parsers.py 47 KB
Newer Older
1
"""parsers.py - parser combinators for for DHParser
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16

Copyright 2016  by Eckhart Arnold (arnold@badw.de)
                Bavarian Academy of Sciences an Humanities (badw.de)

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at

    http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
implied.  See the License for the specific language governing
permissions and limitations under the License.
Eckhart Arnold's avatar
Eckhart Arnold committed
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48

Module ``parsers.py`` contains a number of classes that together
make up parser combinators for left-recursive grammers. For each
element of the extended Backus-Naur-Form as well as for a regular
expression token a class is defined. The set of classes can be used to
define a parser for (ambiguous) left-recursive grammers.


References and Acknowledgements:

Dominikus Herzberg: Objekt-orientierte Parser-Kombinatoren in Python,
Blog-Post, September, 18th 2008 on denkspuren. gedanken, ideen,
anregungen und links rund um informatik-themen, URL:
http://denkspuren.blogspot.de/2008/09/objekt-orientierte-parser-kombinatoren.html

Dominikus Herzberg: Eine einfache Grammatik für LaTeX, Blog-Post,
September, 18th 2008 on denkspuren. gedanken, ideen, anregungen und
links rund um informatik-themen, URL:
http://denkspuren.blogspot.de/2008/09/eine-einfache-grammatik-fr-latex.html

Dominikus Herzberg: Uniform Syntax, Blog-Post, February, 27th 2007 on
denkspuren. gedanken, ideen, anregungen und links rund um
informatik-themen, URL:
http://denkspuren.blogspot.de/2007/02/uniform-syntax.html

Richard A. Frost, Rahmatullah Hafiz and Paul Callaghan: Parser
Combinators for Ambiguous Left-Recursive Grammars, in: P. Hudak and
D.S. Warren (Eds.): PADL 2008, LNCS 4902, pp. 167–181, Springer-Verlag
Berlin Heidelberg 2008.

Juancarlo Añez: grako, a PEG parser generator in Python,
https://bitbucket.org/apalala/grako
49
50
"""

51

Eckhart Arnold's avatar
Eckhart Arnold committed
52
import abc
53
54
import copy
import os
55
from functools import partial
56
57
58
59
try:
    import regex as re
except ImportError:
    import re
60
61
62
63
try:
    from typing import Any, Callable, Collection, Dict, Iterator, List, Set, Tuple, Union
except ImportError:
    from .typing34 import Any, Callable, Dict, Iterator, List, Set, Tuple, Union
64

65
from DHParser.toolkit import is_logging, log_dir, logfile_basename, escape_re, sane_parser_name
Eckhart Arnold's avatar
Eckhart Arnold committed
66
67
from DHParser.syntaxtree import WHITESPACE_PTYPE, TOKEN_PTYPE, ZOMBIE_PARSER, ParserBase, \
    Node, TransformationFunc
68
from DHParser.toolkit import load_if_file, error_messages
Eckhart Arnold's avatar
Eckhart Arnold committed
69

70
71
__all__ = ['ScannerFunc',
           'HistoryRecord',
Eckhart Arnold's avatar
Eckhart Arnold committed
72
           'Parser',
73
           'Grammar',
Eckhart Arnold's avatar
Eckhart Arnold committed
74
75
76
77
78
79
80
81
82
83
           'RX_SCANNER_TOKEN',
           'BEGIN_SCANNER_TOKEN',
           'END_SCANNER_TOKEN',
           'make_token',
           'nil_scanner',
           'ScannerToken',
           'RegExp',
           'RE',
           'Token',
           'mixin_comment',
84
85
86
           # 'UnaryOperator',
           # 'NaryOperator',
           'Synonym',
Eckhart Arnold's avatar
Eckhart Arnold committed
87
88
89
           'Optional',
           'ZeroOrMore',
           'OneOrMore',
90
           'Series',
Eckhart Arnold's avatar
Eckhart Arnold committed
91
92
93
94
95
96
97
98
99
100
101
           'Alternative',
           'FlowOperator',
           'Required',
           'Lookahead',
           'NegativeLookahead',
           'Lookbehind',
           'NegativeLookbehind',
           'Capture',
           'Retrieve',
           'Pop',
           'Forward',
102
           'Compiler',
Eckhart Arnold's avatar
Eckhart Arnold committed
103
           'compile_source']
Eckhart Arnold's avatar
Eckhart Arnold committed
104

105

106
107
108
109
110
111
112
113
114
115
########################################################################
#
# Grammar and parsing infrastructure
#
########################################################################


ScannerFunc = Union[Callable[[str], str], partial]


116
LEFT_RECURSION_DEPTH = 10  # because of pythons recursion depth limit, this
Eckhart Arnold's avatar
Eckhart Arnold committed
117
                           # value ought not to be set too high
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
MAX_DROPOUTS = 25  # stop trying to recover parsing after so many errors


class HistoryRecord:
    """
    Stores debugging information about one completed step in the
    parsing history. 

