The expiration time for new job artifacts in CI/CD pipelines is now 30 days (GitLab default). Previously generated artifacts in already completed jobs will not be affected by the change. The latest artifacts for all jobs in the latest successful pipelines will be kept. More information: https://gitlab.lrz.de/help/user/admin_area/settings/continuous_integration.html#default-artifacts-expiration

parse.py 76.2 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, ParserBase, WHITESPACE_PTYPE, \
41
    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
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
           'RE',
           'Token',
           'mixin_comment',
           # 'UnaryOperator',
           # 'NaryOperator',
           'Synonym',
           'Option',
           'ZeroOrMore',
           'OneOrMore',
           'Series',
           'Alternative',
           'AllOf',
           'SomeOf',
           'Unordered',
           'Lookahead',
           'NegativeLookahead',
           'Lookbehind',
           'NegativeLookbehind',
           'last_value',
           'counterpart',
           'accumulate',
           'Capture',
           'Retrieve',
           'Pop',
76
           'Forward')
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100


########################################################################
#
# 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
101
102
            grammar = parser.grammar
            location = grammar.document_length__ - len(text)
103

Eckhart Arnold's avatar
Eckhart Arnold committed
104
            if grammar.last_rb__loc__ >= location:
105
106
107
108
109
                grammar.rollback_to__(location)

            # if location has already been visited by the current parser,
            # return saved result
            if location in parser.visited:
110
                # no history recording in case of meomized results
111
112
113
114
115
116
117
118
                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__:
119
                if parser.recursion_counter[location] > LEFT_RECURSION_DEPTH:
120
121
122
123
124
125
126
127
128
                    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
129
                # don't clear recursion_locations__ !!!
130
131
132
133
134
135
136
137
138
139
140
141

            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
142
            else:
143
                assert node._pos < 0
Eckhart Arnold's avatar
Eckhart Arnold committed
144
145
146
                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
147
148
149
150
151
152
                        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)
153

154
            # Mind that meomized parser calls will not appear in the history record!
155
156
            if grammar.history_tracking__:
                # don't track returning parsers except in case an error has occurred
157
                # remaining = len(rest)
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
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
                if grammar.moving_forward__ or (node and node.error_flag):  # 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.add_error("maximum recursion depth of parser reached; "
                           "potentially due to too many errors!")
            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]

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

        # add "aspect oriented" wrapper around parser calls
        # for memoizing, left recursion and tracing
244
245
246
247
248
249
250
        if not isinstance(self, Forward):  # should Forward-Parser no 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
251
252
253
254
255
256
257
258
259
260
261
262
263
264

    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.
        """
        return self.__class__(self.name)

    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."""
265
266
267
        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]
268
269
270
271
272
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

    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


336
337
338
339
340
341
class UnknownParserError(KeyError):
    """UnknownParserError is raised if a Grammer object is called with a
    parser that does not exist or if in the course of parsing a parser
    is reffered to that does not exist."""


342
343
344
345
346
347
348
349
350
351
352
353
354
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.

355
    Example for direct instantiation of a grammar:
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

        >>> number = RE('\d+') + RE('\.') + RE('\d+') | RE('\d+')
        >>> 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.

383
    Example::
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420

        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+')
            factor = INTEGER | Token("(") + expression + Token(")")
            term = factor + ZeroOrMore((Token("*") | Token("/")) + factor)
            expression.set(term + ZeroOrMore((Token("+") | Token("-")) + term))
            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:

421
422
423
424
425
426
427
    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.
428

429
430
        wspR__:  regular expression string for right aligned whitespace,
                 which either equals WSP__ or is empty.
431

432
433
434
435
        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.
436

437
438
439
440
441
        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.
442
443
444
445
446
447
448

    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

449
450
451
452
453
454
        whitespace__: A parser for the implicit optional whitespace (or the
                :class:zombie-parser if the default is empty). The default
                whitespace will be used by parsers :class:`Token` and, if no
                other parsers are passed to its constructor, by parser
                :class:`RE`. It can also be place explicitly in the
                EBNF-Grammar via the "~"-sign.
455

456
457
458
459
460
        wsp_left_parser__: The same as ``whitespace`` for
               left-adjacent-whitespace.

        wsp_right_parser__: The same as ``whitespace`` for
               right-adjacent-whitespace.
461
462
463
464
465
466
467
468

        _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
469
470
        document_length__:  the length of the document.

