1 | """data_lang/htm8.py
|
2 |
|
3 | TODO
|
4 |
|
5 | Migrate:
|
6 |
|
7 | - maybe: migrate everything off of TagLexer()
|
8 | - and AttrValueLexer() - this should requires Validate()
|
9 |
|
10 | API:
|
11 | - Get rid of Reset()?
|
12 | - Deprecate tag_lexer.GetTagName() in favor of lx.CanonicalTagName() or
|
13 | _LiteralTagName()
|
14 | - UTF-8 check, like JSON8
|
15 | - re2c
|
16 | - port lexer, which will fix static typing issues
|
17 | - the abstraction needs to support submatch?
|
18 | - for finding the end of a tag, etc.?
|
19 | - and what about no match?
|
20 |
|
21 | - harmonize LexError and ParseError with data_lang/j8.py, which uses
|
22 | error.Decode(msg, ..., cur_line_num)
|
23 |
|
24 | - Copy all errors into doc/ref/chap-errors.md
|
25 | - This helps understand the language
|
26 |
|
27 | - Update doc/htm8.md
|
28 | - list of Algorithms:
|
29 | - lex just the top level
|
30 | - lex both levels
|
31 | - and match tags - this is the level for value.Htm8Frag?
|
32 | - convert to XML!
|
33 | - lazy selection by tag, or attr (id= and class=)
|
34 | - lazy selection by CSS selector expression
|
35 | - convert to DOMTree
|
36 | - sed-like replacement of DOM Tree or element
|
37 | - untrusted HTML filter, e.g. like StackOverflow / Reddit
|
38 | - this is Safe HTM8
|
39 | - should have a zero alloc way to support this, with good errors?
|
40 | - I think most of them silently strip data
|
41 | """
|
42 |
|
43 | import re
|
44 |
|
45 | from typing import Dict, List, Tuple, Optional, IO, Iterator, Any
|
46 |
|
47 | from _devbuild.gen.htm8_asdl import (h8_id, h8_id_t, h8_tag_id, h8_tag_id_t,
|
48 | h8_tag_id_str, attr_name, attr_name_t,
|
49 | attr_name_str, attr_value_e, attr_value_t,
|
50 | h8_val_id)
|
51 | from doctools.util import log
|
52 |
|
53 |
|
54 | class LexError(Exception):
|
55 | """
|
56 | Examples of lex errors:
|
57 |
|
58 | - h8_id.Invalid, like <> or &&
|
59 | - Unclosed <!-- <? <![CDATA[ <script> <style>
|
60 | """
|
61 |
|
62 | def __init__(self, msg, code_str, start_pos):
|
63 | # type: (str, str, int) -> None
|
64 | self.msg = msg
|
65 | self.code_str = code_str
|
66 | self.start_pos = start_pos
|
67 |
|
68 | def __str__(self):
|
69 | # type: () -> str
|
70 | return '(LexError %r %r)' % (
|
71 | self.msg, self.code_str[self.start_pos:self.start_pos + 20])
|
72 |
|
73 |
|
74 | def _FindLineNum(s, error_pos):
|
75 | # type: (str, int) -> int
|
76 | current_pos = 0
|
77 | line_num = 1
|
78 | while True:
|
79 | newline_pos = s.find('\n', current_pos)
|
80 | #log('current = %d, N %d, line %d', current_pos, newline_pos, line_num)
|
81 |
|
82 | if newline_pos == -1: # this is the last line
|
83 | return line_num
|
84 | if newline_pos >= error_pos:
|
85 | return line_num
|
86 | line_num += 1
|
87 | current_pos = newline_pos + 1
|
88 |
|
89 |
|
90 | class ParseError(Exception):
|
91 | """
|
92 | Examples of parse errors
|
93 |
|
94 | - unbalanced tag structure
|
95 | - ul_table.py errors
|
96 | """
|
97 |
|
98 | def __init__(self, msg, s=None, start_pos=-1):
|
99 | # type: (str, Optional[str], int) -> None
|
100 | self.msg = msg
|
101 | self.s = s
|
102 | self.start_pos = start_pos
|
103 |
|
104 | def __str__(self):
|
105 | # type: () -> str
|
106 | if self.s is not None:
|
107 | assert self.start_pos != -1, self.start_pos
|
108 | snippet = (self.s[self.start_pos:self.start_pos + 20])
|
109 |
|
110 | line_num = _FindLineNum(self.s, self.start_pos)
|
111 | else:
|
112 | snippet = ''
|
113 | line_num = -1
|
114 | msg = 'line %d: %r %r' % (line_num, self.msg, snippet)
|
115 | return msg
|
116 |
|
117 |
|
118 | class Output(object):
|
119 | """Output for sed-like "replacement" model.
|
120 |
|
121 | Takes an underlying input buffer and an output file. Maintains a position
|
122 | in the input buffer.
|
123 |
|
124 | Print FROM the input or print new text to the output.
|
125 | """
|
126 |
|
127 | def __init__(self, s, f, left_pos=0, right_pos=-1):
|
128 | # type: (str, IO[str], int, int) -> None
|
129 | self.s = s
|
130 | self.f = f
|
131 | self.pos = left_pos
|
132 | self.right_pos = len(s) if right_pos == -1 else right_pos
|
133 |
|
134 | def SkipTo(self, pos):
|
135 | # type: (int) -> None
|
136 | """Skip to a position."""
