parsers.py 46 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
from typing import Any, Callable, Dict, Iterator, List, Set, Tuple, Union
61

62
from DHParser.toolkit import is_logging, log_dir, logfile_basename, escape_re, sane_parser_name
Eckhart Arnold's avatar
Eckhart Arnold committed
63
64
from DHParser.syntaxtree import WHITESPACE_PTYPE, TOKEN_PTYPE, ZOMBIE_PARSER, ParserBase, \
    Node, TransformationFunc
65
from DHParser.toolkit import load_if_file, error_messages
Eckhart Arnold's avatar
Eckhart Arnold committed
66

67
68
__all__ = ['ScannerFunc',
           'HistoryRecord',
Eckhart Arnold's avatar
Eckhart Arnold committed
69
           'Parser',
70
           'Grammar',
Eckhart Arnold's avatar
Eckhart Arnold committed
71
72
73
74
75
76
77
78
79
80
           'RX_SCANNER_TOKEN',
           'BEGIN_SCANNER_TOKEN',
           'END_SCANNER_TOKEN',
           'make_token',
           'nil_scanner',
           'ScannerToken',
           'RegExp',
           'RE',
           'Token',
           'mixin_comment',
81
82
83
           # 'UnaryOperator',
           # 'NaryOperator',
           'Synonym',
Eckhart Arnold's avatar
Eckhart Arnold committed
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
           'Optional',
           'ZeroOrMore',
           'OneOrMore',
           'Sequence',
           'Alternative',
           'FlowOperator',
           'Required',
           'Lookahead',
           'NegativeLookahead',
           'Lookbehind',
           'NegativeLookbehind',
           'Capture',
           'Retrieve',
           'Pop',
           'Forward',
99
           'Compiler',
Eckhart Arnold's avatar
Eckhart Arnold committed
100
           'compile_source']
Eckhart Arnold's avatar
Eckhart Arnold committed
101

102

103
104
105
106
107
108
109
110
111
112
########################################################################
#
# Grammar and parsing infrastructure
#
########################################################################


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


113
LEFT_RECURSION_DEPTH = 10  # because of pythons recursion depth limit, this
Eckhart Arnold's avatar
Eckhart Arnold committed
114
                           # value ought not to be set too high
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
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"

    def __init__(self, call_stack, node, remaining):
        self.call_stack = call_stack
        self.node = node
        self.remaining = remaining

142
    def err_msg(self) -> str:
143
144
        return self.ERROR + ": " + "; ".join(self.node._errors).replace('\n', '\\')

145
    @property
146
    def stack(self) -> str:
147
148
149
        return "->".join(str(parser) for parser in self.call_stack)

    @property
150
    def status(self) -> str:
151
152
        return self.FAIL if self.node is None else \
            self.err_msg() if self.node._errors else self.MATCH
153
154

    @property
155
    def extent(self) -> Tuple[int, int]:
156
157
158
159
160
        return ((-self.remaining - self.node.len, -self.remaining) if self.node
                else (-self.remaining, None))


def add_parser_guard(parser_func):
161
    def guarded_call(parser: 'Parser', text: str) -> Tuple[Node, str]:
162
163
164
165
166
167
168
169
170
171
172
        try:
            location = len(text)
            # 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
Eckhart Arnold's avatar
Eckhart Arnold committed
173
            grammar = parser.grammar    # grammar may be 'None' for unconnected parsers!
174

175
            if grammar.history_tracking:
176
177
178
179
180
181
                grammar.call_stack.append(parser)
                grammar.moving_forward = True

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

182
            if grammar.history_tracking:
183
184
                # don't track returning parsers except in case an error has occurred
                if grammar.moving_forward or (node and node._errors):
185
186
187
                    grammar.moving_forward = False
                    record = HistoryRecord(grammar.call_stack.copy(), node, len(rest))
                    grammar.history.append(record)
188
                    # print(record.stack, record.status, rest[:20].replace('\n', '|'))
189
190
191
192
193
194
                grammar.call_stack.pop()

            if node is not None:
                # in case of a recursive call saves the result of the first
                # (or left-most) call that matches
                parser.visited[location] = (node, rest)
195
                grammar.last_node = node   # store last node for Lookbehind operator
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
            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