    A parsing step is "completed" when the last one of a nested
    sequence of parser-calls returns. The call stack including
    the last parser call will be frozen in the ``HistoryRecord``-
    object. In addition a reference to the generated leaf node
    (if any) will be stored and the result status of the last
    parser call, which ist either MATCH, FAIL (i.e. no match)
    or ERROR.
    """
    __slots__ = ('call_stack', 'node', 'remaining')

    MATCH = "MATCH"
    ERROR = "ERROR"
    FAIL = "FAIL"

    def __init__(self, call_stack, node, remaining):
        self.call_stack = call_stack
        self.node = node
        self.remaining = remaining

145
    def err_msg(self) -> str:
146
147
        return self.ERROR + ": " + "; ".join(self.node._errors).replace('\n', '\\')

148
    @property
149
    def stack(self) -> str:
150
151
152
        return "->".join(str(parser) for parser in self.call_stack)

    @property
153
    def status(self) -> str:
154
155
        return self.FAIL if self.node is None else \
            self.err_msg() if self.node._errors else self.MATCH
156
157

    @property
158
    def extent(self) -> Tuple[int, int]:
159
160
161
162
163
        return ((-self.remaining - self.node.len, -self.remaining) if self.node
                else (-self.remaining, None))


def add_parser_guard(parser_func):
164
    def guarded_call(parser: 'Parser', text: str) -> Tuple[Node, str]:
165
166
        try:
            location = len(text)
167
168
            grammar = parser.grammar  # grammar may be 'None' for unconnected parsers!

169
170
171
172
173
174
175
176
177
            if not grammar.moving_forward__:
                # rollback variable changes for discarded branch of parsing tree
                while grammar.rollback__ and grammar.rollback__[-1][0] <= location:
                    grammar.rollback__[-1][1]()
                    grammar.rollback__.pop()
            grammar.moving_forward__ = True

            if grammar.history_tracking__:
                grammar.call_stack__.append(parser)
178

179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
            # if location has already been visited by the current parser,
            # return saved result
            if location in parser.visited:
                return parser.visited[location]
            # break left recursion at the maximum allowed depth
            if parser.recursion_counter.setdefault(location, 0) > LEFT_RECURSION_DEPTH:
                return None, text

            parser.recursion_counter[location] += 1

            # run original __call__ method
            node, rest = parser_func(parser, text)

            if node is not None:
                # in case of a recursive call saves the result of the first
                # (or left-most) call that matches
                parser.visited[location] = (node, rest)
196
                grammar.last_node__ = node   # store last node for Lookbehind operator
197
198
199
200
201
202
203
            elif location in parser.visited:
                # if parser did non match but a saved result exits, assume
                # left recursion and use the saved result
                node, rest = parser.visited[location]

            parser.recursion_counter[location] -= 1

204
            if grammar.history_tracking__:
205
                # don't track returning parsers except in case an error has occurred
206
207
208
                if grammar.moving_forward__ or (node and node._errors):
                    record = HistoryRecord(grammar.call_stack__.copy(), node, len(rest))
                    grammar.history__.append(record)
209
                    # print(record.stack, record.status, rest[:20].replace('\n', '|'))
210
211
                grammar.call_stack__.pop()
            grammar.moving_forward__ = False
212

213
214
215
216
217
218
219
220
221
222
223
        except RecursionError:
            node = Node(None, text[:min(10, max(1, text.find("\n")))] + " ...")
            node.add_error("maximum recursion depth of parser reached; "
                           "potentially due to too many errors!")
            rest = ''

        return node, rest

    return guarded_call


Eckhart Arnold's avatar
Eckhart Arnold committed
224
class ParserMetaClass(abc.ABCMeta):
225
226
227
228
229
230
231
232
233
234
    def __init__(cls, name, bases, attrs):
        # The following condition is necessary for classes that don't override
        # the __call__() method, because in these cases the non-overridden
        # __call__()-method would be substituted a second time!
        guarded_parser_call = add_parser_guard(cls.__call__)
        if cls.__call__.__code__ != guarded_parser_call.__code__:
            cls.__call__ = guarded_parser_call
        super(ParserMetaClass, cls).__init__(name, bases, attrs)


Eckhart Arnold's avatar
Eckhart Arnold committed
235
class Parser(ParserBase, metaclass=ParserMetaClass):
236
237
    ApplyFunc = Callable[['Parser'], None]

238
239
    def __init__(self, name: str = '') -> None:
        # assert isinstance(name, str), str(name)
Eckhart Arnold's avatar
Eckhart Arnold committed
240
        super(Parser, self).__init__(name)
241
        self._grammar = None  # type: 'Grammar'
242
243
        self.reset()

Eckhart Arnold's avatar
Eckhart Arnold committed
244
    def __deepcopy__(self, memo):
245
246
        return self.__class__(self.name)

247
    def reset(self):
248
249
250
        self.visited = dict()  # type: Dict[int, Tuple[Node, str]]
        self.recursion_counter = dict()  # type: Dict[int, int]
        self.cycle_detection = set()  # type: Set[Callable]
251
        return self
252

253
    def __call__(self, text: str) -> Tuple[Node, str]:
254
255
256
        return None, text  # default behaviour: don't match

    def __str__(self):
257
        return self.name or self.ptype
258

259
    def __add__(self, other):
260
        return Series(self, other)
261
262
263
264

    def __or__(self, other):
        return Alternative(self, other)

265
    @property
266
    def grammar(self) -> 'Grammar':
267
268
269
        return self._grammar

    @grammar.setter
270
271
    def grammar(self, grammar: 'Grammar'):
        self._grammar = grammar
272
273
274
275
276
        self._grammar_assigned_notifier()

    def _grammar_assigned_notifier(self):
        pass

277
    def apply(self, func: ApplyFunc):
278
        """Applies function `func(parser)` recursively to this parser and all
Eckhart Arnold's avatar
Eckhart Arnold committed
279
        descendants of the tree of parsers. The same function can never
280
281
282
283
284
285
286
287
288
289
        be applied twice between calls of the ``reset()``-method!
        """
        if func in self.cycle_detection:
            return False
        else:
            self.cycle_detection.add(func)
            func(self)
            return True


