parsers.py 49.9 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
47
48

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.

Juancarlo Añez: grako, a PEG parser generator in Python,
https://bitbucket.org/apalala/grako
49
50
"""

51

Eckhart Arnold's avatar
Eckhart Arnold committed
52
import abc
53
54
import copy
import os
55
from functools import partial
56
57
58
59
try:
    import regex as re
except ImportError:
    import re
60
try:
61
    from typing import Any, Callable, Collection, cast, Dict, Iterator, List, Set, Tuple, Union
62
except ImportError:
63
    from .typing34 import Any, Callable, cast, Dict, Iterator, List, Set, Tuple, Union
64

65
from DHParser.toolkit import is_logging, log_dir, logfile_basename, escape_re, sane_parser_name
Eckhart Arnold's avatar
Eckhart Arnold committed
66
67
from DHParser.syntaxtree import WHITESPACE_PTYPE, TOKEN_PTYPE, ZOMBIE_PARSER, ParserBase, \
    Node, TransformationFunc
68
from DHParser.toolkit import load_if_file, error_messages
Eckhart Arnold's avatar
Eckhart Arnold committed
69

70
71
__all__ = ['ScannerFunc',
           'HistoryRecord',
Eckhart Arnold's avatar
Eckhart Arnold committed
72
           'Parser',
73
           'Grammar',
Eckhart Arnold's avatar
Eckhart Arnold committed
74
75
76
77
78
79
80
81
82
83
           'RX_SCANNER_TOKEN',
           'BEGIN_SCANNER_TOKEN',
           'END_SCANNER_TOKEN',
           'make_token',
           'nil_scanner',
           'ScannerToken',
           'RegExp',
           'RE',
           'Token',
           'mixin_comment',
84
85
86
           # 'UnaryOperator',
           # 'NaryOperator',
           'Synonym',
Eckhart Arnold's avatar
Eckhart Arnold committed
87
88
89
           'Optional',
           'ZeroOrMore',
           'OneOrMore',
90
           'Series',
Eckhart Arnold's avatar
Eckhart Arnold committed
91
92
93
94
95
96
97
           'Alternative',
           'FlowOperator',
           'Required',
           'Lookahead',
           'NegativeLookahead',
           'Lookbehind',
           'NegativeLookbehind',
98
99
100
           'last_value',
           'counterpart',
           'accumulate',
Eckhart Arnold's avatar
Eckhart Arnold committed
101
102
103
104
           'Capture',
           'Retrieve',
           'Pop',
           'Forward',
105
           'Compiler',
Eckhart Arnold's avatar
Eckhart Arnold committed
106
           'compile_source']
Eckhart Arnold's avatar
Eckhart Arnold committed
107

108

109
110
111
112
113
114
115
116
117
118
########################################################################
#
# Grammar and parsing infrastructure
#
########################################################################


ScannerFunc = Union[Callable[[str], str], partial]


119
LEFT_RECURSION_DEPTH = 10  # because of pythons recursion depth limit, this
Eckhart Arnold's avatar
Eckhart Arnold committed
120
                           # value ought not to be set too high
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
MAX_DROPOUTS = 25  # stop trying to recover parsing after so many errors


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"

143
144
145
146
    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
147

148
    def err_msg(self) -> str:
149
150
        return self.ERROR + ": " + "; ".join(self.node._errors).replace('\n', '\\')

151
    @property
152
    def stack(self) -> str:
153
154
        def ptos(p):
            return
155
156
157
        return "->".join(str(parser) for parser in self.call_stack)

    @property
158
    def status(self) -> str:
159
160
        return self.FAIL if self.node is None else \
            self.err_msg() if self.node._errors else self.MATCH
161
162

    @property
163
164
165
    def extent(self) -> slice:
        return (slice(-self.remaining - self.node.len, -self.remaining) if self.node
                else slice(-self.remaining, None))
166
167
168


def add_parser_guard(parser_func):
169
    def guarded_call(parser: 'Parser', text: str) -> Tuple[Node, str]:
170
171
        try:
            location = len(text)
172
173
            grammar = parser.grammar  # grammar may be 'None' for unconnected parsers!