        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!")
            node.error_flag = True
            rest = ''

        return node, rest

    return guarded_call


Eckhart Arnold's avatar
Eckhart Arnold committed
215
class ParserMetaClass(abc.ABCMeta):
216
217
218
219
220
221
222
223
224
225
    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
226
class Parser(ParserBase, metaclass=ParserMetaClass):
227
228
    ApplyFunc = Callable[['Parser'], None]

229
230
    def __init__(self, name: str = '') -> None:
        # assert isinstance(name, str), str(name)
Eckhart Arnold's avatar
Eckhart Arnold committed
231
        super(Parser, self).__init__(name)
232
        self._grammar = None  # type: 'Grammar'
233
234
        self.reset()

Eckhart Arnold's avatar
Eckhart Arnold committed
235
    def __deepcopy__(self, memo):
236
237
        return self.__class__(self.name)

238
    def reset(self):
239
240
241
        self.visited = dict()  # type: Dict[int, Tuple[Node, str]]
        self.recursion_counter = dict()  # type: Dict[int, int]
        self.cycle_detection = set()  # type: Set[Callable]
242
        return self
243

244
    def __call__(self, text: str) -> Tuple[Node, str]:
245
246
247
        return None, text  # default behaviour: don't match

    def __str__(self):
248
        return self.name or self.ptype
249

250
251
252
253
254
255
    def __add__(self, other):
        return Sequence(self, other)

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

256
    @property
257
    def grammar(self) -> 'Grammar':
258
259
260
        return self._grammar

    @grammar.setter
261
262
    def grammar(self, grammar: 'Grammar'):
        self._grammar = grammar
263
264
265
266
267
        self._grammar_assigned_notifier()

    def _grammar_assigned_notifier(self):
        pass

268
    def apply(self, func: ApplyFunc):
269
        """Applies function `func(parser)` recursively to this parser and all
Eckhart Arnold's avatar
Eckhart Arnold committed
270
        descendants of the tree of parsers. The same function can never
271
272
273
274
275
276
277
278
279
280
        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


281
class Grammar:
282
283
    root__ = None  # type: Union[Parser, None]
    # root__ must be overwritten with the root-parser by grammar subclass
Eckhart Arnold's avatar
Eckhart Arnold committed
284
    parser_initialization__ = "pending"  # type: str
285
286
287
288
289
290

    @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.
291
            class Grammar(Grammar):
292
293
294
                ...
                symbol = RE('(?!\\d)\\w+')
        After the call of this method symbol.name == "symbol"
295
296
297
298
        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()``
299
300
301
302

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

303
304
305
306
307
        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. 
308
309
310
311
312
        """
        if cls.parser_initialization__ == "done":
            return
        cdict = cls.__dict__
        for entry, parser in cdict.items():
313
            if isinstance(parser, Parser) and sane_parser_name(entry):
314
                if not parser.name:
315
                    parser.name = entry
316
                if (isinstance(parser, Forward) and (not parser.parser.name)):
317
318
319
                    parser.parser.name = entry
        cls.parser_initialization__ = "done"

Eckhart Arnold's avatar
Eckhart Arnold committed
320
321
322
    def __init__(self, root: Parser=None) -> None:
        # if not hasattr(self.__class__, 'parser_initialization__'):
        #     self.__class__.parser_initialization__ = "pending"
323
324
325
326
        if not hasattr(self.__class__, 'wspL__'):
            self.wspL__ = ''
        if not hasattr(self.__class__, 'wspR__'):
            self.wspR__ = ''
Eckhart Arnold's avatar
Eckhart Arnold committed
327
        self.all_parsers = set()  # type: Set[Parser]
328
        self.dirty_flag = False
329
        self.history_tracking = False
330
        self._reset()
Eckhart Arnold's avatar
Eckhart Arnold committed
331
332
333
334
335
336
        # 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.
337
        self.root__ = root if root else copy.deepcopy(self.__class__.root__)
338
        if self.wspL__:
Eckhart Arnold's avatar
Eckhart Arnold committed
339
            self.wsp_left_parser__ = Whitespace(self.wspL__)  # type: ParserBase
340
            self.wsp_left_parser__.grammar = self
341
            self.all_parsers.add(self.wsp_left_parser__)  # don't you forget about me...
342
343
344
        else:
            self.wsp_left_parser__ = ZOMBIE_PARSER
        if self.wspR__:
Eckhart Arnold's avatar
Eckhart Arnold committed
345
            self.wsp_right_parser__ = Whitespace(self.wspR__)  # type: ParserBase
346
            self.wsp_right_parser__.grammar = self
347
            self.all_parsers.add(self.wsp_right_parser__)  # don't you forget about me...
348
349
350
351
        else:
            self.wsp_right_parser__ = ZOMBIE_PARSER
        self.root__.apply(self._add_parser)

