parse.py 88 KB
Newer Older
1
# parse.py - parser combinators 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
Module ``parse`` contains the python classes and functions for
21
22
DHParser's packrat-parser. It's central class is the
``Grammar``-class, which is the base class for any concrete
23
24
25
26
27
28
29
30
31
Grammar. Grammar-objects are callable and parsing is done by
calling a Grammar-object with a source text as argument.

The different parsing functions are callable descendants of class
``Parser``. Usually, they are organized in a tree and defined
within the namespace of a grammar-class. See ``ebnf.EBNFGrammar``
for an example.
"""

32

Eckhart Arnold's avatar
Eckhart Arnold committed
33
from collections import defaultdict, OrderedDict
34
import copy
35
36
37
38

from DHParser.error import Error, linebreaks
from DHParser.log import is_logging, HistoryRecord
from DHParser.preprocess import BEGIN_TOKEN, END_TOKEN, RX_TOKEN_NAME
39
from DHParser.stringview import StringView, EMPTY_STRING_VIEW
di68kap's avatar
di68kap committed
40
41
from DHParser.syntaxtree import Node, RootNode, WHITESPACE_PTYPE, \
    TOKEN_PTYPE, ZOMBIE, ResultType
42
from DHParser.toolkit import sane_parser_name, escape_control_characters, re, typing
Eckhart Arnold's avatar
Eckhart Arnold committed
43
from typing import Callable, cast, List, Tuple, Set, Dict, DefaultDict, Union, Optional, Any
44

45

46
__all__ = ('Parser',
47
           'UnknownParserError',
48
49
           'Grammar',
           'PreprocessorToken',
50
           'Token',
51
           'RegExp',
52
53
           'RE',
           'TKN',
54
           'Whitespace',
55
           'mixin_comment',
eckhart's avatar
eckhart committed
56
57
           'UnaryOperator',
           'NaryOperator',
58
59
60
61
62
63
64
65
66
           'Synonym',
           'Option',
           'ZeroOrMore',
           'OneOrMore',
           'Series',
           'Alternative',
           'AllOf',
           'SomeOf',
           'Unordered',
eckhart's avatar
eckhart committed
67
           'Required',
68
69
70
71
72
73
74
75
76
77
           'Lookahead',
           'NegativeLookahead',
           'Lookbehind',
           'NegativeLookbehind',
           'last_value',
           'counterpart',
           'accumulate',
           'Capture',
           'Retrieve',
           'Pop',
78
           'Forward')
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94


########################################################################
#
# Grammar and parsing infrastructure
#
########################################################################


LEFT_RECURSION_DEPTH = 8  # type: int
# because of python's recursion depth limit, this value ought not to be
# set too high. PyPy allows higher values than CPython
MAX_DROPOUTS = 3  # type: int
# stop trying to recover parsing after so many errors


Eckhart Arnold's avatar
Eckhart Arnold committed
95
96
97
class ParserError(Exception):
    """
    A `ParserError` is thrown for those parser errors that allow the
98
    controlled re-entrance of the parsing process after the error occurred.
Eckhart Arnold's avatar
Eckhart Arnold committed
99
    If a reentry-rule has been configured for the parser where the error
100
    occurred, the parser guard can resume the parsing process.
Eckhart Arnold's avatar
Eckhart Arnold committed
101
102
103
104
105

    Currently, the only case when a `ParserError` is thrown (and not some
    different kind of error like `UnknownParserError`, is when a `Series`-
    detects a missing mandatory element.
    """
106
    def __init__(self, node: Node, rest: StringView, first_throw: bool):
Eckhart Arnold's avatar
Eckhart Arnold committed
107
108
        self.node = node  # type: Node
        self.rest = rest  # type: StringView
109
        self.first_throw = first_throw  # type: bool
Eckhart Arnold's avatar
Eckhart Arnold committed
110
111


112
ResumeList = List[Union[str, Any]]  # list of strings or regular expressiones
Eckhart Arnold's avatar
Eckhart Arnold committed
113
114


115
def reentry_point(rest: StringView, rules: ResumeList) -> int:
Eckhart Arnold's avatar
Eckhart Arnold committed
116
117
118
119
120
121
122
123
124
    """
    Finds the point where parsing should resume after a ParserError has been caught.
    Args:
        rest:  The rest of the parsed text or, in other words, the point where
                a ParserError was thrown.
        rules: A list of strings or regular expressions. The rest of the text is
                searched for each of these. The closest match is the point where
                parsing will be resumed.
    Returns:
125
126
        The integer index of the closest reentry point or -1 if no reentry-point
        was found.
Eckhart Arnold's avatar
Eckhart Arnold committed
127
    """
128
    upper_limit = len(rest) + 1
Eckhart Arnold's avatar
Eckhart Arnold committed
129
130
131
132
    i = upper_limit
    #find closest match
    for rule in rules:
        if isinstance(rule, str):
133
134
            k = rest.find(rule)
            i = min(k if k >= 0 else upper_limit, i)
Eckhart Arnold's avatar
Eckhart Arnold committed
135
136
137
        else:
            m = rest.search(rule)
            if m:
138
                i = min(rest.index(m.start()), i)
139
    # in case no rule matched return -1
140
    if i == upper_limit:
141
142
143
        i = -1
    return i
    # return Node(None, rest[:i]), rest[i:]
Eckhart Arnold's avatar
Eckhart Arnold committed
144
145


eckhart's avatar
eckhart committed
146
147
148
149
ApplyFunc = Callable[['Parser'], None]
FlagFunc = Callable[[ApplyFunc, Set[ApplyFunc]], bool]


di68kap's avatar
di68kap committed
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
class ParserBase:
    """
    ParserBase is the base class for all real and mock parser classes.
    It is defined here, because Node objects require a parser object
    for instantiation.
    """

    __slots__ = 'name', 'ptype', 'tag_name'

    def __init__(self,):  # , pbases=frozenset()):
        self.name = ''  # type: str
        self.ptype = ':' + self.__class__.__name__  # type: str
        self.tag_name = self.ptype  # type: str

    def __repr__(self):
        return self.name + self.ptype

    def __str__(self):
        return self.name + (' = ' if self.name else '') + repr(self)

    def __call__(self, text: StringView) -> Tuple[Optional['Node'], StringView]:
        return None, text

    @property
    def repr(self) -> str:
        """Returns the parser's name if it has a name and repr()"""
        return self.name if self.name else self.__repr__()

    def reset(self):
        """Resets any parser variables. (Should be overridden.)"""
        pass

    @property
    def grammar(self) -> 'Grammar':
        """Returns the Grammar object to which the parser belongs. If not
        yet connected to any Grammar object, None is returned."""
        raise NotImplementedError

    def apply(self, func: Callable):
        """Applies the function `func` recursively to the parser and all
        descendant parsers, if any exist."""
        pass


194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
class Parser(ParserBase):
    """
    (Abstract) Base class for Parser combinator parsers. Any parser
    object that is actually used for parsing (i.e. no mock parsers)
    should should be derived from this class.