174
175
176
177
178
179
180
181
182
            if not grammar.moving_forward__:
                # rollback variable changes for discarded branch of parsing tree
                while grammar.rollback__ and grammar.rollback__[-1][0] <= location:
                    grammar.rollback__[-1][1]()
                    grammar.rollback__.pop()
            grammar.moving_forward__ = True

            if grammar.history_tracking__:
                grammar.call_stack__.append(parser)
183

184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
            # if location has already been visited by the current parser,
            # return saved result
            if location in parser.visited:
                return parser.visited[location]
            # break left recursion at the maximum allowed depth
            if parser.recursion_counter.setdefault(location, 0) > LEFT_RECURSION_DEPTH:
                return None, text

            parser.recursion_counter[location] += 1

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

            if node is not None:
                # in case of a recursive call saves the result of the first
199
200
201
202
                # (or left-most) call that matches; but not for variable manipulating parsers,
                # because caching would interfere with changes of variable state
                if not (grammar.rollback__ and grammar.rollback__[-1][0] <= location):
                    parser.visited[location] = (node, rest)
203
                grammar.last_node__ = node   # store last node for Lookbehind parser
204
205
206
207
208
209
210
            elif location in parser.visited:
                # if parser did non match but a saved result exits, assume
                # left recursion and use the saved result
                node, rest = parser.visited[location]

            parser.recursion_counter[location] -= 1

211
            if grammar.history_tracking__:
212
                # don't track returning parsers except in case an error has occurred
213
214
215
                if grammar.moving_forward__ or (node and node._errors):
                    record = HistoryRecord(grammar.call_stack__.copy(), node, len(rest))
                    grammar.history__.append(record)
216
                    # print(record.stack, record.status, rest[:20].replace('\n', '|'))
217
218
                grammar.call_stack__.pop()
            grammar.moving_forward__ = False
219

220
221
222
223
224
225
226
227
228
229
230
        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
231
class ParserMetaClass(abc.ABCMeta):
232
233
234
235
236
237
238
239
240
241
    def __init__(cls, name, bases, attrs):
        # The following condition 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!
        guarded_parser_call = add_parser_guard(cls.__call__)
        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
242
class Parser(ParserBase, metaclass=ParserMetaClass):
243
244
    ApplyFunc = Callable[['Parser'], None]

245
246
    def __init__(self, name: str = '') -> None:
        # assert isinstance(name, str), str(name)
Eckhart Arnold's avatar
Eckhart Arnold committed
247
        super(Parser, self).__init__(name)
248
        self._grammar = None  # type: 'Grammar'
249
250
        self.reset()

Eckhart Arnold's avatar
Eckhart Arnold committed
251
    def __deepcopy__(self, memo):
252
253
        return self.__class__(self.name)

254
    def reset(self):
255
        self.visited = dict()            # type: Dict[int, Tuple[Node, str]]
256
        self.recursion_counter = dict()  # type: Dict[int, int]
257
        self.cycle_detection = set()     # type: Set[Callable]
258
        return self
259

260
    def __call__(self, text: str) -> Tuple[Node, str]:
261
262
        return None, text  # default behaviour: don't match

263
    def __add__(self, other):
264
        return Series(self, other)
265
266
267
268

    def __or__(self, other):
        return Alternative(self, other)

269
    @property
270
    def grammar(self) -> 'Grammar':
271
272
273
        return self._grammar

    @grammar.setter
274
275
    def grammar(self, grammar: 'Grammar'):
        self._grammar = grammar
276
277
278
279
280
        self._grammar_assigned_notifier()

    def _grammar_assigned_notifier(self):
        pass

281
    def apply(self, func: ApplyFunc):
282
        """Applies function `func(parser)` recursively to this parser and all
Eckhart Arnold's avatar
Eckhart Arnold committed
283
        descendants of the tree of parsers. The same function can never
284
285
286
287
288
289
290
291
292
293
        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


294
class Grammar:
295
296
    root__ = None  # type: Union[Parser, None]
    # root__ must be overwritten with the root-parser by grammar subclass
Eckhart Arnold's avatar
Eckhart Arnold committed
297
    parser_initialization__ = "pending"  # type: str
298
299
300
301
302
303

    @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.
304
            class Grammar(Grammar):
305
306
307
                ...
                symbol = RE('(?!\\d)\\w+')
        After the call of this method symbol.name == "symbol"
308
309
310
311
        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()``
312
313
314
315

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

316
317
318
319
320
        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. 
321
322
323
324
325
        """
        if cls.parser_initialization__ == "done":
            return
        cdict = cls.__dict__
        for entry, parser in cdict.items():
326
            if isinstance(parser, Parser) and sane_parser_name(entry):
327
                if not parser.name:
328
                    parser.name = entry
329
                if (isinstance(parser, Forward) and (not parser.parser.name)):
330
331
332
                    parser.parser.name = entry
        cls.parser_initialization__ = "done"