352
    def __getitem__(self, key):
353
354
355
356
357
        try:
            return self.__dict__[key]
        except KeyError:
            parser = getattr(self, key, None)
            if parser:
Eckhart Arnold's avatar
Eckhart Arnold committed
358
359
360
                # if toolkit.warnings():
                #     raise KeyError(('Parser "%s" inaccesible, because it is not connected '
                #                     'to the root parser "%s" !') % (key,  self.root__.name))
361
                # print('Parser "%s" not connected to root parser.' % key)
Eckhart Arnold's avatar
Eckhart Arnold committed
362
363
                # add parser to grammar object on the fly...
                setattr(self, key, copy.deepcopy(parser))
364
                self[key].apply(self._add_parser)
Eckhart Arnold's avatar
Eckhart Arnold committed
365
                return self[key]
366
            raise KeyError('Unknown parser "%s" !' % key)
367

368
    def _reset(self):
369
370
371
372
373
374
375
376
377
378
379
380
        # variables stored and recalled by Capture and Retrieve parsers
        self.variables = dict()  # type: Dict[str, List[str]]
        self.document = ""  # type: str
        # previously parsed node, needed by Lookbehind parser
        self.last_node = None  # type: Node
        # support for call stack tracing
        self.call_stack = []  # type: List[Parser]
        # snapshots of call stacks
        self.history = []  # type: List[HistoryRecord]
        # also needed for call stack tracing
        self.moving_forward = True

381
    # TODO: Either make sure not to miss out unconnected parsers or raise an error! Actually, the EBNF-Compiler should keep track of this!
Eckhart Arnold's avatar
Eckhart Arnold committed
382
    def _add_parser(self, parser: Parser) -> None:
Eckhart Arnold's avatar
Eckhart Arnold committed
383
        """Adds the particular copy of the parser object to this
384
        particular instance of Grammar.
385
        """
386
387
        if parser.name:
            setattr(self, parser.name, parser)
388
389
390
        self.all_parsers.add(parser)
        parser.grammar = self

Eckhart Arnold's avatar
Eckhart Arnold committed
391
    def __call__(self, document: str, start_parser="root__") -> Node:
392
393
394
395
        """Parses a document with with parser-combinators.

        Args:
            document (str): The source text to be parsed.
396
397
398
            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.)
399
400
401
        Returns:
            Node: The root node ot the parse tree.
        """
402
        # assert isinstance(document, str), type(document)
403
404
405
406
407
408
409
410
        if self.root__ is None:
            raise NotImplementedError()
        if self.dirty_flag:
            self._reset()
            for parser in self.all_parsers:
                parser.reset()
        else:
            self.dirty_flag = True
411
        self.history_tracking = is_logging()
412
        self.document = document
Eckhart Arnold's avatar
Eckhart Arnold committed
413
        parser = self[start_parser] if isinstance(start_parser, str) else start_parser
414
415
        assert parser.grammar == self, "Cannot run parsers from a different grammar object!" \
                                       " %s vs. %s" % (str(self), str(parser.grammar))
416
        stitches = []  # type: List[Node]
417
        rest = document
418
419
        if not rest:
            result, ignore = parser(rest)
420
421
422
423
424
425
426
427
428
429
        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" + \
430
431
432
                                (("! trying to recover" +
                                  (" but stopping history recording at this point."
                                   if self.history_tracking else "..."))
433
434
435
436
                                 if len(stitches) < MAX_DROPOUTS
                                 else " too often! Terminating parser.")
                stitches.append(Node(None, skip))
                stitches[-1].add_error(error_msg)
437
                if self.history_tracking:
438
439
440
441
442
443
444
                    # 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
                    for record in self.history:
                        if record.node and record.node._pos < 0:
                            record.node.pos = 0
445
446
447
                    record = HistoryRecord(self.call_stack.copy(), stitches[-1], len(rest))
                    self.history.append(record)
                    self.history_tracking = False
448
449
450
451
452
453
454
        if stitches:
            if rest:
                stitches.append(Node(None, rest))
            result = Node(None, tuple(stitches))
        result.pos = 0  # calculate all positions
        return result