471
472
473
        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
474
                initialized if :attr:`history_tracking__` is true.
475

476
        _reversed__:  the same text in reverse order - needed by the `Lookbehind`-
477
478
479
                parsers.

        variables__:  A mapping for variable names to a stack of their respective
480
481
                string values - needed by the :class:`Capture`-, :class:`Retrieve`-
                and :class:`Pop`-parsers.
482
483

        rollback__:  A list of tuples (location, rollback-function) that are
484
485
486
487
                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.
488
489
490

        last_rb__loc__:  The last, i.e. most advanced location in the text
                where a variable changing operation occurred. If the parser
491
492
                backtracks to a location at or before last_rb__loc__ (i.e.
                location <= last_rb__loc__) then a rollback of all variable
493
494
                changing operations is necessary that occurred after the
                location to which the parser backtracks. This is done by
495
                calling method :func:`rollback_to__(location)`.
496
497
498
499
500
501
502
503
504
505
506
507
508

        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
509
510
                recording is turned off! (See :func:`add_parser_guard` and its local
                function :func:`guarded_call`)
511
512
513
514

        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
515
516
                if full memoization is enabled. (See :func:`add_parser_guard` and its
                local function :func:`guarded_call`)
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541

        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.
    """
    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
542
        the field's name, e.g.::
543

544
545
546
            class Grammar(Grammar):
                ...
                symbol = RE('(?!\\d)\\w+')
547

548
549
550
551
        After the call of this method symbol.name == "symbol"
        holds. Names assigned via the ``name``-parameter of the
        constructor will not be overwritten. Parser names starting or
        ending with a double underscore like ``root__`` will be
552
        ignored. See :func:`sane_parser_name()`
553
554
555
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
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596

        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):
                    if not parser.name:
                        parser._name = entry
                    if isinstance(parser, Forward) and (not parser.parser._name):
                        parser.parser._name = entry
            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__)

597
598
599
600
601
602
603
604
        if self.WSP__:
            try:
                probe = self.whitespace__
                assert self.whitespace__.regexp.pattern == self.WSP__
            except AttributeError:
                self.whitespace__ = Whitespace(self.WSP__)
            self.whitespace__.grammar = self
            self.all_parsers__.add(self.whitespace__)   # don't you forget about me...
605
        else:
606
607
608
609
610
611
612
            self.whitespace__ = ZOMBIE_PARSER

        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

613
614
615
616
617
618
619
620
621
622
623
624
625
626
        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]
627
            raise UnknownParserError('Unknown parser "%s" !' % key)
628
629
630
631
632


    def _reset__(self):
        self.document__ = EMPTY_STRING_VIEW   # type: StringView
        self._reversed__ = EMPTY_STRING_VIEW  # type: StringView
eckhart's avatar
eckhart committed
633
        self.document_length__ = 0            # type: int
634
635
        self.document_lbreaks__ = []          # type: List[int]
        # variables stored and recalled by Capture and Retrieve parsers
636
        self.variables__ = defaultdict(lambda :[])  # type: DefaultDict[str, List[str]]
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
        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 \
668
                isinstance(self.__dict__[parser.name], parser.__class__), \
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
                ('Cannot add parser "%s" because a field with the same name '
                 'already exists in grammar object!' % parser.name)
            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.
        """
688
689

        def tail_pos(predecessors: Union[List[Node], Tuple[Node, ...]]) -> int:
eckhart's avatar
eckhart committed
690
691
            """Adds the position after the last node in the list of
            predecessors to the node."""
eckhart's avatar
eckhart committed
692
            return predecessors[-1].pos + len(predecessors[-1]) if predecessors else 0
eckhart's avatar
eckhart committed
693

694
695
696
697
698
699
700
701
702
703
704
        # 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
705
        self.document_length__ = len(self.document__)
706
        self.document_lbreaks__ = linebreaks(document) if self.history_tracking__ else []
Eckhart Arnold's avatar
Eckhart Arnold committed
707
        self.last_rb__loc__ = -1  # rollback location
708
709
710
711
712
713
714
715
716
        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
717
                result = Node(None, '').init_pos(0)
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
                result.add_error('Parser "%s" did not match empty document.' % str(parser))
        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
737
                stitches.append(Node(None, skip).init_pos(tail_pos(stitches)))
738
739
                stitches[-1].add_error(error_msg)
                if self.history_tracking__:
eckhart's avatar
eckhart committed
740
741
742
743
744
745
746
                    # # 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)
747
748
749
750
751
752
                    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
753
754
                stitches.append(Node(None, rest))
            result = Node(None, tuple(stitches)).init_pos(0)
755
        if any(self.variables__.values()):
756
757
            error_str = "Capture-retrieve-stack not empty after end of parsing: " \
                + str(self.variables__)
758
759
760
761
762
            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
763
                    error_node = Node(ZOMBIE_PARSER, '').init_pos(tail_pos(result.children))
764
                    error_node.add_error(error_str)
eckhart's avatar
eckhart committed
765
                    result.result = result.children + (error_node,)
766
767
                else:
                    result.add_error(error_str)
eckhart's avatar
eckhart committed
768
        # result.pos = 0  # calculate all positions
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
        # result.collect_errors(self.document__)
        return result


    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
789
        while self.rollback__ and self.rollback__[-1][0] >= location:
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
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)


########################################################################
#
# Token and Regular Expression parser classes (i.e. leaf classes)
#
########################################################################


