parse.py 86.1 KB
Newer Older
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
# parse.py - parser combinators for for DHParser
#
# 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.
#
# Module ``parsers.py`` contains a number of classes that together
# make up parser combinators for left-recursive grammers. For each
# element of the extended Backus-Naur-Form as well as for a regular
# expression token a class is defined. The set of classes can be used to
# define a parser for (ambiguous) left-recursive grammers.
#
#
# References and Acknowledgements:
#
# Dominikus Herzberg: Objekt-orientierte Parser-Kombinatoren in Python,
# Blog-Post, September, 18th 2008 on denkspuren. gedanken, ideen,
# anregungen und links rund um informatik-themen, URL:
# http://denkspuren.blogspot.de/2008/09/objekt-orientierte-parser-kombinatoren.html
#
# Dominikus Herzberg: Eine einfache Grammatik für LaTeX, Blog-Post,
# September, 18th 2008 on denkspuren. gedanken, ideen, anregungen und
# links rund um informatik-themen, URL:
# http://denkspuren.blogspot.de/2008/09/eine-einfache-grammatik-fr-latex.html
#
# Dominikus Herzberg: Uniform Syntax, Blog-Post, February, 27th 2007 on
# denkspuren. gedanken, ideen, anregungen und links rund um
# informatik-themen, URL:
# http://denkspuren.blogspot.de/2007/02/uniform-syntax.html
#
# Richard A. Frost, Rahmatullah Hafiz and Paul Callaghan: Parser
# Combinators for Ambiguous Left-Recursive Grammars, in: P. Hudak and
# D.S. Warren (Eds.): PADL 2008, LNCS 4902, pp. 167–181, Springer-Verlag
# Berlin Heidelberg 2008.
#
# Elizabeth Scott and Adrian Johnstone, GLL Parsing,
# in: Electronic Notes in Theoretical Computer Science 253 (2010) 177–189,
# http://dotat.at/tmp/gll.pdf
#
# Christopher Seaton: A Programming Language Where the Syntax and Semantics
# are Mutuable at Runtime, University of Bristol 2007,
# http://chrisseaton.com/katahdin/katahdin.pdfs
#
# Juancarlo Añez: grako, a PEG parser generator in Python,
# https://bitbucket.org/apalala/grako
#
# Vegard Øye: General Parser Combinators in Racket, 2012,
# https://epsil.github.io/gll/
Eckhart Arnold's avatar
Eckhart Arnold committed
60

61
62

"""
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
Module ``parse`` contains the python classes and functions for
DHParser's packrat parser. It's central class is the
``Grammar``-class, which is the base class for any contrete
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.

Module ``parse`` furthermode contains the base class for a
compiler as well as a generic ccompiler function. Compiler
objects are also callabe receive the Abstract syntax tree (AST)
as argument and yield whatever output the compiler produces. In
most Digital Humanities applications this will be
XML-code. However, it can also be anything else, like binary
code or, as in the case of DHParser's ebnf-compiler, Python
source code.

See module ``ebnf`` for a sample of the implementation of a
compiler object.
"""

87
88
89

import copy
import os
90
from typing import Any, Callable, cast, Dict, List, Set, Tuple, Union, Optional
91

92
from DHParser.error import Error, is_error, linebreaks, adjust_error_locations
93
94
95
96
from DHParser.log import is_logging, logfile_basename, HistoryRecord, log_ST, \
    log_parsing_history
from DHParser.preprocess import BEGIN_TOKEN, END_TOKEN, RX_TOKEN_NAME, \
    PreprocessorFunc, with_source_mapping, strip_tokens
97
98
99
from DHParser.stringview import StringView, EMPTY_STRING_VIEW
from DHParser.syntaxtree import Node, TransformationFunc, ParserBase, WHITESPACE_PTYPE, \
    TOKEN_PTYPE, ZOMBIE_PARSER
100
101
from DHParser.toolkit import sane_parser_name, \
    escape_control_characters, load_if_file, re
102

103
__all__ = ('Parser',
104
           'UnknownParserError',
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
           'Grammar',
           'PreprocessorToken',
           'RegExp',
           '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',
           'Forward',
           'Compiler',
           'compile_source')


########################################################################
#
# 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
159
160
            grammar = parser.grammar
            location = grammar.document_length__ - len(text)
161

Eckhart Arnold's avatar
Eckhart Arnold committed
162
            if grammar.last_rb__loc__ >= location:
163
164
165
166
167
                grammar.rollback_to__(location)