290
class Grammar:
291
292
    root__ = None  # type: Union[Parser, None]
    # root__ must be overwritten with the root-parser by grammar subclass
Eckhart Arnold's avatar
Eckhart Arnold committed
293
    parser_initialization__ = "pending"  # type: str
294
295
296
297
298
299

    @classmethod
    def _assign_parser_names(cls):
        """Initializes the `parser.name` fields of those
        Parser objects that are directly assigned to a class field with
        the field's name, e.g.
300
            class Grammar(Grammar):
301
302
303
                ...
                symbol = RE('(?!\\d)\\w+')
        After the call of this method symbol.name == "symbol"
304
305
306
307
        holds. Names assigned via the ``name``-parameter of the
        constructor will not be overwritten. Parser names starting or
        ending with a double underscore like ``root__`` will be
        ignored. See ``toolkit.sane_parser_name()``
308
309
310
311

        This is done only once, upon the first instantiation of the
        grammar class!

312
313
314
315
316
        Attention: If there exists more than one reference to the same
        parser, only the first one will be chosen for python versions 
        greater or equal 3.6.  For python version <= 3.5 an arbitrarily
        selected reference will be chosen. See PEP 520
        (www.python.org/dev/peps/pep-0520/) for an explanation of why. 
317
318
319
320
321
        """
        if cls.parser_initialization__ == "done":
            return
        cdict = cls.__dict__
        for entry, parser in cdict.items():
322
            if isinstance(parser, Parser) and sane_parser_name(entry):
323
                if not parser.name:
324
                    parser.name = entry
325
                if (isinstance(parser, Forward) and (not parser.parser.name)):
326
327
328
                    parser.parser.name = entry
        cls.parser_initialization__ = "done"

Eckhart Arnold's avatar
Eckhart Arnold committed
329
330
331
    def __init__(self, root: Parser=None) -> None:
        # if not hasattr(self.__class__, 'parser_initialization__'):
        #     self.__class__.parser_initialization__ = "pending"
332
333
334
335
        if not hasattr(self.__class__, 'wspL__'):
            self.wspL__ = ''
        if not hasattr(self.__class__, 'wspR__'):
            self.wspR__ = ''
336
337
338
        self.all_parsers__ = set()  # type: Set[Parser]
        self.dirty_flag__ = False
        self.history_tracking__ = False
339
        self._reset()
Eckhart Arnold's avatar
Eckhart Arnold committed
340
341
342
343
344
345
        # prepare parsers in the class, first
        self._assign_parser_names()
        # then deep-copy the parser tree from class to instance;
        # parsers not connected to the root object will be copied later
        # on demand (see Grammar.__getitem__()). Usually, the need to
        # do so only arises during testing.
346
        self.root__ = root if root else copy.deepcopy(self.__class__.root__)
347
        if self.wspL__:
Eckhart Arnold's avatar
Eckhart Arnold committed
348
            self.wsp_left_parser__ = Whitespace(self.wspL__)  # type: ParserBase
349
            self.wsp_left_parser__.grammar = self
350
            self.all_parsers__.add(self.wsp_left_parser__)  # don't you forget about me...
351
352
353
        else:
            self.wsp_left_parser__ = ZOMBIE_PARSER
        if self.wspR__:
Eckhart Arnold's avatar
Eckhart Arnold committed
354
            self.wsp_right_parser__ = Whitespace(self.wspR__)  # type: ParserBase
355
            self.wsp_right_parser__.grammar = self
356
            self.all_parsers__.add(self.wsp_right_parser__)  # don't you forget about me...
357
358
359
360
        else:
            self.wsp_right_parser__ = ZOMBIE_PARSER
        self.root__.apply(self._add_parser)

361
    def __getitem__(self, key):
362
363
364
        try:
            return self.__dict__[key]
        except KeyError:
365
366
            parser_template = getattr(self, key, None)
            if parser_template:
Eckhart Arnold's avatar
Eckhart Arnold committed
367
                # add parser to grammar object on the fly...
368
369
370
                parser = copy.deepcopy(parser_template)
                parser.apply(self._add_parser)
                # assert self[key] == parser
Eckhart Arnold's avatar
Eckhart Arnold committed
371
                return self[key]
372
            raise KeyError('Unknown parser "%s" !' % key)
373

374
    def _reset(self):
375
        self.document__ = ""       # type: str
376
        # variables stored and recalled by Capture and Retrieve parsers
377
378
        self.variables__ = dict()  # type: Dict[str, List[str]]
        self.rollback__ = []       # type: List[Tuple[int, Callable]]
379
        # previously parsed node, needed by Lookbehind parser
380
        self.last_node__ = None    # type: Node
381
        # support for call stack tracing
382
        self.call_stack__ = []     # type: List[Parser]
383
        # snapshots of call stacks
384
        self.history__ = []        # type: List[HistoryRecord]
385
        # also needed for call stack tracing
386
        self.moving_forward__ = True  # type: bool
387

Eckhart Arnold's avatar
Eckhart Arnold committed
388
    def _add_parser(self, parser: Parser) -> None:
Eckhart Arnold's avatar
Eckhart Arnold committed
389
        """Adds the particular copy of the parser object to this
390
        particular instance of Grammar.
391
        """
392
        if parser.name:
393
394
395
            assert parser.name not in self.__dict__, \
                ('Cannot add parser "%s" because a field with the same name '
                 'already exists in grammar object!' % parser.name)
396
            setattr(self, parser.name, parser)
397
        self.all_parsers__.add(parser)
398
399
        parser.grammar = self

Eckhart Arnold's avatar
Eckhart Arnold committed
400
    def __call__(self, document: str, start_parser="root__") -> Node:
401
402
403
404
        """Parses a document with with parser-combinators.