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()
842
        assert RX_TOKEN_NAME.match(token)
eckhart's avatar
eckhart committed
843
        super().__init__(token)
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865

    def __call__(self, text: StringView) -> Tuple[Optional[Node], StringView]:
        if text[0:1] == BEGIN_TOKEN:
            end = text.find(END_TOKEN, 1)
            if end < 0:
                node = Node(self, '').add_error(
                    'END_TOKEN delimiter missing from preprocessor token. '
                    '(Most likely due to a preprocessor bug!)')  # type: Node
                return node, text[1:]
            elif end == 0:
                node = Node(self, '').add_error(
                    'Preprocessor-token cannot have zero length. '
                    '(Most likely due to a preprocessor bug!)')
                return node, text[2:]
            elif text.find(BEGIN_TOKEN, 1, end) >= 0:
                node = Node(self, text[len(self.name) + 1:end])
                node.add_error(
                    '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:
866
                return Node(self, text[len(self.name) + 2:end]), text[end + 1:]
867
868
869
        return None, text


eckhart's avatar
eckhart committed
870
871
class PlainText(Parser):
    """
872
    Parses plain text strings. (Could be done by RegExp as well, but is faster.)
eckhart's avatar
eckhart committed
873

eckhart's avatar
eckhart committed
874
875
876
877
878
    Example::

        >>> while_token = PlainText("while")
        >>> Grammar(while_token)("while").content
        'while'
eckhart's avatar
eckhart committed
879
880
881
882
883
    """

    def __init__(self, text: str, name: str = '') -> None:
        super().__init__(name)
        self.text = text
884
        self.len = len(text)
eckhart's avatar
eckhart committed
885
886
887
888
889
890

    def __deepcopy__(self, memo):
        return self.__class__(self.text, self.name)

    def __call__(self, text: StringView) -> Tuple[Optional[Node], StringView]:
        if text.startswith(self.text):
891
            return Node(self, self.text, True), text[self.len:]
eckhart's avatar
eckhart committed
892
893
        return None, text

894
895
896
    def __repr__(self):
        return ("'%s'" if self.text.find("'") <= 0 else '"%s"') % self.text

eckhart's avatar
eckhart committed
897

898
899
900
901
902
903
904
905
906
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.

907
908
909
910
911
    Example::

        >>> word = RegExp(r'\w+')
        >>> Grammar(word)("Haus").content
        'Haus'
912

913
    EBNF-Notation:  ``/ ... /``
914

915
    EBNF-Example:   ``word = /\w+/``
916
917
918
    """

    def __init__(self, regexp, name: str = '') -> None:
eckhart's avatar
eckhart committed
919
        super().__init__(name)
920
921
922
923
924
925
926
927
        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
eckhart's avatar
eckhart committed
928
        return self.__class__(regexp, self.name)
929
930

    def __call__(self, text: StringView) -> Tuple[Optional[Node], StringView]:
931
932
933
934
        match = text.match(self.regexp)
        if match:
            capture = match.group(0)
            end = text.index(match.end())
935
936
            # regular expression must never match preprocessor-tokens!
            # TODO: Find a better solution here? e.g. static checking/re-mangling at compile time
937
938
939
940
941
            i = capture.find(BEGIN_TOKEN)
            if i >= 0:
                capture = capture[:i]
                end = i
            return Node(self, capture, True), text[end:]
942
943
944
        return None, text

    def __repr__(self):
945
        return escape_control_characters('/%s/' % self.regexp.pattern)
946
947
948
949
950
951
952
953


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"


954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
#######################################################################
#######################################################################
#
# WARNING: The following code is hard to maintain, because it
# introduces a special case, i.e. a parser with child parsers that is
# not a descandent of the NaryOperator and, because it itneracts
# With the constructor of the Grammar class (see the instantiations of
# the Whitespace-class, there).
#
# That is all the more regrettable, as class RE basically just
# introduces syntactical sugar for
#
#     Series(whitespace__, RegExp('something'), whitespace__)
#
# What to do? Throw the syntactical sugar out? :-( Or find a more
# robust solution for that kind of syntactical sugar? Or just leave
# it be?
#
######################################################################
######################################################################


976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
class RE(Parser):
    r"""
    Regular Expressions with optional leading or trailing whitespace.

    The RE-parser parses pieces of text that match a given regular
    expression. Other than the ``RegExp``-Parser it can also skip
    "implicit whitespace" before or after the matched text.

    The whitespace is in turn defined by a regular expression. It should
    be made sure that this expression also matches the empty string,
    e.g. use r'\s*' or r'[\t ]+', but not r'\s+'. If the respective
    parameters in the constructor are set to ``None`` the default
    whitespace expression from the Grammar object will be used.

    Example (allowing whitespace on the right hand side, but not on
991
992
993
994
995
996
997
998
999
1000
1001
    the left hand side of a regular expression)::

        >>> word = RE(r'\w+', wR=r'\s*')
        >>> parser = Grammar(word)
        >>> result = parser('Haus ')
        >>> result.content
        'Haus '
        >>> result.structure
        '(:RE (:RegExp "Haus") (:Whitespace " "))'
        >>> str(parser(' Haus'))
        ' <<< Error on " Haus" | Parser did not match! Invalid source file?\n    Most advanced: None\n    Last match:    None; >>> '
1002

1003
    EBNF-Notation:  ``/ ... /~`  or  `~/ ... /`  or  `~/ ... /~``
1004

1005
    EBNF-Example:   ``word = /\w+/~``
1006
1007
    """

eckhart's avatar
eckhart committed
1008
    def __init__(self, regexp, wL=None, wR=None, name: str='') -> None:
1009
1010
1011
1012
1013
1014
1015
1016
        r"""Constructor for class RE.