    Since parsers can contain other parsers (see classes UnaryOperator
    and NaryOperator) they form a cyclical directed graph. A root
    parser is a parser from which all other parsers can be reached.
    Usually, there is one root parser which serves as the starting
    point of the parsing process. When speaking of "the root parser"
    it is this root parser object that is meant.

    There are two different types of parsers:

    1. *Named parsers* for which a name is set in field `parser.name`.
       The results produced by these parsers can later be retrieved in
       the AST by the parser name.

    2. *Anonymous parsers* where the name-field just contains the empty
       string. AST-transformation of Anonymous parsers can be hooked
       only to their class name, and not to the individual parser.

    Parser objects are callable and parsing is done by calling a parser
    object with the text to parse.

    If the parser matches it returns a tuple consisting of a node
    representing the root of the concrete syntax tree resulting from the
    match as well as the substring `text[i:]` where i is the length of
    matched text (which can be zero in the case of parsers like
    `ZeroOrMore` or `Option`). If `i > 0` then the parser has "moved
    forward".

    If the parser does not match it returns `(None, text). **Note** that
    this is not the same as an empty match `("", text)`. Any empty match
    can for example be returned by the `ZeroOrMore`-parser in case the
    contained parser is repeated zero times.

    Attributes and Properties:
        visited:  Mapping of places this parser has already been to
                during the current parsing process onto the results the
                parser returned at the respective place. This dictionary
                is used to implement memoizing.

        recursion_counter:  Mapping of places to how often the parser
                has already been called recursively at this place. This
                is needed to implement left recursion. The number of
                calls becomes irrelevant once a resault has been memoized.

        cycle_detection:  The apply()-method uses this variable to make
                sure that one and the same function will not be applied
                (recursively) a second time, if it has already been
                applied to this parser.

        grammar:  A reference to the Grammar object to which the parser
                is attached.
    """

252
    def __init__(self) -> None:
253
        # assert isinstance(name, str), str(name)
254
        super().__init__()
eckhart's avatar
eckhart committed
255
        self._grammar = ZOMBIE_GRAMMAR  # type: Grammar
256
257
        self.reset()

258
259
260
261
262
263
264
265
266
        # # add "aspect oriented" wrapper around parser calls
        # # for memoizing, left recursion and tracing
        # if not isinstance(self, Forward):  # should Forward-Parser not be guarded? Not sure...
        #     guarded_parser_call = add_parser_guard(self.__class__.__call__)
        #     # The following check 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!
        #     if self.__class__.__call__.__code__ != guarded_parser_call.__code__:
        #         self.__class__.__call__ = guarded_parser_call
267
268
269
270
271
272
273
274

    def __deepcopy__(self, memo):
        """Deepcopy method of the parser. Upon instantiation of a Grammar-
        object, parsers will be deep-copied to the Grammar object. If a
        derived parser-class changes the signature of the constructor,
        `__deepcopy__`-method must be replaced (i.e. overridden without
        calling the same method from the superclass) by the derived class.
        """
275
276
        duplicate = self.__class__()
        duplicate.name = self.name
eckhart's avatar
eckhart committed
277
        duplicate.ptype = self.ptype
278
        return duplicate
279
280
281
282
283

    def reset(self):
        """Initializes or resets any parser variables. If overwritten,
        the `reset()`-method of the parent class must be called from the
        `reset()`-method of the derived class."""
284
        self.visited = dict()  # type: Dict[int, Tuple[Optional[Node], StringView]]
eckhart's avatar
eckhart committed
285
        self.recursion_counter = defaultdict(lambda: 0)  # type: DefaultDict[int, int]
eckhart's avatar
eckhart committed
286
        self.cycle_detection = set()  # type: Set[ApplyFunc]
287

288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
    def __call__(self: 'Parser', text: StringView) -> Tuple[Optional[Node], StringView]:
        """Applies the parser to the given text. This is a wrapper method that adds
        the business intelligence that is common to all parsers. The actual parsing is
        done in the overridden method `_parse()`.
        """
        grammar = self._grammar  # read protected member instead property to avoid function call
        location = grammar.document_length__ - len(text)

        try:
            if grammar.last_rb__loc__ >= location:
                grammar.rollback_to__(location)

            # if location has already been visited by the current parser,
            # return saved result
            if location in self.visited:
                # no history recording in case of memoized results
                return self.visited[location]

            # break left recursion at the maximum allowed depth
            if grammar.left_recursion_handling__:
                if self.recursion_counter[location] > LEFT_RECURSION_DEPTH:
                    grammar.recursion_locations__.add(location)
                    return None, text
                self.recursion_counter[location] += 1

            if grammar.history_tracking__:
                grammar.call_stack__.append(self)
                grammar.moving_forward__ = True

            try:
                # PARSER CALL: run _parse() method
                node, rest = self._parse(text)
            except ParserError as error:
                # does this play well with variable setting? add rollback clause here? tests needed...
                gap = len(text) - len(error.rest)
                rules = grammar.resume_rules__.get(self.name, [])
                rest = error.rest[len(error.node):]
                i = reentry_point(rest, rules)
                if i >= 0 or self == grammar.start_parser__:
                    # apply reentry-rule or catch error at root-parser
                    if i < 0:
                        i = 1
                    nd = Node(None, rest[:i])
                    rest = rest[i:]
                    assert error.node.children
                    if error.first_throw:
                        node = error.node
                        node.result += (nd,)
                    else:
337
                        node = Node(self.tag_name, (Node(None, text[:gap]), error.node, nd))
338
339
340
341
                elif error.first_throw:
                    raise ParserError(error.node, error.rest, first_throw=False)
                else:
                    result = (Node(None, text[:gap]), error.node) if gap else error.node  # type: ResultType
342
                    raise ParserError(Node(self.tag_name, result), text, first_throw=False)
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391

            if grammar.left_recursion_handling__:
                self.recursion_counter[location] -= 1
                # don't clear recursion_locations__ !!!

