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

parsers.py 49.8 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
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
        return "->".join((repr(p) if p.ptype == ':RegExp' else p.name or p.ptype)
                         for p in self.call_stack)
155
156

    @property
157
    def status(self) -> str:
158
159
        return self.FAIL if self.node is None else \
            self.err_msg() if self.node._errors else self.MATCH
160
161

    @property
162
163
164
    def extent(self) -> slice:
        return (slice(-self.remaining - self.node.len, -self.remaining) if self.node
                else slice(-self.remaining, None))
165
166
167


def add_parser_guard(parser_func):
168
    def guarded_call(parser: 'Parser', text: str) -> Tuple[Node, str]:
169
170
        try:
            location = len(text)
171
172
            grammar = parser.grammar  # grammar may be 'None' for unconnected parsers!

173
174
175
176
177
178
179
180
181
            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)
182

183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
            # 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
198
199
200
201
                # (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)
202
                grammar.last_node__ = node   # store last node for Lookbehind parser
203
204
205
206
207
208
209
            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

210
            if grammar.history_tracking__:
211
                # don't track returning parsers except in case an error has occurred
212
213
214
                if grammar.moving_forward__ or (node and node._errors):
                    record = HistoryRecord(grammar.call_stack__.copy(), node, len(rest))
                    grammar.history__.append(record)
215
                    # print(record.stack, record.status, rest[:20].replace('\n', '|'))
216
217
                grammar.call_stack__.pop()
            grammar.moving_forward__ = False
218

219
220
221
222
223
224
225
226
227
228
229
        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
230
class ParserMetaClass(abc.ABCMeta):
231
232
233
234
235
236
237
238
239
240
    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
241
class Parser(ParserBase, metaclass=ParserMetaClass):
242
243
    ApplyFunc = Callable[['Parser'], None]

244
245
    def __init__(self, name: str = '') -> None:
        # assert isinstance(name, str), str(name)
Eckhart Arnold's avatar
Eckhart Arnold committed
246
        super(Parser, self).__init__(name)
247
        self._grammar = None  # type: 'Grammar'
248
249
        self.reset()

Eckhart Arnold's avatar
Eckhart Arnold committed
250
    def __deepcopy__(self, memo):
251
252
        return self.__class__(self.name)

253
    def reset(self):
254
        self.visited = dict()            # type: Dict[int, Tuple[Node, str]]
255
        self.recursion_counter = dict()  # type: Dict[int, int]
256
        self.cycle_detection = set()     # type: Set[Callable]
257
        return self
258

259
    def __call__(self, text: str) -> Tuple[Node, str]:
260
261
        return None, text  # default behaviour: don't match

262
    def __add__(self, other):
263
        return Series(self, other)
264
265
266
267

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

268
    @property
269
    def grammar(self) -> 'Grammar':
270
271
272
        return self._grammar

    @grammar.setter
273
274
    def grammar(self, grammar: 'Grammar'):
        self._grammar = grammar
275
276
277
278
279
        self._grammar_assigned_notifier()

    def _grammar_assigned_notifier(self):
        pass

280
    def apply(self, func: ApplyFunc):
281
        """Applies function `func(parser)` recursively to this parser and all
Eckhart Arnold's avatar
Eckhart Arnold committed
282
        descendants of the tree of parsers. The same function can never
283
284
285
286
287
288
289
290
291
292
        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


293
class Grammar:
294
295
    root__ = None  # type: Union[Parser, None]
    # root__ must be overwritten with the root-parser by grammar subclass
Eckhart Arnold's avatar
Eckhart Arnold committed
296
    parser_initialization__ = "pending"  # type: str
297
298
299
300
301
302

    @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.
303
            class Grammar(Grammar):
304
305
306
                ...
                symbol = RE('(?!\\d)\\w+')
        After the call of this method symbol.name == "symbol"
307
308
309
310
        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()``
311
312
313
314

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

315
316
317
318
319
        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. 
320
321
322
323
324
        """
        if cls.parser_initialization__ == "done":
            return
        cdict = cls.__dict__
        for entry, parser in cdict.items():
325
            if isinstance(parser, Parser) and sane_parser_name(entry):
326
                if not parser.name:
327
                    parser.name = entry
328
                if (isinstance(parser, Forward) and (not parser.parser.name)):
329
330
331
                    parser.parser.name = entry
        cls.parser_initialization__ = "done"