        Args:
            regexp (str or regex object):  The regular expression to be
                used for parsing.
            wL (str or regexp):  Left whitespace regular expression,
                i.e. either ``None``, the empty string or a regular
                expression (e.g. "\s*") that defines whitespace. An
1017
1018
                empty string means no whitespace will be skipped; ``None``
                means that the default whitespace will be used.
1019
1020
1021
1022
            wR (str or regexp):  Right whitespace regular expression.
                See above.
            name:  The optional name of the parser.
        """
eckhart's avatar
eckhart committed
1023
        super().__init__(name)
1024
1025
1026
1027
        self.rx_wsl = wL
        self.rx_wsr = wR
        self.wsp_left = Whitespace(wL) if wL else ZOMBIE_PARSER
        self.wsp_right = Whitespace(wR) if wR else ZOMBIE_PARSER
eckhart's avatar
eckhart committed
1028
        self.main = self.create_main_parser(regexp)
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043

    def __deepcopy__(self, memo={}):
        try:
            regexp = copy.deepcopy(self.main.regexp, memo)
        except TypeError:
            regexp = self.main.regexp.pattern
        return self.__class__(regexp, self.rx_wsl, self.rx_wsr, self.name)

    def __call__(self, text: StringView) -> Tuple[Optional[Node], StringView]:
        # assert self.main.regexp.pattern != "@"
        txt = text    # type: StringView
        wsl, txt = self.wsp_left(txt)
        main, txt = self.main(txt)
        if main:
            wsr, txt = self.wsp_right(txt)
eckhart's avatar
eckhart committed
1044
            result = tuple(nd for nd in (wsl, main, wsr) if nd)
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
            return Node(self, result), txt
        return None, text

    def __repr__(self):
        wsl = '~' if self.wsp_left != ZOMBIE_PARSER else ''
        wsr = '~' if self.wsp_right != ZOMBIE_PARSER else ''
        return wsl + '/%s/' % self.main.regexp.pattern + wsr

    def _grammar_assigned_notifier(self):
        if self.grammar:
            # use default whitespace parsers if not otherwise specified
            if self.rx_wsl is None:
                self.wsp_left = self.grammar.wsp_left_parser__
            if self.rx_wsr is None:
                self.wsp_right = self.grammar.wsp_right_parser__

    def apply(self, func: Parser.ApplyFunc) -> bool:
        if super().apply(func):
            if self.rx_wsl:
                self.wsp_left.apply(func)
            if self.rx_wsr:
                self.wsp_right.apply(func)
            self.main.apply(func)
            return True
        return False

eckhart's avatar
eckhart committed
1071
1072
1073
1074
    def create_main_parser(self, arg) -> Parser:
        """Creates the main parser of this compound parser. Can be overridden."""
        return RegExp(arg)

1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089

class Token(RE):
    """
    Class Token parses simple strings. Any regular regular expression
    commands will be interpreted as simple sequence of characters.

    Other than that class Token is essentially a renamed version of
    class RE. Because tokens often have a particular semantic different
    from other REs, parsing them with a separate parser class allows to
    distinguish them by their parser type.
    """
    assert TOKEN_PTYPE == ":Token"

    def __init__(self, token: str, wL=None, wR=None, name: str = '') -> None:
        self.token = token
eckhart's avatar
eckhart committed
1090
        super().__init__(token, wL, wR, name)
1091
1092
1093
1094
1095
1096
1097

    def __deepcopy__(self, memo={}):
        return self.__class__(self.token, self.rx_wsl, self.rx_wsr, self.name)

    def __repr__(self):
        return '"%s"' % self.token if self.token.find('"') < 0 else "'%s'" % self.token

eckhart's avatar
eckhart committed
1098
1099
1100
    def create_main_parser(self, arg) -> Parser:
        return PlainText(arg)

1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149

########################################################################
#
# 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.
    """

    def __init__(self, parser: Parser, name: str = '') -> None:
        super(UnaryOperator, self).__init__(name)
        assert isinstance(parser, Parser), str(parser)
        self.parser = parser  # type: Parser

    def __deepcopy__(self, memo):
        parser = copy.deepcopy(self.parser, memo)
        return self.__class__(parser, self.name)

    def apply(self, func: Parser.ApplyFunc) -> bool:
        if super().apply(func):
            self.parser.apply(func)
            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.
    """

    def __init__(self, *parsers: Parser, name: str = '') -> None:
eckhart's avatar
eckhart committed
1150
        super().__init__(name)
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
        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)
        return self.__class__(*parsers, name=self.name)

    def apply(self, func: Parser.ApplyFunc) -> bool:
        if super().apply(func):
            for parser in self.parsers:
                parser.apply(func)
            return True
        return False


class Option(UnaryOperator):
    r"""
1168
    Parser ``Option`` always matches, even if its child-parser
1169
1170
    did not match.

