parse.py 79.7 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

33
from collections import defaultdict
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
40
from DHParser.syntaxtree import Node, RootNode, ParserBase, WHITESPACE_PTYPE, \
di68kap's avatar
di68kap committed
41
    PLAINTEXT_PTYPE, TOKEN_PTYPE, ZOMBIE_PARSER
42
from DHParser.toolkit import sane_parser_name, escape_control_characters, re, typing
43
from typing import Callable, cast, Dict, DefaultDict, List, Set, Tuple, Union, Optional
44

45

46
__all__ = ('Parser',
47
           'UnknownParserError',
48
49
50
           'Grammar',
           'PreprocessorToken',
           'RegExp',
51
           'Whitespace',
52
53
           '_RE',
           '_Token',
54
55
56
57
58
59
60
61
62
63
64
65
           'mixin_comment',
           # 'UnaryOperator',
           # 'NaryOperator',
           'Synonym',
           'Option',
           'ZeroOrMore',
           'OneOrMore',
           'Series',
           'Alternative',
           'AllOf',
           'SomeOf',
           'Unordered',
eckhart's avatar
eckhart committed
66
           'Required',
67
68
69
70
71
72
73
74
75
76
           'Lookahead',
           'NegativeLookahead',
           'Lookbehind',
           'NegativeLookbehind',
           'last_value',
           'counterpart',
           'accumulate',
           'Capture',
           'Retrieve',
           'Pop',
77
           'Forward')
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101


########################################################################
#
# 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


def add_parser_guard(parser_func):
    """
    Add a wrapper function to a parser functions (i.e. Parser.__call__ method)
    that takes care of memoizing, left recursion and, optionally, tracing
    (aka "history tracking") of parser calls. Returns the wrapped call.
    """
    def guarded_call(parser: 'Parser', text: StringView) -> Tuple[Optional[Node], StringView]:
        try:
Eckhart Arnold's avatar
Eckhart Arnold committed
102
103
            grammar = parser.grammar
            location = grammar.document_length__ - len(text)
104

Eckhart Arnold's avatar
Eckhart Arnold committed
105
            if grammar.last_rb__loc__ >= location:
106
107
108
109
110
                grammar.rollback_to__(location)

            # if location has already been visited by the current parser,
            # return saved result
            if location in parser.visited:
111
                # no history recording in case of meomized results
112
113
114
115
116
117
118
119
                return parser.visited[location]

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

            # break left recursion at the maximum allowed depth
            if grammar.left_recursion_handling__:
120
                if parser.recursion_counter[location] > LEFT_RECURSION_DEPTH:
121
122
123
124
125
126
127
128
129
                    grammar.recursion_locations__.add(location)
                    return None, text
                parser.recursion_counter[location] += 1

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

            if grammar.left_recursion_handling__:
                parser.recursion_counter[location] -= 1
130
                # don't clear recursion_locations__ !!!
131
132
133
134
135
136
137
138
139
140
141
142

            if node is None:
                # retrieve an earlier match result (from left recursion) if it exists
                if location in grammar.recursion_locations__:
                    if location in parser.visited:
                        node, rest = parser.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
                    parser.visited[location] = (None, rest)
eckhart's avatar
eckhart committed
143
            else:
144
                assert node._pos < 0
Eckhart Arnold's avatar
Eckhart Arnold committed
145
146
147
                node._pos = location
                assert node._pos >= 0, str("%i < %i" % (grammar.document_length__, location))
                if (grammar.last_rb__loc__ < location
eckhart's avatar
eckhart committed
148
149
150
151
152
153
                        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
                    parser.visited[location] = (node, rest)
154

155
            # Mind that meomized parser calls will not appear in the history record!
156
157
            if grammar.history_tracking__:
                # don't track returning parsers except in case an error has occurred
158
                # remaining = len(rest)
di68kap's avatar
di68kap committed
159
                if (grammar.moving_forward__ or (node and node.errors)):
160
161
162
163
164
165
166
167
                    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")))]) + " ...")
168
            node._pos = location
eckhart's avatar
eckhart committed
169
            grammar.tree__.new_error(node, "maximum recursion depth of parser reached; "
170
                                     "potentially due to too many errors!")
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
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
            rest = EMPTY_STRING_VIEW

        return node, rest

    return guarded_call


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.
    """

    ApplyFunc = Callable[['Parser'], None]

238
    def __init__(self) -> None:
239
        # assert isinstance(name, str), str(name)
240
        super().__init__()
eckhart's avatar
eckhart committed
241
        self._grammar = None  # type: Optional['Grammar']
242
243
244
245
        self.reset()