            if node is None:
                # retrieve an earlier match result (from left recursion) if it exists
                if location in grammar.recursion_locations__:
                    if location in self.visited:
                        node, rest = self.visited[location]
                        # TODO: maybe add a warning about occurrence of left-recursion here?
                    # don't overwrite any positive match (i.e. node not None) in the cache
                    # and don't add empty entries for parsers returning from left recursive calls!
                elif grammar.memoization__:
                    # otherwise also cache None-results
                    self.visited[location] = (None, rest)
            else:
                assert node._pos < 0
                node._pos = location
                assert node._pos >= 0, str("%i < %i" % (grammar.document_length__, location))
                if (grammar.last_rb__loc__ < location
                        and (grammar.memoization__ or location in grammar.recursion_locations__)):
                    # - variable manipulating parsers will not be entered into the cache,
                    #   because caching would interfere with changes of variable state
                    # - in case of left recursion, the first recursive step that
                    #   matches will store its result in the cache
                    # TODO: need a unit-test concerning interference of variable manipulation and left recursion algorithm?
                    self.visited[location] = (node, rest)

            # Mind that memoized parser calls will not appear in the history record!
            # Does this make sense? Or should it be changed?
            if grammar.history_tracking__:
                # don't track returning parsers except in case an error has occurred
                # remaining = len(rest)
                if (grammar.moving_forward__ or (node and node.errors)):
                    record = HistoryRecord(grammar.call_stack__, node, text)
                    grammar.history__.append(record)
                    # print(record.stack, record.status, rest[:20].replace('\n', '|'))
                grammar.moving_forward__ = False
                grammar.call_stack__.pop()

        except RecursionError:
            node = Node(None, str(text[:min(10, max(1, text.find("\n")))]) + " ...")
            node._pos = location
            grammar.tree__.new_error(node, "maximum recursion depth of parser reached; "
                                           "potentially due to too many errors!")
            rest = EMPTY_STRING_VIEW

        return node, rest
392
393
394
395
396
397
398
399
400
401
402
403

    def __add__(self, other: 'Parser') -> 'Series':
        """The + operator generates a series-parser that applies two
        parsers in sequence."""
        return Series(self, other)

    def __or__(self, other: 'Parser') -> 'Alternative':
        """The | operator generates an alternative parser that applies
        the first parser and, if that does not match, the second parser.
        """
        return Alternative(self, other)

404
405
406
407
408
409
410
    def _parse(self, text: StringView) -> Tuple[Optional[Node], StringView]:
        """Applies the parser to the given `text` and returns a node with
        the results or None as well as the text at the position right behind
        the matching string."""
        raise NotImplementedError
        # return None, text  # default behaviour: don't match

411
    @property
eckhart's avatar
eckhart committed
412
413
    def grammar(self) -> 'Grammar':
        if self._grammar != ZOMBIE_GRAMMAR:
eckhart's avatar
eckhart committed
414
415
416
            return self._grammar
        else:
            raise AssertionError('Grammar has not yet been set!')
417
418
419

    @grammar.setter
    def grammar(self, grammar: 'Grammar'):
eckhart's avatar
eckhart committed
420
        if self._grammar == ZOMBIE_GRAMMAR:
421
422
            self._grammar = grammar
            self._grammar_assigned_notifier()
eckhart's avatar
eckhart committed
423
424
425
        elif self._grammar != grammar:
              raise AssertionError("Parser has already been assigned"
                                   "to a different Grammar object!")
426
427
428
429
430
431

    def _grammar_assigned_notifier(self):
        """A function that notifies the parser object that it has been
        assigned to a grammar."""
        pass

eckhart's avatar
eckhart committed
432
    def _apply(self, func: ApplyFunc, flip: FlagFunc) -> bool:
433
434
        """
        Applies function `func(parser)` recursively to this parser and all
eckhart's avatar
eckhart committed
435
436
437
438
439
440
441
442
443
        descendant parsers, if any exist.

        In order to break cycles, function `flip` is called, which should
        return `True`, if this parser has already been visited. If not, it
        flips the cycle detection flag and returns `False`.

        This is a protected function and should not called from outside
        class Parser or any of its descendants. The entry point for external
        calls is the method `apply()` without underscore!
444
        """
eckhart's avatar
eckhart committed
445
        if flip(func, self.cycle_detection):
446
447
448
449
450
            return False
        else:
            func(self)
            return True

eckhart's avatar
eckhart committed
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
    def apply(self, func: ApplyFunc):
        """
        Applies function `func(parser)` recursively to this parser and all
        descendant parsers, if any exist. Traversal is pre-order.
        """
        def positive_flip(f: ApplyFunc, flagged: Set[Callable]) -> bool:
            """Returns True, if function `f` has already been applied to this
            parser and sets the flag accordingly. Interprets `f in flagged == True`
            as meaning that `f` has already been applied."""
            if f in flagged:
                return True
            else:
                flagged.add(f)
                return False

        def negative_flip(f: ApplyFunc, flagged: Set[Callable]) -> bool:
            """Returns True, if function `f` has already been applied to this
            parser and sets the flag accordingly. Interprets `f in flagged == False`
            as meaning that `f` has already been applied."""
            if f not in flagged:
                return True
            else:
                flagged.remove(f)
                return False

        if func in self.cycle_detection:
            self._apply(func, negative_flip)
        else:
            self._apply(func, positive_flip)

481
482
483
484
485
486
487
488
489
490
491
492
493
494
495

def mixin_comment(whitespace: str, comment: str) -> str:
    """
    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).
    """
    wspc = '(?:' + whitespace + '(?:' + comment + whitespace + ')*)'
    return wspc


496
class UnknownParserError(KeyError):
eckhart's avatar
eckhart committed
497
    """UnknownParserError is raised if a Grammar object is called with a
498
    parser that does not exist or if in the course of parsing a parser
eckhart's avatar
eckhart committed
499
    is referred to that does not exist."""
500
501


502
503
504
505
506
507
508
509
510
511
512
513
514
class Grammar:
    r"""
    Class Grammar directs the parsing process and stores global state
    information of the parsers, i.e. state information that is shared
    accross parsers.

    Grammars are basically collections of parser objects, which are
    connected to an instance object of class Grammar. There exist two
    ways of connecting parsers to grammar objects: Either by passing
    the root parser object to the constructor of a Grammar object
    ("direct instantiation"), or by assigning the root parser to the
    class variable "root__" of a descendant class of class Grammar.

eckhart's avatar
eckhart committed
515
    Example for direct instantiation of a grammar::
516

517
        >>> number = RE('\d+') + RE('\.') + RE('\d+') | RE('\d+')
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
        >>> number_parser = Grammar(number)
        >>> number_parser("3.1416").content
        '3.1416'

    Collecting the parsers that define a grammar in a descendant class of
    class Grammar and assigning the named parsers to class variables
    rather than global variables has several advantages:

    1. It keeps the namespace clean.

    2. The parser names of named parsers do not need to be passed to the
       constructor of the Parser object explicitly, but it suffices to
       assign them to class variables, which results in better
       readability of the Python code.

