parser.py 63.8 KB
Newer Older
1
"""parsers.py - parser combinators for for DHParser
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16

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

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

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

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

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.

47
48
49
50
Elizabeth Scott and Adrian Johnstone, GLL Parsing,
in: Electronic Notes in Theoretical Computer Science 253 (2010) 177–189,
http://dotat.at/tmp/gll.pdf

Eckhart Arnold's avatar
Eckhart Arnold committed
51
52
Juancarlo Añez: grako, a PEG parser generator in Python,
https://bitbucket.org/apalala/grako
53
54
55

Vegard Øye: General Parser Combinators in Racket, 2012,
https://epsil.github.io/gll/
56
57
"""

58

Eckhart Arnold's avatar
Eckhart Arnold committed
59
import abc
60
import copy
61
import os
Eckhart Arnold's avatar
Eckhart Arnold committed
62
63
from functools import partial

64
65
66
67
try:
    import regex as re
except ImportError:
    import re
68
try:
69
    from typing import Any, Callable, cast, Dict, Iterator, List, Set, Tuple, Union
70
71
72
73
    # try:
    #     from typing import Collection
    # except ImportError:
    #     pass
74
except ImportError:
75
    from .typing34 import Any, Callable, cast, Dict, Iterator, List, Set, Tuple, Union
76

77
from DHParser.toolkit import is_logging, log_dir, logfile_basename, escape_re, sane_parser_name
Eckhart Arnold's avatar
Eckhart Arnold committed
78
79
from DHParser.syntaxtree import WHITESPACE_PTYPE, TOKEN_PTYPE, ZOMBIE_PARSER, ParserBase, \
    Node, TransformationFunc
80
from DHParser.toolkit import load_if_file, error_messages
Eckhart Arnold's avatar
Eckhart Arnold committed
81

82
__all__ = ('PreprocessorFunc',
83
           'HistoryRecord',
Eckhart Arnold's avatar
Eckhart Arnold committed
84
           'Parser',
85
           'Grammar',
86
87
88
           'RX_PREPROCESSOR_TOKEN',
           'BEGIN_TOKEN',
           'END_TOKEN',
Eckhart Arnold's avatar
Eckhart Arnold committed
89
           'make_token',
90
91
           'nil_preprocessor',
           'PreprocessorToken',
Eckhart Arnold's avatar
Eckhart Arnold committed
92
93
94
95
           'RegExp',
           'RE',
           'Token',
           'mixin_comment',
96
97
98
           # 'UnaryOperator',
           # 'NaryOperator',
           'Synonym',
Eckhart Arnold's avatar
Eckhart Arnold committed
99
100
101
           'Optional',
           'ZeroOrMore',
           'OneOrMore',
102
           'Series',
Eckhart Arnold's avatar
Eckhart Arnold committed
103
104
105
106
107
108
109
           'Alternative',
           'FlowOperator',
           'Required',
           'Lookahead',
           'NegativeLookahead',
           'Lookbehind',
           'NegativeLookbehind',
110
111
112
           'last_value',
           'counterpart',
           'accumulate',
Eckhart Arnold's avatar
Eckhart Arnold committed
113
114
115
116
           'Capture',
           'Retrieve',
           'Pop',
           'Forward',
117
           'Compiler',
118
           'compile_source')
Eckhart Arnold's avatar
Eckhart Arnold committed
119

120

121

122
123
124
125
126
127
128
########################################################################
#
# Grammar and parsing infrastructure
#
########################################################################


129
PreprocessorFunc = Union[Callable[[str], str], partial]
130
131


132
133
134
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
Eckhart Arnold's avatar
Eckhart Arnold committed
135
MAX_DROPOUTS = 3  # type: int
136
# stop trying to recover parsing after so many errors
137

138

139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
class HistoryRecord:
    """
    Stores debugging information about one completed step in the
    parsing history. 

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

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

158
159
160
161
    def __init__(self, call_stack: List['Parser'], node: Node, remaining: int) -> None:
        self.call_stack = call_stack    # type: List['Parser']
        self.node = node                # type: Node
        self.remaining = remaining      # type: int
162

163
    def err_msg(self) -> str:
164
165
        return self.ERROR + ": " + "; ".join(self.node._errors).replace('\n', '\\')

166
    @property
167
    def stack(self) -> str:
168
169
        return "->".join((repr(p) if p.ptype == ':RegExp' else p.name or p.ptype)
                         for p in self.call_stack)
170
171

    @property
172
    def status(self) -> str:
173
174
        return self.FAIL if self.node is None else \
            self.err_msg() if self.node._errors else self.MATCH
175
176

    @property
177
178
179
    def extent(self) -> slice:
        return (slice(-self.remaining - self.node.len, -self.remaining) if self.node
                else slice(-self.remaining, None))
180
181
182


def add_parser_guard(parser_func):
183
184
185
186
187
    """
    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.
    """
188
    def guarded_call(parser: 'Parser', text: str) -> Tuple[Node, str]:
189
190
        try:
            location = len(text)
191
192
            grammar = parser.grammar  # grammar may be 'None' for unconnected parsers!

193
            if not grammar.moving_forward__:
194
                # rollback variable changes from discarded parser passes
195
196
197
                if grammar.last_rb__loc__ <= location:
                    grammar.rollback_to__(location)
                grammar.moving_forward__ = True
198
                grammar.left_recursion_encountered__ = False
199
200
201

            if grammar.history_tracking__:
                grammar.call_stack__.append(parser)
202

203
204
205
            # if location has already been visited by the current parser,
            # return saved result
            if location in parser.visited:
206
                return parser.visited[location]
207

208
209
            # break left recursion at the maximum allowed depth
            if parser.recursion_counter.setdefault(location, 0) > LEFT_RECURSION_DEPTH:
210
                grammar.left_recursion_encountered__ = True
211
212
213
214
215
216
217
                return None, text

            parser.recursion_counter[location] += 1

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

218
            if node is None:
Eckhart Arnold's avatar
Eckhart Arnold committed
219
220
221
222
223
224
225
                # retrieve an earlier match result (from left recursion)
                # if it exists
                node, rest = parser.visited.get(location, (None, rest))
                # 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!
                if node is None and not grammar.left_recursion_encountered__:
                    # ortherwise also cache None-results
226
227
                    parser.visited[location] = None, rest
            else:
228
                # variable manipulating parsers will be excluded, though,
229
                # because caching would interfere with changes of variable state
230
                if grammar.last_rb__loc__ > location:
Eckhart Arnold's avatar
Eckhart Arnold committed
231
232
                    # in case of left recursion, the first recursive step that
                    # matches will store its result in the cache
233
                    parser.visited[location] = (node, rest)
234
235
236

            parser.recursion_counter[location] -= 1

237
            if grammar.history_tracking__:
238
                # don't track returning parsers except in case an error has occurred
239
240
241
                if grammar.moving_forward__ or (node and node._errors):
                    record = HistoryRecord(grammar.call_stack__.copy(), node, len(rest))
                    grammar.history__.append(record)
242
                    # print(record.stack, record.status, rest[:20].replace('\n', '|'))
243
244
                grammar.call_stack__.pop()
            grammar.moving_forward__ = False
245

246
247
248
249
250
251
252
253
254
255
256
        except RecursionError:
            node = Node(None, text[:min(10, max(1, text.find("\n")))] + " ...")
            node.add_error("maximum recursion depth of parser reached; "
                           "potentially due to too many errors!")
            rest = ''

        return node, rest

    return guarded_call


Eckhart Arnold's avatar
Eckhart Arnold committed
257
class ParserMetaClass(abc.ABCMeta):
258
259
260
261
262
    """
    ParserMetaClass adds a wrapper to the __call__ method of parser
    objects during initialization that takes care of memoizing,
    left recursion and tracing.
    """
263
    def __init__(cls, name, bases, attrs):
264
265
        guarded_parser_call = add_parser_guard(cls.__call__)
        # The following check is necessary for classes that don't override
266
267
268
269
270
271
272
        # the __call__() method, because in these cases the non-overridden
        # __call__()-method would be substituted a second time!
        if cls.__call__.__code__ != guarded_parser_call.__code__:
            cls.__call__ = guarded_parser_call
        super(ParserMetaClass, cls).__init__(name, bases, attrs)


Eckhart Arnold's avatar
Eckhart Arnold committed
273
class Parser(ParserBase, metaclass=ParserMetaClass):
274
275
276
277
    """
    (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.
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294

    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.
Eckhart Arnold's avatar
Eckhart Arnold committed
295
296
297
298
299
300
301
302
303

    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 `Optional`).
    If `i > 0` then the parser has "moved forward". If the parser does
    not match it returns `(None, text).
304
305
    """

306
307
    ApplyFunc = Callable[['Parser'], None]

308
309
    def __init__(self, name: str = '') -> None:
        # assert isinstance(name, str), str(name)
Eckhart Arnold's avatar
Eckhart Arnold committed
310
        super(Parser, self).__init__(name)
311
        self._grammar = None  # type: 'Grammar'
312
313
        self.reset()

Eckhart Arnold's avatar
Eckhart Arnold committed
314
    def __deepcopy__(self, memo):
Eckhart Arnold's avatar
Eckhart Arnold committed
315
316
317
318
319
320
        """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.
        """
321
322
        return self.__class__(self.name)

323
    def reset(self):
Eckhart Arnold's avatar
Eckhart Arnold committed
324
325
326
        """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."""
327
        self.visited = dict()            # type: Dict[int, Tuple[Node, str]]
328
        self.recursion_counter = dict()  # type: Dict[int, int]
329
        self.cycle_detection = set()     # type: Set[Callable]
330
        return self
331

332
    def __call__(self, text: str) -> Tuple[Node, str]:
Eckhart Arnold's avatar
Eckhart Arnold committed
333
334
335
        """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."""
336
337
        return None, text  # default behaviour: don't match

338
    def __add__(self, other: 'Parser') -> 'Series':
339
340
        """The + operator generates a series-parser that applies two
        parsers in sequence."""
341
        return Series(self, other)
342

343
    def __or__(self, other: 'Parser') -> 'Alternative':
344
345
346
        """The | operator generates an alternative parser that applies
        the first parser and, if that does not match, the second parser.
        """
347
348
        return Alternative(self, other)

349
    @property
350
    def grammar(self) -> 'Grammar':
351
352
353
        return self._grammar

    @grammar.setter
354
    def grammar(self, grammar: 'Grammar'):
Eckhart Arnold's avatar
Eckhart Arnold committed
355
356
357
358
359
360
        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!"
361
362

    def _grammar_assigned_notifier(self):
363
364
        """A function that notifies the parser object that it has been
        assigned to a grammar."""
365
366
        pass

367
    def apply(self, func: ApplyFunc):
368
369
        """
        Applies function `func(parser)` recursively to this parser and all
Eckhart Arnold's avatar
Eckhart Arnold committed
370
        descendant parsers if any exist. The same function can never
371
372
373
374
375
376
377
378
379
380
        be applied twice between calls of the ``reset()``-method!
        """
        if func in self.cycle_detection:
            return False
        else:
            self.cycle_detection.add(func)
            func(self)
            return True


381
def mixin_comment(whitespace: str, comment: str) -> str:
382
383
    """
    Returns a regular expression that merges comment and whitespace
384
385
386
387
388
389
390
391
392
393
394
    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


395
class Grammar:
396
    """
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
    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)
Eckhart Arnold's avatar
Eckhart Arnold committed
412
        >>> number_parser("3.1416").content()
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
        '3.1416'

    Collecting the parsers that define a grammar in a descentand 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.

    3. The parsers in 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.
434
435

    Example:
436

437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
        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,
455
456
457
    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.
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
    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 objects contain a few special fields for implicit
    whitespace and comments that should be overwritten, if the defaults
    (no comments, horizontal right aligned whitespace) don't fit:
    COMMENT__   - regular expression string for matching comments
    wspL__      - regular expression string for left aligned whitespace
    wspR__      - regular expression string for right aligned whitespace
475
    root__      - the root parser of the grammar
476
477
    """

478
479
    root__ = None  # type: Union[Parser, None]
    # root__ must be overwritten with the root-parser by grammar subclass
Eckhart Arnold's avatar
Eckhart Arnold committed
480
    parser_initialization__ = "pending"  # type: str
481
482
483
484
485
    # some default values
    COMMENT__ = r''  # r'#.*(?:\n|$)'
    WSP__ = mixin_comment(whitespace=r'[\t ]*', comment=COMMENT__)
    wspL__ = ''
    wspR__ = WSP__
486