Eckhart Arnold's avatar
Eckhart Arnold committed
455
    def log_parsing_history(self, log_file_name: str='') -> None:
456
457
458
459
460
461
        """Writes a log of the parsing history of the most recently parsed
        document. 
        """
        def prepare_line(record):
            excerpt = self.document.__getitem__(slice(*record.extent))[:25].replace('\n', '\\n')
            excerpt = "'%s'" % excerpt if len(excerpt) < 25 else "'%s...'" % excerpt
462
            return record.stack, record.status, excerpt
463
464

        def write_log(history, log_name):
Eckhart Arnold's avatar
Eckhart Arnold committed
465
            path = os.path.join(log_dir(), log_name + "_parser.log")
466
467
468
469
470
471
            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
472
473
474
475
476
477
478
        if not log_file_name:
            name = self.__class__.__name__
            log_file_name = name[:-7] if name.lower().endswith('grammar') else name
        full_history, match_history, errors_only = [], [], []
        for record in self.history:
            line = ";  ".join(prepare_line(record))
            full_history.append(line)
479
            if record.node and record.node.parser.ptype != WHITESPACE_PTYPE:
Eckhart Arnold's avatar
Eckhart Arnold committed
480
481
482
483
484
485
                match_history.append(line)
                if record.node.errors:
                    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')
486
487


Eckhart Arnold's avatar
Eckhart Arnold committed
488
489
def dsl_error_msg(parser: Parser, error_str: str) -> str:
    """Returns an error message for errors in the parser configuration,
490
491
492
    e.g. errors that result in infinite loops.

    Args:
Eckhart Arnold's avatar
Eckhart Arnold committed
493
        parser (Parser):  The parser where the error was noticed. Note
494
            that this is not necessarily the parser that caused the
Eckhart Arnold's avatar
Eckhart Arnold committed
495
            error but only where the error became aparent.
496
497
498
499
500
        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.
    """
501
    msg = ["DSL parser specification error:", error_str, 'Caught by parser "%s".' % str(parser)]
502
503
504
505
506
507
508
    if parser.grammar.history:
        msg.extend(["\nCall stack:", parser.grammar.history[-1].stack])
    else:
        msg.extend(["\nEnable history tracking in Grammar object to display call stack."])
    return " ".join(msg)


509
510
511
512
513
514
515
516
517
518
519
520
########################################################################
#
# Token and Regular Expression parser classes (i.e. leaf classes)
#
########################################################################


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


521
def make_token(token: str, argument: str = '') -> str:
522
523
524
525
526
527
528
529
530
531
532
533
534
    """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


535
def nil_scanner(text: str) -> str:
Eckhart Arnold's avatar
Eckhart Arnold committed
536
    return text
537
538
539


class ScannerToken(Parser):
540
541
542
543
544
545
546
547
548
549
    """
    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.
    """
550
551
552

    def __init__(self, scanner_token: str) -> None:
        assert scanner_token and scanner_token.isupper()
553
        assert RX_SCANNER_TOKEN.match(scanner_token)
554
        super(ScannerToken, self).__init__(scanner_token)
555