            # if location has already been visited by the current parser,
            # return saved result
            if location in parser.visited:
168
                # no history recording in case of meomized results
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
                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__:
                if parser.recursion_counter.setdefault(location, 0) > LEFT_RECURSION_DEPTH:
                    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

            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
199
            else:
200
                assert node._pos < 0
Eckhart Arnold's avatar
Eckhart Arnold committed
201
202
203
                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
204
205
206
207
208
209
                        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)
210

211
            # Mind that meomized parser calls will not appear in the history record!
212
213
            if grammar.history_tracking__:
                # don't track returning parsers except in case an error has occurred
214
                # remaining = len(rest)
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
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
                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)
        super(Parser, self).__init__(name)
        self._grammar = None  # type: 'Grammar'
        self.reset()

        # add "aspect oriented" wrapper around parser calls
        # for memoizing, left recursion and tracing
301
302
303
304
305
306
307
        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
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
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392

    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."""
        self.visited = dict()            # type: Dict[int, Tuple[Optional[Node], StringView]]
        self.recursion_counter = dict()  # type: Dict[int, int]
        self.cycle_detection = set()     # type: Set[Callable]

    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


393
394
395
396
397
398
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."""


399
400
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
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
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.

    Example for direct instantian of a grammar:

        >>> 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.

    Example:

        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:

478
479
480
481
482
483
484
    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.
485

486
487
        wspR__:  regular expression string for right aligned whitespace,
                 which either equals WSP__ or is empty.
488

489
490
491
492
        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.
493

494
495
496
497
498
        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.
499
500
501
502
503
504
505
506

    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

        wsp_left_parser__:  A parser for the default left-adjacent-whitespace
507
                or the :class:zombie-parser if the
508
                default is empty. The default whitespace will be used by parsers
509
510
                :class:`Token` and, if no other parsers are passed to its constructor,
                by parser :class:`RE`.
511
512
513
514

        wsp_right_parser__: The same for the default right-adjacent-whitespace.
                Both wsp_left_parser__ and wsp_right_parser__ merely serve the
                purpose to avoid having to specify the default whitespace
515
                explicitly every time an :class:`RE`-parser-object is created.
516
517
518
519
520
521
522
523

        _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
524
525
        document_length__:  the length of the document.

526
527
528
        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
529
                initialized if :attr:`history_tracking__` is true.
530

531
        _reversed__:  the same text in reverse order - needed by the `Lookbehind`-
532
533
534
                parsers.

        variables__:  A mapping for variable names to a stack of their respective
535
536
                string values - needed by the :class:`Capture`-, :class:`Retrieve`-
                and :class:`Pop`-parsers.
537
538

        rollback__:  A list of tuples (location, rollback-function) that are
539
540
541
542
                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.
543
544
545

        last_rb__loc__:  The last, i.e. most advanced location in the text
                where a variable changing operation occurred. If the parser
546
547
                backtracks to a location at or before last_rb__loc__ (i.e.
                location <= last_rb__loc__) then a rollback of all variable
548
549
                changing operations is necessary that occurred after the
                location to which the parser backtracks. This is done by
550
                calling method :func:`rollback_to__(location)`.
551
552
553
554
555
556
557
558
559
560
561
562
563

        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
564
565
                recording is turned off! (See :func:`add_parser_guard` and its local
                function :func:`guarded_call`)
566
567
568
569

        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
570
571
                if full memoization is enabled. (See :func:`add_parser_guard` and its
                local function :func:`guarded_call`)
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
597

        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
        the field's name, e.g.
598

599
600
601
            class Grammar(Grammar):
                ...
                symbol = RE('(?!\\d)\\w+')
602

603
604
605
606
        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
607
        ignored. See :func:`sane_parser_name()`
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
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
668
669
670
671
672
673
674
675
676
677

        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__)

        if self.wspL__:
            self.wsp_left_parser__ = Whitespace(self.wspL__)  # type: ParserBase
            self.wsp_left_parser__.grammar = self
            self.all_parsers__.add(self.wsp_left_parser__)  # don't you forget about me...
        else:
            self.wsp_left_parser__ = ZOMBIE_PARSER
        if self.wspR__:
            self.wsp_right_parser__ = Whitespace(self.wspR__)  # type: ParserBase
            self.wsp_right_parser__.grammar = self
            self.all_parsers__.add(self.wsp_right_parser__)  # don't you forget about me...
        else:
            self.wsp_right_parser__ = ZOMBIE_PARSER
        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]
678
            raise UnknownParserError('Unknown parser "%s" !' % key)
679
680
681
682
683