1171
    If the child-parser did not match ``Option`` returns a node
1172
1173
    with no content and does not move forward in the text.

1174
    If the child-parser did match, ``Option`` returns the a node
1175
1176
1177
1178
    with the node returnd by the child-parser as its single
    child and the text at the position where the child-parser
    left it.

1179
1180
1181
1182
1183
1184
1185
1186
1187
    Examples::

        >>> number = Option(Token('-')) + RegExp(r'\d+') + Option(RegExp(r'\.\d+'))
        >>> Grammar(number)('3.14159').content
        '3.14159'
        >>> Grammar(number)('3.14159').structure
        '(:Series (:Option) (:RegExp "3") (:Option (:RegExp ".14159")))'
        >>> Grammar(number)('-1').content
        '-1'
1188

1189
    EBNF-Notation: ``[ ... ]``
1190

1191
    EBNF-Example:  ``number = ["-"]  /\d+/  [ /\.\d+/ ]``
1192
1193
1194
    """

    def __init__(self, parser: Parser, name: str = '') -> None:
eckhart's avatar
eckhart committed
1195
        super().__init__(parser, name)
1196
1197
        # assert isinstance(parser, Parser)
        assert not isinstance(parser, Option), \
1198
            "Redundant nesting of options: %s(%s)" % (str(name), str(parser.name))
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
        # assert not isinstance(parser, Required), \
        #     "Nesting options with required elements is contradictory: " \
        #     "%s(%s)" % (str(name), str(parser.name))

    def __call__(self, text: StringView) -> Tuple[Optional[Node], StringView]:
        node, text = self.parser(text)
        if node:
            return Node(self, node), text
        return Node(self, ()), text

    def __repr__(self):
        return '[' + (self.parser.repr[1:-1] if isinstance(self.parser, Alternative)
                      and not self.parser.name else self.parser.repr) + ']'


class ZeroOrMore(Option):
    r"""
    `ZeroOrMore` applies a parser repeatedly as long as this parser
    matches. Like `Option` the `ZeroOrMore` parser always matches. In
    case of zero repetitions, the empty match `((), text)` is returned.

1220
1221
1222
1223
1224
1225
1226
    Examples::

        >>> sentence = ZeroOrMore(RE(r'\w+,?')) + Token('.')
        >>> Grammar(sentence)('Wo viel der Weisheit, da auch viel des Grämens.').content
        'Wo viel der Weisheit, da auch viel des Grämens.'
        >>> Grammar(sentence)('.').content  # an empty sentence also matches
        '.'
1227

1228
    EBNF-Notation: ``{ ... }``
1229

1230
    EBNF-Example:  ``sentence = { /\w+,?/ } "."``
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
    """

    def __call__(self, text: StringView) -> Tuple[Optional[Node], StringView]:
        results = ()  # type: Tuple[Node, ...]
        n = len(text) + 1
        while text and len(text) < n:
            n = len(text)
            node, text = self.parser(text)
            if not node:
                break
            if len(text) == n:
                node.add_error(dsl_error_msg(self, 'Infinite Loop detected.'))
            results += (node,)
        return Node(self, results), text

    def __repr__(self):
        return '{' + (self.parser.repr[1:-1] if isinstance(self.parser, Alternative)
                      and not self.parser.name else self.parser.repr) + '}'


class OneOrMore(UnaryOperator):
    r"""
    `OneOrMore` applies a parser repeatedly as long as this parser
    matches. Other than `ZeroOrMore` which always matches, at least
    one match is required by `OneOrMore`.

1257
1258
1259
1260
1261
1262
1263
    Examples::

        >>> sentence = OneOrMore(RE(r'\w+,?')) + Token('.')
        >>> Grammar(sentence)('Wo viel der Weisheit, da auch viel des Grämens.').content
        'Wo viel der Weisheit, da auch viel des Grämens.'
        >>> str(Grammar(sentence)('.'))  # an empty sentence also matches
        ' <<< Error on "." | Parser did not match! Invalid source file?\n    Most advanced: None\n    Last match:    None; >>> '
1264

1265
    EBNF-Notation: ``{ ... }+``
1266

1267
    EBNF-Example:  ``sentence = { /\w+,?/ }+``
1268
1269
1270
    """

    def __init__(self, parser: Parser, name: str = '') -> None:
eckhart's avatar
eckhart committed
1271
        super().__init__(parser, name)
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
        assert not isinstance(parser, Option), \
            "Use ZeroOrMore instead of nesting OneOrMore and Option: " \
            "%s(%s)" % (str(name), str(parser.name))

    def __call__(self, text: StringView) -> Tuple[Optional[Node], StringView]:
        results = ()  # type: Tuple[Node, ...]
        text_ = text  # type: StringView
        n = len(text) + 1
        while text_ and len(text_) < n:
            n = len(text_)
            node, text_ = self.parser(text_)
            if not node:
                break
            if len(text_) == n:
                node.add_error(dsl_error_msg(self, 'Infinite Loop detected.'))
            results += (node,)
        if results == ():
            return None, text
        return Node(self, results), text_

    def __repr__(self):
        return '{' + (self.parser.repr[1:-1] if isinstance(self.parser, Alternative)
                      and not self.parser.name else self.parser.repr) + '}+'


class Series(NaryOperator):
    r"""
    Matches if each of a series of parsers matches exactly in the order of
    the series.

