1 | """data_lang/htm8.py
|
2 |
|
3 | TODO
|
4 |
|
5 | API:
|
6 | - Get rid of AttrValueLexer - this should be in the TagLexer
|
7 | - this also means that unquoted values can be more similar
|
8 | - We can use a single lexer mode for everything inside <>
|
9 | - the SPACE is the only difference
|
10 | - Deprecate tag_lexer.GetTagName() in favor of lx.CanonicalTagName() or
|
11 | _LiteralTagName()
|
12 | - UTF-8 check, like JSON8
|
13 | - re2c
|
14 | - port lexer, which will fix static typing issues
|
15 | - the abstraction needs to support submatch?
|
16 | - for finding the end of a tag, etc.?
|
17 |
|
18 | - LexError and ParseError need details
|
19 | - harmonize with data_lang/j8.py, which uses error.Decode(msg, ...,
|
20 | cur_line_num)
|
21 |
|
22 | - Copy all errors into doc/ref/chap-errors.md
|
23 | - This helps understand the language
|
24 |
|
25 | - Update doc/htm8.md
|
26 | - list of Algorithms:
|
27 | - lex just the top level
|
28 | - lex both levels
|
29 | - and match tags - this is the level for value.Htm8Frag?
|
30 | - convert to XML!
|
31 | - lazy selection by tag, or attr (id= and class=)
|
32 | - lazy selection by CSS selector expression
|
33 | - convert to DOMTree
|
34 | - sed-like replacement of DOM Tree or element
|
35 | - untrusted HTML filter, e.g. like StackOverflow / Reddit
|
36 | - this is Safe HTM8
|
37 | - should have a zero alloc way to support this, with good errors?
|
38 | - I think most of them silently strip data
|
39 | """
|
40 |
|
41 | import re
|
42 |
|
43 | from typing import Dict, List, Tuple, Optional, IO, Iterator, Any
|
44 |
|
45 | from _devbuild.gen.htm8_asdl import (h8_id, h8_id_t, h8_tag_id, h8_tag_id_t,
|
46 | h8_tag_id_str, attr_name, attr_name_t,
|
47 | attr_value_e, attr_value_t, h8_val_id)
|
48 | from doctools.util import log
|
49 |
|
50 |
|
51 | class LexError(Exception):
|
52 | """
|
53 | Examples of lex errors:
|
54 |
|
55 | - h8_id.Invalid, like <> or &&
|
56 | - Unclosed <!-- <? <![CDATA[ <script> <style>
|
57 | """
|
58 |
|
59 | def __init__(self, s, start_pos):
|
60 | # type: (str, int) -> None
|
61 | self.s = s
|
62 | self.start_pos = start_pos
|
63 |
|
64 | def __str__(self):
|
65 | # type: () -> str
|
66 | return '(LexError %r)' % (self.s[self.start_pos:self.start_pos + 20])
|
67 |
|
68 |
|
69 | def _FindLineNum(s, error_pos):
|
70 | # type: (str, int) -> int
|
71 | current_pos = 0
|
72 | line_num = 1
|
73 | while True:
|
74 | newline_pos = s.find('\n', current_pos)
|
75 | #log('current = %d, N %d, line %d', current_pos, newline_pos, line_num)
|
76 |
|
77 | if newline_pos == -1: # this is the last line
|
78 | return line_num
|
79 | if newline_pos >= error_pos:
|
80 | return line_num
|
81 | line_num += 1
|
82 | current_pos = newline_pos + 1
|
83 |
|
84 |
|
85 | class ParseError(Exception):
|
86 | """
|
87 | Examples of parse errors
|
88 |
|
89 | - unbalanced tag structure
|
90 | - ul_table.py errors
|
91 | """
|
92 |
|
93 | def __init__(self, msg, s=None, start_pos=-1):
|
94 | # type: (str, Optional[str], int) -> None
|
95 | self.msg = msg
|
96 | self.s = s
|
97 | self.start_pos = start_pos
|
98 |
|
99 | def __str__(self):
|
100 | # type: () -> str
|
101 | if self.s is not None:
|
102 | assert self.start_pos != -1, self.start_pos
|
103 | snippet = (self.s[self.start_pos:self.start_pos + 20])
|
104 |
|
105 | line_num = _FindLineNum(self.s, self.start_pos)
|
106 | else:
|
107 | snippet = ''
|
108 | line_num = -1
|
109 | msg = 'line %d: %r %r' % (line_num, self.msg, snippet)
|
110 | return msg
|
111 |
|
112 |
|
113 | class Output(object):
|
114 | """Takes an underlying input buffer and an output file. Maintains a
|
115 | position in the input buffer.
|
116 |
|
117 | Print FROM the input or print new text to the output.
|
118 | """
|
119 |
|
120 | def __init__(self, s, f, left_pos=0, right_pos=-1):
|
121 | # type: (str, IO[str], int, int) -> None
|
122 | self.s = s
|
123 | self.f = f
|
124 | self.pos = left_pos
|
125 | self.right_pos = len(s) if right_pos == -1 else right_pos
|
126 |
|
127 | def SkipTo(self, pos):
|
128 | # type: (int) -> None
|
129 | """Skip to a position."""