|
137 | self.pos = pos
|
138 |
|
139 | def PrintUntil(self, pos):
|
140 | # type: (int) -> None
|
141 | """Print until a position."""
|
142 | piece = self.s[self.pos:pos]
|
143 | self.f.write(piece)
|
144 | self.pos = pos
|
145 |
|
146 | def PrintTheRest(self):
|
147 | # type: () -> None
|
148 | """Print until the end of the string."""
|
149 | self.PrintUntil(self.right_pos)
|
150 |
|
151 | def Print(self, s):
|
152 | # type: (str) -> None
|
153 | """Print text to the underlying buffer."""
|
154 | self.f.write(s)
|
155 |
|
156 |
|
157 | def MakeLexer(rules):
|
158 | return [(re.compile(pat, re.VERBOSE), i) for (pat, i) in rules]
|
159 |
|
160 |
|
161 | #
|
162 | # Lexers
|
163 | #
|
164 |
|
165 | _NAME = r'[a-zA-Z][a-zA-Z0-9:_\-]*' # must start with letter
|
166 |
|
167 | CHAR_LEX = [
|
168 | # Characters
|
169 | # https://www.w3.org/TR/xml/#sec-references
|
170 | (r'&\# [0-9]+ ;', h8_id.DecChar),
|
171 | (r'&\# x[0-9a-fA-F]+ ;', h8_id.HexChar),
|
172 | (r'& %s ;' % _NAME, h8_id.CharEntity),
|
173 | # Allow unquoted, and quoted
|
174 | (r'&', h8_id.BadAmpersand),
|
175 | ]
|
176 |
|
177 | HTM8_LEX = CHAR_LEX + [
|
178 | # TODO: CommentBegin, ProcessingBegin, CDataBegin could have an additional
|
179 | # action associated with them? The ending substring
|
180 | (r'<!--', h8_id.CommentBegin),
|
181 |
|
182 | # Processing instruction are used for the XML header:
|
183 | # <?xml version="1.0" encoding="UTF-8"?>
|
184 | # They are technically XML-only, but in HTML5, they are another kind of
|
185 | # comment:
|
186 | #
|
187 | # https://developer.mozilla.org/en-US/docs/Web/API/ProcessingInstruction
|
188 | #
|
189 | (r'<\?', h8_id.ProcessingBegin),
|
190 | # Not necessary in HTML5, but occurs in XML
|
191 | (r'<!\[CDATA\[', h8_id.CDataBegin), # <![CDATA[
|
192 |
|
193 | # Markup declarations
|
194 | # - In HTML5, there is only <!DOCTYPE html>
|
195 | # - XML has 4 more declarations: <!ELEMENT ...> ATTLIST ENTITY NOTATION
|
196 | # - these seem to be part of DTD
|
197 | # - it's useful to skip these, and be able to parse the rest of the document
|
198 | # - Note: < is allowed?
|
199 | (r'<! [^>\x00]+ >', h8_id.Decl),
|
200 |
|
201 | # Tags
|
202 | # Notes:
|
203 | # - We look for a valid tag name, but we don't validate attributes.
|
204 | # That's done in the tag lexer.
|
205 | # - We don't allow leading whitespace
|
206 | (r'</ (%s) >' % _NAME, h8_id.EndTag),
|
207 | # self-closing <br/> comes before StartTag
|
208 | # could/should these be collapsed into one rule?
|
209 | (r'< (%s) [^>\x00]* />' % _NAME, h8_id.StartEndTag), # end </a>
|
210 | (r'< (%s) [^>\x00]* >' % _NAME, h8_id.StartTag), # start <a>
|
211 |
|
212 | # HTML5 allows unescaped > in raw data, but < is not allowed.
|
213 | # https://stackoverflow.com/questions/10462348/right-angle-bracket-in-html
|
214 | #
|
215 | # - My early blog has THREE errors when disallowing >
|
216 | # - So do some .wwz files
|
217 | (r'[^&<>\x00]+', h8_id.RawData),
|
218 | (r'>', h8_id.BadGreaterThan),
|
219 | # NUL is the end, an accomodation for re2c. Like we do in frontend/match.
|
220 | (r'\x00', h8_id.EndOfStream),
|
221 | # This includes < - it is not BadLessThan because it's NOT recoverable
|
222 | (r'.', h8_id.Invalid),
|
223 | ]
|
224 |
|
225 | # Old notes:
|
226 | #
|
227 | # Non-greedy matches are regular and can be matched in linear time
|
228 | # with RE2.