1302
1303
1304
1305
1306
1307
1308
    Example::

        >>> variable_name = RegExp('(?!\d)\w') + RE('\w*')
        >>> Grammar(variable_name)('variable_1').content
        'variable_1'
        >>> str(Grammar(variable_name)('1_variable'))
        ' <<< Error on "1_variable" | Parser did not match! Invalid source file?\n    Most advanced: None\n    Last match:    None; >>> '
1309

1310
    EBNF-Notation: ``... ...``    (sequence of parsers separated by a blank or new line)
1311

1312
    EBNF-Example:  ``series = letter letter_or_digit``
1313
1314
1315
1316
1317
    """
    RX_ARGUMENT = re.compile(r'\s(\S)')
    NOPE = 1000

    def __init__(self, *parsers: Parser, mandatory: int = NOPE, name: str = '') -> None:
eckhart's avatar
eckhart committed
1318
        super().__init__(*parsers, name=name)
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
        length = len(self.parsers)
        assert 1 <= length < Series.NOPE, \
            'Length %i of series exceeds maximum length of %i' % (length, Series.NOPE)
        if mandatory < 0:
            mandatory += length
        assert 0 <= mandatory < length or mandatory == Series.NOPE
        self.mandatory = mandatory

    def __deepcopy__(self, memo):
        parsers = copy.deepcopy(self.parsers, memo)
        return self.__class__(*parsers, mandatory=self.mandatory, name=self.name)

    def __call__(self, text: StringView) -> Tuple[Optional[Node], StringView]:
        results = ()  # type: Tuple[Node, ...]
        text_ = text  # type: StringView
        pos = 0
        for parser in self.parsers:
            node, text_ = parser(text_)
            if not node:
                if pos < self.mandatory:
                    return None, text
                else:
                    # Provide useful error messages
                    match = text.search(Series.RX_ARGUMENT)
                    i = max(1, text.index(match.regs[1][0])) if match else 1
1344
1345
                    node = Node(self, text_[:i]).init_pos(self.grammar.document_length__
                                                          - len(text_))
eckhart's avatar
eckhart committed
1346
                    node.add_error('%s expected; "%s" found!'
di68kap's avatar
di68kap committed
1347
                                   % (parser.repr, text_[:10].replace('\n', '\\n ')),
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
                                   code=Error.MANDATORY_CONTINUATION)
                    text_ = text_[i:]
            results += (node,)
            # if node.error_flag:  # break on first error
            #    break
            pos += 1
        assert len(results) <= len(self.parsers)
        return Node(self, results), text_

    def __repr__(self):
        return " ".join([parser.repr for parser in self.parsers[:self.mandatory]]
                        + (['§'] if self.mandatory != Series.NOPE else [])
                        + [parser.repr for parser in self.parsers[self.mandatory:]])

    # The following operator definitions add syntactical sugar, so one can write:
    # `RE('\d+') + Optional(RE('\.\d+)` instead of `Series(RE('\d+'), Optional(RE('\.\d+))`

    @staticmethod
    def combined_mandatory(left: Parser, right: Parser):
        """
        Returns the position of the first mandatory element (if any) when
        parsers `left` and `right` are joined to a sequence.
        """
        left_mandatory, left_length = (left.mandatory, len(left.parsers)) \
            if isinstance(left, Series) else (Series.NOPE, 1)
        if left_mandatory != Series.NOPE:
            return left_mandatory
        right_mandatory = right.mandatory if isinstance(right, Series) else Series.NOPE
        if right_mandatory != Series.NOPE:
            return right_mandatory + left_length
        return Series.NOPE

    def __add__(self, other: Parser) -> 'Series':
        other_parsers = cast('Series', other).parsers if isinstance(other, Series) \
1382
            else cast(Tuple[Parser, ...], (other,))  # type: Tuple[Parser, ...]
1383
1384
1385
1386
1387
        return Series(*(self.parsers + other_parsers),
                      mandatory=self.combined_mandatory(self, other))

    def __radd__(self, other: Parser) -> 'Series':
        other_parsers = cast('Series', other).parsers if isinstance(other, Series) \
1388
            else cast(Tuple[Parser, ...], (other,))  # type: Tuple[Parser, ...]
1389
1390
1391
1392
1393
        return Series(*(other_parsers + self.parsers),
                      mandatory=self.combined_mandatory(other, self))

    def __iadd__(self, other: Parser) -> 'Series':
        other_parsers = cast('Series', other).parsers if isinstance(other, Series) \
1394
            else cast(Tuple[Parser, ...], (other,))  # type: Tuple[Parser, ...]
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
        self.parsers += other_parsers
        self.mandatory = self.combined_mandatory(self, other)
        return self


class Alternative(NaryOperator):
    r"""
    Matches if one of several alternatives matches. Returns
    the first match.