        Args:
            document (str): The source text to be parsed.
405
406
407
            start_parser (str): The name of the parser with which to
                start. This is useful for testing particular parsers
                (i.e. particular parts of the EBNF-Grammar.)
408
409
410
        Returns:
            Node: The root node ot the parse tree.
        """
411
        # assert isinstance(document, str), type(document)
412
413
        if self.root__ is None:
            raise NotImplementedError()
414
        if self.dirty_flag__:
415
            self._reset()
416
            for parser in self.all_parsers__:
417
418
                parser.reset()
        else:
419
420
421
            self.dirty_flag__ = True
        self.history_tracking__ = is_logging()
        self.document__ = document
Eckhart Arnold's avatar
Eckhart Arnold committed
422
        parser = self[start_parser] if isinstance(start_parser, str) else start_parser
423
424
        assert parser.grammar == self, "Cannot run parsers from a different grammar object!" \
                                       " %s vs. %s" % (str(self), str(parser.grammar))
425
        stitches = []  # type: List[Node]
426
        rest = document
427
428
        if not rest:
            result, ignore = parser(rest)
429
430
431
432
433
434
435
436
437
438
        while rest and len(stitches) < MAX_DROPOUTS:
            result, rest = parser(rest)
            if rest:
                fwd = rest.find("\n") + 1 or len(rest)
                skip, rest = rest[:fwd], rest[fwd:]
                if result is None:
                    error_msg = "Parser did not match! Invalid source file?"
                else:
                    stitches.append(result)
                    error_msg = "Parser stopped before end" + \
439
440
                                (("! trying to recover" +
                                  (" but stopping history recording at this point."
441
                                   if self.history_tracking__ else "..."))
442
443
444
445
                                 if len(stitches) < MAX_DROPOUTS
                                 else " too often! Terminating parser.")
                stitches.append(Node(None, skip))
                stitches[-1].add_error(error_msg)
446
                if self.history_tracking__:
447
448
449
450
                    # some parsers may have matched and left history records with nodes != None.
                    # Because these are not connected to the stiched root node, their pos
                    # properties will not be initialized by setting the root node's pos property
                    # to zero. Therefore, their pos properties need to be initialized here
451
                    for record in self.history__:
452
453
                        if record.node and record.node._pos < 0:
                            record.node.pos = 0
454
455
456
                    record = HistoryRecord(self.call_stack__.copy(), stitches[-1], len(rest))
                    self.history__.append(record)
                    self.history_tracking__ = False
457
458
459
460
        if stitches:
            if rest:
                stitches.append(Node(None, rest))
            result = Node(None, tuple(stitches))
461
        if any(self.variables__.values()):
462
            result.add_error("Capture-retrieve-stack not empty after end of parsing: "
463
                             + str(self.variables__))
464
465
466
        result.pos = 0  # calculate all positions
        return result

Eckhart Arnold's avatar
Eckhart Arnold committed
467
    def log_parsing_history(self, log_file_name: str='') -> None:
468
469
470
471
        """Writes a log of the parsing history of the most recently parsed
        document. 
        """
        def prepare_line(record):
472
            excerpt = self.document__.__getitem__(slice(*record.extent))[:25].replace('\n', '\\n')
473
            excerpt = "'%s'" % excerpt if len(excerpt) < 25 else "'%s...'" % excerpt
474
            return record.stack, record.status, excerpt
475
476

        def write_log(history, log_name):
Eckhart Arnold's avatar
Eckhart Arnold committed
477
            path = os.path.join(log_dir(), log_name + "_parser.log")
478
479
480
481
482
483
            if history:
                with open(path, "w", encoding="utf-8") as f:
                    f.write("\n".join(history))
            elif os.path.exists(path):
                os.remove(path)

Eckhart Arnold's avatar
Eckhart Arnold committed
484
485
486
487
        if not log_file_name:
            name = self.__class__.__name__
            log_file_name = name[:-7] if name.lower().endswith('grammar') else name
        full_history, match_history, errors_only = [], [], []
488
        for record in self.history__:
Eckhart Arnold's avatar
Eckhart Arnold committed
489
490
            line = ";  ".join(prepare_line(record))
            full_history.append(line)
491
            if record.node and record.node.parser.ptype != WHITESPACE_PTYPE:
Eckhart Arnold's avatar
Eckhart Arnold committed
492
                match_history.append(line)
493
                if record.node.error_flag:
Eckhart Arnold's avatar
Eckhart Arnold committed
494
495
496
497
                    errors_only.append(line)
        write_log(full_history, log_file_name + '_full')
        write_log(match_history, log_file_name + '_match')
        write_log(errors_only, log_file_name + '_errors')
498
499


Eckhart Arnold's avatar
Eckhart Arnold committed
500
501
def dsl_error_msg(parser: Parser, error_str: str) -> str:
    """Returns an error message for errors in the parser configuration,
502
503
504
    e.g. errors that result in infinite loops.