    def _reset__(self):
        self.document__ = EMPTY_STRING_VIEW   # type: StringView
        self._reversed__ = EMPTY_STRING_VIEW  # type: StringView
eckhart's avatar
eckhart committed
684
        self.document_length__ = 0            # type: int
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
        self.document_lbreaks__ = []          # type: List[int]
        # variables stored and recalled by Capture and Retrieve parsers
        self.variables__ = dict()             # type: Dict[str, List[str]]
        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 \
719
                isinstance(self.__dict__[parser.name], parser.__class__), \
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
                ('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.
        """
739
740

        def tail_pos(predecessors: Union[List[Node], Tuple[Node, ...]]) -> int:
eckhart's avatar
eckhart committed
741
742
            """Adds the position after the last node in the list of
            predecessors to the node."""
eckhart's avatar
eckhart committed
743
            return predecessors[-1].pos + len(predecessors[-1]) if predecessors else 0
eckhart's avatar
eckhart committed
744

745
746
747
748
749
750
751
752
753
754
755
        # 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
756
        self.document_length__ = len(self.document__)
757
        self.document_lbreaks__ = linebreaks(document) if self.history_tracking__ else []
Eckhart Arnold's avatar
Eckhart Arnold committed
758
        self.last_rb__loc__ = -1  # rollback location
759
760
761
762
763
764
765
766
767
        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
768
                result = Node(None, '').init_pos(0)
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
                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
788
                stitches.append(Node(None, skip).init_pos(tail_pos(stitches)))
789
790
                stitches[-1].add_error(error_msg)
                if self.history_tracking__:
eckhart's avatar
eckhart committed
791
792
793
794
795
796
797
                    # # 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)
798
799
800
801
802
803
                    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
804
805
                stitches.append(Node(None, rest))
            result = Node(None, tuple(stitches)).init_pos(0)
806
        if any(self.variables__.values()):
807
808
            error_str = "Capture-retrieve-stack not empty after end of parsing: " \
                + str(self.variables__)
809
810
811
812
813
            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
814
                    error_node = Node(ZOMBIE_PARSER, '').init_pos(tail_pos(result.children))
815
                    error_node.add_error(error_str)
eckhart's avatar
eckhart committed
816
                    result.result = result.children + (error_node,)
817
818
                else:
                    result.add_error(error_str)
eckhart's avatar
eckhart committed
819
        # result.pos = 0  # calculate all positions
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
        # 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
840
        while self.rollback__ and self.rollback__[-1][0] >= location:
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
            _, 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()
893
        assert RX_TOKEN_NAME.match(token)
eckhart's avatar
eckhart committed
894
        super().__init__(token)
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916

    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:
917
                return Node(self, text[len(self.name) + 2:end]), text[end + 1:]
918
919
920
        return None, text


eckhart's avatar
eckhart committed
921
922
class PlainText(Parser):
    """
923
    Parses plain text strings. (Could be done by RegExp as well, but is faster.)
eckhart's avatar
eckhart committed
924
925
926
927
928
929
930
931
932
933

    Example:
    >>> while_token = PlainText("while")
    >>> Grammar(while_token)("while").content
    'while'
    """

    def __init__(self, text: str, name: str = '') -> None:
        super().__init__(name)
        self.text = text
934
        self.len = len(text)
eckhart's avatar
eckhart committed
935
936
937
938
939
940

    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):
941
            return Node(self, self.text, True), text[self.len:]
eckhart's avatar
eckhart committed
942
943
944
        return None, text


945
946
947
948
949
950
951
952
953
954
955
956
957
958
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.

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

959
960
    EBNF-Notation:  ``/ ... /``
    EBNF-Example:   ``word = /\w+/``
961
962
963
    """

    def __init__(self, regexp, name: str = '') -> None:
eckhart's avatar
eckhart committed
964
        super().__init__(name)
965
966
967
968
969
970
971
972
        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
973
        return self.__class__(regexp, self.name)
974
975

    def __call__(self, text: StringView) -> Tuple[Optional[Node], StringView]:
976
977
978
979
        match = text.match(self.regexp)
        if match:
            capture = match.group(0)
            end = text.index(match.end())
980
981
            # regular expression must never match preprocessor-tokens!
            # TODO: Find a better solution here? e.g. static checking/re-mangling at compile time
982
983
984
985
986
            i = capture.find(BEGIN_TOKEN)
            if i >= 0:
                capture = capture[:i]
                end = i
            return Node(self, capture, True), text[end:]
987
988
989
        return None, text

    def __repr__(self):
990
        return escape_control_characters('/%s/' % self.regexp.pattern)
991
992
993
994
995
996
997
998
999
1000


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"


class RE(Parser):
    r"""