|
130 | self.pos = pos
|
131 |
|
132 | def PrintUntil(self, pos):
|
133 | # type: (int) -> None
|
134 | """Print until a position."""
|
135 | piece = self.s[self.pos:pos]
|
136 | self.f.write(piece)
|
137 | self.pos = pos
|
138 |
|
139 | def PrintTheRest(self):
|
140 | # type: () -> None
|
141 | """Print until the end of the string."""
|
142 | self.PrintUntil(self.right_pos)
|
143 |
|
144 | def Print(self, s):
|
145 | # type: (str) -> None
|
146 | """Print text to the underlying buffer."""
|
147 | self.f.write(s)
|
148 |
|
149 |
|
150 | def MakeLexer(rules):
|
151 | return [(re.compile(pat, re.VERBOSE), i) for (pat, i) in rules]
|
152 |
|
153 |
|
154 | #
|
155 | # Eggex
|
156 | #
|
157 | # Tag = / ~['>']+ /
|
158 |
|
159 | # Is this valid? A single character?
|
160 | # Tag = / ~'>'* /
|
161 |
|
162 | # Maybe better: / [NOT '>']+/
|
163 | # capital letters not allowed there?
|
164 | #
|
165 | # But then this is confusing:
|
166 | # / [NOT ~digit]+/
|
167 | #
|
168 | # / [NOT digit] / is [^\d]
|
169 | # / ~digit / is \D
|
170 | #
|
171 | # Or maybe:
|
172 | #
|
173 | # / [~ digit]+ /
|
174 | # / [~ '>']+ /
|
175 | # / [NOT '>']+ /
|
176 |
|
177 | # End = / '</' Tag '>' /
|
178 | # StartEnd = / '<' Tag '/>' /
|
179 | # Start = / '<' Tag '>' /
|
180 | #
|
181 | # EntityRef = / '&' dot{* N} ';' /
|
182 |
|
183 | # Tag name, or attribute name
|
184 | # colon is used in XML
|
185 |
|
186 | # https://www.w3.org/TR/xml/#NT-Name
|
187 | # Hm there is a lot of unicode stuff. We are simplifying parsing
|
188 |
|
189 | _NAME = r'[a-zA-Z][a-zA-Z0-9:_\-]*' # must start with letter
|
190 |
|
191 | CHAR_LEX = [
|
192 | # Characters
|
193 | # https://www.w3.org/TR/xml/#sec-references
|
194 | (r'&\# [0-9]+ ;', h8_id.DecChar),
|
195 | (r'&\# x[0-9a-fA-F]+ ;', h8_id.HexChar),
|
196 | (r'& %s ;' % _NAME, h8_id.CharEntity),
|
197 | # Allow unquoted, and quoted
|
198 | (r'&', h8_id.BadAmpersand),
|
199 | ]
|
200 |
|
201 | HTM8_LEX = CHAR_LEX + [
|
202 | # TODO: CommentBegin, ProcessingBegin, CDataBegin could have an additional
|
203 | # action associated with them? The ending substring
|
204 | (r'<!--', h8_id.CommentBegin),
|
205 |
|
206 | # Processing instruction are used for the XML header:
|
207 | # <?xml version="1.0" encoding="UTF-8"?>
|
208 | # They are technically XML-only, but in HTML5, they are another kind of
|
209 | # comment:
|
210 | #
|
211 | # https://developer.mozilla.org/en-US/docs/Web/API/ProcessingInstruction
|
212 | #
|
213 | (r'<\?', h8_id.ProcessingBegin),
|
214 | # Not necessary in HTML5, but occurs in XML
|
215 | (r'<!\[CDATA\[', h8_id.CDataBegin), # <![CDATA[
|
216 |
|
217 | # Markup declarations
|
218 | # - In HTML5, there is only <!DOCTYPE html>
|
219 | # - XML has 4 more declarations: <!ELEMENT ...> ATTLIST ENTITY NOTATION
|
220 | # - these seem to be part of DTD
|
221 | # - it's useful to skip these, and be able to parse the rest of the document
|
222 | # - Note: < is allowed?
|
223 | (r'<! [^>\x00]+ >', h8_id.Decl),
|
224 |
|
225 | # Tags
|
226 | # Notes:
|
227 | # - We look for a valid tag name, but we don't validate attributes.
|
228 | # That's done in the tag lexer.
|
229 | # - We don't allow leading whitespace
|
230 | (r'</ (%s) >' % _NAME, h8_id.EndTag),
|
231 | # self-closing <br/> comes before StartTag
|
232 | # could/should these be collapsed into one rule?
|
233 | (r'< (%s) [^>\x00]* />' % _NAME, h8_id.StartEndTag), # end </a>
|
234 | (r'< (%s) [^>\x00]* >' % _NAME, h8_id.StartTag), # start <a>
|
235 |
|
236 | # HTML5 allows unescaped > in raw data, but < is not allowed.
|
237 | # https://stackoverflow.com/questions/10462348/right-angle-bracket-in-html
|
238 | #
|
239 | # - My early blog has THREE errors when disallowing >
|
240 | # - So do some .wwz files
|
241 | (r'[^&<>\x00]+', h8_id.RawData),
|
242 | (r'>', h8_id.BadGreaterThan),
|
243 | # NUL is the end, an accomodation for re2c. Like we do in frontend/match.