487

488
    @classmethod
489
    def _assign_parser_names__(cls):
490
491
        """
        Initializes the `parser.name` fields of those
492
493
        Parser objects that are directly assigned to a class field with
        the field's name, e.g.
494
            class Grammar(Grammar):
495
496
497
                ...
                symbol = RE('(?!\\d)\\w+')
        After the call of this method symbol.name == "symbol"
498
499
500
501
        holds. Names assigned via the ``name``-parameter of the
        constructor will not be overwritten. Parser names starting or
        ending with a double underscore like ``root__`` will be
        ignored. See ``toolkit.sane_parser_name()``
502
503
504
505

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

506
507
508
509
510
        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. 
511
512
513
514
515
        """
        if cls.parser_initialization__ == "done":
            return
        cdict = cls.__dict__
        for entry, parser in cdict.items():
516
            if isinstance(parser, Parser) and sane_parser_name(entry):
517
                if not parser.name:
518
                    parser.name = entry
519
                if (isinstance(parser, Forward) and (not parser.parser.name)):
520
521
522
                    parser.parser.name = entry
        cls.parser_initialization__ = "done"

523

Eckhart Arnold's avatar
Eckhart Arnold committed
524
525
526
    def __init__(self, root: Parser=None) -> None:
        # if not hasattr(self.__class__, 'parser_initialization__'):
        #     self.__class__.parser_initialization__ = "pending"
527
528
529
530
        # if not hasattr(self.__class__, 'wspL__'):
        #     self.wspL__ = ''
        # if not hasattr(self.__class__, 'wspR__'):
        #     self.wspR__ = ''
531
532
533
        self.all_parsers__ = set()  # type: Set[Parser]
        self.dirty_flag__ = False
        self.history_tracking__ = False
534
        self._reset__()
535

Eckhart Arnold's avatar
Eckhart Arnold committed
536
        # prepare parsers in the class, first
537
        self._assign_parser_names__()
538

Eckhart Arnold's avatar
Eckhart Arnold committed
539
540
541
542
        # 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.
Eckhart Arnold's avatar
Eckhart Arnold committed
543
        self.root__ = copy.deepcopy(root) if root else copy.deepcopy(self.__class__.root__)
544

545
        if self.wspL__:
Eckhart Arnold's avatar
Eckhart Arnold committed
546
            self.wsp_left_parser__ = Whitespace(self.wspL__)  # type: ParserBase
547
            self.wsp_left_parser__.grammar = self
548
            self.all_parsers__.add(self.wsp_left_parser__)  # don't you forget about me...
549
550
551
        else:
            self.wsp_left_parser__ = ZOMBIE_PARSER
        if self.wspR__:
Eckhart Arnold's avatar
Eckhart Arnold committed
552
            self.wsp_right_parser__ = Whitespace(self.wspR__)  # type: ParserBase
553
            self.wsp_right_parser__.grammar = self
554
            self.all_parsers__.add(self.wsp_right_parser__)  # don't you forget about me...
555
556
        else:
            self.wsp_right_parser__ = ZOMBIE_PARSER
557
        self.root__.apply(self._add_parser__)
558

559

560
    def __getitem__(self, key):
561
562
563
        try:
            return self.__dict__[key]
        except KeyError:
564
565
            parser_template = getattr(self, key, None)
            if parser_template:
Eckhart Arnold's avatar
Eckhart Arnold committed
566
                # add parser to grammar object on the fly...
567
                parser = copy.deepcopy(parser_template)
568
                parser.apply(self._add_parser__)
569
                # assert self[key] == parser
Eckhart Arnold's avatar
Eckhart Arnold committed
570
                return self[key]
571
            raise KeyError('Unknown parser "%s" !' % key)
572

573

574
    def _reset__(self):
575
        self.document__ = ""          # type: str
Eckhart Arnold's avatar
Eckhart Arnold committed
576
        self._reversed__ = ""         # type: str
577
        # variables stored and recalled by Capture and Retrieve parsers
578
579
        self.variables__ = dict()     # type: Dict[str, List[str]]
        self.rollback__ = []          # type: List[Tuple[int, Callable]]
580
        self.last_rb__loc__ = -1  # type: int
581
        # support for call stack tracing
582
        self.call_stack__ = []        # type: List[Parser]
583
        # snapshots of call stacks
584
        self.history__ = []           # type: List[HistoryRecord]
585
        # also needed for call stack tracing
586
        self.moving_forward__ = True  # type: bool
587
        self.left_recursion_encountered__ = False  # type: bool
588

589

Eckhart Arnold's avatar
Eckhart Arnold committed
590
591
592
593
594
595
596
    @property
    def reversed__(self) -> str:
        if not self._reversed__:
            self._reversed__ = self.document__[::-1]
        return self._reversed__


597
    def _add_parser__(self, parser: Parser) -> None:
598
599
        """
        Adds the particular copy of the parser object to this
600
        particular instance of Grammar.
601
        """
602
        if parser.name:
Eckhart Arnold's avatar
Eckhart Arnold committed
603
604
605
            # prevent overwriting instance variables or parsers of a different class
            assert parser.name not in self.__dict__ or \
                   isinstance(self.__dict__[parser.name], parser.__class__), \
606
607
                ('Cannot add parser "%s" because a field with the same name '
                 'already exists in grammar object!' % parser.name)
608
            setattr(self, parser.name, parser)
609
        self.all_parsers__.add(parser)
610
611
        parser.grammar = self

612

Eckhart Arnold's avatar
Eckhart Arnold committed
613
    def __call__(self, document: str, start_parser="root__") -> Node:
614
615
        """
        Parses a document with with parser-combinators.
616
617
618

        Args:
            document (str): The source text to be parsed.
619
620
621
            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.)
622
623
624
        Returns:
            Node: The root node ot the parse tree.
        """
625
        # assert isinstance(document, str), type(document)
626
627
        if self.root__ is None:
            raise NotImplementedError()
628
        if self.dirty_flag__:
629
            self._reset__()
630
            for parser in self.all_parsers__:
631
632
                parser.reset()
        else:
633
634
635
            self.dirty_flag__ = True
        self.history_tracking__ = is_logging()
        self.document__ = document
636
        self.last_rb__loc__ = len(document) + 1  # rollback location
Eckhart Arnold's avatar
Eckhart Arnold committed
637
        parser = self[start_parser] if isinstance(start_parser, str) else start_parser
638
639
        assert parser.grammar == self, "Cannot run parsers from a different grammar object!" \
                                       " %s vs. %s" % (str(self), str(parser.grammar))
640
        stitches = []  # type: List[Node]
641
        rest = document
642
643
        if not rest:
            result, ignore = parser(rest)
644
645
646
647
648
649
650
651
652
653
        while rest and len(stitches) < MAX_DROPOUTS:
            result, rest = parser(rest)
            if rest:
                fwd = rest.find("\n") + 1 or len(rest)
                skip, rest = rest[:fwd], rest[fwd:]
                if result is None:
                    error_msg = "Parser did not match! Invalid source file?"
                else:
                    stitches.append(result)
                    error_msg = "Parser stopped before end" + \
654
655
                                (("! trying to recover" +
                                  (" but stopping history recording at this point."
656
                                   if self.history_tracking__ else "..."))
657
658
659
660
                                 if len(stitches) < MAX_DROPOUTS
                                 else " too often! Terminating parser.")
                stitches.append(Node(None, skip))
                stitches[-1].add_error(error_msg)
661
                if self.history_tracking__:
662
663
664
665
                    # some parsers may have matched and left history records with nodes != None.
                    # Because these are not connected to the stiched root node, their pos
                    # properties will not be initialized by setting the root node's pos property
                    # to zero. Therefore, their pos properties need to be initialized here
666
                    for record in self.history__:
667
668
                        if record.node and record.node._pos < 0:
                            record.node.pos = 0
669
670
671
                    record = HistoryRecord(self.call_stack__.copy(), stitches[-1], len(rest))
                    self.history__.append(record)
                    self.history_tracking__ = False
672
673
674
675
        if stitches:
            if rest:
                stitches.append(Node(None, rest))
            result = Node(None, tuple(stitches))
676
        if any(self.variables__.values()):
677
            result.add_error("Capture-retrieve-stack not empty after end of parsing: "
678
                             + str(self.variables__))
679
680
681
        result.pos = 0  # calculate all positions
        return result

682

683
    def push_rollback__(self, location, func):
684
685
        """
        Adds a rollback function that either removes or re-adds
686
687
688
689
        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.
        """
690
691
692
        self.rollback__.append((location, func))
        self.last_rb__loc__ = location