    3. The parsers in the class do not necessarily need to be connected
       to one single root parser, which is helpful for testing and
       building up a parser successively of several components.

    As a consequence, though, it is highly recommended that a Grammar
    class should not define any other variables or methods with names
    that are legal parser names. A name ending with a double
    underscore '__' is *not* a legal parser name and can safely be
    used.

543
    Example::
544
545
546
547
548
549
550
551
552

        class Arithmetic(Grammar):
            # special fields for implicit whitespace and comment configuration
            COMMENT__ = r'#.*(?:\n|$)'  # Python style comments
            wspR__ = mixin_comment(whitespace=r'[\t ]*', comment=COMMENT__)

            # parsers
            expression = Forward()
            INTEGER = RE('\\d+')
553
554
555
            factor = INTEGER | TKN("(") + expression + TKN(")")
            term = factor + ZeroOrMore((TKN("*") | TKN("/")) + factor)
            expression.set(term + ZeroOrMore((TKN("+") | TKN("-")) + term))
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
            root__ = expression

    Upon instantiation the parser objects are deep-copied to the
    Grammar object and assigned to object variables of the same name.
    Any parser that is directly assigned to a class variable is a
    'named' parser and its field `parser.name` contains the variable
    name after instantiation of the Grammar class. All other parsers,
    i.e. parsers that are defined within a `named` parser, remain
    "anonymous parsers" where `parser.name` is the empty string, unless
    a name has been passed explicitly upon instantiation.
    If one and the same parser is assigned to several class variables
    such as, for example the parser `expression` in the example above,
    the first name sticks.

    Grammar objects are callable. Calling a grammar object with a UTF-8
    encoded document, initiates the parsing of the document with the
    root parser. The return value is the concrete syntax tree. Grammar
    objects can be reused (i.e. called again) after parsing. Thus, it
    is not necessary to instantiate more than one Grammar object per
    thread.

    Grammar classes contain a few special class fields for implicit
    whitespace and comments that should be overwritten, if the defaults
    (no comments, horizontal right aligned whitespace) don't fit:

581
582
583
584
585
    Attributes:
        root__:  The root parser of the grammar. Theoretically, all parsers of the
                 grammar should be reachable by the root parser. However, for testing
                 of yet incomplete grammars class Grammar does not assume that this
                 is the case.
586

Eckhart Arnold's avatar
Eckhart Arnold committed
587
        resume_rules__: A mapping of parser names to a list of regular expressions or search
588
                strings that act as rules to find the the reentry point if a ParserError
Eckhart Arnold's avatar
Eckhart Arnold committed
589
590
                was thrown during the execution of the parser with the respective name.

591
592
593
594
595
        parser_initializiation__:  Before the parser class (!) has been initialized,
                 which happens upon the first time it is instantiated (see
                 :func:_assign_parser_names()` for an explanation), this class
                 field contains a value other than "done". A value of "done" indicates
                 that the class has already been initialized.
596

eckhart's avatar
eckhart committed
597
598
599
600
        python__src__:  For the purpose of debugging and inspection, this field can
                 take the python src of the concrete grammar class
                 (see `dsl.grammar_provider`).

601
602
603
    Attributes:
        all_parsers__:  A set of all parsers connected to this grammar object

eckhart's avatar
eckhart committed
604
605
606
607
        start_parser__:  During parsing, the parser with which the parsing process
                was started (see method `__call__`) or `None` if no parsing process
                is running.

608
609
610
611
612
613
614
615
616
617
        history_tracking__:  A flag indicating that the parsing history shall
                be tracked

        _dirty_flag__:  A flag indicating that the Grammar has been called at
                least once so that the parsing-variables need to be reset
                when it is called again.

        document__:  the text that has most recently been parsed or that is
                currently being parsed.

eckhart's avatar
eckhart committed
618
619
        document_length__:  the length of the document.

620
621
622
        document_lbreaks__:  list of linebreaks within the document, starting
                with -1 and ending with EOF. This helps generating line
                and column number for history recording and will only be
623
                initialized if :attr:`history_tracking__` is true.
624

625
626
627
628
629
        tree__: The root-node of the parsing tree. This variable is available
               for error-reporting already during parsing  via
               ``self.grammar.tree__.add_error``, but it references the full
               parsing tree only after parsing has been finished.

630
        _reversed__:  the same text in reverse order - needed by the `Lookbehind`-
631
632
633
                parsers.

        variables__:  A mapping for variable names to a stack of their respective
634
635
                string values - needed by the :class:`Capture`-, :class:`Retrieve`-
                and :class:`Pop`-parsers.
636
637

        rollback__:  A list of tuples (location, rollback-function) that are
638
639
640
641
                deposited by the :class:`Capture`- and :class:`Pop`-parsers.
                If the parsing process reaches a dead end then all
                rollback-functions up to the point to which it retreats will be
                called and the state of the variable stack restored accordingly.
642
643
644

        last_rb__loc__:  The last, i.e. most advanced location in the text
                where a variable changing operation occurred. If the parser
645
646
                backtracks to a location at or before last_rb__loc__ (i.e.
                location <= last_rb__loc__) then a rollback of all variable
647
648
                changing operations is necessary that occurred after the
                location to which the parser backtracks. This is done by
649
                calling method :func:`rollback_to__(location)`.
650
651
652
653
654
655
656
657
658
659
660
661
662

        call_stack__:  A stack of all parsers that have been called. This
                is required for recording the parser history (for debugging)
                and, eventually, i.e. one day in the future, for tracing through
                the parsing process.

        history__:  A list of parser-call-stacks. A parser-call-stack is
                appended to the list each time a parser either matches, fails
                or if a parser-error occurs.

        moving_forward__: This flag indicates that the parsing process is currently
                moving forward . It is needed to reduce noise in history recording
                and should not be considered as having a valid value if history
663
664
                recording is turned off! (See :func:`add_parser_guard` and its local
                function :func:`guarded_call`)
665
666
667
668

        recursion_locations__:  Stores the locations where left recursion was
                detected. Needed to provide minimal memoization for the left
                recursion detection algorithm, but, strictly speaking, superfluous
669
670
                if full memoization is enabled. (See :func:`add_parser_guard` and its
                local function :func:`guarded_call`)
671
672
673
674
675
676
677
678
679
680

        memoization__:  Turns full memoization on or off. Turning memoization off
                results in less memory usage and sometimes reduced parsing time.
                In some situations it may drastically increase parsing time, so
                it is safer to leave it on. (Default: on)

        left_recursion_handling__:  Turns left recursion handling on or off.
                If turned off, a recursion error will result in case of left
                recursion.
    """
eckhart's avatar
eckhart committed
681
    python_src__ = ''  # type: str
682
683
684
    root__ = ZOMBIE_PARSER  # type: ParserBase
    # root__ must be overwritten with the root-parser by grammar subclass
    parser_initialization__ = "pending"  # type: str
eckhart's avatar
eckhart committed
685
    resume_rules__ = dict()  # type: Dict[str, ResumeList]
686
    # some default values
687
688
    # COMMENT__ = r''  # type: str  # r'#.*(?:\n|$)'
    # WSP_RE__ = mixin_comment(whitespace=r'[\t ]*', comment=COMMENT__)  # type: str
689
690
691
692
693
694
695