|
244 | (r'\x00', h8_id.EndOfStream),
|
245 | # This includes < - it is not BadLessThan because it's NOT recoverable
|
246 | (r'.', h8_id.Invalid),
|
247 | ]
|
248 |
|
249 | # Old notes:
|
250 | #
|
251 | # Non-greedy matches are regular and can be matched in linear time
|
252 | # with RE2.
|
253 | #
|
254 | # https://news.ycombinator.com/item?id=27099798
|
255 | #
|
256 |
|
257 | # This person tried to do it with a regex:
|
258 | #
|
259 | # https://skeptric.com/html-comment-regexp/index.html
|
260 |
|
261 | # . is any char except newline
|
262 | # https://re2c.org/manual/manual_c.html
|
263 |
|
264 | # Discarded options
|
265 | #(r'<!-- .*? -->', h8_id.Comment),
|
266 |
|
267 | # Hack from Claude: \s\S instead of re.DOTALL. I don't like this
|
268 | #(r'<!-- [\s\S]*? -->', h8_id.Comment),
|
269 | #(r'<!-- (?:.|[\n])*? -->', h8_id.Comment),
|
270 |
|
271 | HTM8_LEX_COMPILED = MakeLexer(HTM8_LEX)
|
272 |
|
273 |
|
274 | class Lexer(object):
|
275 |
|
276 | def __init__(self, s, left_pos=0, right_pos=-1, no_special_tags=False):
|
277 | # type: (str, int, int, bool) -> None
|
278 | self.s = s
|
279 | self.pos = left_pos
|
280 | self.right_pos = len(s) if right_pos == -1 else right_pos
|
281 | self.no_special_tags = no_special_tags
|
282 |
|
283 | # string -> compiled regex pattern object
|
284 | self.cache = {} # type: Dict[str, Any]
|
285 |
|
286 | # either </script> or </style> - we search until we see that
|
287 | self.search_state = None # type: Optional[str]
|
288 |
|
289 | # Position of tag name, if applicable
|
290 | # - Set after you get a StartTag, EndTag, or StartEndTag
|
291 | # - Unset on other tags
|
292 | self.tag_pos_left = -1
|
293 | self.tag_pos_right = -1
|
294 |
|
295 | def _Read(self):
|
296 | # type: () -> Tuple[h8_id_t, int]
|
297 | if self.pos == self.right_pos:
|
298 | return h8_id.EndOfStream, self.pos
|
299 |
|
300 | assert self.pos < self.right_pos, self.pos
|
301 |
|
302 | if self.search_state is not None and not self.no_special_tags:
|
303 | # TODO: case-insensitive search for </SCRIPT> <SCRipt> ?
|
304 | #
|
305 | # Another strategy: enter a mode where we find ONLY the end tag
|
306 | # regex, and any data that's not <, and then check the canonical
|
307 | # tag name for 'script' or 'style'.
|
308 | pos = self.s.find(self.search_state, self.pos)
|
309 | if pos == -1:
|
310 | # unterminated <script> or <style>
|
311 | raise LexError(self.s, self.pos)
|
312 | self.search_state = None
|
313 | # beginning
|
314 | return h8_id.HtmlCData, pos
|
315 |
|
316 | # Find the first match.
|
317 | # Note: frontend/match.py uses _LongestMatch(), which is different!
|
318 | # TODO: reconcile them. This lexer should be expressible in re2c.
|
319 |
|
320 | for pat, tok_id in HTM8_LEX_COMPILED:
|
321 | m = pat.match(self.s, self.pos)
|
322 | if m:
|
323 | if tok_id in (h8_id.StartTag, h8_id.EndTag, h8_id.StartEndTag):
|
324 | self.tag_pos_left = m.start(1)
|
325 | self.tag_pos_right = m.end(1)
|
326 | else:
|
327 | # Reset state
|
328 | self.tag_pos_left = -1
|
329 | self.tag_pos_right = -1
|
330 |
|
331 | if tok_id == h8_id.CommentBegin:
|
332 | pos = self.s.find('-->', self.pos)
|
333 | if pos == -1:
|
334 | # unterminated <!--
|
335 | raise LexError(self.s, self.pos)
|
336 | return h8_id.Comment, pos + 3 # -->
|
337 |
|
338 | if tok_id == h8_id.ProcessingBegin:
|
339 | pos = self.s.find('?>', self.pos)
|
340 | if pos == -1:
|
341 | # unterminated <?