    Args:
Eckhart Arnold's avatar
Eckhart Arnold committed
505
        parser (Parser):  The parser where the error was noticed. Note
506
            that this is not necessarily the parser that caused the
Eckhart Arnold's avatar
Eckhart Arnold committed
507
            error but only where the error became aparent.
508
509
510
511
512
        error_str (str):  A short string describing the error.
    Returns:  
        str: An error message including the call stack if history 
        tacking has been turned in the grammar object.
    """
513
    msg = ["DSL parser specification error:", error_str, 'Caught by parser "%s".' % str(parser)]
514
515
    if parser.grammar.history__:
        msg.extend(["\nCall stack:", parser.grammar.history__[-1].stack])
516
517
518
519
520
    else:
        msg.extend(["\nEnable history tracking in Grammar object to display call stack."])
    return " ".join(msg)


521
522
523
524
525
526
527
528
529
530
531
532
########################################################################
#
# Token and Regular Expression parser classes (i.e. leaf classes)
#
########################################################################


RX_SCANNER_TOKEN = re.compile('\w+')
BEGIN_SCANNER_TOKEN = '\x1b'
END_SCANNER_TOKEN = '\x1c'


533
def make_token(token: str, argument: str = '') -> str:
534
535
536
537
538
539
540
541
542
543
544
545
546
    """Turns the ``token`` and ``argument`` into a special token that
    will be caught by the `ScannerToken`-parser.

    This function is a support function that should be used by scanners
    to inject scanner tokens into the source text.
    """
    assert RX_SCANNER_TOKEN.match(token)
    assert argument.find(BEGIN_SCANNER_TOKEN) < 0
    assert argument.find(END_SCANNER_TOKEN) < 0

    return BEGIN_SCANNER_TOKEN + token + argument + END_SCANNER_TOKEN


547
def nil_scanner(text: str) -> str:
Eckhart Arnold's avatar
Eckhart Arnold committed
548
    return text
549
550
551


class ScannerToken(Parser):
552
553
554
555
556
557
558
559
560
561
    """
    Parses tokens that have been inserted by a Scanner.
    
    Scanners can generate Tokens with the ``make_token``-function.
    These tokens start and end with magic characters that can only be
    matched by the ScannerToken Parser. Scanner tokens can be used to
    insert BEGIN - END delimiters at the beginning or ending of an 
    indented block. Otherwise indented block are difficult to handle 
    with parsing expression grammars.
    """
562
563
564

    def __init__(self, scanner_token: str) -> None:
        assert scanner_token and scanner_token.isupper()
565
        assert RX_SCANNER_TOKEN.match(scanner_token)
566
        super(ScannerToken, self).__init__(scanner_token)
567

568
    def __call__(self, text: str) -> Tuple[Node, str]:
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
        if text[0:1] == BEGIN_SCANNER_TOKEN:
            end = text.find(END_SCANNER_TOKEN, 1)
            if end < 0:
                node = Node(self, '').add_error(
                    'END_SCANNER_TOKEN delimiter missing from scanner token. '
                    '(Most likely due to a scanner bug!)')
                return node, text[1:]
            elif end == 0:
                node = Node(self, '').add_error(
                    'Scanner token cannot have zero length. '
                    '(Most likely due to a scanner bug!)')
                return node, text[2:]
            elif text.find(BEGIN_SCANNER_TOKEN, 1, end) >= 0:
                node = Node(self, text[len(self.name) + 1:end])
                node.add_error(
                    'Scanner tokens must not be nested or contain '
                    'BEGIN_SCANNER_TOKEN delimiter as part of their argument. '
                    '(Most likely due to a scanner bug!)')
                return node, text[end:]
            if text[1:len(self.name) + 1] == self.name:
                return Node(self, text[len(self.name) + 1:end]), \
                       text[end + 1:]
        return None, text


class RegExp(Parser):
595
596
    """
    Regular expression parser.
597
598
599
600
601
602
    
    The RegExp-parser parses text that matches a regular expression.
    RegExp can also be considered as the "atomic parser", because all
    other parsers delegate part of the parsing job to other parsers,
    but do not match text directly.
    """
603
604

    def __init__(self, regexp, name: str = '') -> None:
605
        super(RegExp, self).__init__(name)
606
607
608
        self.regexp = re.compile(regexp) if isinstance(regexp, str) else regexp

    def __deepcopy__(self, memo):
Eckhart Arnold's avatar
Eckhart Arnold committed
609
        # `regex` supports deep copies, but not `re`
610
        try:
Eckhart Arnold's avatar
Eckhart Arnold committed
611
            regexp = copy.deepcopy(self.regexp, memo)
612
613
        except TypeError:
            regexp = self.regexp.pattern
614
        return RegExp(regexp, self.name)
615

616
    def __call__(self, text: str) -> Tuple[Node, str]:
617
618
619
620
621
622
        match = text[0:1] != BEGIN_SCANNER_TOKEN and self.regexp.match(text)  # ESC starts a scanner token.
        if match:
            end = match.end()
            return Node(self, text[:end]), text[end:]
        return None, text

623
624
625
    def __str__(self):
        return self.name or self.ptype + ' /%s/' % self.regexp.pattern

626

627
628
class Whitespace(RegExp):
    assert WHITESPACE_PTYPE == ":Whitespace"
629
630


631
632
class RE(Parser):
    """Regular Expressions with optional leading or trailing whitespace.
633
634
635
636
637
638
639
640
641
642
    
    The RE-parser parses pieces of text that match a given regular
    expression. Other than the ``RegExp``-Parser it can also skip 
    "implicit whitespace" before or after the matched text.
    
    The whitespace is in turn defined by a regular expression. It
    should be made sure that this expression also matches the empty
    string, e.g. use r'\s*' or r'[\t ]+', but not r'\s+'. If the
    respective parameters in the constructor are set to ``None`` the
    default whitespace expression from the Grammar object will be used.
643
    """
644
    def __init__(self, regexp, wL=None, wR=None, name=''):
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
        """Constructor for class RE.
                