Eckhart Arnold's avatar
Eckhart Arnold committed
333
334
335
    def __init__(self, root: Parser=None) -> None:
        # if not hasattr(self.__class__, 'parser_initialization__'):
        #     self.__class__.parser_initialization__ = "pending"
336
337
338
339
        if not hasattr(self.__class__, 'wspL__'):
            self.wspL__ = ''
        if not hasattr(self.__class__, 'wspR__'):
            self.wspR__ = ''
340
341
342
        self.all_parsers__ = set()  # type: Set[Parser]
        self.dirty_flag__ = False
        self.history_tracking__ = False
343
        self._reset()
Eckhart Arnold's avatar
Eckhart Arnold committed
344
345
346
347
348
349
        # 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.
350
        self.root__ = root if root else copy.deepcopy(self.__class__.root__)
351
        if self.wspL__:
Eckhart Arnold's avatar
Eckhart Arnold committed
352
            self.wsp_left_parser__ = Whitespace(self.wspL__)  # type: ParserBase
353
            self.wsp_left_parser__.grammar = self
354
            self.all_parsers__.add(self.wsp_left_parser__)  # don't you forget about me...
355
356
357
        else:
            self.wsp_left_parser__ = ZOMBIE_PARSER
        if self.wspR__:
Eckhart Arnold's avatar
Eckhart Arnold committed
358
            self.wsp_right_parser__ = Whitespace(self.wspR__)  # type: ParserBase
359
            self.wsp_right_parser__.grammar = self
360
            self.all_parsers__.add(self.wsp_right_parser__)  # don't you forget about me...
361
362
363
364
        else:
            self.wsp_right_parser__ = ZOMBIE_PARSER
        self.root__.apply(self._add_parser)

365
    def __getitem__(self, key):
366
367
368
        try:
            return self.__dict__[key]
        except KeyError:
369
370
            parser_template = getattr(self, key, None)
            if parser_template:
Eckhart Arnold's avatar
Eckhart Arnold committed
371
                # add parser to grammar object on the fly...
372
373
374
                parser = copy.deepcopy(parser_template)
                parser.apply(self._add_parser)
                # assert self[key] == parser
Eckhart Arnold's avatar
Eckhart Arnold committed
375
                return self[key]
376
            raise KeyError('Unknown parser "%s" !' % key)
377

378
    def _reset(self):
379
        self.document__ = ""          # type: str
380
        # variables stored and recalled by Capture and Retrieve parsers
381
382
        self.variables__ = dict()     # type: Dict[str, List[str]]
        self.rollback__ = []          # type: List[Tuple[int, Callable]]
383
        # previously parsed node, needed by Lookbehind parser
384
        self.last_node__ = None       # type: Node
385
        # support for call stack tracing
386
        self.call_stack__ = []        # type: List[Parser]
387
        # snapshots of call stacks
388
        self.history__ = []           # type: List[HistoryRecord]
389
        # also needed for call stack tracing
390
        self.moving_forward__ = True  # type: bool
391

Eckhart Arnold's avatar
Eckhart Arnold committed
392
    def _add_parser(self, parser: Parser) -> None:
Eckhart Arnold's avatar
Eckhart Arnold committed
393
        """Adds the particular copy of the parser object to this
394
        particular instance of Grammar.
395
        """
396
        if parser.name:
Eckhart Arnold's avatar
Eckhart Arnold committed
397
398
399
            # 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__), \
400
401
                ('Cannot add parser "%s" because a field with the same name '
                 'already exists in grammar object!' % parser.name)
402
            setattr(self, parser.name, parser)
403
        self.all_parsers__.add(parser)
404
405
        parser.grammar = self

Eckhart Arnold's avatar
Eckhart Arnold committed
406
    def __call__(self, document: str, start_parser="root__") -> Node:
407
408
409
410
        """Parses a document with with parser-combinators.

        Args:
            document (str): The source text to be parsed.
411
412
413
            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.)
414
415
416
        Returns:
            Node: The root node ot the parse tree.
        """
417
        # assert isinstance(document, str), type(document)
418
419
        if self.root__ is None:
            raise NotImplementedError()
420
        if self.dirty_flag__:
421
            self._reset()
422
            for parser in self.all_parsers__:
423
424
                parser.reset()
        else:
425
426
427
            self.dirty_flag__ = True
        self.history_tracking__ = is_logging()
        self.document__ = document
Eckhart Arnold's avatar
Eckhart Arnold committed
428
        parser = self[start_parser] if isinstance(start_parser, str) else start_parser
429
430
        assert parser.grammar == self, "Cannot run parsers from a different grammar object!" \
                                       " %s vs. %s" % (str(self), str(parser.grammar))
431
        stitches = []  # type: List[Node]
432
        rest = document
433
434
        if not rest:
            result, ignore = parser(rest)
435
436
437
438
439
440
441
442
443
444
        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" + \
445
446
                                (("! trying to recover" +
                                  (" but stopping history recording at this point."
447
                                   if self.history_tracking__ else "..."))
448
449
450
451
                                 if len(stitches) < MAX_DROPOUTS
                                 else " too often! Terminating parser.")
                stitches.append(Node(None, skip))
                stitches[-1].add_error(error_msg)
452
                if self.history_tracking__:
453
454
455
456
                    # 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
457
                    for record in self.history__:
458
459
                        if record.node and record.node._pos < 0:
                            record.node.pos = 0
460
461
462
                    record = HistoryRecord(self.call_stack__.copy(), stitches[-1], len(rest))
                    self.history__.append(record)
                    self.history_tracking__ = False
463
464
465
466
        if stitches:
            if rest:
                stitches.append(Node(None, rest))
            result = Node(None, tuple(stitches))
467
        if any(self.variables__.values()):
468
            result.add_error("Capture-retrieve-stack not empty after end of parsing: "
469
                             + str(self.variables__))
470
471
472
        result.pos = 0  # calculate all positions
        return result

Eckhart Arnold's avatar
Eckhart Arnold committed
473
    def log_parsing_history(self, log_file_name: str='') -> None:
474
475
476
477
        """Writes a log of the parsing history of the most recently parsed
        document. 
        """
        def prepare_line(record):
478
            excerpt = self.document__.__getitem__(record.extent)[:25].replace('\n', '\\n')
479
            excerpt = "'%s'" % excerpt if len(excerpt) < 25 else "'%s...'" % excerpt
480
            return record.stack, record.status, excerpt
481
482

        def write_log(history, log_name):
Eckhart Arnold's avatar
Eckhart Arnold committed
483
            path = os.path.join(log_dir(), log_name + "_parser.log")
484
485
486
487
488
489
            if history:
                with open(path, "w", encoding="utf-8") as f:
                    f.write("\n".join(history))
            elif os.path.exists(path):
                os.remove(path)

Eckhart Arnold's avatar
Eckhart Arnold committed
490
491
492
493
        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 = [], [], []
494
        for record in self.history__:
Eckhart Arnold's avatar
Eckhart Arnold committed
495
496
            line = ";  ".join(prepare_line(record))
            full_history.append(line)
497
            if record.node and record.node.parser.ptype != WHITESPACE_PTYPE:
Eckhart Arnold's avatar
Eckhart Arnold committed
498
                match_history.append(line)
499
                if record.node.error_flag:
Eckhart Arnold's avatar
Eckhart Arnold committed
500
501
502
503
                    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')
504
505


Eckhart Arnold's avatar
Eckhart Arnold committed
506
507
def dsl_error_msg(parser: Parser, error_str: str) -> str:
    """Returns an error message for errors in the parser configuration,
508
509
510
    e.g. errors that result in infinite loops.

    Args:
Eckhart Arnold's avatar
Eckhart Arnold committed
511
        parser (Parser):  The parser where the error was noticed. Note
512
            that this is not necessarily the parser that caused the
Eckhart Arnold's avatar
Eckhart Arnold committed
513
            error but only where the error became aparent.
514
515
516
517
518
        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.
    """
519
    msg = ["DSL parser specification error:", error_str, 'Caught by parser "%s".' % str(parser)]
520
521
    if parser.grammar.history__:
        msg.extend(["\nCall stack:", parser.grammar.history__[-1].stack])
522
523
524
525
526
    else:
        msg.extend(["\nEnable history tracking in Grammar object to display call stack."])
    return " ".join(msg)


527
528
529
530
531
532
533
534
535
536
537
538
########################################################################
#
# Token and Regular Expression parser classes (i.e. leaf classes)
#
########################################################################


RX_SCANNER_TOKEN = re.compile('\w+')
BEGIN_SCANNER_TOKEN = '\x1b'
END_SCANNER_TOKEN = '\x1c'


539
def make_token(token: str, argument: str = '') -> str:
540
541
542
543
544
545
546
547
548
549
550
551
552
    """Turns the ``token`` and ``argument`` into a special token that
    will be caught by the `ScannerToken`-parser.

    This function is a support function that should be used by scanners
    to inject scanner tokens into the source text.
    """
    assert RX_SCANNER_TOKEN.match(token)
    assert argument.find(BEGIN_SCANNER_TOKEN) < 0
    assert argument.find(END_SCANNER_TOKEN) < 0

    return BEGIN_SCANNER_TOKEN + token + argument + END_SCANNER_TOKEN


553
def nil_scanner(text: str) -> str:
Eckhart Arnold's avatar
Eckhart Arnold committed
554
    return text
555
556
557


class ScannerToken(Parser):
558
559
560
561
562
563
564
565
566
567
    """
    Parses tokens that have been inserted by a Scanner.
    
    Scanners can generate Tokens with the ``make_token``-function.
    These tokens start and end with magic characters that can only be
    matched by the ScannerToken Parser. Scanner tokens can be used to
    insert BEGIN - END delimiters at the beginning or ending of an 
    indented block. Otherwise indented block are difficult to handle 
    with parsing expression grammars.
    """
568
569
570

    def __init__(self, scanner_token: str) -> None:
        assert scanner_token and scanner_token.isupper()
571
        assert RX_SCANNER_TOKEN.match(scanner_token)
572
        super(ScannerToken, self).__init__(scanner_token)
573

574
    def __call__(self, text: str) -> Tuple[Node, str]:
575
576
577
578
579
        if text[0:1] == BEGIN_SCANNER_TOKEN:
            end = text.find(END_SCANNER_TOKEN, 1)
            if end < 0:
                node = Node(self, '').add_error(
                    'END_SCANNER_TOKEN delimiter missing from scanner token. '
580
                    '(Most likely due to a scanner bug!)')  # type: Node
581
582
583
584
                return node, text[1:]
            elif end == 0:
                node = Node(self, '').add_error(
                    'Scanner token cannot have zero length. '
585
                    '(Most likely due to a scanner bug!)')  # type: Node
586
587
588
589
590
591
                return node, text[2:]
            elif text.find(BEGIN_SCANNER_TOKEN, 1, end) >= 0:
                node = Node(self, text[len(self.name) + 1:end])
                node.add_error(
                    'Scanner tokens must not be nested or contain '
                    'BEGIN_SCANNER_TOKEN delimiter as part of their argument. '
592
                    '(Most likely due to a scanner bug!)')  # type: Node
593
594
595
596
597
598
599
600
                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):
601
602
    """
    Regular expression parser.
603
604
605
606
607
608
    
    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.
    """
609
610

    def __init__(self, regexp, name: str = '') -> None:
611
        super(RegExp, self).__init__(name)
612
613
614
        self.regexp = re.compile(regexp) if isinstance(regexp, str) else regexp

    def __deepcopy__(self, memo):
Eckhart Arnold's avatar
Eckhart Arnold committed
615
        # `regex` supports deep copies, but not `re`
616
        try:
Eckhart Arnold's avatar
Eckhart Arnold committed
617
            regexp = copy.deepcopy(self.regexp, memo)
618
619
        except TypeError:
            regexp = self.regexp.pattern
620
        return RegExp(regexp, self.name)
621

622
    def __call__(self, text: str) -> Tuple[Node, str]:
623
624
625
626
627
628
        match = text[0:1] != BEGIN_SCANNER_TOKEN and self.regexp.match(text)  # ESC starts a scanner token.
        if match:
            end = match.end()
            return Node(self, text[:end]), text[end:]
        return None, text

629
    def __repr__(self):
630
        return '/%s/' % self.regexp.pattern
631

632
633
634
    def __str__(self):
        return repr(self)

635

636
637
class Whitespace(RegExp):
    assert WHITESPACE_PTYPE == ":Whitespace"
638
639


640
641
class RE(Parser):
    """Regular Expressions with optional leading or trailing whitespace.
642
643
644
645
646
647
648
649
650
651
    
    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.
652
    """
653
    def __init__(self, regexp, wL=None, wR=None, name=''):
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
        """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.
        """
669
        super(RE, self).__init__(name)
670
671
        self.wL = wL
        self.wR = wR
672
673
        self.wspLeft = Whitespace(wL) if wL else ZOMBIE_PARSER
        self.wspRight = Whitespace(wR) if wR else ZOMBIE_PARSER
674
675
        self.main = RegExp(regexp)

Eckhart Arnold's avatar
Eckhart Arnold committed
676
677
678
679
680
    def __deepcopy__(self, memo={}):
        try:
            regexp = copy.deepcopy(self.main.regexp, memo)
        except TypeError:
            regexp = self.main.regexp.pattern
681
        return self.__class__(regexp, self.wL, self.wR, self.name)
Eckhart Arnold's avatar
Eckhart Arnold committed
682

683
    def __call__(self, text: str) -> Tuple[Node, str]:
684
        # assert self.main.regexp.pattern != "@"
685
        t = text    # type: str
686
687
688
689
690
691
692
693
694
        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

695
    def __repr__(self):
696
697
698
        wL = '~' if self.wL else ''
        wR = '~' if self.wR else ''
        return wL + '/%s/' % self.main.regexp.pattern + wR
699

700
701
    def __str__(self):
        return self.name or repr(self)
702
703
704

    def _grammar_assigned_notifier(self):
        if self.grammar:
705
            # use default whitespace parsers if not otherwise specified
706
707
708
709
710
            if self.wL is None:
                self.wspLeft = self.grammar.wsp_left_parser__
            if self.wR is None:
                self.wspRight = self.grammar.wsp_right_parser__

711
    def apply(self, func: Parser.ApplyFunc):
712
713
714
715
716
717
718
719
        if super(RE, self).apply(func):
            if self.wL:
                self.wspLeft.apply(func)
            if self.wR:
                self.wspRight.apply(func)
            self.main.apply(func)


720
class Token(RE):
721
722
723
724
725
726
727
728
729
    """Class Token parses simple strings. Any regular regular
    expression commands will be interpreted as simple sequence of
    characters.

    Other than that class Token is essentially a renamed version of
    class RE. Because tokens often have a particular semantic different
    from other REs, parsing them with a separate parser class allows to
    distinguish them by their parser type.
    """
730
731
    assert TOKEN_PTYPE == ":Token"

732
    def __init__(self, token: str, wL=None, wR=None, name: str = '') -> None:
733
734
735
736
737
        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)
738

739
    def __repr__(self):
740
        return '"%s"' % self.token if self.token.find('"') < 0 else "'%s'" % self.token
741

742

743
def mixin_comment(whitespace: str, comment: str) -> str:
744
745
746
747
748
749
750
    """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).
751
752
753
754
755
756
757
758
759
760
761
762
763
    """
    wspc = '(?:' + whitespace + '(?:' + comment + whitespace + ')*)'
    return wspc


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


class UnaryOperator(Parser):
764
    def __init__(self, parser: Parser, name: str = '') -> None:
765
        super(UnaryOperator, self).__init__(name)
766
        # assert isinstance(parser, Parser)
767
        self.parser = parser  # type: Parser
768

Eckhart Arnold's avatar
Eckhart Arnold committed
769
770
    def __deepcopy__(self, memo):
        parser = copy.deepcopy(self.parser, memo)
771
        return self.__class__(parser, self.name)
Eckhart Arnold's avatar
Eckhart Arnold committed
772

773
    def apply(self, func: Parser.ApplyFunc):
774
775
776
777
778
        if super(UnaryOperator, self).apply(func):
            self.parser.apply(func)


class NaryOperator(Parser):
779
    def __init__(self, *parsers: Parser, name: str = '') -> None:
780
        super(NaryOperator, self).__init__(name)
781
        # assert all([isinstance(parser, Parser) for parser in parsers]), str(parsers)
782
        self.parsers = parsers  # type: Tuple[Parser, ...]
783

Eckhart Arnold's avatar
Eckhart Arnold committed
784
785
    def __deepcopy__(self, memo):
        parsers = copy.deepcopy(self.parsers, memo)
786
        return self.__class__(*parsers, name=self.name)
Eckhart Arnold's avatar
Eckhart Arnold committed
787

788
    def apply(self, func: Parser.ApplyFunc):
789
790
791
792
793
        if super(NaryOperator, self).apply(func):
            for parser in self.parsers:
                parser.apply(func)


794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
class Synonym(UnaryOperator):
    """Simply calls another parser and encapsulates the result in
    another node if that parser matches.

    This parser is needed to support synonyms in EBNF, e.g.
        jahr       = JAHRESZAHL
        JAHRESZAHL = /\d\d\d\d/
    Otherwise the first line could not be represented by any parser
    class, in which case it would be unclear whether the parser
    RE('\d\d\d\d') carries the name 'JAHRESZAHL' or 'jahr'
    """
    def __call__(self, text: str) -> Tuple[Node, str]:
        node, text = self.parser(text)
        if node:
            return Node(self, node), text
        return None, text

811
    def __repr__(self):
812
        return self.name or self.parser.repr
813

814

815
class Optional(UnaryOperator):
816
    def __init__(self, parser: Parser, name: str = '') -> None:
817
        super(Optional, self).__init__(parser, name)
818
        # assert isinstance(parser, Parser)
819
        assert not isinstance(parser, Optional), \
820
            "Redundant nesting of options: %s(%s)" % \
821
822
            (str(name), str(parser.name))
        assert not isinstance(parser, Required), \
823
            "Nesting options with required elements is contradictory: " \
824
825
            "%s(%s)" % (str(name), str(parser.name))

826
    def __call__(self, text: str) -> Tuple[Node, str]:
827
828
829
830
831
        node, text = self.parser(text)
        if node:
            return Node(self, node), text
        return Node(self, ()), text

832
    def __repr__(self):
833
834
        return '[' + (self.parser.repr[1:-1] if isinstance(self.parser, Alternative)
                      and not self.parser.name else self.parser.repr) + ']'
835
836

class ZeroOrMore(Optional):
837
838
    def __call__(self, text: str) -> Tuple[Node, str]:
        results = ()  # type: Tuple[Node, ...]
839
840
841
        n = len(text) + 1
        while text and len(text) < n:
            n = len(text)
842
843
844
            node, text = self.parser(text)
            if not node:
                break
845
            if len(text) == n:
di68kap's avatar
di68kap committed
846
                node.add_error(dsl_error_msg(self, 'Infinite Loop detected.'))
847
848
849
            results += (node,)
        return Node(self, results), text

850
    def __repr__(self):
851
852
        return '{' + (self.parser.repr[1:-1] if isinstance(self.parser, Alternative)
                      and not self.parser.name else self.parser.repr) + '}'
853

854
855

class OneOrMore(UnaryOperator):
856
    def __init__(self, parser: Parser, name: str = '') -> None:
857
        super(OneOrMore, self).__init__(parser, name)
858
859
860
861
        assert not isinstance(parser, Optional), \
            "Use ZeroOrMore instead of nesting OneOrMore and Optional: " \
            "%s(%s)" % (str(name), str(parser.name))

862
863
864
    def __call__(self, text: str) -> Tuple[Node, str]:
        results = ()  # type: Tuple[Node, ...]
        text_ = text  # type: str
865
866
867
        n = len(text) + 1
        while text_ and len(text_) < n:
            n = len(text_)
868
869
870
            node, text_ = self.parser(text_)
            if not node:
                break
871
            if len(text_) == n:
872
                node.add_error(dsl_error_msg(self, 'Infinite Loop detected.'))
873
874
875
876
877
            results += (node,)
        if results == ():
            return None, text
        return Node(self, results), text_

878
    def __repr__(self):
879
880
        return '{' + (self.parser.repr[1:-1] if isinstance(self.parser, Alternative)
                      and not self.parser.name else self.parser.repr) + '}+'
881

882

883
class Series(NaryOperator):
884
    def __init__(self, *parsers: Parser, name: str = '') -> None:
885
        super(Series, self).__init__(*parsers, name=name)
886
887
        assert len(self.parsers) >= 1

888
889
890
    def __call__(self, text: str) -> Tuple[Node, str]:
        results = ()  # type: Tuple[Node, ...]
        text_ = text  # type: str
891
892
893
        for parser in self.parsers:
            node, text_ = parser(text_)
            if not node:
894
895
                return None, text
            results += (node,)
896
897
898
899
900
            if node.error_flag:
                break
        assert len(results) <= len(self.parsers)
        return Node(self, results), text_

901
902
903
    def __repr__(self):
        return " ".join(parser.repr for parser in self.parsers)

904
905
906
    def __add__(self, other: Parser) -> 'Series':
        other_parsers = cast('Series', other).parsers if isinstance(other, Series) else (other,)
        return Series(*(self.parsers + other_parsers))
907

908
909
910
    def __radd__(self, other: Parser) -> 'Series':
        other_parsers = cast('Series', other).parsers if isinstance(other, Series) else (other,)
        return Series(*(other_parsers + self.parsers))
911

912
913
914
915
    def __iadd__(self, other):
        other_parsers = cast('Series', other).parsers if isinstance(other, Series) else (other,)
        self.parsers += other_parsers
        return self
916

917
918

class Alternative(NaryOperator):
919
920
921
922
923
924
925
    """Matches if at least one of several alternatives matches. Returns
    the first match.

