OILS / lazylex / html.py View on Github | oils.pub

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