        Args:
            regexp (str or regex object):  The regular expression to be
                used for parsing. 
            wL (str or regexp):  Left whitespace regular expression, 
                i.e. either ``None``, the empty string or a regular
                expression (e.g. "\s*") that defines whitespace. An 
                empty string means no whitespace will be skipped,
                ``None`` means that the default whitespace will be 
                used.
            wR (str or regexp):  Right whitespace regular expression.
                See above.
            name:  The optional name of the parser.
        """
660
        super(RE, self).__init__(name)
661
662
        self.wL = wL
        self.wR = wR
663
664
        self.wspLeft = Whitespace(wL) if wL else ZOMBIE_PARSER
        self.wspRight = Whitespace(wR) if wR else ZOMBIE_PARSER
665
666
        self.main = RegExp(regexp)

Eckhart Arnold's avatar
Eckhart Arnold committed
667
668
669
670
671
    def __deepcopy__(self, memo={}):
        try:
            regexp = copy.deepcopy(self.main.regexp, memo)
        except TypeError:
            regexp = self.main.regexp.pattern
672
        return self.__class__(regexp, self.wL, self.wR, self.name)
Eckhart Arnold's avatar
Eckhart Arnold committed
673

674
    def __call__(self, text: str) -> Tuple[Node, str]:
675
676
677
678
679
680
681
682
683
684
685
        # assert self.main.regexp.pattern != "@"
        t = text
        wL, t = self.wspLeft(t)
        main, t = self.main(t)
        if main:
            wR, t = self.wspRight(t)
            result = tuple(nd for nd in (wL, main, wR)
                           if nd and nd.result != '')
            return Node(self, result), t
        return None, text

686
    def __str__(self):
687
        if self.ptype == TOKEN_PTYPE:
688
689
690
            return 'Token "%s"' % self.main.regexp.pattern.replace('\\', '')
        return self.name or ('RE ' + ('~' if self.wL else '')
                             + '/%s/' % self.main.regexp.pattern + ('~' if self.wR else ''))
691
692
693

    def _grammar_assigned_notifier(self):
        if self.grammar:
694
            # use default whitespace parsers if not otherwise specified
695
696
697
698
699
            if self.wL is None:
                self.wspLeft = self.grammar.wsp_left_parser__
            if self.wR is None:
                self.wspRight = self.grammar.wsp_right_parser__

700
    def apply(self, func: Parser.ApplyFunc):
701
702
703
704
705
706
707
708
        if super(RE, self).apply(func):
            if self.wL:
                self.wspLeft.apply(func)
            if self.wR:
                self.wspRight.apply(func)
            self.main.apply(func)


709
710
711
class Token(RE):
    assert TOKEN_PTYPE == ":Token"

712
    def __init__(self, token: str, wL=None, wR=None, name: str = '') -> None:
713
714
715
716
717
        self.token = token
        super(Token, self).__init__(escape_re(token), wL, wR, name)

    def __deepcopy__(self, memo={}):
        return self.__class__(self.token, self.wL, self.wR, self.name)
718
719


720
def mixin_comment(whitespace: str, comment: str) -> str:
721
722
723
724
725
726
727
    """Returns a regular expression that merges comment and whitespace
    regexps. Thus comments cann occur whereever whitespace is allowed
    and will be skipped just as implicit whitespace.
    
    Note, that because this works on the level of regular expressions,
    nesting comments is not possible. It also makes it much harder to
    use directives inside comments (which isn't recommended, anyway).
728
729
730
731
732
733
734
735
736
737
738
739
740
    """
    wspc = '(?:' + whitespace + '(?:' + comment + whitespace + ')*)'
    return wspc


########################################################################
#
# Combinator parser classes (i.e. trunk classes of the parser tree)
#
########################################################################


class UnaryOperator(Parser):
741
    def __init__(self, parser: Parser, name: str = '') -> None:
742
        super(UnaryOperator, self).__init__(name)
743
        # assert isinstance(parser, Parser)
744
        self.parser = parser  # type: Parser
745

Eckhart Arnold's avatar
Eckhart Arnold committed
746
747
    def __deepcopy__(self, memo):
        parser = copy.deepcopy(self.parser, memo)
748
        return self.__class__(parser, self.name)
Eckhart Arnold's avatar
Eckhart Arnold committed
749

750
    def apply(self, func: Parser.ApplyFunc):
751
752
753
754
755
        if super(UnaryOperator, self).apply(func):
            self.parser.apply(func)


class NaryOperator(Parser):
756
    def __init__(self, *parsers: Parser, name: str = '') -> None:
757
        super(NaryOperator, self).__init__(name)
758
        # assert all([isinstance(parser, Parser) for parser in parsers]), str(parsers)
759
        self.parsers = parsers  # type: Collection
760

Eckhart Arnold's avatar
Eckhart Arnold committed
761
762
    def __deepcopy__(self, memo):
        parsers = copy.deepcopy(self.parsers, memo)
763
        return self.__class__(*parsers, name=self.name)
Eckhart Arnold's avatar
Eckhart Arnold committed
764

765
    def apply(self, func: Parser.ApplyFunc):
766
767
768
769
770
        if super(NaryOperator, self).apply(func):
            for parser in self.parsers:
                parser.apply(func)


771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
class Synonym(UnaryOperator):
    """Simply calls another parser and encapsulates the result in
    another node if that parser matches.