    This parser represents the EBNF-operator "|" with the qualification
    that both the symmetry and the ambiguity of the EBNF-or-operator
1407
    are broken by selecting the first match.::
1408

1409
1410
1411
1412
        # the order of the sub-expression matters!
        >>> number = RE('\d+') | RE('\d+') + RE('\.') + RE('\d+')
        >>> str(Grammar(number)("3.1416"))
        '3 <<< Error on ".141" | Parser stopped before end! trying to recover... >>> '
1413

1414
1415
1416
1417
        # the most selective expression should be put first:
        >>> number = RE('\d+') + RE('\.') + RE('\d+') | RE('\d+')
        >>> Grammar(number)("3.1416").content
        '3.1416'
1418

1419
    EBNF-Notation: ``... | ...``
1420

1421
    EBNF-Example:  ``sentence = /\d+\.\d+/ | /\d+/``
1422
1423
1424
    """

    def __init__(self, *parsers: Parser, name: str = '') -> None:
eckhart's avatar
eckhart committed
1425
        super().__init__(*parsers, name=name)
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
        assert len(self.parsers) >= 1
        # only the last alternative may be optional. Could this be checked at compile time?
        assert all(not isinstance(p, Option) for p in self.parsers[:-1]), \
            "Parser-specification Error: only the last alternative may be optional!"

    def __call__(self, text: StringView) -> Tuple[Optional[Node], StringView]:
        for parser in self.parsers:
            node, text_ = parser(text)
            if node:
                return Node(self, node), text_
        return None, text

    def __repr__(self):
        return '(' + ' | '.join(parser.repr for parser in self.parsers) + ')'

    def reset(self):
        super(Alternative, self).reset()
        return self

    # The following operator definitions add syntactical sugar, so one can write:
    # `RE('\d+') + RE('\.') + RE('\d+') | RE('\d+')` instead of:
    # `Alternative(Series(RE('\d+'), RE('\.'), RE('\d+')), RE('\d+'))`

    def __or__(self, other: Parser) -> 'Alternative':
        other_parsers = cast('Alternative', other).parsers if isinstance(other, Alternative) \
1451
            else cast(Tuple[Parser, ...], (other,))  # type: Tuple[Parser, ...]
1452
1453
1454
1455
        return Alternative(*(self.parsers + other_parsers))

    def __ror__(self, other: Parser) -> 'Alternative':
        other_parsers = cast('Alternative', other).parsers if isinstance(other, Alternative) \
1456
            else cast(Tuple[Parser, ...], (other,))  # type: Tuple[Parser, ...]
1457
1458
1459
1460
        return Alternative(*(other_parsers + self.parsers))

    def __ior__(self, other: Parser) -> 'Alternative':
        other_parsers = cast('Alternative', other).parsers if isinstance(other, Alternative) \
1461
            else cast(Tuple[Parser, ...], (other,))  # type: Tuple[Parser, ...]
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
        self.parsers += other_parsers
        return self


class AllOf(NaryOperator):
    """
    Matches if all elements of a list of parsers match. Each parser must
    match exactly once. Other than in a sequence, the order in which
    the parsers match is arbitrary, however.

1472
1473
1474
1475
1476
1477
1478
    Example::

        >>> prefixes = AllOf(Token("A"), Token("B"))
        >>> Grammar(prefixes)('A B').content
        'A B'
        >>> Grammar(prefixes)('B A').content
        'B A'
1479

1480
    EBNF-Notation: ``<... ...>``    (sequence of parsers enclosed by angular brackets)
1481

1482
    EBNF-Example:  ``set = <letter letter_or_digit>``
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
    """

    def __init__(self, *parsers: Parser, name: str = '') -> None:
        if len(parsers) == 1 and isinstance(parsers[0], Series):
            assert isinstance(parsers[0], Series), \
                "Parser-specification Error: No single arguments other than a Series " \
                "allowed as arguments for AllOf-Parser !"
            series = cast(Series, parsers[0])
            assert series.mandatory == Series.NOPE, \
                "AllOf cannot contain mandatory (§) elements!"
            parsers = series.parsers
        super().__init__(*parsers, name=name)

    def __call__(self, text: StringView) -> Tuple[Optional[Node], StringView]:
        results = ()  # type: Tuple[Node, ...]
        text_ = text  # type: StringView
        parsers = list(self.parsers)  # type: List[Parser]
        while parsers:
            for i, parser in enumerate(parsers):
                node, text__ = parser(text_)
                if node:
                    results += (node,)
                    text_ = text__
                    del parsers[i]
                    break
            else:
                return None, text
        assert len(results) <= len(self.parsers)
        return Node(self, results), text_

    def __repr__(self):
        return '<' + ' '.join(parser.repr for parser in self.parsers) + '>'


class SomeOf(NaryOperator):
    """
    Matches if at least one element of a list of parsers match. No parser
    must match more than once . Other than in a sequence, the order in which
    the parsers match is arbitrary, however.

1523
1524
1525
1526
1527
1528
1529
1530
1531
    Example::

        >>> prefixes = SomeOf(Token("A"), Token("B"))
        >>> Grammar(prefixes)('A B').content
        'A B'
        >>> Grammar(prefixes)('B A').content
        'B A'
        >>> Grammar(prefixes)('B').content
        'B'
1532

1533
    EBNF-Notation: ``<... ...>``    (sequence of parsers enclosed by angular brackets)
1534

1535
    EBNF-Example:  ``set = <letter letter_or_digit>``
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
    """

    def __init__(self, *parsers: Parser, name: str = '') -> None:
        if len(parsers) == 1:
            assert isinstance(parsers[0], Alternative), \
                "Parser-specification Error: No single arguments other than a Alternative " \
                "allowed as arguments for SomeOf-Parser !"
            alternative = cast(Alternative, parsers[0])
            parsers = alternative.parsers
        super().__init__(*parsers, name=name)

    def __call__(self, text: StringView) -> Tuple[Optional[Node], StringView]:
        results = ()  # type: Tuple[Node, ...]
        text_ = text  # type: StringView
        parsers = list(self.parsers)  # type: List[Parser]
        while parsers:
            for i, parser in enumerate(parsers):
                node, text__ = parser(text_)
                if node:
                    results += (node,)
                    text_ = text__
                    del parsers[i]
                    break
            else:
                parsers = []
        assert len(results) <= len(self.parsers)
        if results:
            return Node(self, results), text_
        else:
            return None, text

    def __repr__(self):
        return '<' + ' | '.join(parser.repr for parser in self.parsers) + '>'


def Unordered(parser: NaryOperator, name: str = '') -> NaryOperator:
    """
    Returns an AllOf- or SomeOf-parser depending on whether `parser`
    is a Series (AllOf) or an Alternative (SomeOf).
    """
    if isinstance(parser, Series):
        return AllOf(parser, name=name)
    elif isinstance(parser, Alternative):
        return SomeOf(parser, name=name)
    else:
        raise AssertionError("Unordered can take only Series or Alternative as parser.")


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

class FlowOperator(UnaryOperator):
    """
    Base class for all flow operator parsers like Lookahead and Lookbehind.
    """
    def sign(self, bool_value) -> bool:
        """Returns the value. Can be overriden to return the inverted bool."""
        return bool_value


# class Required(FlowOperator):
#     """OBSOLETE. Use mandatory-parameter of Series-parser instead!
#     """
#     RX_ARGUMENT = re.compile(r'\s(\S)')
#
#     def __call__(self, text: StringView) -> Tuple[Optional[Node], StringView]:
#         node, text_ = self.parser(text)
#         if not node:
#             m = text.search(Required.RX_ARGUMENT)  # re.search(r'\s(\S)', text)
#             i = max(1, text.index(m.regs[1][0])) if m else 1
#             node = Node(self, text[:i])
#             text_ = text[i:]
#             node.add_error('%s expected; "%s" found!' % (str(self.parser), text[:10]),
#                            code=Error.MANDATORY_CONTINUATION)
#         return node, text_
#
#     def __repr__(self):
#         return '§' + self.parser.repr


class Lookahead(FlowOperator):
    """
    Matches, if the contained parser would match for the following text,
    but does not consume any text.
    """
    def __call__(self, text: StringView) -> Tuple[Optional[Node], StringView]:
        node, _ = self.parser(text)
        if self.sign(node is not None):
            return Node(self, ''), text
        else:
            return None, text

    def __repr__(self):
        return '&' + self.parser.repr


class NegativeLookahead(Lookahead):
    """
    Matches, if the contained parser would *not* match for the following
    text.
    """
    def __repr__(self):
        return '!' + self.parser.repr

    def sign(self, bool_value) -> bool:
        return not bool_value


class Lookbehind(FlowOperator):
    """
    Matches, if the contained parser would match backwards. Requires
1650
1651
1652
    the contained parser to be a RegExp, RE, PlainText or Token parser.

    EXPERIMENTAL
1653
1654
1655
1656
1657
    """
    def __init__(self, parser: Parser, name: str = '') -> None:
        p = parser
        while isinstance(p, Synonym):
            p = p.parser
eckhart's avatar
eckhart committed
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
        assert isinstance(p, RegExp) or isinstance(p, PlainText) or isinstance(p, RE), str(type(p))
        self.regexp = None
        self.text = None
        if isinstance(p, RE):
            if isinstance(cast(RE, p).main, RegExp):
                self.regexp = cast(RegExp, cast(RE, p).main).regexp
            else:  # p.main is of type PlainText
                self.text = cast(PlainText, cast(RE, p).main