Eckhart Arnold's avatar
Eckhart Arnold committed
332
333
334
    def __init__(self, root: Parser=None) -> None:
        # if not hasattr(self.__class__, 'parser_initialization__'):
        #     self.__class__.parser_initialization__ = "pending"
335
336
337
338
        if not hasattr(self.__class__, 'wspL__'):
            self.wspL__ = ''
        if not hasattr(self.__class__, 'wspR__'):
            self.wspR__ = ''
339
340
341
        self.all_parsers__ = set()  # type: Set[Parser]
        self.dirty_flag__ = False
        self.history_tracking__ = False
342
        self._reset()
Eckhart Arnold's avatar
Eckhart Arnold committed
343
344
345
346
347
348
        # 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.
349
        self.root__ = root if root else copy.deepcopy(self.__class__.root__)
350
        if self.wspL__:
Eckhart Arnold's avatar
Eckhart Arnold committed
351
            self.wsp_left_parser__ = Whitespace(self.wspL__)  # type: ParserBase
352
            self.wsp_left_parser__.grammar = self
353
            self.all_parsers__.add(self.wsp_left_parser__)  # don't you forget about me...
354
355
356
        else:
            self.wsp_left_parser__ = ZOMBIE_PARSER
        if self.wspR__:
Eckhart Arnold's avatar
Eckhart Arnold committed
357
            self.wsp_right_parser__ = Whitespace(self.wspR__)  # type: ParserBase
358
            self.wsp_right_parser__.grammar = self
359
            self.all_parsers__.add(self.wsp_right_parser__)  # don't you forget about me...
360
361
362
363
        else:
            self.wsp_right_parser__ = ZOMBIE_PARSER
        self.root__.apply(self._add_parser)

364
    def __getitem__(self, key):
365
366
367
        try:
            return self.__dict__[key]
        except KeyError:
368
369
            parser_template = getattr(self, key, None)
            if parser_template:
Eckhart Arnold's avatar
Eckhart Arnold committed
370
                # add parser to grammar object on the fly...
371
372
373
                parser = copy.deepcopy(parser_template)
                parser.apply(self._add_parser)
                # assert self[key] == parser
Eckhart Arnold's avatar
Eckhart Arnold committed
374
                return self[key]
375
            raise KeyError('Unknown parser "%s" !' % key)
376

377
    def _reset(self):
378
        self.document__ = ""          # type: str
379
        # variables stored and recalled by Capture and Retrieve parsers
380
381
        self.variables__ = dict()     # type: Dict[str, List[str]]
        self.rollback__ = []          # type: List[Tuple[int, Callable]]
382
        # previously parsed node, needed by Lookbehind parser
383
        self.last_node__ = None       # type: Node
384
        # support for call stack tracing
385
        self.call_stack__ = []        # type: List[Parser]
386
        # snapshots of call stacks
387
        self.history__ = []           # type: List[HistoryRecord]
388
        # also needed for call stack tracing
389
        self.moving_forward__ = True  # type: bool
390

Eckhart Arnold's avatar
Eckhart Arnold committed
391
    def _add_parser(self, parser: Parser) -> None:
Eckhart Arnold's avatar
Eckhart Arnold committed
392
        """Adds the particular copy of the parser object to this
393
        particular instance of Grammar.
394
        """
395
        if parser.name:
Eckhart Arnold's avatar
Eckhart Arnold committed
396
397
398
            # 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__), \
399
400
                ('Cannot add parser "%s" because a field with the same name '
                 'already exists in grammar object!' % parser.name)
401
            setattr(self, parser.name, parser)
402
        self.all_parsers__.add(parser)
403
404
        parser.grammar = self

Eckhart Arnold's avatar
Eckhart Arnold committed
405
    def __call__(self, document: str, start_parser="root__") -> Node:
406
407
408
409
        """Parses a document with with parser-combinators.