    This parser is needed to support synonyms in EBNF, e.g.
        jahr       = JAHRESZAHL
        JAHRESZAHL = /\d\d\d\d/
    Otherwise the first line could not be represented by any parser
    class, in which case it would be unclear whether the parser
    RE('\d\d\d\d') carries the name 'JAHRESZAHL' or 'jahr'
    """
    def __call__(self, text: str) -> Tuple[Node, str]:
        node, text = self.parser(text)
        if node:
            return Node(self, node), text
        return None, text


789
class Optional(UnaryOperator):
790
    def __init__(self, parser: Parser, name: str = '') -> None:
791
        super(Optional, self).__init__(parser, name)
792
        # assert isinstance(parser, Parser)
793
        assert not isinstance(parser, Optional), \
794
            "Redundant nesting of options: %s(%s)" % \
795
796
            (str(name), str(parser.name))
        assert not isinstance(parser, Required), \
797
            "Nesting options with required elements is contradictory: " \
798
799
            "%s(%s)" % (str(name), str(parser.name))

800
    def __call__(self, text: str) -> Tuple[Node, str]:
801
802
803
804
805
806
807
        node, text = self.parser(text)
        if node:
            return Node(self, node), text
        return Node(self, ()), text


class ZeroOrMore(Optional):
808
809
    def __call__(self, text: str) -> Tuple[Node, str]:
        results = ()  # type: Tuple[Node, ...]
810
811
812
        n = len(text) + 1
        while text and len(text) < n:
            n = len(text)
813
814
815
            node, text = self.parser(text)
            if not node:
                break
816
            if len(text) == n:
di68kap's avatar
di68kap committed
817
                node.add_error(dsl_error_msg(self, 'Infinite Loop detected.'))
818
819
820
821
822
            results += (node,)
        return Node(self, results), text


class OneOrMore(UnaryOperator):
823
    def __init__(self, parser: Parser, name: str = '') -> None:
824
        super(OneOrMore, self).__init__(parser, name)
825
826
827
828
        assert not isinstance(parser, Optional), \
            "Use ZeroOrMore instead of nesting OneOrMore and Optional: " \
            "%s(%s)" % (str(name), str(parser.name))

829
830
831
    def __call__(self, text: str) -> Tuple[Node, str]:
        results = ()  # type: Tuple[Node, ...]
        text_ = text  # type: str
832
833
834
        n = len(text) + 1
        while text_ and len(text_) < n:
            n = len(text_)
835
836
837
            node, text_ = self.parser(text_)
            if not node:
                break
838
            if len(text_) == n:
839
                node.add_error(dsl_error_msg(self, 'Infinite Loop detected.'))
840
841
842
843
844
845
            results += (node,)
        if results == ():
            return None, text
        return Node(self, results), text_


846
class Series(NaryOperator):
847
    def __init__(self, *parsers: Parser, name: str = '') -> None:
848
        super(Series, self).__init__(*parsers, name=name)
849
850
        assert len(self.parsers) >= 1

851
852
853
    def __call__(self, text: str) -> Tuple[Node, str]:
        results = ()  # type: Tuple[Node, ...]
        text_ = text  # type: str
854
855
856
        for parser in self.parsers:
            node, text_ = parser(text_)
            if not node:
857
858
                return None, text
            results += (node,)
859
860
861
862
863
            if node.error_flag:
                break
        assert len(results) <= len(self.parsers)
        return Node(self, results), text_

864
865
    def __add__(self, other: 'Series') -> 'Series':
        return Series(*(tuple(self.parsers) + (other,)))
866

867
868
    def __radd__(self, other: 'Series') -> 'Series':
        return Series(other, *self.parsers)
869
870

        # def __iadd__(self, other):
871
        #     if isinstance(other, Series):
872
873
874
875
876
        #         self.parsers = self.parsers + other.parsers
        #     else:
        #         self.parsers = self.parsers + (other,)
        #     return self

877
878

class Alternative(NaryOperator):
879
880
881
882
883
884
885
    """Matches if at least one of several alternatives matches. Returns
    the first match.

    This parser represents the EBNF-operator "|" with the qualification
    that both the symmetry and the ambiguity of the EBNF-or-operator
    are broken by selecting the first match.

Eckhart Arnold's avatar
Eckhart Arnold committed
886
887
    # the order of the sub-expression matters!
    >>> number = RE('\d+') | RE('\d+') + RE('\.') + RE('\d+')
888
889
    >>> Grammar(number)("3.1416").show()
    '3 <<< Error on ".1416" | Parser stopped before end! trying to recover... >>> '
890
891

    # the most selective expression should be put first:
Eckhart Arnold's avatar
Eckhart Arnold committed
892
    >>> number = RE('\d+') + RE('\.') + RE('\d+') | RE('\d+')
893
    >>> Grammar(number)("3.1416").show()
Eckhart Arnold's avatar
Eckhart Arnold committed
894
    '3.1416'
895
    """
896
897

    def __init__(self, *parsers: Parser, name: str = '') -> None:
898
        super(Alternative, self).__init__(*parsers, name=name)
899
900
901
        assert len(self.parsers) >= 1
        assert all(not isinstance(p, Optional) for p in self.parsers)

902
    def __call__(self, text: str) -> Tuple[Node, str]:
903
904
905
906
907
908
        for parser in self.parsers:
            node, text_ = parser(text)
            if node:
                return Node(self, node), text_
        return None, text

909
    def __or__(self, other):
Eckhart Arnold's avatar
Eckhart Arnold committed
910
        return Alternative(*(self.parsers + (other,)))
911
912
913
914
915

    def __ror__(self, other):
        return Alternative(other, *self.parsers)

        # def __ior__(self, other):
916
        #     if isinstance(other, Series):
917
918
919
920
921
        #         self.parsers = self.parsers + other.parsers
        #     else:
        #         self.parsers = self.parsers + (other,)
        #     return self