|
229 | #
|
230 | # https://news.ycombinator.com/item?id=27099798
|
231 | #
|
232 |
|
233 | # This person tried to do it with a regex:
|
234 | #
|
235 | # https://skeptric.com/html-comment-regexp/index.html
|
236 |
|
237 | # . is any char except newline
|
238 | # https://re2c.org/manual/manual_c.html
|
239 |
|
240 | # Discarded options
|
241 | #(r'<!-- .*? -->', h8_id.Comment),
|
242 |
|
243 | # Hack from Claude: \s\S instead of re.DOTALL. I don't like this
|
244 | #(r'<!-- [\s\S]*? -->', h8_id.Comment),
|
245 | #(r'<!-- (?:.|[\n])*? -->', h8_id.Comment),
|
246 |
|
247 | HTM8_LEX_COMPILED = MakeLexer(HTM8_LEX)
|
248 |
|
249 |
|
250 | class Lexer(object):
|
251 |
|
252 | def __init__(self, s, left_pos=0, right_pos=-1, no_special_tags=False):
|
253 | # type: (str, int, int, bool) -> None
|
254 | self.s = s
|
255 | self.pos = left_pos
|
256 | self.right_pos = len(s) if right_pos == -1 else right_pos
|
257 | self.no_special_tags = no_special_tags
|
258 |
|
259 | # string -> compiled regex pattern object
|
260 | self.cache = {} # type: Dict[str, Any]
|
261 |
|
262 | # either </script> or </style> - we search until we see that
|
263 | self.search_state = None # type: Optional[str]
|
264 |
|
265 | # Position of tag name, if applicable
|
266 | # - Set after you get a StartTag, EndTag, or StartEndTag
|
267 | # - Unset on other tags
|
268 | self.tag_pos_left = -1
|
269 | self.tag_pos_right = -1
|
270 |
|
271 | def _Read(self):
|
272 | # type: () -> Tuple[h8_id_t, int]
|
273 | if self.pos == self.right_pos:
|
274 | return h8_id.EndOfStream, self.pos
|
275 |
|
276 | assert self.pos < self.right_pos, self.pos
|
277 |
|
278 | if self.search_state is not None and not self.no_special_tags:
|
279 | # TODO: case-insensitive search for </SCRIPT> <SCRipt> ?
|
280 | #
|
281 | # Another strategy: enter a mode where we find ONLY the end tag
|
282 | # regex, and any data that's not <, and then check the canonical
|
283 | # tag name for 'script' or 'style'.
|
284 | pos = self.s.find(self.search_state, self.pos)
|
285 | if pos == -1:
|
286 | raise LexError('Unterminated <script> or <style>', self.s,
|
287 | self.pos)
|
288 | self.search_state = None
|
289 | # beginning
|
290 | return h8_id.HtmlCData, pos
|
291 |
|
292 | # Find the first match.
|
293 | # Note: frontend/match.py uses _LongestMatch(), which is different!
|
294 | # TODO: reconcile them. This lexer should be expressible in re2c.
|
295 |
|
296 | for pat, tok_id in HTM8_LEX_COMPILED:
|
297 | m = pat.match(self.s, self.pos)
|
298 | if m:
|
299 | if tok_id in (h8_id.StartTag, h8_id.EndTag, h8_id.StartEndTag):
|
300 | self.tag_pos_left = m.start(1)
|
301 | self.tag_pos_right = m.end(1)
|
302 | else:
|
303 | # Reset state
|
304 | self.tag_pos_left = -1
|
305 | self.tag_pos_right = -1
|
306 |
|
307 | if tok_id == h8_id.CommentBegin:
|
308 | pos = self.s.find('-->', self.pos)
|
309 | if pos == -1:
|
310 | raise LexError('Unterminated <!--', self.s, self.pos)
|
311 | return h8_id.Comment, pos + 3 # -->
|
312 |
|
313 | if tok_id == h8_id.ProcessingBegin:
|
314 | pos = self.s.find('?>', self.pos)
|
315 | if pos == -1:
|
316 | raise LexError('Unterminated <?', self.s, self.pos)
|
317 | return h8_id.Processing, pos + 2 # ?>
|
318 |
|
319 | if tok_id == h8_id.CDataBegin:
|
320 | pos = self.s.find(']]>', self.pos)
|
321 | if pos == -1:
|
322 | # unterminated <![CDATA[
|
323 | raise LexError('Unterminated <![CDATA[', self.s,
|
324 | self.pos)
|
325 | return h8_id.CData, pos + 3 # ]]>
|
326 |
|
327 | if tok_id == h8_id.StartTag:
|
328 | # TODO: reduce allocations
|
329 | if (self.TagNameEquals('script') or
|
330 | self.TagNameEquals('style')):
|
331 | # <SCRipt a=b> -> </SCRipt>
|
332 | self.search_state = '</' + self._LiteralTagName() + '>'
|
333 |
|
334 | return tok_id, m.end()
|
335 | else:
|
336 | raise AssertionError('h8_id.Invalid rule should have matched')
|
337 |
|
338 | def TagNamePos(self):
|
339 | # type: () -> int
|
340 | """The right position of the tag pos"""
|
341 | assert self.tag_pos_right != -1, self.tag_pos_right
|
342 | return self.tag_pos_right
|
343 |
|
344 | def TagNameEquals(self, expected):
|
345 | # type: (str) -> bool
|
346 | assert self.tag_pos_left != -1, self.tag_pos_left
|
347 | assert self.tag_pos_right != -1, self.tag_pos_right
|
348 |
|
349 | # TODO: In C++, this does not need an allocation. Can we test
|
350 | # directly?
|
351 | return expected == self.CanonicalTagName()
|
352 |
|
353 | def _LiteralTagName(self):
|
354 | # type: () -> str
|
355 | assert self.tag_pos_left != -1, self.tag_pos_left
|
356 | assert self.tag_pos_right != -1, self.tag_pos_right
|
357 |
|
358 | return self.s[self.tag_pos_left:self.tag_pos_right]
|
359 |
|
360 | def CanonicalTagName(self):
|
361 | # type: () -> str
|
362 | tag_name = self._LiteralTagName()
|
363 | # Most tags are already lower case, so avoid allocation with this conditional
|
364 | # TODO: this could go in the mycpp runtime?