556
    def __call__(self, text: str) -> Tuple[Node, str]:
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
        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. '
                    '(Most likely due to a scanner bug!)')
                return node, text[1:]
            elif end == 0:
                node = Node(self, '').add_error(
                    'Scanner token cannot have zero length. '
                    '(Most likely due to a scanner bug!)')
                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. '
                    '(Most likely due to a scanner bug!)')
                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):
583
584
    """
    Regular expression parser.
585
586
587
588
589
590
    
    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.
    """
591
592

    def __init__(self, regexp, name: str = '') -> None:
593
        super(RegExp, self).__init__(name)
594
595
596
        self.regexp = re.compile(regexp) if isinstance(regexp, str) else regexp

    def __deepcopy__(self, memo):
Eckhart Arnold's avatar
Eckhart Arnold committed
597
        # `regex` supports deep copies, but not `re`
598
        try:
Eckhart Arnold's avatar
Eckhart Arnold committed
599
            regexp = copy.deepcopy(self.regexp, memo)
600
601
        except TypeError:
            regexp = self.regexp.pattern
602
        return RegExp(regexp, self.name)
603

604
    def __call__(self, text: str) -> Tuple[Node, str]:
605
606
607
608
609
610
        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

611
612
613
    def __str__(self):
        return self.name or self.ptype + ' /%s/' % self.regexp.pattern

614

615
616
class Whitespace(RegExp):
    assert WHITESPACE_PTYPE == ":Whitespace"
617
618


619
620
class RE(Parser):
    """Regular Expressions with optional leading or trailing whitespace.
621
622
623
624
625
626
627
628
629
630
    
    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.
631
    """
632
    def __init__(self, regexp, wL=None, wR=None, name=''):
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
        """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.
        """
648
        super(RE, self).__init__(name)
649
650
        self.wL = wL
        self.wR = wR
651
652
        self.wspLeft = Whitespace(wL) if wL else ZOMBIE_PARSER
        self.wspRight = Whitespace(wR) if wR else ZOMBIE_PARSER
653
654
        self.main = RegExp(regexp)

Eckhart Arnold's avatar
Eckhart Arnold committed
655
656
657
658
659
    def __deepcopy__(self, memo={}):
        try:
            regexp = copy.deepcopy(self.main.regexp, memo)
        except TypeError:
            regexp = self.main.regexp.pattern
660
        return self.__class__(regexp, self.wL, self.wR, self.name)
Eckhart Arnold's avatar
Eckhart Arnold committed
661

662
    def __call__(self, text: str) -> Tuple[Node, str]:
663
664
665
666
667
668
669
670
671
672
673
        # assert self.main.regexp.pattern != "@"
        t = text
        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

674
    def __str__(self):
675
        if self.ptype == TOKEN_PTYPE:
676
677
678
            return 'Token "%s"' % self.main.regexp.pattern.replace('\\', '')
        return self.name or ('RE ' + ('~' if self.wL else '')
                             + '/%s/' % self.main.regexp.pattern + ('~' if self.wR else ''))
679
680
681

    def _grammar_assigned_notifier(self):
        if self.grammar:
682
            # use default whitespace parsers if not otherwise specified
683
684
685
686
687
            if self.wL is None:
                self.wspLeft = self.grammar.wsp_left_parser__
            if self.wR is None:
                self.wspRight = self.grammar.wsp_right_parser__

688
    def apply(self, func: Parser.ApplyFunc):
689
690
691
692
693
694
695
696
        if super(RE, self).apply(func):
            if self.wL:
                self.wspLeft.apply(func)
            if self.wR:
                self.wspRight.apply(func)
            self.main.apply(func)


697
698
699
class Token(RE):
    assert TOKEN_PTYPE == ":Token"

700
    def __init__(self, token: str, wL=None, wR=None, name: str = '') -> None:
701
702
703
704
705
        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)
706
707


708
def mixin_comment(whitespace: str, comment: str) -> str:
709
710
711
712
713
714
715
    """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).
716
717
718
719
720
721
722
723
724
725
726
727
728
    """
    wspc = '(?:' + whitespace + '(?:' + comment + whitespace + ')*)'
    return wspc