    @classmethod
    def _assign_parser_names__(cls):
        """
        Initializes the `parser.name` fields of those
        Parser objects that are directly assigned to a class field with
696
        the field's name, e.g.::
697

698
699
            class Grammar(Grammar):
                ...
700
                symbol = RE('(?!\\d)\\w+')
701

di68kap's avatar
di68kap committed
702
703
704
        After the call of this method symbol.name == "symbol" holds.
        Parser names starting or ending with a double underscore like
        ``root__`` will be ignored. See :func:`sane_parser_name()`
705
706
707
708
709
710
711
712
713
714
715
716
717
718

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

        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.
        """
        if cls.parser_initialization__ != "done":
            cdict = cls.__dict__
            for entry, parser in cdict.items():
                if isinstance(parser, Parser) and sane_parser_name(entry):
di68kap's avatar
di68kap committed
719
                    if isinstance(parser, Forward):
di68kap's avatar
di68kap committed
720
                        if not cast(Forward, parser).parser.name:
di68kap's avatar
di68kap committed
721
722
                            cast(Forward, parser).parser.name = entry
                    else:   # if not parser.name:
723
                        parser.name = entry
724
725
726
727
728
            cls.parser_initialization__ = "done"


    def __init__(self, root: Parser = None) -> None:
        self.all_parsers__ = set()             # type: Set[ParserBase]
eckhart's avatar
eckhart committed
729
        self.start_parser__ = None             # type: Optional[ParserBase]
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
        self._dirty_flag__ = False             # type: bool
        self.history_tracking__ = False        # type: bool
        self.memoization__ = True              # type: bool
        self.left_recursion_handling__ = True  # type: bool
        self._reset__()

        # 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.
        self.root__ = copy.deepcopy(root) if root else copy.deepcopy(self.__class__.root__)
        self.root__.apply(self._add_parser__)


    def __getitem__(self, key):
        try:
            return self.__dict__[key]
        except KeyError:
            parser_template = getattr(self, key, None)
            if parser_template:
                # add parser to grammar object on the fly...
                parser = copy.deepcopy(parser_template)
                parser.apply(self._add_parser__)
                # assert self[key] == parser
                return self[key]
758
            raise UnknownParserError('Unknown parser "%s" !' % key)
759
760
761


    def _reset__(self):
762
        self.tree__ = RootNode()              # type: RootNode
763
764
        self.document__ = EMPTY_STRING_VIEW   # type: StringView
        self._reversed__ = EMPTY_STRING_VIEW  # type: StringView
eckhart's avatar
eckhart committed
765
        self.document_length__ = 0            # type: int
766
767
        self.document_lbreaks__ = []          # type: List[int]
        # variables stored and recalled by Capture and Retrieve parsers
eckhart's avatar
eckhart committed
768
        self.variables__ = defaultdict(lambda: [])  # type: DefaultDict[str, List[str]]
769
770
771
        self.rollback__ = []                  # type: List[Tuple[int, Callable]]
        self.last_rb__loc__ = -1              # type: int
        # support for call stack tracing
eckhart's avatar
eckhart committed
772
        self.call_stack__ = []                # type: List[ParserBase]
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
        # snapshots of call stacks
        self.history__ = []                   # type: List[HistoryRecord]
        # also needed for call stack tracing
        self.moving_forward__ = False         # type: bool
        self.recursion_locations__ = set()    # type: Set[int]


    @property
    def reversed__(self) -> StringView:
        """
        Returns a reversed version of the currently parsed document. As
        about the only case where this is needed is the Lookbehind-parser,
        this is done lazily.
        """
        if not self._reversed__:
788
            self._reversed__ = StringView(self.document__.get_text()[::-1])
789
790
791
792
793
794
795
796
797
798
799
        return self._reversed__


    def _add_parser__(self, parser: Parser) -> None:
        """
        Adds the particular copy of the parser object to this
        particular instance of Grammar.
        """
        if parser.name:
            # prevent overwriting instance variables or parsers of a different class
            assert parser.name not in self.__dict__ or \
800
                isinstance(self.__dict__[parser.name], parser.__class__), \
801
                ('Cannot add parser "%s" because a field with the same name '
di68kap's avatar
di68kap committed
802
803
                 'already exists in grammar object: %s!'
                 % (parser.name, str(self.__dict__[parser.name])))
804
            setattr(self, parser.name, parser)
805
        parser.tag_name = parser.name or parser.ptype
806
807
808
809
        self.all_parsers__.add(parser)
        parser.grammar = self


eckhart's avatar
eckhart committed
810
811
812
813
    def __call__(self,
                 document: str,
                 start_parser: Union[str, Parser] = "root__",
                 track_history: bool = False) -> RootNode:
814
815
816
817
818
        """
        Parses a document with with parser-combinators.

        Args:
            document (str): The source text to be parsed.
di68kap's avatar
di68kap committed
819
820
            start_parser (str or Parser): The name of the parser with which
                to start. This is useful for testing particular parsers
821
                (i.e. particular parts of the EBNF-Grammar.)
di68kap's avatar
di68kap committed
822
823
824
825
826
            track_history (bool): If true, the parsing history will be
                recorded in self.history__. If logging is turned on (i.e.
                DHParser.log.is_logging() returns true), the parsing history
                will always be recorded, even if `False` is passed to
                the `track_history` parameter.
827
        Returns:
di68kap's avatar
di68kap committed
828
            Node: The root node to the parse tree.
829
        """
830

eckhart's avatar
eckhart committed
831
        def tail_pos(predecessors: Union[List[Node], Tuple[Node, ...], None]) -> int:
eckhart's avatar
eckhart committed
832
833
            """Adds the position after the last node in the list of
            predecessors to the node."""
eckhart's avatar
eckhart committed
834
            return predecessors[-1].pos + len(predecessors[-1]) if predecessors else 0
eckhart's avatar
eckhart committed
835

836
837
838
839
840
841
842
843
844
        # assert isinstance(document, str), type(document)
        if self.root__ is None:
            raise NotImplementedError()
        if self._dirty_flag__:
            self._reset__()
            for parser in self.all_parsers__:
                parser.reset()
        else:
            self._dirty_flag__ = True
di68kap's avatar
di68kap committed
845
846
847
848
        self.history_tracking__ = track_history or is_logging()
        # safe tracking state, because history_tracking__ might be set to false, later,
        # but original tracking state is needed for additional error information.
        track_history = self.history_tracking__
849
        self.document__ = StringView(document)
eckhart's avatar
eckhart committed
850
        self.document_length__ = len(self.document__)
851
        self.document_lbreaks__ = linebreaks(document) if self.history_tracking__ else []
Eckhart Arnold's avatar
Eckhart Arnold committed
852
        self.last_rb__loc__ = -1  # rollback location
853
        parser = self[start_parser] if isinstance(start_parser, str) else start_parser
eckhart's avatar
eckhart committed
854
        self.start_parser__ = parser
855
856
857
858
859
860
861
862
        assert parser.grammar == self, "Cannot run parsers from a different grammar object!" \
                                       " %s vs. %s" % (str(self), str(parser.grammar))
        result = None  # type: Optional[Node]
        stitches = []  # type: List[Node]
        rest = self.document__
        if not rest:
            result, _ = parser(rest)
            if result is None:
eckhart's avatar
eckhart committed
863
                result = Node(None, '').init_pos(0)
eckhart's avatar
eckhart committed
864
                self.tree__.new_error(result,
di68kap's avatar
di68kap committed
865
866
                                      'Parser "%s" did not match empty document.' % str(parser),
                                      Error.PARSER_DID_NOT_MATCH)
867
868
869
870
871
872
        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:
di68kap's avatar
di68kap committed
873
874
875
876
877
878
                    err_info = '' if not track_history else \
                               '\n    Most advanced: %s\n    Last match:    %s;' % \
                               (str(HistoryRecord.most_advanced_match(self.history__)),
                                str(HistoryRecord.last_match(self.history__)))
                    # Check if a Lookahead-Parser did match. Needed for testing, because
                    # in a test case this is not necessarily an error.
eckhart's avatar
eckhart committed
879
                    last_record = self.history__[-2] if len(self.history__) > 1 else None  # type: Optional[HistoryRecord]
di68kap's avatar
di68kap committed
880
881
                    if last_record and parser != self.root__ \
                            and last_record.status == HistoryRecord.MATCH \
eckhart's avatar
eckhart committed
882
883
                            and last_record.node.pos \
                            + len(last_record.node) >= len(self.document__) \
di68kap's avatar
di68kap committed
884
885
886
887
888
889
890
                            and any(isinstance(parser, Lookahead)
                                    for parser in last_record.call_stack):
                        error_msg = 'Parser did not match except for lookahead! ' + err_info
                        error_code = Error.PARSER_LOOKAHEAD_MATCH_ONLY
                    else:
                        error_msg = 'Parser did not match!' + err_info
                        error_code = Error.PARSER_DID_NOT_MATCH
891
892
                else:
                    stitches.append(result)
eckhart's avatar
eckhart committed
893
894
895
896
897
898
                    error_msg = "Parser stopped before end" \
                        + (("! trying to recover"
                            + (" but stopping history recording at this point."
                               if self.history_tracking__ else "..."))
                            if len(stitches) < MAX_DROPOUTS
                            else " too often! Terminating parser.")
di68kap's avatar
di68kap committed
899
                    error_code = Error.PARSER_STOPPED_BEFORE_END
eckhart's avatar
eckhart committed
900
                stitches.append(Node(None, skip).init_pos(tail_pos(stitches)))
di68kap's avatar
di68kap committed
901
                self.tree__.new_error(stitches[-1], error_msg, error_code)
902
                if self.history_tracking__:
eckhart's avatar
eckhart committed
903
904
905
906
907
908
909
                    # # some parsers may have matched and left history records with nodes != None.
                    # # Because these are not connected to the stitched 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
                    # for record in self.history__:
                    #     if record.node and record.node._pos < 0:
                    #         record.node.init_pos(0)
910
911
912
913
914
915
                    record = HistoryRecord(self.call_stack__.copy(), stitches[-1], rest)
                    self.history__.append(record)
                    # stop history tracking when parser returned too early
                    self.history_tracking__ = False
        if stitches:
            if rest:
Eckhart Arnold's avatar
Eckhart Arnold committed
916
                stitches.append(Node(None, rest))
917
918
919
920
921
922
            #try:
            result = Node(None, tuple(stitches)).init_pos(0)
            # except AssertionError as error:
            #     # some debugging output
            #     print(Node(None, tuple(stitches)).as_sxpr())
            #     raise error
923
        if any(self.variables__.values()):
di68kap's avatar
di68kap committed
924
            error_msg = "Capture-retrieve-stack not empty after end of parsing: " \
925
                + str(self.variables__)
di68kap's avatar
di68kap committed
926
            error_code = Error.CAPTURE_STACK_NOT_EMPTY
927
928
929
930
931
            if result:
                if result.children:
                    # add another child node at the end to ensure that the position
                    # of the error will be the end of the text. Otherwise, the error
                    # message above ("...after end of parsing") would appear illogical.
932
                    error_node = Node(ZOMBIE, '').init_pos(tail_pos(result.children))
di68kap's avatar
di68kap committed
933
                    self.tree__.new_error(error_node, error_msg, error_code)
eckhart's avatar
eckhart committed
934
                    result.result = result.children + (error_node,)
935
                else:
di68kap's avatar
di68kap committed
936
                    self.tree__.new_error(result, error_msg, error_code)
eckhart's avatar
eckhart committed
937
        # result.pos = 0  # calculate all positions
938
        # result.collect_errors(self.document__)
eckhart's avatar
eckhart committed
939
940
        if result:
            self.tree__.swallow(result)
eckhart's avatar
eckhart committed
941
        self.start_parser__ = None
942
        return self.tree__
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960


    def push_rollback__(self, location, func):
        """
        Adds a rollback function that either removes or re-adds
        values on the variable stack (`self.variables`) that have been
        added (or removed) by Capture or Pop Parsers, the results of
        which have been dismissed.
        """
        self.rollback__.append((location, func))
        self.last_rb__loc__ = location


    def rollback_to__(self, location):
        """
        Rolls back the variable stacks (`self.variables`) to its
        state at an earlier location in the parsed document.
        """
Eckhart Arnold's avatar
Eckhart Arnold committed
961
        while self.rollback__ and self.rollback__[-1][0] >= location:
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
            _, rollback_func = self.rollback__.pop()
            # assert not loc > self.last_rb__loc__, \
            #     "Rollback confusion: line %i, col %i < line %i, col %i" % \
            #     (*line_col(self.document__, len(self.document__) - loc),
            #      *line_col(self.document__, len(self.document__) - self.last_rb__loc__))
            rollback_func()
        self.last_rb__loc__ == self.rollback__[-1][0] if self.rollback__ \
            else (len(self.document__) + 1)


def dsl_error_msg(parser: Parser, error_str: str) -> str:
    """
    Returns an error message for errors in the parser configuration,
    e.g. errors that result in infinite loops.

    Args:
        parser (Parser):  The parser where the error was noticed. Note
            that this is not necessarily the parser that caused the
            error but only where the error became apparent.
        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.
    """
    msg = ["DSL parser specification error:", error_str, 'Caught by parser "%s".' % str(parser)]
    if parser.grammar.history__:
        msg.extend(["\nCall stack:", parser.grammar.history__[-1].stack])
    else:
        msg.extend(["\nEnable history tracking in Grammar object to display call stack."])
    return " ".join(msg)


eckhart's avatar
eckhart committed
994
995
996
ZOMBIE_GRAMMAR = Grammar()


997
998
########################################################################
#
999
# _Token and Regular Expression parser classes (i.e. leaf classes)
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
#
########################################################################


class PreprocessorToken(Parser):
    """
    Parses tokens that have been inserted by a preprocessor.

    Preprocessors can generate Tokens with the ``make_token``-function.
    These tokens start and end with magic characters that can only be
    matched by the PreprocessorToken Parser. Such tokens can be used to
    insert BEGIN - END delimiters at the beginning or ending of a
    quoted block, for example.
    """

    def __init__(self, token: str) -> None:
        assert token and token.isupper()
1017
        assert RX_TOKEN_NAME.match(token)
1018
1019
        super().__init__()
        self.name = token
1020

1021
1022
1023
    def __deepcopy__(self, memo):
        duplicate = self.__class__(self.name)
        duplicate.name = self.name
eckhart's avatar
eckhart committed
1024
        duplicate.ptype = self.ptype
1025
1026
        return duplicate

1027
    def _parse(self, text: StringView) -> Tuple[Optional[Node], StringView]:
1028
1029
1030
        if text[0:1] == BEGIN_TOKEN:
            end = text.find(END_TOKEN, 1)
            if end < 0:
1031
                node = Node(self.tag_name, '')
eckhart's avatar
eckhart committed
1032
1033
                self.grammar.tree__.new_error(
                    node,
1034
1035
                    'END_TOKEN delimiter missing from preprocessor token. '
                    '(Most likely due to a preprocessor bug!)')  # type: Node
1036
                return node, text[1:]
1037
            elif end == 0:
1038
                node = Node(self.tag_name, '')
eckhart's avatar
eckhart committed
1039
1040
                self.grammar.tree__.new_error(
                    node,
1041
1042
                    'Preprocessor-token cannot have zero length. '
                    '(Most likely due to a preprocessor bug!)')
1043
                return node, text[2:]
1044
            elif text.find(BEGIN_TOKEN, 1, end) >= 0:
1045
                node = Node(self.tag_name, text[len(self.name) + 1:end])
eckhart's avatar
eckhart committed
1046
1047
                self.grammar.tree__.new_error(
                    node,
1048
1049
1050
1051
1052
                    'Preprocessor-tokens must not be nested or contain '
                    'BEGIN_TOKEN delimiter as part of their argument. '
                    '(Most likely due to a preprocessor bug!)')
                return node, text[end:]
            if text[1:len(self.name) + 1] == self.name:
1053
                return Node(self.tag_name, text[len(self.name) + 2:end]), text[end + 1:]
1054
1055
1056
        return None, text


1057
class Token(Parser):
eckhart's avatar
eckhart committed
1058
    """
1059
    Parses plain text strings. (Could be done by RegExp as well, but is faster.)
eckhart's avatar
eckhart committed
1060

eckhart's avatar
eckhart committed
1061
1062
    Example::

1063
        >>> while_token = Token("while")
eckhart's avatar
eckhart committed
1064
1065
        >>> Grammar(while_token)("while").content
        'while'
eckhart's avatar
eckhart committed
1066
    """
1067
    assert TOKEN_PTYPE == ":Token"
eckhart's avatar
eckhart committed
1068

1069
1070
    def __init__(self, text: str) -> None:
        super().__init__()
eckhart's avatar
eckhart committed
1071
        self.text = text
1072
        self.len = len(text)
eckhart's avatar
eckhart committed
1073
1074

    def __deepcopy__(self, memo):
1075
1076
        duplicate = self.__class__(self.text)
        duplicate.name = self.name
1077
        duplicate.ptype = self.ptype
1078
        return duplicate
eckhart's avatar
eckhart committed
1079

1080
    def _parse(self, text: StringView) -> Tuple[Optional[Node], StringView]:
eckhart's avatar
eckhart committed
1081
        if text.startswith(self.text):
1082
            return Node(self.tag_name, self.text, True), text[self.len:]
eckhart's avatar
eckhart committed
1083
1084
        return None, text

1085
1086
1087
    def __repr__(self):
        return ("'%s'" if self.text.find("'") <= 0 else '"%s"') % self.text

eckhart's avatar
eckhart committed
1088

1089
1090
1091
1092
1093
1094
1095
1096
1097
class RegExp(Parser):
    r"""
    Regular expression parser.

    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.

1098
1099
1100
1101
1102
    Example::

        >>> word = RegExp(r'\w+')
        >>> Grammar(word)("Haus").content
        'Haus'
1103

1104
    EBNF-Notation:  ``/ ... /``
1105

1106
    EBNF-Example:   ``word = /\w+/``
1107
1108
    """

1109
1110
    def __init__(self, regexp) -> None:
        super().__init__()
1111
1112
1113
1114
1115
1116
1117
1118
        self.regexp = re.compile(regexp) if isinstance(regexp, str) else regexp

    def __deepcopy__(self, memo):
        # `regex` supports deep copies, but not `re`
        try:
            regexp = copy.deepcopy(self.regexp, memo)
        except TypeError:
            regexp = self.regexp.pattern
1119
1120
        duplicate = self.__class__(regexp)
        duplicate.name = self.name
eckhart's avatar
eckhart committed
1121
        duplicate.ptype = self.ptype
1122
        return duplicate
1123

1124
    def _parse(self, text: StringView) -> Tuple[Optional[Node], StringView]:
1125
1126
1127
1128
        match = text.match(self.regexp)
        if match:
            capture = match.group(0)
            end = text.index(match.end())
1129
1130
            # regular expression must never match preprocessor-tokens!
            # TODO: Find a better solution here? e.g. static checking/re-mangling at compile time
1131
1132
1133
1134
            i = capture.find(BEGIN_TOKEN)
            if i >= 0:
                capture = capture[:i]
                end = i
1135
            return Node(self.tag_name, capture, True), text[end:]
1136
1137
1138
        return None, text

    def __repr__(self):
1139
        return escape_control_characters('/%s/' % self.regexp.pattern)
1140
1141


eckhart's avatar
eckhart committed
1142
def withWS(parser_factory, wsL='', wsR=r'\s*'):
1143
    """Syntactic Sugar for 'Series(Whitespace(wsL), parser_factory(), Whitespace(wsR))'.
1144
    """
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
    if wsL and isinstance(wsL, str):
        wsL = Whitespace(wsL)
    if wsR and isinstance(wsR, str):
        wsR = Whitespace(wsR)
    if wsL and wsR:
        return Series(wsL, parser_factory(), wsR)
    elif wsL:
        return Series(wsL, parser_factory())
    elif wsR:
        return Series(parser_factory(), wsR)
    else:
        return parser_factory()
1157

1158

eckhart's avatar
eckhart committed
1159
def RE(regexp, wsL='', wsR=r'\s*'):
1160
    """Syntactic Sugar for 'Series(Whitespace(wsL), RegExp(regexp), Whitespace(wsR))'"""
eckhart's avatar
eckhart committed
1161
    return withWS(lambda: RegExp(regexp), wsL, wsR)
1162
1163


eckhart's avatar
eckhart committed
1164
def TKN(token, wsL='', wsR=r'\s*'):
1165
1166
    """Syntactic Sugar for 'Series(Whitespace(wsL), Token(token), Whitespace(wsR))'"""
    return withWS(lambda: Token(token), wsL, wsR)
1167
1168


1169
1170
1171
1172
class Whitespace(RegExp):
    """An variant of RegExp that signifies through its class name that it
    is a RegExp-parser for whitespace."""
    assert WHITESPACE_PTYPE == ":Whitespace"
eckhart's avatar
eckhart committed
1173

eckhart's avatar
eckhart committed
1174
1175
1176
    def __repr__(self):
        return '~'

1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196

########################################################################
#
# Containing parser classes, i.e. parsers that contain other parsers
# to which they delegate parsing
#
########################################################################


class UnaryOperator(Parser):
    """
    Base class of all unary parser operators, i.e. parser that contains
    one and only one other parser, like the optional parser for example.

    The UnaryOperator base class supplies __deepcopy__ and apply
    methods for unary parser operators. The __deepcopy__ method needs
    to be overwritten, however, if the constructor of a derived class
    has additional parameters.
    """

1197
1198
    def __init__(self, parser: Parser) -> None:
        super(UnaryOperator, self).__init__()
1199
1200
1201
1202
1203
        assert isinstance(parser, Parser), str(parser)
        self.parser = parser  # type: Parser

    def __deepcopy__(self, memo):
        parser = copy.deepcopy(self.parser, memo)
1204
1205
        duplicate = self.__class__(parser)
        duplicate.name = self.name
eckhart's avatar
eckhart committed
1206
        duplicate.ptype = self.ptype
1207
        return duplicate
1208

eckhart's avatar
eckhart committed
1209
1210
1211
    def _apply(self, func: ApplyFunc, flip: FlagFunc) -> bool:
        if super()._apply(func, flip):
            self.parser._apply(func, flip)
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
            return True
        return False


class NaryOperator(Parser):
    """
    Base class of all Nnary parser operators, i.e. parser that
    contains one or more other parsers, like the alternative
    parser for example.

    The NnaryOperator base class supplies __deepcopy__ and apply methods
    for unary parser operators. The __deepcopy__ method needs to be
    overwritten, however, if the constructor of a derived class has
    additional parameters.
    """

1228
1229
    def __init__(self, *parsers: Parser) -> None:
        super().__init__()
1230
1231
1232
1233
1234
        assert all([isinstance(parser, Parser) for parser in parsers]), str(parsers)
        self.parsers = parsers  # type: Tuple[Parser, ...]

    def __deepcopy__(self, memo):
        parsers = copy.deepcopy(self.parsers, memo)
1235
1236
        duplicate = self.__class__(*parsers)
        duplicate.name = self.name
eckhart's avatar
eckhart committed
1237
        duplicate.ptype = self.ptype
1238
        return duplicate
1239

eckhart's avatar
eckhart committed
1240
1241
    def _apply(self, func: ApplyFunc, flip: FlagFunc) -> bool:
        if super()._apply(func, flip):
1242
            for parser in self.parsers:
eckhart's avatar
eckhart committed
1243
                parser._apply(func, flip)
1244
1245
1246
1247
1248
1249
            return True
        return False


class Option(UnaryOperator):
    r"""
1250
    Parser ``Option`` always matches, even if its child-parser
1251
1252
    did not match.

1253
    If the child-parser did not match ``Option`` returns a node
1254
1255
    with no content and does not move forward in the text.

1256
    If the child-parser did match, ``Option`` returns the a node
eckhart's avatar
eckhart committed
1257
    with the node returned by the child-parser as its single
1258
1259
1260
    child and the text at the position where the child-parser
    left it.

1261
1262
    Examples::

1263
        >>> number = Option(TKN('-')) + RegExp(r'\d+') + Option(RegExp(r'\.\d+'))
1264
1265
1266
1267
1268
1269
        >>> Grammar(number)('3.14159').content
        '3.14159'
        >>> Grammar(number)('3.14159').structure
        '(:Series (:Option) (:RegExp "3") (:Option (:RegExp ".14159")))'
        >>> Grammar(number)('-1').content
        '-1'
1270

1271
    EBNF-Notation: ``[ ... ]``
1272

1273
    EBNF-Example:  ``number = ["-"]  /\d+/  [ /\.\d+/ ]``
1274
1275
    """

1276
1277
    def __init__(self, parser: Parser) -> None:
        super().__init__(parser)
1278
1279
        # assert isinstance(parser, Parser)
        assert not isinstance(parser, Option), \
eckhart's avatar
eckhart committed
1280
            "Redundant nesting of options: %s%s" % (str(parser.name), str(self.ptype))
1281

1282
    def _parse(self, text: StringView) -> Tuple[Optional[Node], StringView]:
1283
1284
        node, text = self.parser(text)
        if node:
1285
1286
            return Node(self.tag_name, node), text
        return Node(self.tag_name, ()), text
eckhart's avatar