693

694
    def rollback_to__(self, location):
695
696
        """
        Rolls back the variable stacks (`self.variables`) to its
697
698
        state at an earlier location in the parsed document.
        """
699
700
701
702
703
704
705
        while self.rollback__ and self.rollback__[-1][0] <= location:
            loc, rollback_func = self.rollback__.pop()
            assert not loc > self.last_rb__loc__
            rollback_func()
        self.last_rb__loc__ == self.rollback__[-1][0] if self.rollback__ \
            else (len(self.document__) + 1)

706

707
    def log_parsing_history__(self, log_file_name: str = '') -> None:
708
709
        """
        Writes a log of the parsing history of the most recently parsed
710
711
712
        document. 
        """
        def prepare_line(record):
713
            excerpt = self.document__.__getitem__(record.extent)[:25].replace('\n', '\\n')
714
            excerpt = "'%s'" % excerpt if len(excerpt) < 25 else "'%s...'" % excerpt
715
            return record.stack, record.status, excerpt
716
717

        def write_log(history, log_name):
Eckhart Arnold's avatar
Eckhart Arnold committed
718
            path = os.path.join(log_dir(), log_name + "_parser.log")
719
720
721
722
723
724
            if history:
                with open(path, "w", encoding="utf-8") as f:
                    f.write("\n".join(history))
            elif os.path.exists(path):
                os.remove(path)

725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
        if is_logging():
            assert self.history__, \
                "Parser did not yet run or logging was turned off when running parser!"
            if not log_file_name:
                name = self.__class__.__name__
                log_file_name = name[:-7] if name.lower().endswith('grammar') else name
            full_history, match_history, errors_only = [], [], []
            for record in self.history__:
                line = ";  ".join(prepare_line(record))
                full_history.append(line)
                if record.node and record.node.parser.ptype != WHITESPACE_PTYPE:
                    match_history.append(line)
                    if record.node.error_flag:
                        errors_only.append(line)
            write_log(full_history, log_file_name + '_full')
            write_log(match_history, log_file_name + '_match')
            write_log(errors_only, log_file_name + '_errors')
742
743


Eckhart Arnold's avatar
Eckhart Arnold committed
744
def dsl_error_msg(parser: Parser, error_str: str) -> str:
745
746
    """
    Returns an error message for errors in the parser configuration,
747
748
749
    e.g. errors that result in infinite loops.

    Args:
Eckhart Arnold's avatar
Eckhart Arnold committed
750
        parser (Parser):  The parser where the error was noticed. Note
751
            that this is not necessarily the parser that caused the
752
            error but only where the error became apparent.
753
754
755
756
757
        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.
    """
758
    msg = ["DSL parser specification error:", error_str, 'Caught by parser "%s".' % str(parser)]
759
760
    if parser.grammar.history__:
        msg.extend(["\nCall stack:", parser.grammar.history__[-1].stack])
761
762
763
764
765
    else:
        msg.extend(["\nEnable history tracking in Grammar object to display call stack."])
    return " ".join(msg)


766

767
768
769
770
771
772
773
########################################################################
#
# Token and Regular Expression parser classes (i.e. leaf classes)
#
########################################################################


774
775
776
RX_PREPROCESSOR_TOKEN = re.compile('\w+')
BEGIN_TOKEN = '\x1b'
END_TOKEN = '\x1c'
777
778


779
def make_token(token: str, argument: str = '') -> str:
780
781
    """
    Turns the ``token`` and ``argument`` into a special token that
782
    will be caught by the `PreprocessorToken`-parser.
783

784
785
    This function is a support function that should be used by
    preprocessors to inject preprocessor tokens into the source text.
786
    """
787
788
789
    assert RX_PREPROCESSOR_TOKEN.match(token)
    assert argument.find(BEGIN_TOKEN) < 0
    assert argument.find(END_TOKEN) < 0
790

791
    return BEGIN_TOKEN + token + argument + END_TOKEN
792
793


794
def nil_preprocessor(text: str) -> str:
Eckhart Arnold's avatar
Eckhart Arnold committed
795
    return text
796
797


798
class PreprocessorToken(Parser):
799
    """
800
    Parses tokens that have been inserted by a preprocessor.
801
    
802
    Preprocessors can generate Tokens with the ``make_token``-function.
803
    These tokens start and end with magic characters that can only be
804
805
806
    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.
807
    """
808

809
810
811
812
    def __init__(self, token: str) -> None:
        assert token and token.isupper()
        assert RX_PREPROCESSOR_TOKEN.match(token)
        super(PreprocessorToken, self).__init__(token)
813