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


class UnaryOperator(Parser):
729
    def __init__(self, parser: Parser, name: str = '') -> None:
730
        super(UnaryOperator, self).__init__(name)
731
        # assert isinstance(parser, Parser)
732
        self.parser = parser  # type: Parser
733

Eckhart Arnold's avatar
Eckhart Arnold committed
734
735
    def __deepcopy__(self, memo):
        parser = copy.deepcopy(self.parser, memo)
736
        return self.__class__(parser, self.name)
Eckhart Arnold's avatar
Eckhart Arnold committed
737

738
    def apply(self, func: Parser.ApplyFunc):
739
740
741
742
743
        if super(UnaryOperator, self).apply(func):
            self.parser.apply(func)


class NaryOperator(Parser):
744
    def __init__(self, *parsers: Parser, name: str = '') -> None:
745
        super(NaryOperator, self).__init__(name)
746
        # assert all([isinstance(parser, Parser) for parser in parsers]), str(parsers)
747
        self.parsers = parsers  # type: Container  ## [Parser]
748

Eckhart Arnold's avatar
Eckhart Arnold committed
749
750
    def __deepcopy__(self, memo):
        parsers = copy.deepcopy(self.parsers, memo)
751
        return self.__class__(*parsers, name=self.name)
Eckhart Arnold's avatar
Eckhart Arnold committed
752

753
    def apply(self, func: Parser.ApplyFunc):
754
755
756
757
758
        if super(NaryOperator, self).apply(func):
            for parser in self.parsers:
                parser.apply(func)


759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
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


777
class Optional(UnaryOperator):
778
    def __init__(self, parser: Parser, name: str = '') -> None:
779
        super(Optional, self).__init__(parser, name)
780
        # assert isinstance(parser, Parser)
781
        assert not isinstance(parser, Optional), \
782
            "Redundant nesting of options: %s(%s)" % \
783
784
            (str(name), str(parser.name))
        assert not isinstance(parser, Required), \
785
            "Nesting options with required elements is contradictory: " \
786
787
            "%s(%s)" % (str(name), str(parser.name))

788
    def __call__(self, text: str) -> Tuple[Node, str]:
789
790
791
792
793
794
795
        node, text = self.parser(text)
        if node:
            return Node(self, node), text
        return Node(self, ()), text


class ZeroOrMore(Optional):
796
797
    def __call__(self, text: str) -> Tuple[Node, str]:
        results = ()  # type: Tuple[Node, ...]
798
799
800
        n = len(text) + 1
        while text and len(text) < n:
            n = len(text)
801
802
803
            node, text = self.parser(text)
            if not node:
                break
804
            if len(text) == n:
di68kap's avatar
di68kap committed
805
                node.add_error(dsl_error_msg(self, 'Infinite Loop detected.'))
806
807
808
809
810
            results += (node,)
        return Node(self, results), text


class OneOrMore(UnaryOperator):
811
    def __init__(self, parser: Parser, name: str = '') -> None:
812
        super(OneOrMore, self).__init__(parser, name)
813
814
815
816
        assert not isinstance(parser, Optional), \
            "Use ZeroOrMore instead of nesting OneOrMore and Optional: " \
            "%s(%s)" % (str(name), str(parser.name))

817
818
819
    def __call__(self, text: str) -> Tuple[Node, str]:
        results = ()  # type: Tuple[Node, ...]
        text_ = text  # type: str
820
821
822
        n = len(text) + 1
        while text_ and len(text_) < n:
            n = len(text_)
823
824
825
            node, text_ = self.parser(text_)
            if not node:
                break
826
            if len(text_) == n:
827
                node.add_error(dsl_error_msg(self, 'Infinite Loop detected.'))
828
829
830
831
832
833
834
            results += (node,)
        if results == ():
            return None, text
        return Node(self, results), text_


class Sequence(NaryOperator):
835
    def __init__(self, *parsers: Parser, name: str = '') -> None:
836
        super(Sequence, self).__init__(*parsers, name=name)
837
838
        assert len(self.parsers) >= 1

839
840
841
    def __call__(self, text: str) -> Tuple[Node, str]:
        results = ()  # type: Tuple[Node, ...]
        text_ = text  # type: str
842
843
844
845
846
847
848
849
850
851
852
        for parser in self.parsers:
            node, text_ = parser(text_)
            if not node:
                return node, text
            if node.result:  # Nodes with zero-length result are silently omitted
                results += (node,)
            if node.error_flag:
                break
        assert len(results) <= len(self.parsers)
        return Node(self, results), text_

853
    def __add__(self, other: 'Sequence') -> 'Sequence':
854
855
        return Sequence(*self.parsers, other)

856
    def __radd__(self, other: 'Sequence') -> 'Sequence':
857
858
859
860
861
862
863
864
865
        return Sequence(other, *self.parsers)

        # def __iadd__(self, other):
        #     if isinstance(other, Sequence):
        #         self.parsers = self.parsers + other.parsers
        #     else:
        #         self.parsers = self.parsers + (other,)
        #     return self

866
867

class Alternative(NaryOperator):
868
869
870
871
872
873
874
    """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
875
876
    # the order of the sub-expression matters!
    >>> number = RE('\d+') | RE('\d+') + RE('\.') + RE('\d+')
877
878
    >>> Grammar(number)("3.1416").show()
    '3 <<< Error on ".1416" | Parser stopped before end! trying to recover... >>> '
879
880

    # the most selective expression should be put first:
Eckhart Arnold's avatar
Eckhart Arnold committed
881
    >>> number = RE('\d+') + RE('\.') + RE('\d+') | RE('\d+')
882
    >>> Grammar(number)("3.1416").show()
Eckhart Arnold's avatar
Eckhart Arnold committed
883
    '3.1416'
884
    """
885
886

    def __init__(self, *parsers: Parser, name: str = '') -> None:
887
        super(Alternative, self).__init__(*parsers, name=name)
888
889
890
        assert len(self.parsers) >= 1
        assert all(not isinstance(p, Optional) for p in self.parsers)

891
    def __call__(self, text: str) -> Tuple[Node, str]:
892
893
894
895
896
897
        for parser in self.parsers:
            node, text_ = parser(text)
            if node:
                return Node(self, node), text_
        return None, text

898
899
900
901
902
903
904
905
906
907
908
909
910
    def __or__(self, other):
        return Alternative(*self.parsers, other)

    def __ror__(self, other):
        return Alternative(other, *self.parsers)

        # def __ior__(self, other):
        #     if isinstance(other, Sequence):
        #         self.parsers = self.parsers + other.parsers
        #     else:
        #         self.parsers = self.parsers + (other,)
        #     return self

911
912
913
914
915
916
917
918
919

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


class FlowOperator(UnaryOperator):
920
    def __init__(self, parser: Parser, name: str = '') -> None:
921
        super(FlowOperator, self).__init__(parser, name)
922
923
924


class Required(FlowOperator):
925
    # Add constructor that checks for logical errors, like `Required(Optional(...))` constructs ?
926
    def __call__(self, text: str) -> Tuple[Node, str]:
927
928
929
930
931
932
933
934
935
936
937
938
939
        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_


class Lookahead(FlowOperator):
940
    def __init__(self, parser: Parser, name: str = '') -> None:
941
        super(Lookahead, self).__init__(parser, name)
942

943
    def __call__(self, text: str) -> Tuple[Node, str]:
944
945
946
947
948
949
        node, text_ = self.parser(text)
        if self.sign(node is not None):
            return Node(self, ''), text
        else:
            return None, text

950
    def sign(self, bool_value) -> bool:
951
952
953
954
        return bool_value


class NegativeLookahead(Lookahead):
955
    def sign(self, bool_value) -> bool:
956
957
958
        return not bool_value


959
def iter_right_branch(node) -> Iterator[Node]:
960
961
962
963
964
965
966
967
968
969
970
971
    """Iterates over the right branch of `node` starting with node itself.
    Iteration is stopped if either there are no child nodes any more or
    if the parser of a node is a Lookahead parser. (Reason is: Since
    lookahead nodes do not advance the parser, it does not make sense
    to look back to them.)
    """
    while node and not isinstance(node.parser, Lookahead):  # the second condition should not be necessary
        yield node  # for well-formed EBNF code
        node = node.children[-1] if node.children else None


class Lookbehind(FlowOperator):
972
    def __init__(self, parser: Parser, name: str = '') -> None:
973
        super(Lookbehind, self).__init__(parser, name)
974
975
        print("WARNING: Lookbehind Operator is experimental!")