922
923
924
925
926
927
928
929
930

########################################################################
#
# Flow control operators
#
########################################################################


class FlowOperator(UnaryOperator):
931
    def __init__(self, parser: Parser, name: str = '') -> None:
932
        super(FlowOperator, self).__init__(parser, name)
933
934
935


class Required(FlowOperator):
936
    # Add constructor that checks for logical errors, like `Required(Optional(...))` constructs ?
937
    def __call__(self, text: str) -> Tuple[Node, str]:
938
939
940
941
942
943
944
945
946
947
948
949
950
        node, text_ = self.parser(text)
        if not node:
            m = re.search(r'\s(\S)', text)
            i = max(1, m.regs[1][0]) if m else 1
            node = Node(self, text[:i])
            text_ = text[i:]
            # assert False, "*"+text[:i]+"*"
            node.add_error('%s expected; "%s..." found!' %
                           (str(self.parser), text[:10]))
        return node, text_


class Lookahead(FlowOperator):
951
    def __init__(self, parser: Parser, name: str = '') -> None:
952
        super(Lookahead, self).__init__(parser, name)
953

954
    def __call__(self, text: str) -> Tuple[Node, str]:
955
956
957
958
959
960
        node, text_ = self.parser(text)
        if self.sign(node is not None):
            return Node(self, ''), text
        else:
            return None, text

961
    def sign(self, bool_value) -> bool:
962
963
964
965
        return bool_value


class NegativeLookahead(Lookahead):
966
    def sign(self, bool_value) -> bool:
967
968
969
        return not bool_value


970
def iter_right_branch(node) -> Iterator[Node]:
971
972
973
974
975
976
977
978
979
980
981
982
    """Iterates over the right branch of `node` starting with node itself.
    Iteration is stopped if either there are no child nodes any more or
    if the parser of a node is a Lookahead parser. (Reason is: Since
    lookahead nodes do not advance the parser, it does not make sense
    to look back to them.)
    """
    while node and not isinstance(node.parser, Lookahead):  # the second condition should not be necessary
        yield node  # for well-formed EBNF code
        node = node.children[-1] if node.children else None


class Lookbehind(FlowOperator):
983
    def __init__(self, parser: Parser, name: str = '') -> None:
984
        super(Lookbehind, self).__init__(parser, name)
985
986
        print("WARNING: Lookbehind Operator is experimental!")

987
    def __call__(self, text: str) -> Tuple[Node, str]:
988
        if isinstance(self.grammar.last_node__, Lookahead):
989
990
991
992
993
994
995
            return Node(self, '').add_error('Lookbehind right after Lookahead '
                                            'does not make sense!'), text
        if self.sign(self.condition()):
            return Node(self, ''), text
        else:
            return None, text

996
    def sign(self, bool_value) -> bool:
997
998
999
1000
        return bool_value

    def condition(self):
        node = None
1001
        for node in iter_right_branch(self.grammar.last_node__):
1002
1003
1004
1005
1006
1007
1008
1009
1010
            if node.parser.name == self.parser.name:
                return True
        if node and isinstance(self.parser, RegExp) and \
                self.parser.regexp.match(str(node)):  # Is there really a use case for this?
            return True
        return False


class NegativeLookbehind(Lookbehind):
1011
    def sign(self, bool_value) -> bool:
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
        return not bool_value


########################################################################
#
# Capture and Retrieve operators (for passing variables in the parser)
#
########################################################################


class Capture(UnaryOperator):
1023
1024
    """STILL EXPERIMENTAL!"""

1025
    def __init__(self, parser: Parser, name: str = '') -> None:
1026
        super(Capture, self).__init__(parser, name)
1027

1028
    def __call__(self, text: str) -> Tuple[Node, str]:
1029
        node, text_ = self.parser(text)
1030
        if node:
1031
            stack = self.grammar.variables__.setdefault(self.name, [])
1032
            stack.append(str(node))
1033
1034
            self.grammar.rollback__.append((len(text), lambda : stack.pop()))
            return Node(self, node), text_
1035
1036
        else:
            return None, text
1037
1038


1039
1040
1041
RetrieveFilter = Callable[[List[str]], str]


1042
def last_value(stack: List[str]) -> str:
1043
1044
1045
    return stack[-1]


1046
def counterpart(stack: List[str]) -> str:
1047
1048
1049
1050
    value = stack[-1]
    return value.replace("(", ")").replace("[", "]").replace("{", "}").replace(">", "<")


1051
1052
def accumulate(stack: List[str]) -> str:
    return "".join(stack) if len(stack) > 1 else stack[-1]  # provoke IndexError if stack empty
1053
1054


1055
class Retrieve(Parser):
1056
    """STILL EXPERIMENTAL!"""
1057
    def __init__(self, symbol: Parser, filter: RetrieveFilter = None, name: str = '') -> None:
1058
        super(Retrieve, self).__init__(name)
1059
        self.symbol = symbol
1060
        self.filter = filter if filter else last_value
1061

Eckhart Arnold's avatar
Eckhart Arnold committed
1062
    def __deepcopy__(self, memo):
1063
        return self.__class__(self.symbol, self.filter, self.name)
Eckhart Arnold's avatar
Eckhart Arnold committed
1064

1065
    def __call__(self, text: str) -> Tuple[Node, str]:
1066
1067
1068
        return self.call(text)  # allow call method to be called from subclass circumventing the parser guard

    def call(self, text: str) -> Tuple[Node, str]:
1069
        try:
1070
            stack = self.grammar.variables__[self.symbol.name]
1071
            value = self.filter(stack)
1072
1073
        except (KeyError, IndexError):
            return Node(self, '').add_error(dsl_error_msg(self,
1074