814
    def __call__(self, text: str) -> Tuple[Node, str]:
815
816
        if text[0:1] == BEGIN_TOKEN:
            end = text.find(END_TOKEN, 1)
817
818
            if end < 0:
                node = Node(self, '').add_error(
819
820
                    'END_TOKEN delimiter missing from preprocessor token. '
                    '(Most likely due to a preprocessor bug!)')  # type: Node
821
822
823
                return node, text[1:]
            elif end == 0:
                node = Node(self, '').add_error(
824
825
                    'Preprocessor-token cannot have zero length. '
                    '(Most likely due to a preprocessor bug!)')
826
                return node, text[2:]
827
            elif text.find(BEGIN_TOKEN, 1, end) >= 0:
828
829
                node = Node(self, text[len(self.name) + 1:end])
                node.add_error(
830
831
832
                    'Preprocessor-tokens must not be nested or contain '
                    'BEGIN_TOKEN delimiter as part of their argument. '
                    '(Most likely due to a preprocessor bug!)')
833
834
835
836
837
838
839
840
                return node, text[end:]
            if text[1:len(self.name) + 1] == self.name:
                return Node(self, text[len(self.name) + 1:end]), \
                       text[end + 1:]
        return None, text


class RegExp(Parser):
Eckhart Arnold's avatar
Eckhart Arnold committed
841
    """Regular expression parser.
842
843
844
845
846
    
    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.
Eckhart Arnold's avatar
Eckhart Arnold committed
847
848
849
850
851
852
853
854

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

    EBNF-Notation:  `/ ... /`
    EBNF-Example:   `word = /\w+/`
855
    """
856
857

    def __init__(self, regexp, name: str = '') -> None:
858
        super(RegExp, self).__init__(name)
859
860
861
        self.regexp = re.compile(regexp) if isinstance(regexp, str) else regexp

    def __deepcopy__(self, memo):
Eckhart Arnold's avatar
Eckhart Arnold committed
862
        # `regex` supports deep copies, but not `re`
863
        try:
Eckhart Arnold's avatar
Eckhart Arnold committed
864
            regexp = copy.deepcopy(self.regexp, memo)
865
866
        except TypeError:
            regexp = self.regexp.pattern
867
        return RegExp(regexp, self.name)
868

869
    def __call__(self, text: str) -> Tuple[Node, str]:
870
        match = text[0:1] != BEGIN_TOKEN and self.regexp.match(text)  # ESC starts a preprocessor token.
871
872
873
874
875
        if match:
            end = match.end()
            return Node(self, text[:end]), text[end:]
        return None, text

876
    def __repr__(self):
877
        return '/%s/' % self.regexp.pattern
878

879

880
881
class Whitespace(RegExp):
    assert WHITESPACE_PTYPE == ":Whitespace"
882
883


884
class RE(Parser):
885
886
    """
    Regular Expressions with optional leading or trailing whitespace.
887
888
889
890
891
892
893
894
895
896
    
    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.
Eckhart Arnold's avatar
Eckhart Arnold committed
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911

    Example (allowing whitespace on the right hand side, but not on
    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 " "))'
    >>> parser(' Haus').content()
    ' <<< Error on " Haus" | Parser did not match! Invalid source file? >>> '

    EBNF-Notation:  `/ ... /~`  or  `~/ ... /`  or  `~/ ... /~`
    EBNF-Example:   `word = /\w+/~`
912
    """
913
    def __init__(self, regexp, wL=None, wR=None, name=''):
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
        """Constructor for class RE.
                
        Args:
            regexp (str or regex object):  The regular expression to be
                used for parsing. 
            wL (str or regexp):  Left whitespace regular expression, 
                i.e. either ``None``, the empty string or a regular
                expression (e.g. "\s*") that defines whitespace. An 
                empty string means no whitespace will be skipped,
                ``None`` means that the default whitespace will be 
                used.
            wR (str or regexp):  Right whitespace regular expression.
                See above.
            name:  The optional name of the parser.
        """
929
        super(RE, self).__init__(name)
930
931
        self.wL = wL
        self.wR = wR
932
933
        self.wspLeft = Whitespace(wL) if wL else ZOMBIE_PARSER
        self.wspRight = Whitespace(wR) if wR else ZOMBIE_PARSER
934
935
        self.main = RegExp(regexp)