        Args:
            document (str): The source text to be parsed.
410
411
412
            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.)
413
414
415
        Returns:
            Node: The root node ot the parse tree.
        """
416
        # assert isinstance(document, str), type(document)
417
418
        if self.root__ is None:
            raise NotImplementedError()
419
        if self.dirty_flag__:
420
            self._reset()
421
            for parser in self.all_parsers__:
422
423
                parser.reset()
        else:
424
425
426
            self.dirty_flag__ = True
        self.history_tracking__ = is_logging()
        self.document__ = document
Eckhart Arnold's avatar
Eckhart Arnold committed
427
        parser = self[start_parser] if isinstance(start_parser, str) else start_parser
428
429
        assert parser.grammar == self, "Cannot run parsers from a different grammar object!" \
                                       " %s vs. %s" % (str(self), str(parser.grammar))
430
        stitches = []  # type: List[Node]
431
        rest = document
432
433
        if not rest:
            result, ignore = parser(rest)
434
435
436
437
438
439
440
441
442
443
        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" + \
444
445
                                (("! trying to recover" +
                                  (" but stopping history recording at this point."
446
                                   if self.history_tracking__ else "..."))
447
448
449
450
                                 if len(stitches) < MAX_DROPOUTS
                                 else " too often! Terminating parser.")
                stitches.append(Node(None, skip))
                stitches[-1].add_error(error_msg)
451
                if self.history_tracking__:
452
453
454
455
                    # 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
456
                    for record in self.history__:
457
458
                        if record.node and record.node._pos < 0:
                            record.node.pos = 0
459
460
461
                    record = HistoryRecord(self.call_stack__.copy(), stitches[-1], len(rest))
                    self.history__.append(record)
                    self.history_tracking__ = False
462
463
464
465
        if stitches:
            if rest:
                stitches.append(Node(None, rest))
            result = Node(None, tuple(stitches))
466
        if any(self.variables__.values()):
467
            result.add_error("Capture-retrieve-stack not empty after end of parsing: "
468
                             + str(self.variables__))
469
470
471
        result.pos = 0  # calculate all positions
        return result

Eckhart Arnold's avatar
Eckhart Arnold committed
472
    def log_parsing_history(self, log_file_name: str='') -> None:
473
474
475
476
        """Writes a log of the parsing history of the most recently parsed
        document. 
        """
        def prepare_line(record):
477
            excerpt = self.document__.__getitem__(record.extent)[:25].replace('\n', '\\n')
478
            excerpt = "'%s'" % excerpt if len(excerpt) < 25 else "'%s...'" % excerpt
479
            return record.stack, record.status, excerpt
480
481

        def write_log(history, log_name):
Eckhart Arnold's avatar
Eckhart Arnold committed
482
            path = os.path.join(log_dir(), log_name + "_parser.log")
483
484
485
486
487
488
            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
489
490
491
492
        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 = [], [], []
493
        for record in self.history__:
Eckhart Arnold's avatar
Eckhart Arnold committed
494
495
            line = ";  ".join(prepare_line(record))
            full_history.append(line)
496
            if record.node and record.node.parser.ptype != WHITESPACE_PTYPE:
Eckhart Arnold's avatar
Eckhart Arnold committed
497
                match_history.append(line)
498
                if record.node.error_flag:
Eckhart Arnold's avatar
Eckhart Arnold committed
499
500
501
502
                    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')
503
504


Eckhart Arnold's avatar
Eckhart Arnold committed
505
506
def dsl_error_msg(parser: Parser, error_str: str) -> str:
    """Returns an error message for errors in the parser configuration,
507
508
509
    e.g. errors that result in infinite loops.

    Args:
Eckhart Arnold's avatar
Eckhart Arnold committed
510
        parser (Parser):  The parser where the error was noticed. Note
511
            that this is not necessarily the parser that caused the
Eckhart Arnold's avatar
Eckhart Arnold committed
512
            error but only where the error became aparent.
513
514
515
516
517
        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.
    """
518
    msg = ["DSL parser specification error:", error_str, 'Caught by parser "%s".' % str(parser)]
519
520
    if parser.grammar.history__:
        msg.extend(["\nCall stack:", parser.grammar.history__[-1].stack])
521
522
523
524
525
    else:
        msg.extend(["\nEnable history tracking in Grammar object to display call stack."])
    return " ".join(msg)


526
527
528
529
530
531
532
533
534
535
536
537
########################################################################
#
# Token and Regular Expression parser classes (i.e. leaf classes)
#
########################################################################


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


538
def make_token(token: str, argument: str = '') -> str:
539
540
541
542
543
544
545
546
547
548
549
550
551
    """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


552
def nil_scanner(text: str) -> str:
Eckhart Arnold's avatar
Eckhart Arnold committed
553
    return text
554
555
556


class ScannerToken(Parser):
557
558
559
560
561
562
563
564
565
566
    """
    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.
    """
567
568
569

    def __init__(self, scanner_token: str) -> None:
        assert scanner_token and scanner_token.isupper()
570
        assert RX_SCANNER_TOKEN.match(scanner_token)
571
        super(ScannerToken, self).__init__(scanner_token)
572