        # add "aspect oriented" wrapper around parser calls
        # for memoizing, left recursion and tracing
di68kap's avatar
di68kap committed
246
        if not isinstance(self, Forward):  # should Forward-Parser not be guarded? Not sure...
247
248
249
250
251
252
            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
253
254
255
256
257
258
259
260

    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.
        """
261
262
        duplicate = self.__class__()
        duplicate.name = self.name
di68kap's avatar
di68kap committed
263
        # duplicate.ptype =  self.ptype
264
        return duplicate
265
266
267
268
269

    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."""
270
271
272
        self.visited = dict()  # type: Dict[int, Tuple[Optional[Node], StringView]]
        self.recursion_counter = defaultdict(lambda :0)  # type: DefaultDict[int, int]
        self.cycle_detection = set()  # type: Set[Callable]
273
274
275
276
277
278
279
280
281
282
283
284
285
286
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
337
338
339
340

    def __call__(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."""
        return None, text  # default behaviour: don't match

    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)

    @property
    def grammar(self) -> 'Grammar':
        return self._grammar

    @grammar.setter
    def grammar(self, grammar: 'Grammar'):
        if self._grammar is None:
            self._grammar = grammar
            self._grammar_assigned_notifier()
        else:
            assert self._grammar == grammar, \
                "Parser has already been assigned to a different Grammar object!"

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

    def apply(self, func: ApplyFunc) -> bool:
        """
        Applies function `func(parser)` recursively to this parser and all
        descendant parsers if any exist. The same function can never
        be applied twice between calls of the ``reset()``-method!
        Returns `True`, if function has been applied, `False` if function
        had been applied earlier already and thus has not been applied again.
        """
        if func in self.cycle_detection:
            return False
        else:
            assert not self.visited, "No calls to Parser.apply() during or " \
                                     "after ongoing parsing process. (Call Parser.reset() first.)"
            self.cycle_detection.add(func)
            func(self)
            return True


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


341
class UnknownParserError(KeyError):
eckhart's avatar
eckhart committed
342
    """UnknownParserError is raised if a Grammar object is called with a
343
    parser that does not exist or if in the course of parsing a parser
eckhart's avatar
eckhart committed
344
    is referred to that does not exist."""
345
346


347
348
349
350
351
352
353
354
355
356
357
358
359
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
360
    Example for direct instantiation of a grammar::
361

362
        >>> number = _RE('\d+') + _RE('\.') + _RE('\d+') | _RE('\d+')
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
        >>> 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.

388
    Example::
389
390
391
392
393
394
395
396
397

        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+')
398
399
400
            factor = INTEGER | _Token("(") + expression + _Token(")")
            term = factor + ZeroOrMore((_Token("*") | _Token("/")) + factor)
            expression.set(term + ZeroOrMore((_Token("+") | _Token("-")) + term))
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
            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:

426
427
428
429
430
431
432
    Attributes:
        COMMENT__:  regular expression string for matching comments

        WSP__:   regular expression for whitespace and comments

        wspL__:  regular expression string for left aligned whitespace,
                 which either equals WSP__ or is empty.
433

434
435
        wspR__:  regular expression string for right aligned whitespace,
                 which either equals WSP__ or is empty.
436

437
438
439
440
        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.
441

442
443
444
445
446
        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.
447

eckhart's avatar
eckhart committed
448
449
450
451
        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`).

452
453
454
455
456
457
    Attributes:
        all_parsers__:  A set of all parsers connected to this grammar object

        history_tracking__:  A flag indicating that the parsing history shall
                be tracked

458
459
        whitespace__: A parser for the implicit optional whitespace (or the
                :class:zombie-parser if the default is empty). The default
460
                whitespace will be used by parsers :class:`_Token` and, if no
461
                other parsers are passed to its constructor, by parser
462
                :class:`_RE`. It can also be place explicitly in the
463
                EBNF-Grammar via the "~"-sign.
464

465
466
467
468
469
        wsp_left_parser__: The same as ``whitespace`` for
               left-adjacent-whitespace.

        wsp_right_parser__: The same as ``whitespace`` for
               right-adjacent-whitespace.
470
471
472
473
474
475
476
477

        _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
478
479
        document_length__:  the length of the document.

480
481
482
        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
483
                initialized if :attr:`history_tracking__` is true.
484

485
486
487
488
489
        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.