Eckhart Arnold's avatar
Eckhart Arnold committed
936
937
938
939
940
    def __deepcopy__(self, memo={}):
        try:
            regexp = copy.deepcopy(self.main.regexp, memo)
        except TypeError:
            regexp = self.main.regexp.pattern
941
        return self.__class__(regexp, self.wL, self.wR, self.name)
Eckhart Arnold's avatar
Eckhart Arnold committed
942

943
    def __call__(self, text: str) -> Tuple[Node, str]:
944
        # assert self.main.regexp.pattern != "@"
945
        t = text    # type: str
946
947
948
949
950
951
952
953
954
        wL, t = self.wspLeft(t)
        main, t = self.main(t)
        if main:
            wR, t = self.wspRight(t)
            result = tuple(nd for nd in (wL, main, wR)
                           if nd and nd.result != '')
            return Node(self, result), t
        return None, text

955
    def __repr__(self):
956
957
        wL = '~' if self.wspLeft != ZOMBIE_PARSER else ''
        wR = '~' if self.wspRight != ZOMBIE_PARSER else ''
958
        return wL + '/%s/' % self.main.regexp.pattern + wR
959

960
961
    def _grammar_assigned_notifier(self):
        if self.grammar:
962
            # use default whitespace parsers if not otherwise specified
963
964
965
966
967
            if self.wL is None:
                self.wspLeft = self.grammar.wsp_left_parser__
            if self.wR is None:
                self.wspRight = self.grammar.wsp_right_parser__

968
    def apply(self, func: Parser.ApplyFunc):
969
970
971
972
973
974
975
976
        if super(RE, self).apply(func):
            if self.wL:
                self.wspLeft.apply(func)
            if self.wR:
                self.wspRight.apply(func)
            self.main.apply(func)


977
class Token(RE):
978
979
    """
    Class Token parses simple strings. Any regular regular
980
981
982
983
984
985
986
987
    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.
    """
988
989
    assert TOKEN_PTYPE == ":Token"

990
    def __init__(self, token: str, wL=None, wR=None, name: str = '') -> None:
991
992
993
994
995
        self.token = token
        super(Token, self).__init__(escape_re(token), wL, wR, name)

    def __deepcopy__(self, memo={}):
        return self.__class__(self.token, self.wL, self.wR, self.name)
996

997
    def __repr__(self):
998
        return '"%s"' % self.token if self.token.find('"') < 0 else "'%s'" % self.token
999

1000
1001
1002
1003
1004
1005
1006
1007
1008

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


class UnaryOperator(Parser):
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
    """
    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.
    """
1019
    def __init__(self, parser: Parser, name: str = '') -> None:
1020
        super(UnaryOperator, self).__init__(name)
1021
        # assert isinstance(parser, Parser)
1022
        self.parser = parser  # type: Parser
1023

Eckhart Arnold's avatar
Eckhart Arnold committed
1024
1025
    def __deepcopy__(self, memo):
        parser = copy.deepcopy(self.parser, memo)
1026
        return self.__class__(parser, self.name)
Eckhart Arnold's avatar
Eckhart Arnold committed
1027

1028
    def apply(self, func: Parser.ApplyFunc):
1029
1030
1031
1032
1033
        if super(UnaryOperator, self).apply(func):
            self.parser.apply(func)


class NaryOperator(Parser):
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
    """
    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.
    """
1044
    def __init__(self, *parsers: Parser, name: str = '') -> None:
1045
        super(NaryOperator, self).__init__(name)
1046
        # assert all([isinstance(parser, Parser) for parser in parsers]), str(parsers)
1047
        self.parsers = parsers  # type: Tuple[Parser, ...]
1048

Eckhart Arnold's avatar
Eckhart Arnold committed
1049
1050
    def __deepcopy__(self, memo):
        parsers = copy.deepcopy(self.parsers, memo)
1051
        return self.__class__(*parsers, name=self.name)
Eckhart Arnold's avatar
Eckhart Arnold committed
1052

1053
    def apply(self, func: Parser.ApplyFunc):
1054
1055
1056
1057
1058
1059
        if super(NaryOperator, self).apply(func):
            for parser in self.parsers:
                parser.apply(func)


class Optional(UnaryOperator):
Eckhart Arnold's avatar
Eckhart Arnold committed
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079