573
    def __call__(self, text: str) -> Tuple[Node, str]:
574
575
576
577
578
        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. '
579
                    '(Most likely due to a scanner bug!)')  # type: Node
580
581
582
583
                return node, text[1:]
            elif end == 0:
                node = Node(self, '').add_error(
                    'Scanner token cannot have zero length. '
584
                    '(Most likely due to a scanner bug!)')  # type: Node
585
586
587
588
589
590
                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. '
591
                    '(Most likely due to a scanner bug!)')  # type: Node
592
593
594
595
596
597
598
599
                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):
600
601
    """
    Regular expression parser.
602
603
604
605
606
607
    
    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.
    """
608
609

    def __init__(self, regexp, name: str = '') -> None:
610
        super(RegExp, self).__init__(name)
611
612
613
        self.regexp = re.compile(regexp) if isinstance(regexp, str) else regexp

    def __deepcopy__(self, memo):
Eckhart Arnold's avatar
Eckhart Arnold committed
614
        # `regex` supports deep copies, but not `re`
615
        try:
Eckhart Arnold's avatar
Eckhart Arnold committed
616
            regexp = copy.deepcopy(self.regexp, memo)
617
618
        except TypeError:
            regexp = self.regexp.pattern
619
        return RegExp(regexp, self.name)
620

621
    def __call__(self, text: str) -> Tuple[Node, str]:
622
623
624
625
626
627
        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

628
    def __repr__(self):
629
        return '/%s/' % self.regexp.pattern
630

631

632
633
class Whitespace(RegExp):
    assert WHITESPACE_PTYPE == ":Whitespace"
634
635


636
637
class RE(Parser):
    """Regular Expressions with optional leading or trailing whitespace.
638
639
640
641
642
643
644
645
646
647
    
    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.
648
    """
649
    def __init__(self, regexp, wL=None, wR=None, name=''):
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
        """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.
        """
665
        super(RE, self).__init__(name)
666
667
        self.wL = wL
        self.wR = wR
668
669
        self.wspLeft = Whitespace(wL) if wL else ZOMBIE_PARSER
        self.wspRight = Whitespace(wR) if wR else ZOMBIE_PARSER
670
671
        self.main = RegExp(regexp)

Eckhart Arnold's avatar
Eckhart Arnold committed
672
673
674
675
676
    def __deepcopy__(self, memo={}):
        try:
            regexp = copy.deepcopy(self.main.regexp, memo)
        except TypeError:
            regexp = self.main.regexp.pattern
677
        return self.__class__(regexp, self.wL, self.wR, self.name)
Eckhart Arnold's avatar
Eckhart Arnold committed
678

679
    def __call__(self, text: str) -> Tuple[Node, str]:
680
        # assert self.main.regexp.pattern != "@"
681
        t = text    # type: str
682
683
684
685
686
687
688
689
690
        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

691
    def __repr__(self):
692
693
694
        wL = '~' if self.wL else ''
        wR = '~' if self.wR else ''
        return wL + '/%s/' % self.main.regexp.pattern + wR
695

696
697
    def _grammar_assigned_notifier(self):
        if self.grammar:
698
            # use default whitespace parsers if not otherwise specified
699
700
701
702
703
            if self.wL is None:
                self.wspLeft = self.grammar.wsp_left_parser__
            if self.wR is None:
                self.wspRight = self.grammar.wsp_right_parser__

704
    def apply(self, func: Parser.ApplyFunc):
705
706
707
708
709
710
711
712
        if super(RE, self).apply(func):
            if self.wL:
                self.wspLeft.apply(func)
            if self.wR:
                self.wspRight.apply(func)
            self.main.apply(func)


713
class Token(RE):
714
715
716
717
718
719
720
721
722
    """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.
    """
723
724
    assert TOKEN_PTYPE == ":Token"

725
    def __init__(self, token: str, wL=None, wR=None, name: str = '') -> None:
726
727
728
729
730
        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)
731

732
    def __repr__(self):
733
        return '"%s"' % self.token if self.token.find('"') < 0 else "'%s'" % self.token
734

735

736
def mixin_comment(whitespace: str, comment: str) -> str:
737
738
739
740
741
742
743
    """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).
744
745
746
747
748
749
750
751
752
753
754
755
756
    """
    wspc = '(?:' + whitespace + '(?:' + comment + whitespace + ')*)'
    return wspc