490
        _reversed__:  the same text in reverse order - needed by the `Lookbehind`-
491
492
493
                parsers.

        variables__:  A mapping for variable names to a stack of their respective
494
495
                string values - needed by the :class:`Capture`-, :class:`Retrieve`-
                and :class:`Pop`-parsers.
496
497

        rollback__:  A list of tuples (location, rollback-function) that are
498
499
500
501
                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.
502
503
504

        last_rb__loc__:  The last, i.e. most advanced location in the text
                where a variable changing operation occurred. If the parser
505
506
                backtracks to a location at or before last_rb__loc__ (i.e.
                location <= last_rb__loc__) then a rollback of all variable
507
508
                changing operations is necessary that occurred after the
                location to which the parser backtracks. This is done by
509
                calling method :func:`rollback_to__(location)`.
510
511
512
513
514
515
516
517
518
519
520
521
522

        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
523
524
                recording is turned off! (See :func:`add_parser_guard` and its local
                function :func:`guarded_call`)
525
526
527
528

        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
529
530
                if full memoization is enabled. (See :func:`add_parser_guard` and its
                local function :func:`guarded_call`)
531
532
533
534
535
536
537
538
539
540

        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
541
    python_src__ = ''  # type: str
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
    root__ = ZOMBIE_PARSER  # type: ParserBase
    # root__ must be overwritten with the root-parser by grammar subclass
    parser_initialization__ = "pending"  # type: str
    # some default values
    COMMENT__ = r''  # type: str  # r'#.*(?:\n|$)'
    WSP__ = mixin_comment(whitespace=r'[\t ]*', comment=COMMENT__)  # type: str
    wspL__ = ''     # type: str
    wspR__ = WSP__  # type: str


    @classmethod
    def _assign_parser_names__(cls):
        """
        Initializes the `parser.name` fields of those
        Parser objects that are directly assigned to a class field with
557
        the field's name, e.g.::
558

559
560
            class Grammar(Grammar):
                ...
561
                symbol = _RE('(?!\\d)\\w+')
562

di68kap's avatar
di68kap committed
563
564
565
        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()`
566
567
568
569
570
571
572
573
574
575
576
577
578
579

        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
580
581
582
583
                    if isinstance(parser, Forward):
                        if  not cast(Forward, parser).parser.name:
                            cast(Forward, parser).parser.name = entry
                    else:   # if not parser.name:
584
                        parser.name = entry
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
            cls.parser_initialization__ = "done"


    def __init__(self, root: Parser = None) -> None:
        # if not hasattr(self.__class__, 'parser_initialization__'):
        #     self.__class__.parser_initialization__ = "pending"
        # if not hasattr(self.__class__, 'wspL__'):
        #     self.wspL__ = ''
        # if not hasattr(self.__class__, 'wspR__'):
        #     self.wspR__ = ''
        self.all_parsers__ = set()             # type: Set[ParserBase]
        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__)

611
612
        if self.WSP__:
            try:
eckhart's avatar
eckhart committed
613
                probe = self.whitespace__  # type: RegExp
614
615
                assert self.whitespace__.regexp.pattern == self.WSP__
            except AttributeError:
eckhart's avatar
eckhart committed
616
                self.whitespace__ = Whitespace(self.WSP__)  # type: RegExp
617
618
            self.whitespace__.grammar = self
            self.all_parsers__.add(self.whitespace__)   # don't you forget about me...
619
        else:
eckhart's avatar
eckhart committed
620
            self.whitespace__ = cast(RegExp, ZOMBIE_PARSER)
621
622
623
624
625
626

        assert not self.wspL__ or self.wspL__ == self.WSP__
        assert not self.wspR__ or self.wspR__ == self.WSP__
        self.wsp_left_parser__ = self.whitespace__ if self.wspL__ else ZOMBIE_PARSER
        self.wsp_right_parser__ = self.whitespace__ if self.wspR__ else ZOMBIE_PARSER