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

Eckhart Arnold's avatar
Eckhart Arnold committed
926
927
    # the order of the sub-expression matters!
    >>> number = RE('\d+') | RE('\d+') + RE('\.') + RE('\d+')
928
929
    >>> Grammar(number)("3.1416").show()
    '3 <<< Error on ".1416" | Parser stopped before end! trying to recover... >>> '
930
931

    # the most selective expression should be put first:
Eckhart Arnold's avatar
Eckhart Arnold committed
932
    >>> number = RE('\d+') + RE('\.') + RE('\d+') | RE('\d+')
933
    >>> Grammar(number)("3.1416").show()
Eckhart Arnold's avatar
Eckhart Arnold committed
934
    '3.1416'
935
    """
936
937

    def __init__(self, *parsers: Parser, name: str = '') -> None:
938
        super(Alternative, self).__init__(*parsers, name=name)
939
940
941
        assert len(self.parsers) >= 1
        assert all(not isinstance(p, Optional) for p in self.parsers)

942
    def __call__(self, text: str) -> Tuple[Node, str]:
943
944
945
946
947
948
        for parser in self.parsers:
            node, text_ = parser(text)
            if node:
                return Node(self, node), text_
        return None, text

949
    def __repr__(self):
950
        return '(' + ' | '.join(parser.repr for parser in self.parsers) + ')'
951

952
953
954
955
    def __or__(self, other: Parser) -> 'Alternative':
        other_parsers = cast('Alternative', other).parsers \
                        if isinstance(other, Alternative) else (other,)
        return Alternative(*(self.parsers + other_parsers))
956

957
958
959
960
    def __ror__(self, other: Parser) -> 'Alternative':
        other_parsers = cast('Alternative', other).parsers \
                        if isinstance(other, Alternative) else (other,)
        return Alternative(*(other_parsers + self.parsers))
961

962
963
964
965
966
    def __ior__(self, other):
        other_parsers = cast('Alternative', other.parsers) \
                        if isinstance(other, Alternative) else (other,)
        self.parsers += other_parsers
        return self
967

968
969
970
971
972
973
974
975
976

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


class FlowOperator(UnaryOperator):
977
    def __init__(self, parser: Parser, name: str = '') -> None:
978
        super(FlowOperator, self).__init__(parser, name)
979
980
981


class Required(FlowOperator):
982
    # Add constructor that checks for logical errors, like `Required(Optional(...))` constructs ?
983
    def __call__(self, text: str) -> Tuple[Node, str]:
984
985
986
987
988
989
990
991
992
993
994
        node, text_ = self.parser(text)
        if not node:
            m = re.search(r'\s(\S)', text)
            i = max(1, m.regs[1][0]) if m else 1
            node = Node(self, text[:i])
            text_ = text[i:]
            # assert False, "*"+text[:i]+"*"
            node.add_error('%s expected; "%s..." found!' %
                           (str(self.parser), text[:10]))
        return node, text_

995
    def __repr__(self):
996
        return '§' + self.parser.repr
997

998
999

class Lookahead(FlowOperator):
1000
    def __init__(self, parser: Parser, name: str = '') -> None:
1001
        super(Lookahead, self).__init__(parser, name)
1002

1003
    def __call__(self, text: str) -> Tuple[Node, str]:
1004
1005
1006
1007
1008
1009
        node, text_ = self.parser(text)
        if self.sign(node is not None):
            return Node(self, ''), text
        else:
            return None, text

1010
    def __repr__(self):
1011
        return '&' + self.parser.repr
1012

1013
    def sign(self, bool_value) -> bool:
1014
1015
1016
1017
        return bool_value


class NegativeLookahead(Lookahead):
1018
    def __repr__(self):
1019
        return '!' + self.parser.repr
1020

1021
    def sign(self, bool_value) -> bool:
1022
1023
1024
        return not bool_value


1025
def iter_right_branch(node) -> Iterator[Node]:
1026
1027
    """
    Iterates over the right branch of `node` starting with node itself.