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


class UnaryOperator(Parser):
757
    def __init__(self, parser: Parser, name: str = '') -> None:
758
        super(UnaryOperator, self).__init__(name)
759
        # assert isinstance(parser, Parser)
760
        self.parser = parser  # type: Parser
761

Eckhart Arnold's avatar
Eckhart Arnold committed
762
763
    def __deepcopy__(self, memo):
        parser = copy.deepcopy(self.parser, memo)
764
        return self.__class__(parser, self.name)
Eckhart Arnold's avatar
Eckhart Arnold committed
765

766
    def apply(self, func: Parser.ApplyFunc):
767
768
769
770
771
        if super(UnaryOperator, self).apply(func):
            self.parser.apply(func)


class NaryOperator(Parser):
772
    def __init__(self, *parsers: Parser, name: str = '') -> None:
773
        super(NaryOperator, self).__init__(name)
774
        # assert all([isinstance(parser, Parser) for parser in parsers]), str(parsers)
775
        self.parsers = parsers  # type: Tuple[Parser, ...]
776

Eckhart Arnold's avatar
Eckhart Arnold committed
777
778
    def __deepcopy__(self, memo):
        parsers = copy.deepcopy(self.parsers, memo)
779
        return self.__class__(*parsers, name=self.name)
Eckhart Arnold's avatar
Eckhart Arnold committed
780

781
    def apply(self, func: Parser.ApplyFunc):
782
783
784
785
786
        if super(NaryOperator, self).apply(func):
            for parser in self.parsers:
                parser.apply(func)


787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
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

804
    def __repr__(self):
805
        return self.name or self.parser.repr
806

807

808
class Optional(UnaryOperator):
809
    def __init__(self, parser: Parser, name: str = '') -> None:
810
        super(Optional, self).__init__(parser, name)
811
        # assert isinstance(parser, Parser)
812
        assert not isinstance(parser, Optional), \
813
            "Redundant nesting of options: %s(%s)" % \
814
815
            (str(name), str(parser.name))
        assert not isinstance(parser, Required), \
816
            "Nesting options with required elements is contradictory: " \
817
818
            "%s(%s)" % (str(name), str(parser.name))

819
    def __call__(self, text: str) -> Tuple[Node, str]:
820
821
822
823
824
        node, text = self.parser(text)
        if node:
            return Node(self, node), text
        return Node(self, ()), text

825
    def __repr__(self):
826
827
        return '[' + (self.parser.repr[1:-1] if isinstance(self.parser, Alternative)
                      and not self.parser.name else self.parser.repr) + ']'
828
829

class ZeroOrMore(Optional):
830
831
    def __call__(self, text: str) -> Tuple[Node, str]:
        results = ()  # type: Tuple[Node, ...]
832
833
834
        n = len(text) + 1
        while text and len(text) < n:
            n = len(text)
835
836
837
            node, text = self.parser(text)
            if not node:
                break
838
            if len(text) == n:
di68kap's avatar
di68kap committed
839
                node.add_error(dsl_error_msg(self, 'Infinite Loop detected.'))
840
841
842
            results += (node,)
        return Node(self, results), text

843
    def __repr__(self):
844
845
        return '{' + (self.parser.repr[1:-1] if isinstance(self.parser, Alternative)
                      and not self.parser.name else self.parser.repr) + '}'
846

847
848

class OneOrMore(UnaryOperator):
849
    def __init__(self, parser: Parser, name: str = '') -> None:
850
        super(OneOrMore, self).__init__(parser, name)
851
852
853
854
        assert not isinstance(parser, Optional), \
            "Use ZeroOrMore instead of nesting OneOrMore and Optional: " \
            "%s(%s)" % (str(name), str(parser.name))

855
856
857
    def __call__(self, text: str) -> Tuple[Node, str]:
        results = ()  # type: Tuple[Node, ...]
        text_ = text  # type: str
858
859
860
        n = len(text) + 1
        while text_ and len(text_) < n:
            n = len(text_)
861
862
863
            node, text_ = self.parser(text_)
            if not node:
                break
864
            if len(text_) == n:
865
                node.add_error(dsl_error_msg(self, 'Infinite Loop detected.'))
866
867
868
869
870
            results += (node,)
        if results == ():
            return None, text
        return Node(self, results), text_