627
628
629
630
631
632
633
634
635
636
637
638
639
640
        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]
641
            raise UnknownParserError('Unknown parser "%s" !' % key)
642
643
644


    def _reset__(self):
645
        self.tree__ = RootNode()              # type: RootNode
646
647
        self.document__ = EMPTY_STRING_VIEW   # type: StringView
        self._reversed__ = EMPTY_STRING_VIEW  # type: StringView
eckhart's avatar
eckhart committed
648
        self.document_length__ = 0            # type: int
649
650
        self.document_lbreaks__ = []          # type: List[int]
        # variables stored and recalled by Capture and Retrieve parsers
651
        self.variables__ = defaultdict(lambda :[])  # type: DefaultDict[str, List[str]]
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
        self.rollback__ = []                  # type: List[Tuple[int, Callable]]
        self.last_rb__loc__ = -1              # type: int
        # support for call stack tracing
        self.call_stack__ = []                # type: List[Parser]
        # 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__:
            self._reversed__ = StringView(self.document__.text[::-1])
        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 \
683
                isinstance(self.__dict__[parser.name], parser.__class__), \
684
                ('Cannot add parser "%s" because a field with the same name '
di68kap's avatar
di68kap committed
685
686
                 'already exists in grammar object: %s!'
                 % (parser.name, str(self.__dict__[parser.name])))
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
            setattr(self, parser.name, parser)
        self.all_parsers__.add(parser)
        parser.grammar = self


    def __call__(self, document: str, start_parser="root__") -> Node:
        """
        Parses a document with with parser-combinators.

        Args:
            document (str): The source text to be parsed.
            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.)
        Returns:
            Node: The root node ot the parse tree.
        """
704
705

        def tail_pos(predecessors: Union[List[Node], Tuple[Node, ...]]) -> int:
eckhart's avatar
eckhart committed
706
707
            """Adds the position after the last node in the list of
            predecessors to the node."""
eckhart's avatar
eckhart committed
708
            return predecessors[-1].pos + len(predecessors[-1]) if predecessors else 0
eckhart's avatar
eckhart committed
709

710
711
712
713
714
715
716
717
718
719
720
        # 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
        self.history_tracking__ = is_logging()
        self.document__ = StringView(document)
eckhart's avatar
eckhart committed
721
        self.document_length__ = len(self.document__)
722
        self.document_lbreaks__ = linebreaks(document) if self.history_tracking__ else []
Eckhart Arnold's avatar
Eckhart Arnold committed
723
        self.last_rb__loc__ = -1  # rollback location
724
725
726
727
728
729
730
731
732
        parser = self[start_parser] if isinstance(start_parser, str) else start_parser
        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
733
                result = Node(None, '').init_pos(0)
eckhart's avatar
eckhart committed
734
                self.tree__.new_error(result,
735
                                      'Parser "%s" did not match empty document.' % str(parser))
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
        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?' \
                                '\n    Most advanced: %s\n    Last match:    %s;' % \
                                (str(HistoryRecord.most_advanced_match(self.history__)),
                                 str(HistoryRecord.last_match(self.history__)))
                else:
                    stitches.append(result)
                    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.")
eckhart's avatar
eckhart committed
754
                stitches.append(Node(None, skip).init_pos(tail_pos(stitches)))
eckhart's avatar
eckhart committed
755
                self.tree__.new_error(stitches[-1], error_msg)
756
                if self.history_tracking__:
eckhart's avatar
eckhart committed
757
758
759
760
761
762
763
                    # # 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)
764
765
766
767
768
769
                    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
770
771
                stitches.append(Node(None, rest))
            result = Node(None, tuple(stitches)).init_pos(0)
772
        if any(self.variables__.values()):
773
774
            error_str = "Capture-retrieve-stack not empty after end of parsing: " \
                + str(self.variables__)
775
776
777
778
779
            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.
eckhart's avatar
eckhart committed
780
                    error_node = Node(ZOMBIE_PARSER, '').init_pos(tail_pos(result.children))
eckhart's avatar
eckhart committed
781
                    self.tree__.new_error(error_node, error_str)
eckhart's avatar
eckhart committed
782
                    result.result = result.children + (error_node,)
783
                else:
eckhart's avatar
eckhart committed
784
                    self.tree__.new_error(result, error_str)
eckhart's avatar
eckhart committed
785
        # result.pos = 0  # calculate all positions
786
        # result.collect_errors(self.document__)
787
788
        self.tree__.swallow(result)
        return self.tree__
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806


    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
807
        while self.rollback__ and self.rollback__[-1][0] >= location:
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
            _, 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)


########################################################################
#
842
# _Token and Regular Expression parser classes (i.e. leaf classes)
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
#
########################################################################


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()
860
        assert RX_TOKEN_NAME.match(token)
861
862
        super().__init__()
        self.name = token
863

864
865
866
867
868
869
    def __deepcopy__(self, memo):
        duplicate = self.__class__(self.name)
        duplicate.name = self.name
        # duplicate.ptype =  self.ptype
        return duplicate

870
871
872
873
    def __call__(self, text: StringView) -> Tuple[Optional[Node], StringView]:
        if text[0:1] == BEGIN_TOKEN:
            end = text.find(END_TOKEN, 1)
            if end < 0:
874
                node = Node(self, '')
eckhart's avatar
eckhart committed
875
                self.grammar.tree__.new_error(node,
876
877
                    'END_TOKEN delimiter missing from preprocessor token. '
                    '(Most likely due to a preprocessor bug!)')  # type: Node
878
                return node, text[1:]
879
            elif end == 0:
880
                node = Node(self, '')
eckhart's avatar
eckhart committed
881
                self.grammar.tree__.new_error(node,
882
883
                    'Preprocessor-token cannot have zero length. '
                    '(Most likely due to a preprocessor bug!)')
884
                return node, text[2:]
885
886
            elif text.find(BEGIN_TOKEN, 1, end) >= 0:
                node = Node(self, text[len(self.name) + 1:end])
eckhart's avatar
eckhart committed
887
                self.grammar.tree__.new_error(node,
888
889
890
891
892
                    '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:
893
                return Node(self, text[len(self.name) + 2:end]), text[end + 1:]
894
895
896
        return None, text


eckhart's avatar
eckhart committed
897
898
class PlainText(Parser):
    """
899
    Parses plain text strings. (Could be done by RegExp as well, but is faster.)
eckhart's avatar
eckhart committed
900

eckhart's avatar
eckhart committed
901
902
903
904
905
    Example::

        >>> while_token = PlainText("while")
        >>> Grammar(while_token)("while").content
        'while'
eckhart's avatar
eckhart committed
906
    """
di68kap's avatar
di68kap committed
907
    assert PLAINTEXT_PTYPE == ":PlainText"
eckhart's avatar
eckhart committed
908

909
910
    def __init__(self, text: str) -> None:
        super().__init__()
eckhart's avatar
eckhart committed
911
        self.text = text
912
        self.len = len(text)
eckhart's avatar
eckhart committed
913
914

    def __deepcopy__(self, memo):
915
916
        duplicate = self.__class__(self.text)
        duplicate.name = self.name
di68kap's avatar
di68kap committed
917
        # duplicate.ptype =  self.ptype
918
        return duplicate
eckhart's avatar
eckhart committed
919
920
921

    def __call__(self, text: StringView) -> Tuple[Optional[Node], StringView]:
        if text.startswith(self.text):
922
            return Node(self, self.text, True), text[self.len:]
eckhart's avatar
eckhart committed
923
924
        return None, text

925
926
927
    def __repr__(self):
        return ("'%s'" if self.text.find("'") <= 0 else '"%s"') % self.text

eckhart's avatar
eckhart committed
928

929
930
931
932
933
934
935
936
937
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.

938
939
940
941
942
    Example::

        >>> word = RegExp(r'\w+')
        >>> Grammar(word)("Haus").content
        'Haus'
943

944
    EBNF-Notation:  ``/ ... /``
945

946
    EBNF-Example:   ``word = /\w+/``
947
948
    """

949
950
    def __init__(self, regexp) -> None:
        super().__init__()
951
952
953
954
955
956
957
958
        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
959
960
        duplicate = self.__class__(regexp)
        duplicate.name = self.name
di68kap's avatar
di68kap committed
961
        # duplicate.ptype =  self.ptype
962
        return duplicate
963
964

    def __call__(self, text: StringView) -> Tuple[Optional[Node], StringView]:
965
966
967
968
        match = text.match(self.regexp)
        if match:
            capture = match.group(0)
            end = text.index(match.end())
969
970
            # regular expression must never match preprocessor-tokens!
            # TODO: Find a better solution here? e.g. static checking/re-mangling at compile time
971
972
973
974
975
            i = capture.find(BEGIN_TOKEN)
            if i >= 0:
                capture = capture[:i]
                end = i
            return Node(self, capture, True), text[end:]
976
977
978
        return None, text

    def __repr__(self):
979
        return escape_control_characters('/%s/' % self.regexp.pattern)
980
981
982
983
984
985
986
987


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"


988
989
990
991
992
993
994
995
996
997
998
# def RE(regexp, wL=None, wR=None) -> 'Series':
#     def rxp(regex):
#         return regex if isinstance(regex, RegExp) else RegExp(regex)
#     if wL is None and wR is None:
#         return rxp(regexp)
#     elif wL is None:
#         return Series(rxp(regexp), rxp(wR))
#     elif wR is None:
#         return Series(rxp(wL), rxp(regexp))
#     else:
#         return Series(rxp(wL), rxp(regexp), rxp(wR))
999
1000