|
365 | if tag_name.islower():
|
366 | return tag_name
|
367 | else:
|
368 | return tag_name.lower()
|
369 |
|
370 | def Read(self):
|
371 | # type: () -> Tuple[h8_id_t, int]
|
372 | tok_id, end_pos = self._Read()
|
373 | self.pos = end_pos # advance
|
374 | return tok_id, end_pos
|
375 |
|
376 | def LookAhead(self, regex):
|
377 | # type: (str) -> bool
|
378 | """
|
379 | Currently used for ul_table.py. But taking a dynamic regex string is
|
380 | not the right interface.
|
381 | """
|
382 | # Cache the regex compilation. This could also be LookAheadFor(THEAD)
|
383 | # or something.
|
384 | pat = self.cache.get(regex)
|
385 | if pat is None:
|
386 | pat = re.compile(regex)
|
387 | self.cache[regex] = pat
|
388 |
|
389 | m = pat.match(self.s, self.pos)
|
390 | return m is not None
|
391 |
|
392 |
|
393 | A_NAME_LEX = [
|
394 | # Leading whitespace is required, to separate attributes.
|
395 | #
|
396 | # If the = is not present, then we set the lexer in a state for
|
397 | # attr_value_e.Missing.
|
398 | (r'\s+ (%s) \s* (=)? \s*' % _NAME, attr_name.Ok),
|
399 | # unexpected EOF
|
400 |
|
401 | # The closing > or /> is treated as end of stream, and it's not an error.
|
402 | (r'\s* /? >', attr_name.Done),
|
403 |
|
404 | # NUL should not be possible, because the top-level
|
405 |
|
406 | # This includes < - it is not BadLessThan because it's NOT recoverable
|
407 | (r'.', attr_name.Invalid),
|
408 | ]
|
409 |
|
410 | A_NAME_LEX_COMPILED = MakeLexer(A_NAME_LEX)
|
411 |
|
412 | # Here we just loop on regular tokens
|
413 | #
|
414 | # Examples:
|
415 | # <a href = unquoted&foo >
|
416 | # <a href = unquoted&foo > # BadAmpersand is allowed I guess
|
417 | # <a href ="unquoted&foo" > # double quoted
|
418 | # <a href ='unquoted&foo' > # single quoted
|
419 | # <a href = what"foo" > # HTML5 allows this, but we could disallow it if
|
420 | # it's not common. It opens up the j"" and $"" extensions
|
421 | # <a href = what'foo' > # ditto
|
422 |
|
423 | # TODO: get rid of OLD copy
|
424 | _UNQUOTED_VALUE_OLD = r'''[^ \t\r\n<>&"'\x00]*'''
|
425 | _UNQUOTED_VALUE = r'''[^ \t\r\n<>&"'\x00]+'''
|
426 |
|
427 | # What comes after = ?
|
428 | A_VALUE_LEX = [
|
429 | (r'"', h8_val_id.DoubleQuote),
|
430 | (r"'", h8_val_id.SingleQuote),
|
431 | (_UNQUOTED_VALUE, h8_val_id.UnquotedVal),
|
432 | (r'.', h8_val_id.NoMatch),
|
433 | ]
|
434 |
|
435 | A_VALUE_LEX_COMPILED = MakeLexer(A_VALUE_LEX)
|
436 |
|
437 | # What's inside "" or '' ?
|
438 | QUOTED_VALUE_LEX = CHAR_LEX + [
|
439 | (r'"', h8_id.DoubleQuote),
|
440 | (r"'", h8_id.SingleQuote),
|
441 | (r'<', h8_id.BadLessThan), # BadAmpersand is in CharLex
|
442 |
|
443 | # TODO: think about whitespace for efficient class= queries?
|
444 | #(r'[ \r\n\t]', h8_id.Whitespace), # terminates unquoted values
|
445 | (r'''[^"'<>&\x00]+''', h8_id.RawData),
|
446 | # This includes > - it is not BadGreaterThan because it's NOT recoverable
|
447 | (r'.', h8_id.Invalid),
|
448 | ]
|
449 |
|
450 | QUOTED_VALUE_LEX_COMPILED = MakeLexer(QUOTED_VALUE_LEX)
|
451 |
|
452 |
|
453 | class AttrLexer(object):
|
454 | """
|
455 | Typical usage:
|
456 |
|
457 | while True:
|
458 | n, start_pos, end_pos = attr_lx.ReadName()
|
459 | if n == attr_name.Ok:
|
460 | if attr_lx.AttrNameEquals('div'):
|
461 | print('div')
|
462 |
|
463 | # TODO: also pass Optional[List[]] out_tokens?
|
464 | v, start_pos, end_pos = attr_lx.ReadValue()
|
465 | """
|
466 |
|
467 | def __init__(self, s):
|
468 | # type: (str) -> None
|
469 | self.s = s
|
470 |
|
471 | self.tok_id = h8_id.Invalid # Uninitialized
|
472 | self.tag_name_pos = -1 # Invalid
|
473 | self.tag_end_pos = -1
|
474 | self.must_not_exceed_pos = -1
|
475 |
|
476 | self.pos = -1
|
477 |
|
478 | self.name_start = -1
|
479 | self.name_end = -1
|
480 | self.next_value_is_missing = False
|
481 |
|
482 | self.init_t = -1
|
483 | self.init_e = -1
|
484 |
|
485 | def Init(self, tok_id, tag_name_pos, end_pos):
|
486 | # type: (h8_id_t, int, int) -> None
|
487 | """Initialize so we can read names and values.
|
488 |
|
489 | Example:
|
490 | 'x <a y>' # tag_name_pos=4, end_pos=6
|
491 | 'x <a>' # tag_name_pos=4, end_pos=4
|
492 |
|
493 | The Init() method is used to reuse instances of the AttrLexer object.
|
494 | """
|
495 | assert tag_name_pos >= 0, tag_name_pos
|
496 | assert end_pos >= 0, end_pos
|
497 |
|
498 | #log('TAG NAME POS %d', tag_name_pos)
|
499 |
|
500 | self.tok_id = tok_id
|
501 | self.tag_name_pos = tag_name_pos
|
502 | self.end_pos = end_pos
|
503 |
|
504 | # Check for ambiguous <img src=/>
|
505 | if tok_id == h8_id.StartTag:
|
506 | self.must_not_exceed_pos = end_pos - 1 # account for >
|
507 | elif tok_id == h8_id.StartEndTag:
|
508 | self.must_not_exceed_pos = end_pos - 2 # account for />
|
509 | else:
|
510 | raise AssertionError(tok_id)
|
511 |
|
512 | self.pos = tag_name_pos
|
513 |
|
514 | # For Reset()
|
515 | self.init_t = tag_name_pos
|
516 | self.init_e = end_pos
|
517 |
|
518 | def Reset(self):
|
519 | # type: () -> None
|
520 |
|
521 | # TODO: maybe GetAttrRaw() should call this directly? But not any of
|
522 | # the AllAttrs() methods?
|
523 | self.tag_name_pos = self.init_t
|
524 | self.end_pos = self.init_e
|
525 | self.pos = self.init_t
|
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 | #log('ReadName() matching %r at %d', self.s, self.pos)
|
542 | if m:
|
543 | #log('ReadName() tag_name_pos %d pos, %d %s', self.tag_name_pos, self.pos, m.groups())
|
544 | if a == attr_name.Invalid:
|
545 | #log('m.groups %s', m.groups())
|
546 | return attr_name.Invalid, -1, -1
|
547 |
|
548 | self.pos = m.end(0) # Advance if it's not invalid
|
549 |
|
550 | if a == attr_name.Ok:
|
551 | #log('%r', m.groups())
|
552 | self.name_start = m.start(1)
|
553 | self.name_end = m.end(1)
|
554 | # Is the equals sign missing? Set state.
|
555 | if m.group(2) is None:
|
556 | self.next_value_is_missing = True
|
557 | # HACK: REWIND, since we don't want to consume whitespace
|
558 | self.pos = self.name_end
|
559 | else:
|
560 | self.next_value_is_missing = False
|
561 | return attr_name.Ok, self.name_start, self.name_end
|
562 | else:
|
563 | # Reset state - e.g. you must call AttrNameEquals
|
564 | self.name_start = -1
|
565 | self.name_end = -1
|
566 |
|
567 | if a == attr_name.Done:
|
568 | return attr_name.Done, -1, -1
|
569 | else:
|
570 | context = self.s[self.pos:]
|
571 | #log('s %r %d', self.s, self.pos)
|
572 | raise AssertionError('h8_id.Invalid rule should have matched %r' %
|
573 | context)
|
574 |
|
575 | def _CanonicalAttrName(self):
|
576 | # type: () -> str
|
577 | """Return the lower case attribute name.
|
578 |
|
579 | Must call after ReadName()
|
580 | """
|
581 | assert self.name_start >= 0, self.name_start
|
582 | assert self.name_end >= 0, self.name_end
|
583 |
|
584 | attr_name = self.s[self.name_start:self.name_end]
|
585 | if attr_name.islower():
|
586 | return attr_name
|
587 | else:
|
588 | return attr_name.lower()
|
589 |
|
590 | def AttrNameEquals(self, expected):
|
591 | # type: (str) -> bool
|
592 | """
|
593 | Must call after ReadName()
|
594 |
|
595 | TODO: This can be optimized to be "in place", with zero allocs.
|
596 | """
|
597 | return expected == self._CanonicalAttrName()
|
598 |
|
599 | def _QuotedRead(self):
|
600 | # type: () -> Tuple[h8_id_t, int]
|
601 |
|
602 | for pat, tok_id in QUOTED_VALUE_LEX_COMPILED:
|
603 | m = pat.match(self.s, self.pos)
|
604 | if m:
|
605 | end_pos = m.end(0) # Advance
|
606 | #log('_QuotedRead %r', self.s[self.pos:end_pos])
|
607 | return tok_id, end_pos
|
608 | else:
|
609 | context = self.s[self.pos:self.pos + 10]
|
610 | raise AssertionError('h8_id.Invalid rule should have matched %r' %
|
611 | context)
|
612 |
|
613 | def ReadValue(self, tokens_out=None):
|
614 | # type: (Optional[List[Tuple[h8_id, int]]]) -> Tuple[attr_value_t, int, int]
|
615 | """Read the attribute value.