|
342 | raise LexError(self.s, self.pos)
|
343 | return h8_id.Processing, pos + 2 # ?>
|
344 |
|
345 | if tok_id == h8_id.CDataBegin:
|
346 | pos = self.s.find(']]>', self.pos)
|
347 | if pos == -1:
|
348 | # unterminated <![CDATA[
|
349 | raise LexError(self.s, self.pos)
|
350 | return h8_id.CData, pos + 3 # ]]>
|
351 |
|
352 | if tok_id == h8_id.StartTag:
|
353 | # TODO: reduce allocations
|
354 | if (self.TagNameEquals('script') or
|
355 | self.TagNameEquals('style')):
|
356 | # <SCRipt a=b> -> </SCRipt>
|
357 | self.search_state = '</' + self._LiteralTagName() + '>'
|
358 |
|
359 | return tok_id, m.end()
|
360 | else:
|
361 | raise AssertionError('h8_id.Invalid rule should have matched')
|
362 |
|
363 | def TagNamePos(self):
|
364 | """The right position of the tag pos"""
|
365 | assert self.tag_pos_right != -1, self.tag_pos_right
|
366 | return self.tag_pos_right
|
367 |
|
368 | def TagNameEquals(self, expected):
|
369 | # type: (str) -> bool
|
370 | assert self.tag_pos_left != -1, self.tag_pos_left
|
371 | assert self.tag_pos_right != -1, self.tag_pos_right
|
372 |
|
373 | # TODO: In C++, this does not need an allocation. Can we test
|
374 | # directly?
|
375 | return expected == self.CanonicalTagName()
|
376 |
|
377 | def _LiteralTagName(self):
|
378 | # type: () -> str
|
379 | assert self.tag_pos_left != -1, self.tag_pos_left
|
380 | assert self.tag_pos_right != -1, self.tag_pos_right
|
381 |
|
382 | return self.s[self.tag_pos_left:self.tag_pos_right]
|
383 |
|
384 | def CanonicalTagName(self):
|
385 | # type: () -> str
|
386 | tag_name = self._LiteralTagName()
|
387 | # Most tags are already lower case, so avoid allocation with this conditional
|
388 | # TODO: this could go in the mycpp runtime?
|
389 | if tag_name.islower():
|
390 | return tag_name
|
391 | else:
|
392 | return tag_name.lower()
|
393 |
|
394 | def Read(self):
|
395 | # type: () -> Tuple[h8_id_t, int]
|
396 | tok_id, end_pos = self._Read()
|
397 | self.pos = end_pos # advance
|
398 | return tok_id, end_pos
|
399 |
|
400 | def LookAhead(self, regex):
|
401 | # type: (str) -> bool
|
402 | """
|
403 | Currently used for ul_table.py. But taking a dynamic regex string is
|
404 | not the right interface.
|
405 | """
|
406 | # Cache the regex compilation. This could also be LookAheadFor(THEAD)
|
407 | # or something.
|
408 | pat = self.cache.get(regex)
|
409 | if pat is None:
|
410 | pat = re.compile(regex)
|
411 | self.cache[regex] = pat
|
412 |
|
413 | m = pat.match(self.s, self.pos)
|
414 | return m is not None
|
415 |
|
416 |
|
417 | A_NAME_LEX = [
|
418 | # Leading whitespace is required, to separate attributes.
|
419 | #
|
420 | # If the = is not present, then we set the lexer in a state for
|
421 | # attr_value_e.Missing.
|
422 | (r'\s+ (%s) \s* (=)? \s*' % _NAME, attr_name.Ok),
|
423 | # unexpected EOF
|
424 |
|
425 | # The closing > or /> is treated as end of stream, and it's not an error.
|
426 | (r'\s* /? >', attr_name.Done),
|
427 |
|
428 | # NUL should not be possible, because the top-level
|
429 |
|
430 | # This includes < - it is not BadLessThan because it's NOT recoverable
|
431 | (r'.', attr_name.Invalid),
|
432 | ]
|
433 |
|
434 | A_NAME_LEX_COMPILED = MakeLexer(A_NAME_LEX)
|
435 |
|
436 | # Here we just loop on regular tokens
|
437 | #
|
438 | # Examples:
|
439 | # <a href = unquoted&foo >
|
440 | # <a href = unquoted&foo > # BadAmpersand is allowed I guess
|
441 | # <a href ="unquoted&foo" > # double quoted
|
442 | # <a href ='unquoted&foo' > # single quoted
|
443 | # <a href = what"foo" > # HTML5 allows this, but we could disallow it if
|
444 | # it's not common. It opens up the j"" and $"" extensions
|
445 | # <a href = what'foo' > # ditto
|
446 | #
|
447 | # Problem: <a href=foo/> - this is hard to recognize
|
448 | # Because is the unquoted value "foo/" or "foo" ?
|
449 |
|
450 | # Be very lenient - just no whitespace or special HTML chars
|
451 | # I don't think this is more lenient than HTML5, though we should check.
|
452 | #
|
453 | # Bug fix: Also disallow /
|
454 |
|
455 | # TODO: get rid of OLD copy
|
456 | _UNQUOTED_VALUE_OLD = r'''[^ \t\r\n<>&/"'\x00]*'''
|
457 | _UNQUOTED_VALUE = r'''[^ \t\r\n<>&/"'\x00]+'''
|
458 |
|
459 | # Restrictive definition, similar to _NAME
|
460 | # I was trying to capture #ble.sh and so forth
|
461 | # I also had unquoted //github.com, etc.