871
    def __repr__(self):
872
873
        return '{' + (self.parser.repr[1:-1] if isinstance(self.parser, Alternative)
                      and not self.parser.name else self.parser.repr) + '}+'
874

875

876
class Series(NaryOperator):
877
    def __init__(self, *parsers: Parser, name: str = '') -> None:
878
        super(Series, self).__init__(*parsers, name=name)
879
880
        assert len(self.parsers) >= 1

881
882
883
    def __call__(self, text: str) -> Tuple[Node, str]:
        results = ()  # type: Tuple[Node, ...]
        text_ = text  # type: str
884
885
886
        for parser in self.parsers:
            node, text_ = parser(text_)
            if not node:
887
888
                return None, text
            results += (node,)
889
890
891
892
893
            if node.error_flag:
                break
        assert len(results) <= len(self.parsers)
        return Node(self, results), text_

894
895
896
    def __repr__(self):
        return " ".join(parser.repr for parser in self.parsers)

897
898
899
    def __add__(self, other: Parser) -> 'Series':
        other_parsers = cast('Series', other).parsers if isinstance(other, Series) else (other,)
        return Series(*(self.parsers + other_parsers))
900

901
902
903
    def __radd__(self, other: Parser) -> 'Series':
        other_parsers = cast('Series', other).parsers if isinstance(other, Series) else (other,)
        return Series(*(other_parsers + self.parsers))
904

905
906
907
908
    def __iadd__(self, other):
        other_parsers = cast('Series', other).parsers if isinstance(other, Series) else (other,)
        self.parsers += other_parsers
        return self
909

910
911

class Alternative(NaryOperator):
912
913
914
915
916
917
918
    """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
919
920
    # the order of the sub-expression matters!
    >>> number = RE('\d+') | RE('\d+') + RE('\.') + RE('\d+')
921
922
    >>> Grammar(number)("3.1416").show()
    '3 <<< Error on ".1416" | Parser stopped before end! trying to recover... >>> '
923
924

    # the most selective expression should be put first:
Eckhart Arnold's avatar
Eckhart Arnold committed
925
    >>> number = RE('\d+') + RE('\.') + RE('\d+') | RE('\d+')
926
    >>> Grammar(number)("3.1416").show()
Eckhart Arnold's avatar
Eckhart Arnold committed
927
    '3.1416'
928
    """
929
930

    def __init__(self, *parsers: Parser, name: str = '') -> None:
931
        super(Alternative, self).__init__(*parsers, name=name)
932
933
934
        assert len(self.parsers) >= 1
        assert all(not isinstance(p, Optional) for p in self.parsers)

935
    def __call__(self, text: str) -> Tuple[Node, str]:
936
937
938
939
940
941
        for parser in self.parsers:
            node, text_ = parser(text)
            if node:
                return Node(self, node), text_
        return None, text

942
    def __repr__(self):
943
        return '(' + ' | '.join(parser.repr for parser in self.parsers) + ')'
944

945
946
947
948
    def __or__(self, other: Parser) -> 'Alternative':
        other_parsers = cast('Alternative', other).parsers \
                        if isinstance(other, Alternative) else (other,)
        return Alternative(*(self.parsers + other_parsers))
949

950
951
952
953
    def __ror__(self, other: Parser) -> 'Alternative':
        other_parsers = cast('Alternative', other).parsers \
                        if isinstance(other, Alternative) else (other,)
        return Alternative(*(other_parsers + self.parsers))
954

955
956
957
958
959
    def __ior__(self, other):
        other_parsers = cast('Alternative', other.parsers) \
                        if isinstance(other, Alternative) else (other,)
        self.parsers += other_parsers
        return self
960

961
962
963
964
965
966
967
968
969

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


class FlowOperator(UnaryOperator):
970
    def __init__(self, parser: Parser, name: str = '') -> None:
971
        super(FlowOperator, self).__init__(parser, name)
972
973
974


class Required(FlowOperator):
975
    # Add constructor that checks for logical errors, like `Required(Optional(...))` constructs ?
976
    def __call__(self, text: str) -> Tuple[Node, str]:
977
978
979
980
981
982
983
984
985
986
987
        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_

988
    def __repr__(self):
989
        return '§' + self.parser.repr
990

991
992

class Lookahead(FlowOperator):
993
    def __init__(self, parser: Parser, name: str = '') -> None:
994
        super(Lookahead, self).__init__(parser, name)
995

996
    def __call__(self, text: str) -> Tuple[Node, str]: