pos, end_pos
|
| 585 |
|
| 586 | pos = end_pos
|
| 587 |
|
| 588 | raise ParseError('No end tag %r' % tag_name)
|
| 589 |
|
| 590 |
|
| 591 | CHAR_ENTITY = {
|
| 592 | 'amp': '&',
|
| 593 | 'lt': '<',
|
| 594 | 'gt': '>',
|
| 595 | 'quot': '"',
|
| 596 | }
|
| 597 |
|
| 598 |
|
| 599 | def ToText(s, left_pos=0, right_pos=-1):
|
| 600 | """Given HTML, return text by unquoting > and < etc.
|
| 601 |
|
| 602 | Used by:
|
| 603 | doctools/oils_doc.py: PygmentsPlugin
|
| 604 | doctools/help_gen.py: HelpIndexCards
|
| 605 |
|
| 606 | In the latter case, we cold process some tags, like:
|
| 607 |
|
| 608 | - Blue Link (not clickable, but still useful)
|
| 609 | - Red X
|
| 610 |
|
| 611 | That should be html.ToAnsi.
|
| 612 | """
|
| 613 | f = StringIO()
|
| 614 | out = Output(s, f, left_pos, right_pos)
|
| 615 |
|
| 616 | pos = left_pos
|
| 617 | for tok_id, end_pos in ValidTokens(s, left_pos, right_pos):
|
| 618 | if tok_id == Tok.RawData:
|
| 619 | out.SkipTo(pos)
|
| 620 | out.PrintUntil(end_pos)
|
| 621 |
|
| 622 | elif tok_id == Tok.CharEntity: # &
|
| 623 |
|
| 624 | entity = s[pos + 1:end_pos - 1]
|
| 625 |
|
| 626 | out.SkipTo(pos)
|
| 627 | out.Print(CHAR_ENTITY[entity])
|
| 628 | out.SkipTo(end_pos)
|
| 629 |
|
| 630 | # Not handling these yet
|
| 631 | elif tok_id == Tok.HexChar:
|
| 632 | raise AssertionError('Hex Char %r' % s[pos:pos + 20])
|
| 633 |
|
| 634 | elif tok_id == Tok.DecChar:
|
| 635 | raise AssertionError('Dec Char %r' % s[pos:pos + 20])
|
| 636 |
|
| 637 | pos = end_pos
|
| 638 |
|
| 639 | out.PrintTheRest()
|
| 640 | return f.getvalue()
|
| 641 |
|
| 642 |
|
| 643 | # https://developer.mozilla.org/en-US/docs/Glossary/Void_element
|
| 644 | VOID_ELEMENTS = [
|
| 645 | 'area',
|
| 646 | 'base',
|
| 647 | 'br',
|
| 648 | 'col',
|
| 649 | 'embed',
|
| 650 | 'hr',
|
| 651 | 'img',
|
| 652 | 'input',
|
| 653 | 'link',
|
| 654 | 'meta',
|
| 655 | 'param',
|
| 656 | 'source',
|
| 657 | 'track',
|
| 658 | 'wbr',
|
| 659 | ]
|
| 660 |
|
| 661 |
|
| 662 | def main(argv):
|
| 663 | action = argv[1]
|
| 664 |
|
| 665 | if action in ('lex-tags', 'lex-attrs', 'lex-attr-values', 'well-formed'):
|
| 666 | num_tokens = 0
|
| 667 | num_start_tags = 0
|
| 668 | num_start_end_tags = 0
|
| 669 | num_attrs = 0
|
| 670 | max_tag_stack = 0
|
| 671 |
|
| 672 | errors = []
|
| 673 | i = 0
|
| 674 | for line in sys.stdin:
|
| 675 | name = line.strip()
|
| 676 | with open(name) as f:
|
| 677 | contents = f.read()
|
| 678 |
|
| 679 | tag_lexer = TagLexer(contents)
|
| 680 | lx = Lexer(contents)
|
| 681 | tokens = []
|
| 682 | start_pos = 0
|
| 683 | tag_stack = []
|
| 684 | try:
|
| 685 | while True:
|
| 686 | tok_id, end_pos = lx.Read()
|
| 687 |
|
| 688 | if tok_id == Tok.Invalid:
|
| 689 | raise LexError(contents, start_pos)
|
| 690 | if tok_id == Tok.EndOfStream:
|
| 691 | break
|
| 692 |
|
| 693 | tokens.append((tok_id, end_pos))
|
| 694 |
|
| 695 | if tok_id == Tok.StartEndTag:
|
| 696 | num_start_end_tags += 1
|
| 697 | if action in ('lex-attrs', 'lex-attr-values',
|
| 698 | 'well-formed'):
|
| 699 | tag_lexer.Reset(start_pos, end_pos)
|
| 700 | all_attrs = tag_lexer.AllAttrsRaw()
|
| 701 | num_attrs += len(all_attrs)
|
| 702 | elif tok_id == Tok.StartTag:
|
| 703 | num_start_tags += 1
|
| 704 | if action in ('lex-attrs', 'lex-attr-values',
|
| 705 | 'well-formed'):
|
| 706 | tag_lexer.Reset(start_pos, end_pos)
|
| 707 | all_attrs = tag_lexer.AllAttrsRaw()
|
| 708 |
|
| 709 | tag_name = lx.TagName()
|
| 710 | # Don't bother to check
|
| 711 | if tag_name not in VOID_ELEMENTS:
|
| 712 | tag_stack.append(tag_name)
|
| 713 | max_tag_stack = max(max_tag_stack, len(tag_stack))
|
| 714 | elif tok_id == Tok.EndTag:
|
| 715 | try:
|
| 716 | expected = tag_stack.pop()
|
| 717 | except IndexError:
|
| 718 | raise ParseError('Tag stack empty',
|
| 719 | s=contents,
|
| 720 | start_pos=start_pos)
|
| 721 |
|
| 722 | actual = lx.TagName()
|
| 723 | if expected != actual:
|
| 724 | raise ParseError(
|
| 725 | 'Expected closing tag %r, got %r' %
|
| 726 | (expected, actual),
|
| 727 | s=contents,
|
| 728 | start_pos=start_pos)
|
| 729 |
|
| 730 | start_pos = end_pos
|
| 731 | except LexError as e:
|
| 732 | log('Lex error in %r: %s', name, e)
|
| 733 | errors.append((name, e))
|
| 734 | except ParseError as e:
|
| 735 | log('Parse error in %r: %s', name, e)
|
| 736 | errors.append((name, e))
|
| 737 | else:
|
| 738 | num_tokens += len(tokens)
|
| 739 |
|
| 740 | #print('%d %s' % (len(tokens), name))
|
| 741 | i += 1
|
| 742 |
|
| 743 | log('')
|
| 744 | log(
|
| 745 | ' %d tokens, %d start/end tags, %d start tags, %d attrs, %d max tag stack depth in %d files',
|
| 746 | num_tokens, num_start_end_tags, num_start_tags, num_attrs,
|
| 747 | max_tag_stack, i)
|
| 748 | log(' %d errors', len(errors))
|
| 749 | if 0:
|
| 750 | for name, e in errors:
|
| 751 | log('Error in %r: %s', name, e)
|
| 752 |
|
| 753 | else:
|
| 754 | raise RuntimeError('Invalid action %r' % action)
|
| 755 |
|
| 756 |
|
| 757 | if __name__ == '__main__':
|
| 758 | main(sys.argv)
|