|
616 |
|
617 | In general, it is escaped or "raw"
|
618 |
|
619 | Can only be called after a SUCCESSFUL ReadName().
|
620 | Assuming ReadName() returned a value, this should NOT fail.
|
621 | """
|
622 | # ReadName() invariant
|
623 | assert self.name_start >= 0, self.name_start
|
624 | assert self.name_end >= 0, self.name_end
|
625 |
|
626 | self.name_start = -1
|
627 | self.name_end = -1
|
628 |
|
629 | if self.next_value_is_missing:
|
630 | # Do not advance self.pos
|
631 | #log('-> MISSING pos %d : %r', self.pos, self.s[self.pos:])
|
632 | return attr_value_e.Missing, -1, -1
|
633 |
|
634 | # Now read " ', unquoted or empty= is valid too.
|
635 | for pat, a in A_VALUE_LEX_COMPILED:
|
636 | m = pat.match(self.s, self.pos)
|
637 | if m:
|
638 | first_end_pos = m.end(0)
|
639 | # We shouldn't go past the end
|
640 | assert first_end_pos <= self.end_pos, \
|
641 | 'first_end_pos = %d should be less than self.end_pos = %d' % (first_end_pos, self.end_pos)
|
642 | #log('m %s', m.groups())
|
643 |
|
644 | # Note: Unquoted value can't contain & etc. now, so there
|
645 | # is no unquoting, and no respecting tokens_raw.
|
646 | if a == h8_val_id.UnquotedVal:
|
647 | if first_end_pos > self.must_not_exceed_pos:
|
648 | #log('first_end_pos %d', first_end_pos)
|
649 | #log('must_not_exceed_pos %d', self.must_not_exceed_pos)
|
650 | raise LexError(
|
651 | 'Ambiguous slash: last attribute should be quoted',
|
652 | self.s, first_end_pos)
|
653 | self.pos = first_end_pos # Advance
|
654 | return attr_value_e.Unquoted, m.start(0), first_end_pos
|
655 |
|
656 | # TODO: respect tokens_out
|
657 | if a == h8_val_id.DoubleQuote:
|
658 | self.pos = first_end_pos
|
659 | while True:
|
660 | tok_id, q_end_pos = self._QuotedRead()
|
661 | #log('self.pos %d q_end_pos %d', self.pos, q_end_pos)
|
662 | if tok_id == h8_id.Invalid:
|
663 | raise LexError(
|
664 | 'ReadValue() got invalid token (DQ)', self.s,
|
665 | self.pos)
|
666 | if tok_id == h8_id.DoubleQuote:
|
667 | right_pos = self.pos
|
668 | self.pos = q_end_pos # Advance past "
|
669 | return attr_value_e.DoubleQuoted, first_end_pos, right_pos
|
670 | self.pos = q_end_pos # Advance _QuotedRead
|
671 |
|
672 | # TODO: respect tokens_out
|
673 | if a == h8_val_id.SingleQuote:
|
674 | self.pos = first_end_pos
|
675 | while True:
|
676 | tok_id, q_end_pos = self._QuotedRead()
|
677 | if tok_id == h8_id.Invalid:
|
678 | raise LexError(
|
679 | 'ReadValue() got invalid token (SQ)', self.s,
|
680 | self.pos)
|
681 | if tok_id == h8_id.SingleQuote:
|
682 | right_pos = self.pos
|
683 | self.pos = q_end_pos # Advance past "
|
684 | return attr_value_e.SingleQuoted, first_end_pos, right_pos
|
685 | self.pos = q_end_pos # Advance _QuotedRead
|
686 |
|
687 | if a == h8_val_id.NoMatch:
|
688 | # <a foo = >
|
689 | return attr_value_e.Empty, -1, -1
|
690 | else:
|
691 | raise AssertionError('h8_val_id.NoMatch rule should have matched')
|
692 |
|
693 |
|
694 | def GetAttrRaw(attr_lx, name):
|
695 | # type: (AttrLexer, str) -> Optional[str]
|
696 | while True:
|
697 | n, name_start, name_end = attr_lx.ReadName()
|
698 | #log('==> ReadName %s %d %d', attr_name_str(n), name_start, name_end)
|
699 | if n == attr_name.Ok:
|
700 | if attr_lx.AttrNameEquals(name):
|
701 | v, val_start, val_end = attr_lx.ReadValue()
|
702 | return attr_lx.s[val_start:val_end]
|
703 | else:
|
704 | # Problem with stateful API: You are forced to either ReadValue()
|
705 | # or SkipVlaue()
|
706 | attr_lx.ReadValue()
|
707 | elif n == attr_name.Done:
|
708 | break
|
709 | elif n == attr_name.Invalid:
|
710 | raise LexError('GetAttrRaw() got invalid token', attr_lx.s,
|
711 | attr_lx.pos)
|
712 | else:
|
713 | raise AssertionError()
|
714 |
|
715 | return None
|
716 |
|
717 |
|
718 | def AllAttrsRawSlice(attr_lx):
|
719 | # type: (AttrLexer) -> List[Tuple[int, int, attr_value_t, int, int]]
|
720 | result = []
|
721 | while True:
|
722 | n, name_start, name_end = attr_lx.ReadName()
|
723 | if 0:
|
724 | log(' AllAttrsRaw ==> ReadName %s %d %d %r', attr_name_str(n),
|
725 | name_start, name_end, attr_lx.s[attr_lx.pos:attr_lx.pos + 10])
|
726 | if n == attr_name.Ok:
|
727 | #name = attr_lx.s[name_start:name_end]
|
728 | #log(' Name %r', name)
|
729 |
|
730 | v, val_start, val_end = attr_lx.ReadValue()
|
731 | #val = attr_lx.s[val_start:val_end]
|
732 | #log(' ReadValue %r', val)
|
733 | result.append((name_start, name_end, v, val_start, val_end))
|
734 | elif n == attr_name.Done:
|
735 | break
|
736 | elif n == attr_name.Invalid:
|
737 | raise LexError('AllAttrsRaw() got invalid token', attr_lx.s,
|
738 | attr_lx.pos)
|
739 | else:
|
740 | raise AssertionError()
|
741 |
|
742 | return result
|
743 |
|
744 |
|
745 | def AllAttrsRaw(attr_lx):
|
746 | # type: (AttrLexer) -> List[Tuple[str, str]]
|
747 | """
|
748 | Get a list of pairs [('class', 'foo'), ('href', '?foo=1&bar=2')]
|
749 |
|
750 | The quoted values may be escaped. We would need another lexer to
|
751 | unescape them.
|
752 | """
|
753 | slices = AllAttrsRawSlice(attr_lx)
|
754 | pairs = []
|
755 | s = attr_lx.s
|
756 | for name_start, name_end, val_id, val_start, val_end in slices:
|
757 | n = s[name_start:name_end]
|
758 | v = s[val_start:val_end]
|
759 | pairs.append((n, v))
|
760 | return pairs
|
761 |
|
762 |
|
763 | #
|
764 | # OLD API - REMOVE THIS
|
765 | #
|
766 |
|
767 | # Tag names:
|
768 | # Match <a or </a
|
769 | # Match <h2, but not <2h
|
770 | #
|
771 | # HTML 5 doesn't restrict tag names at all
|
772 | # https://html.spec.whatwg.org/#toc-syntax
|
773 | #
|
774 | # XML allows : - .
|
775 | # https://www.w3.org/TR/xml/#NT-NameChar
|
776 |
|
777 | # Namespaces for MathML, SVG
|
778 | # XLink, XML, XMLNS
|
779 | #
|
780 | # https://infra.spec.whatwg.org/#namespaces
|
781 | #
|
782 | # Allow - for td-attrs
|
783 |
|
784 | # TODO: we don't need to capture the tag name here? That's done at the top
|
785 | # level
|
786 | _TAG_RE = re.compile(r'/? \s* (%s)' % _NAME, re.VERBOSE)
|
787 |
|
788 | _TAG_LAST_RE = re.compile(r'\s* /? >', re.VERBOSE)
|
789 |
|
790 | # To match href="foo"
|
791 | # Note: in HTML5 and XML, single quoted attributes are also valid
|
792 |
|
793 | # <button disabled> is standard usage
|
794 |
|
795 | # NOTE: This used to allow whitespace around =
|
796 | # <a foo = "bar"> makes sense in XML
|
797 | # But then you also have
|
798 | # <a foo= bar> - which is TWO attributes, in HTML5
|
799 | # So the space is problematic
|
800 |
|
801 | _ATTR_RE = re.compile(
|
802 | r'''
|
803 | \s+ # Leading whitespace is required
|
804 | (%s) # Attribute name
|
805 | (?: # Optional attribute value
|
806 | \s* = \s* # Spaces allowed around =
|
807 | (?:
|
808 | " ([^>"\x00]*) " # double quoted value
|
809 | | ' ([^>'\x00]*) ' # single quoted value
|
810 | | (%s) # Attribute value
|
811 | )
|
812 | )?
|
813 | ''' % (_NAME, _UNQUOTED_VALUE_OLD), re.VERBOSE)
|
814 |
|
815 |
|
816 | class TagLexer(object):
|
817 | """
|
818 | Given a tag like <a href="..."> or <link type="..." />, the TagLexer
|
819 | provides a few operations:
|
820 |
|
821 | - What is the tag?
|
822 | - Iterate through the attributes, giving (name, value_start_pos, value_end_pos)
|
823 | """
|
824 |
|
825 | def __init__(self, s):
|
826 | # type: (str) -> None
|
827 | self.s = s
|
828 | self.start_pos = -1 # Invalid
|
829 | self.end_pos = -1
|
830 |
|
831 | def Reset(self, start_pos, end_pos):
|
832 | # type: (int, int) -> None
|
833 | """Reuse instances of this object."""