|
462 |
|
463 | # _UNQUOTED_VALUE = r'''[a-zA-Z0-9:_\-]+'''
|
464 | #
|
465 | # For now, I guess we live with <a href=?foo/>
|
466 |
|
467 | A_VALUE_LEX = CHAR_LEX + [
|
468 | (r'"', h8_val_id.DoubleQuote),
|
469 | (r"'", h8_val_id.SingleQuote),
|
470 | (_UNQUOTED_VALUE, h8_val_id.UnquotedVal),
|
471 |
|
472 | #(r'[ \r\n\t]', h8_id.Whitespace), # terminates unquoted values
|
473 | #(r'[^ \r\n\t&>\x00]', h8_id.RawData),
|
474 | #(r'[>\x00]', h8_id.EndOfStream),
|
475 | # e.g. < is an error
|
476 | (r'.', h8_val_id.NoMatch),
|
477 | ]
|
478 |
|
479 | A_VALUE_LEX_COMPILED = MakeLexer(A_VALUE_LEX)
|
480 |
|
481 |
|
482 | class AttrLexer(object):
|
483 | """
|
484 | Typical usage:
|
485 |
|
486 | while True:
|
487 | n, start_pos, end_pos = attr_lx.ReadName()
|
488 | if n == attr_name.Ok:
|
489 | if attr_lx.AttrNameEquals('div'):
|
490 | print('div')
|
491 |
|
492 | # TODO: also pass Optional[List[]] out_tokens?
|
493 | v, start_pos, end_pos = attr_lx.ReadRawValue()
|
494 | """
|
495 |
|
496 | def __init__(self, s):
|
497 | # type: (str) -> None
|
498 | self.s = s
|
499 | self.tag_name_pos = -1 # Invalid
|
500 | self.tag_end_pos = -1
|
501 | self.pos = -1
|
502 |
|
503 | self.name_start = -1
|
504 | self.name_end = -1
|
505 | self.next_value_is_missing = False
|
506 |
|
507 | def Init(self, tag_name_pos, end_pos):
|
508 | # type: (int, int) -> None
|
509 | """Initialize so we can read names and values.
|
510 |
|
511 | Example:
|
512 | 'x <a y>' # tag_name_pos=4, end_pos=6
|
513 | 'x <a>' # tag_name_pos=4, end_pos=4
|
514 |
|
515 | The Reset() method is used to reuse instances of the AttrLexer object.
|
516 | """
|
517 | assert tag_name_pos >= 0, tag_name_pos
|
518 | assert end_pos >= 0, end_pos
|
519 |
|
520 | log('TAG NAME POS %d', tag_name_pos)
|
521 |
|
522 | self.tag_name_pos = tag_name_pos
|
523 | self.end_pos = end_pos
|
524 |
|
525 | self.pos = tag_name_pos
|
526 |
|
527 | def ReadName(self):
|
528 | # type: () -> Tuple[attr_name_t, int, int]
|
529 | """Reads the attribute name
|
530 |
|
531 | EOF case:
|
532 | <a>
|
533 | <a >
|
534 |
|
535 | Error case:
|
536 | <a !>
|
537 | <a foo=bar !>
|
538 | """
|
539 | for pat, a in A_NAME_LEX_COMPILED:
|
540 | m = pat.match(self.s, self.pos)
|
541 | if m:
|
542 | self.pos = m.end(0) # Advance
|
543 |
|
544 | if a == attr_name.Ok:
|
545 | #log('%r', m.groups())
|
546 | self.name_start = m.start(1)
|
547 | self.name_end = m.end(1)
|
548 | # Set state based on =
|
549 | if m.group(2) is None:
|
550 | self.next_value_is_missing = True
|
551 | return attr_name.Ok, self.name_start, self.name_end
|
552 | else:
|
553 | # Reset state - e.g. you must call AttrNameEquals
|
554 | self.name_start = -1
|
555 | self.name_end = -1
|
556 | self.next_value_is_missing = False
|
557 |
|
558 | if a == attr_name.Invalid:
|
559 | return attr_name.Invalid, -1, -1
|
560 | if a == attr_name.Done:
|
561 | return attr_name.Done, -1, -1
|
562 | else:
|
563 | raise AssertionError('h8_id.Invalid rule should have matched')
|
564 |
|
565 | def _CanonicalAttrName(self):
|
566 | # type: () -> str
|
567 | assert self.name_start >= 0, self.name_start
|
568 | assert self.name_end >= 0, self.name_end
|
569 |
|
570 | attr_name = self.s[self.name_start:self.name_end]
|
571 | if attr_name.islower():
|
572 | return attr_name
|
573 | else:
|
574 | return attr_name.lower()
|
575 |
|
576 | def AttrNameEquals(self, expected):
|
577 | # type: (str) -> bool
|
578 | """
|
579 | TODO: Must call this after ReadName() ?
|
580 | Because that can FAIL.
|
581 | """
|
582 | return expected == self._CanonicalAttrName()
|
583 |
|
584 | def _QuotedRead(self):
|
585 | # type: () -> Tuple[h8_id, end_pos]
|
586 |
|
587 | for pat, tok_id in QUOTED_VALUE_LEX_COMPILED:
|
588 | m = pat.match(self.s, self.pos)
|
589 | if m:
|
590 | end_pos = m.end(0) # Advance
|
591 | log('_QuotedRead %r', self.s[self.pos:end_pos])
|
592 | return tok_id, end_pos
|
593 | else:
|
594 | context = self.s[self.pos:self.pos + 10]
|
595 | raise AssertionError('h8_id.Invalid rule should have matched %r' %
|
596 | context)
|
597 |
|
598 | def ReadRawValue(self):
|
599 | # type: () -> Tuple[attr_value_t, int, int]
|
600 | """Read the attribute value.
|
601 |
|
602 | In general, it is escaped or "raw"
|
603 |
|
604 | Can only be called after a SUCCESSFUL ReadName().
|
605 | Assuming ReadName() returned a value, this should NOT fail.
|
606 | """
|
607 | # ReadName() invariant
|
608 | assert self.name_start >= 0, self.name_start
|
609 | assert self.name_end >= 0, self.name_end
|
610 |
|
611 | self.name_start = -1
|
612 | self.name_end = -1
|
613 |
|
614 | if self.next_value_is_missing:
|
615 | return attr_value_e.Missing, -1, -1
|
616 |
|
617 | # Now read " ', unquoted or empty= is valid too.
|
618 | for pat, a in A_VALUE_LEX_COMPILED:
|
619 | m = pat.match(self.s, self.pos)
|
620 | if m:
|
621 | self.pos = m.end(0) # Advance
|
622 |
|
623 | #log('m %s', m.groups())
|
624 | if a == h8_val_id.UnquotedVal:
|
625 | return attr_value_e.Unquoted, m.start(0), m.end(0)
|
626 |
|
627 | if a == h8_val_id.DoubleQuote:
|
628 | left_inner = self.pos
|
629 | while True:
|
630 | tok_id, q_end_pos = self._QuotedRead()
|
631 | if tok_id == h8_id.Invalid:
|
632 | raise LexError(self.s, self.pos)
|
633 | if tok_id == h8_id.DoubleQuote:
|
634 | return attr_value_e.DoubleQuoted, left_inner, self.pos
|
635 | self.pos = q_end_pos # advance
|
636 |
|
637 | if a == h8_val_id.SingleQuote:
|
638 |
|
639 | left_inner = self.pos
|
640 | while True:
|
641 | tok_id, q_end_pos = self._QuotedRead()
|
642 | if tok_id == h8_id.Invalid:
|
643 | raise LexError(self.s, self.pos)
|
644 | if tok_id == h8_id.SingleQuote:
|
645 | return attr_value_e.SingleQuoted, left_inner, self.pos
|
646 | self.pos = q_end_pos # advance
|
647 |
|
648 | if a == h8_val_id.NoMatch:
|
649 | # <a foo = >
|
650 | return attr_value_e.Empty, -1, -1
|
651 | else:
|
652 | raise AssertionError('h8_val_id.NoMatch rule should have matched')
|
653 |
|
654 | def SkipValue(self):
|
655 | # type: () -> None
|
656 | # Just ignore it and return
|
657 | self.ReadRawValue()
|
658 |
|
659 | def ReadValueAndDecode(self):
|
660 | # type: () -> str
|
661 | """Read the attribute vlaue
|
662 | """
|
663 | # TODO: tokenize it
|
664 | pass
|
665 |
|
666 |
|
667 | # Tag names:
|
668 | # Match <a or </a
|
669 | # Match <h2, but not <2h
|
670 | #
|
671 | # HTML 5 doesn't restrict tag names at all
|
672 | # https://html.spec.whatwg.org/#toc-syntax
|
673 | #
|
674 | # XML allows : - .
|
675 | # https://www.w3.org/TR/xml/#NT-NameChar
|
676 |
|
677 | # Namespaces for MathML, SVG
|
678 | # XLink, XML, XMLNS
|
679 | #
|
680 | # https://infra.spec.whatwg.org/#namespaces
|
681 | #
|
682 | # Allow - for td-attrs
|
683 |
|
684 | # TODO: we don't need to capture the tag name here? That's done at the top
|
685 | # level
|
686 | _TAG_RE = re.compile(r'/? \s* (%s)' % _NAME, re.VERBOSE)
|
687 |
|
688 | _TAG_LAST_RE = re.compile(r'\s* /? >', re.VERBOSE)
|
689 |
|
690 | # To match href="foo"
|
691 | # Note: in HTML5 and XML, single quoted attributes are also valid
|
692 |
|
693 | # <button disabled> is standard usage
|
694 |
|
695 | # NOTE: This used to allow whitespace around =
|
696 | # <a foo = "bar"> makes sense in XML
|
697 | # But then you also have
|
698 | # <a foo= bar> - which is TWO attributes, in HTML5
|
699 | # So the space is problematic
|
700 |
|
701 | _ATTR_RE = re.compile(
|
702 | r'''
|
703 | \s+ # Leading whitespace is required
|
704 | (%s) # Attribute name
|
705 | (?: # Optional attribute value
|
706 | \s* = \s* # Spaces allowed around =
|
707 | (?:
|
708 | " ([^>"\x00]*) " # double quoted value
|
709 | | ' ([^>'\x00]*) ' # single quoted value
|
710 | | (%s) # Attribute value
|
711 | )
|
712 | )?
|
713 | ''' % (_NAME, _UNQUOTED_VALUE_OLD), re.VERBOSE)
|
714 |
|
715 |
|
716 | class TagLexer(object):
|
717 | """
|
718 | Given a tag like <a href="..."> or <link type="..." />, the TagLexer
|
719 | provides a few operations:
|
720 |
|
721 | - What is the tag?
|
722 | - Iterate through the attributes, giving (name, value_start_pos, value_end_pos)
|
723 | """
|
724 |
|
725 | def __init__(self, s):
|
726 | # type: (str) -> None
|
727 | self.s = s
|
728 | self.start_pos = -1 # Invalid
|
729 | self.end_pos = -1
|
730 |
|
731 | def Reset(self, start_pos, end_pos):
|
732 | # type: (int, int) -> None
|
733 | """Reuse instances of this object."""
|
734 | assert start_pos >= 0, start_pos
|
735 | assert end_pos >= 0, end_pos
|
736 |
|
737 | self.start_pos = start_pos
|
738 | self.end_pos = end_pos
|
739 |
|
740 | def WholeTagString(self):
|
741 | # type: () -> str
|
742 | """Return the entire tag string, e.g. <a href='foo'>"""
|
743 | return self.s[self.start_pos:self.end_pos]
|
744 |
|
745 | def GetTagName(self):
|
746 | # type: () -> str
|
747 | # First event
|
748 | tok_id, start, end = next(self.Tokens())
|
749 | return self.s[start:end]
|
750 |
|
751 | def GetSpanForAttrValue(self, attr_name):
|
752 | # type: (str) -> Tuple[int, int]
|
753 | """
|
754 | Used by oils_doc.py, for href shortcuts
|
755 | """
|
756 | # Algorithm: search for QuotedValue or UnquotedValue after AttrName
|
757 | # TODO: Could also cache these
|
758 |
|
759 | events = self.Tokens()
|
760 | val = (-1, -1)
|
761 | try:
|
762 | while True:
|
763 | tok_id, start, end = next(events)
|
764 | if tok_id == h8_tag_id.AttrName:
|
765 | name = self.s[start:end]
|
766 | if name == attr_name:
|
767 | # The value should come next
|
768 | tok_id, start, end = next(events)
|
769 | assert tok_id in (
|
770 | h8_tag_id.QuotedValue, h8_tag_id.UnquotedValue,
|
771 | h8_tag_id.MissingValue), h8_tag_id_str(tok_id)
|
772 | val = start, end
|
773 | break
|
774 |
|
775 | except StopIteration:
|
776 | pass
|
777 | return val
|
778 |
|
779 | def GetAttrRaw(self, attr_name):
|
780 | # type: (str) -> Optional[str]
|
781 | """
|
782 | Return the value, which may be UNESCAPED.
|
783 | """
|
784 | start, end = self.GetSpanForAttrValue(attr_name)
|
785 | if start == -1:
|
786 | return None
|
787 | return self.s[start:end]
|
788 |
|
789 | def AllAttrsRawSlice(self):
|
790 | # type: () -> List[Tuple[str, int, int]]
|
791 | """
|
792 | Get a list of pairs [('class', 3, 5), ('href', 9, 12)]
|
793 | """
|
794 | slices = []
|
795 | events = self.Tokens()
|
796 | try:
|
797 | while True:
|
798 | tok_id, start, end = next(events)
|
799 | if tok_id == h8_tag_id.AttrName:
|
800 | name = self.s[start:end]
|
801 |
|
802 | # The value should come next
|
803 | tok_id, start, end = next(events)
|
804 | assert tok_id in (
|
805 | h8_tag_id.QuotedValue, h8_tag_id.UnquotedValue,
|
806 | h8_tag_id.MissingValue), h8_tag_id_str(tok_id)
|
807 | # Note: quoted values may have &
|
808 | # We would need ANOTHER lexer to unescape them, but we
|
809 | # don't need that for ul-table
|
810 | slices.append((name, start, end))
|
811 | except StopIteration:
|
812 | pass
|
813 | return slices
|
814 |
|
815 | def AllAttrsRaw(self):
|
816 | # type: () -> List[Tuple[str, str]]
|
817 | """
|
818 | Get a list of pairs [('class', 'foo'), ('href', '?foo=1&bar=2')]
|
819 |
|
820 | The quoted values may be escaped. We would need another lexer to
|
821 | unescape them.
|
822 | """
|
823 | slices = self.AllAttrsRawSlice()
|
824 | pairs = []
|
825 | for name, start, end in slices:
|
826 | pairs.append((name, self.s[start:end]))
|
827 | return pairs
|
828 |
|
829 | def Tokens(self):
|
830 | # type: () -> Iterator[Tuple[h8_tag_id_t, int, int]]
|
831 | """
|
832 | Yields a sequence of tokens: Tag (AttrName AttrValue?)*
|
833 |
|
834 | Where each Token is (Type, start_pos, end_pos)
|
835 |
|
836 | Note that start and end are NOT redundant! We skip over some unwanted
|
837 | characters.
|
838 | """
|
839 | m = _TAG_RE.match(self.s, self.start_pos + 1)
|
840 | if not m:
|
841 | raise RuntimeError("Couldn't find HTML tag in %r" %
|
842 | self.WholeTagString())
|
843 | yield h8_tag_id.TagName, m.start(1), m.end(1)
|
844 |
|
845 | pos = m.end(0)
|
846 | #log('POS %d', pos)
|
847 |
|
848 | while True:
|
849 | # don't search past the end
|
850 | m = _ATTR_RE.match(self.s, pos, self.end_pos)
|
851 | if not m:
|
852 | #log('BREAK pos %d', pos)
|
853 | break
|
854 | #log('AttrName %r', m.group(1))
|
855 |
|
856 | yield h8_tag_id.AttrName, m.start(1), m.end(1)
|
857 |
|
858 | #log('m.groups() %r', m.groups())
|
859 | if m.group(2) is not None:
|
860 | # double quoted
|
861 | yield h8_tag_id.QuotedValue, m.start(2), m.end(2)
|
862 | elif m.group(3) is not None:
|
863 | # single quoted - TODO: could have different token types
|
864 | yield h8_tag_id.QuotedValue, m.start(3), m.end(3)
|
865 | elif m.group(4) is not None:
|
866 | yield h8_tag_id.UnquotedValue, m.start(4), m.end(4)
|
867 | else:
|
868 | # <button disabled>
|
869 | end = m.end(0)
|
870 | yield h8_tag_id.MissingValue, end, end
|
871 |
|
872 | # Skip past the "
|
873 | pos = m.end(0)
|
874 |
|
875 | #log('TOK %r', self.s)
|
876 |
|
877 | m = _TAG_LAST_RE.match(self.s, pos)
|
878 | #log('_TAG_LAST_RE match %r', self.s[pos:])
|
879 | if not m:
|
880 | # Extra data at end of tag. TODO: add messages for all these.
|
881 | raise LexError(self.s, pos)
|
882 |
|
883 |
|
884 | # This is similar but not identical to
|
885 | # " ([^>"\x00]*) " # double quoted value
|
886 | # | ' ([^>'\x00]*) ' # single quoted value
|
887 | #
|
888 | # Note: for unquoted values, & isn't allowed, and thus & and c and
|
889 | # ™ are not allowed. We could relax that?
|
890 | ATTR_VALUE_LEX = CHAR_LEX + [
|
891 | (r'[^>&\x00]+', h8_id.RawData),
|
892 | (r'.', h8_id.Invalid),
|
893 | ]
|
894 |
|
895 | ATTR_VALUE_LEX_COMPILED = MakeLexer(ATTR_VALUE_LEX)
|
896 |
|
897 | QUOTED_VALUE_LEX = CHAR_LEX + [
|
898 | (r'"', h8_id.DoubleQuote),
|
899 | (r"'", h8_id.SingleQuote),
|
900 | (r'<', h8_id.BadLessThan), # BadAmpersand is in CharLex
|
901 | (r'''[^"'<>&\x00]+''', h8_id.RawData),
|
902 | # This includes > - it is not BadGreaterThan because it's NOT recoverable
|
903 | (r'.', h8_id.Invalid),
|
904 | ]
|
905 |
|
906 | QUOTED_VALUE_LEX_COMPILED = MakeLexer(QUOTED_VALUE_LEX)
|
907 |
|
908 |
|
909 | class AttrValueLexer(object):
|
910 | """
|
911 | <a href="foo=99&bar">
|
912 | <a href='foo=99&bar'>
|
913 | <a href=unquoted>
|
914 | """
|
915 |
|
916 | def __init__(self, s):
|
917 | # type: (str) -> None
|
918 | self.s = s
|
919 | self.start_pos = -1 # Invalid
|
920 | self.end_pos = -1
|
921 |
|
922 | def Reset(self, start_pos, end_pos):
|
923 | # type: (int, int) -> None
|
924 | """Reuse instances of this object."""
|
925 | assert start_pos >= 0, start_pos
|
926 | assert end_pos >= 0, end_pos
|
927 |
|
928 | self.start_pos = start_pos
|
929 | self.end_pos = end_pos
|
930 |
|
931 | def NumTokens(self):
|
932 | # type: () -> int
|
933 | num_tokens = 0
|
934 | pos = self.start_pos
|
935 | for tok_id, end_pos in self.Tokens():
|
936 | if tok_id == h8_id.Invalid:
|
937 | raise LexError(self.s, pos)
|
938 | pos = end_pos
|
939 | #log('pos %d', pos)
|
940 | num_tokens += 1
|
941 | return num_tokens
|
942 |
|
943 | def Tokens(self):
|
944 | # type: () -> Iterator[Tuple[h8_id_t, int]]
|
945 | pos = self.start_pos
|
946 | while pos < self.end_pos:
|
947 | # Find the first match, like above.
|
948 | # Note: frontend/match.py uses _LongestMatch(), which is different!
|
949 | # TODO: reconcile them. This lexer should be expressible in re2c.
|
950 | for pat, tok_id in ATTR_VALUE_LEX_COMPILED:
|
951 | m = pat.match(self.s, pos)
|
952 | if m:
|
953 | if 0:
|
954 | tok_str = m.group(0)
|
955 | log('token = %r', tok_str)
|
956 |
|
957 | end_pos = m.end(0)
|
958 | yield tok_id, end_pos
|
959 | pos = end_pos
|
960 | break
|
961 | else:
|
962 | raise AssertionError('h8_id.Invalid rule should have matched')
|