|
834 | assert start_pos >= 0, start_pos
|
835 | assert end_pos >= 0, end_pos
|
836 |
|
837 | self.start_pos = start_pos
|
838 | self.end_pos = end_pos
|
839 |
|
840 | def WholeTagString(self):
|
841 | # type: () -> str
|
842 | """Return the entire tag string, e.g. <a href='foo'>"""
|
843 | return self.s[self.start_pos:self.end_pos]
|
844 |
|
845 | def GetTagName(self):
|
846 | # type: () -> str
|
847 | # First event
|
848 | tok_id, start, end = next(self.Tokens())
|
849 | return self.s[start:end]
|
850 |
|
851 | def GetSpanForAttrValue(self, attr_name):
|
852 | # type: (str) -> Tuple[int, int]
|
853 | """
|
854 | Used by oils_doc.py, for href shortcuts
|
855 | """
|
856 | # Algorithm: search for QuotedValue or UnquotedValue after AttrName
|
857 | # TODO: Could also cache these
|
858 |
|
859 | events = self.Tokens()
|
860 | val = (-1, -1)
|
861 | try:
|
862 | while True:
|
863 | tok_id, start, end = next(events)
|
864 | if tok_id == h8_tag_id.AttrName:
|
865 | name = self.s[start:end]
|
866 | if name == attr_name:
|
867 | # The value should come next
|
868 | tok_id, start, end = next(events)
|
869 | assert tok_id in (
|
870 | h8_tag_id.QuotedValue, h8_tag_id.UnquotedValue,
|
871 | h8_tag_id.MissingValue), h8_tag_id_str(tok_id)
|
872 | val = start, end
|
873 | break
|
874 |
|
875 | except StopIteration:
|
876 | pass
|
877 | return val
|
878 |
|
879 | def GetAttrRaw(self, attr_name):
|
880 | # type: (str) -> Optional[str]
|
881 | """
|
882 | Return the value, which may be UNESCAPED.
|
883 | """
|
884 | start, end = self.GetSpanForAttrValue(attr_name)
|
885 | if start == -1:
|
886 | return None
|
887 | return self.s[start:end]
|
888 |
|
889 | def AllAttrsRawSlice(self):
|
890 | # type: () -> List[Tuple[str, int, int]]
|
891 | """
|
892 | Get a list of pairs [('class', 3, 5), ('href', 9, 12)]
|
893 | """
|
894 | slices = []
|
895 | events = self.Tokens()
|
896 | try:
|
897 | while True:
|
898 | tok_id, start, end = next(events)
|
899 | if tok_id == h8_tag_id.AttrName:
|
900 | name = self.s[start:end]
|
901 |
|
902 | # The value should come next
|
903 | tok_id, start, end = next(events)
|
904 | assert tok_id in (
|
905 | h8_tag_id.QuotedValue, h8_tag_id.UnquotedValue,
|
906 | h8_tag_id.MissingValue), h8_tag_id_str(tok_id)
|
907 | # Note: quoted values may have &
|
908 | # We would need ANOTHER lexer to unescape them, but we
|
909 | # don't need that for ul-table
|
910 | slices.append((name, start, end))
|
911 | except StopIteration:
|
912 | pass
|
913 | return slices
|
914 |
|
915 | def AllAttrsRaw(self):
|
916 | # type: () -> List[Tuple[str, str]]
|
917 | """
|
918 | Get a list of pairs [('class', 'foo'), ('href', '?foo=1&bar=2')]
|
919 |
|
920 | The quoted values may be escaped. We would need another lexer to
|
921 | unescape them.
|
922 | """
|
923 | slices = self.AllAttrsRawSlice()
|
924 | pairs = []
|
925 | for name, start, end in slices:
|
926 | pairs.append((name, self.s[start:end]))
|
927 | return pairs
|
928 |
|
929 | def Tokens(self):
|
930 | # type: () -> Iterator[Tuple[h8_tag_id_t, int, int]]
|
931 | """
|
932 | Yields a sequence of tokens: Tag (AttrName AttrValue?)*
|
933 |
|
934 | Where each Token is (Type, start_pos, end_pos)
|
935 |
|
936 | Note that start and end are NOT redundant! We skip over some unwanted
|
937 | characters.
|
938 | """
|
939 | m = _TAG_RE.match(self.s, self.start_pos + 1)
|
940 | if not m:
|
941 | raise RuntimeError("Couldn't find HTML tag in %r" %
|
942 | self.WholeTagString())
|
943 | yield h8_tag_id.TagName, m.start(1), m.end(1)
|
944 |
|
945 | pos = m.end(0)
|
946 | #log('POS %d', pos)
|
947 |
|
948 | while True:
|
949 | # don't search past the end
|
950 | m = _ATTR_RE.match(self.s, pos, self.end_pos)
|
951 | if not m:
|
952 | #log('BREAK pos %d', pos)
|
953 | break
|
954 | #log('AttrName %r', m.group(1))
|
955 |
|
956 | yield h8_tag_id.AttrName, m.start(1), m.end(1)
|
957 |
|
958 | #log('m.groups() %r', m.groups())
|
959 | if m.group(2) is not None:
|
960 | # double quoted
|
961 | yield h8_tag_id.QuotedValue, m.start(2), m.end(2)
|
962 | elif m.group(3) is not None:
|
963 | # single quoted - TODO: could have different token types
|
964 | yield h8_tag_id.QuotedValue, m.start(3), m.end(3)
|
965 | elif m.group(4) is not None:
|
966 | yield h8_tag_id.UnquotedValue, m.start(4), m.end(4)
|
967 | else:
|
968 | # <button disabled>
|
969 | end = m.end(0)
|
970 | yield h8_tag_id.MissingValue, end, end
|
971 |
|
972 | # Skip past the "
|
973 | pos = m.end(0)
|
974 |
|
975 | #log('TOK %r', self.s)
|
976 |
|
977 | m = _TAG_LAST_RE.match(self.s, pos)
|
978 | #log('_TAG_LAST_RE match %r', self.s[pos:])
|
979 | if not m:
|
980 | raise LexError('Extra data at end of tag', self.s, pos)
|