976
    def __call__(self, text: str) -> Tuple[Node, str]:
977
978
979
980
981
982
983
984
        if isinstance(self.grammar.last_node, Lookahead):
            return Node(self, '').add_error('Lookbehind right after Lookahead '
                                            'does not make sense!'), text
        if self.sign(self.condition()):
            return Node(self, ''), text
        else:
            return None, text

985
    def sign(self, bool_value) -> bool:
986
987
988
989
990
991
992
993
994
995
996
997
998
999
        return bool_value

    def condition(self):
        node = None
        for node in iter_right_branch(self.grammar.last_node):
            if node.parser.name == self.parser.name:
                return True
        if node and isinstance(self.parser, RegExp) and \
                self.parser.regexp.match(str(node)):  # Is there really a use case for this?
            return True
        return False


class NegativeLookbehind(Lookbehind):
1000
    def sign(self, bool_value) -> bool:
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
        return not bool_value


########################################################################
#
# Capture and Retrieve operators (for passing variables in the parser)
#
########################################################################


class Capture(UnaryOperator):
1012
    def __init__(self, parser: Parser, name: str = '') -> None:
1013
        super(Capture, self).__init__(parser, name)
1014

1015
    def __call__(self, text: str) -> Tuple[Node, str]:
1016
1017
1018
1019
        node, text = self.parser(text)
        if node:
            stack = self.grammar.variables.setdefault(self.name, [])
            stack.append(str(node))
1020
1021
1022
            return Node(self, node), text
        else:
            return None, text
1023
1024


1025
1026
1027
1028
RetrieveFilter = Callable[[List[str]], str]


def nop_filter(stack: List[str]) -> str:
1029
1030
1031
    return stack[-1]


1032
def counterpart_filter(stack: List[str]) -> str:
1033
1034
1035
1036
    value = stack[-1]
    return value.replace("(", ")").replace("[", "]").replace("{", "}").replace(">", "<")


1037
def accumulating_filter(stack: List[str]) -> str:
1038
1039
1040
    return "".join(stack)


1041
class Retrieve(Parser):
1042
    def __init__(self, symbol: Parser, filter: RetrieveFilter = None, name: str = '') -> None:
1043
1044
        if not name:
            name = symbol.name
1045
        super(Retrieve, self).__init__(name)
1046
        self.symbol = symbol
1047
        self.filter = filter if filter else nop_filter
1048

Eckhart Arnold's avatar
Eckhart Arnold committed
1049
    def __deepcopy__(self, memo):
1050
        return self.__class__(self.symbol, self.filter, self.name)
Eckhart Arnold's avatar
Eckhart Arnold committed
1051

1052
    def __call__(self, text: str) -> Tuple[Node, str]:
1053
1054
        try:
            stack = self.grammar.variables[self.symbol.name]
1055
            value = self.filter(stack)
1056
1057
1058
1059
            self.pick_value(stack)
        except (KeyError, IndexError):
            return Node(self, '').add_error(dsl_error_msg(self,
                "%s undefined or exhausted" % self.symbol.name)), text
1060
1061
1062
1063
1064
        if text.startswith(value):
            return Node(self, value), text[len(value):]
        else:
            return None, text

1065
    def pick_value(self, stack: List[str]) -> str:
1066
1067
1068
1069
        return stack[-1]


class Pop(Retrieve):
1070
    def pick_value(self, stack: List[str]) -> str:
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
        return stack.pop()


########################################################################
#
# Forward class (for recursive symbols)
#
########################################################################


class Forward(Parser):
    def __init__(self):
        Parser.__init__(self)
        self.parser = None
1085
        self.cycle_reached = False
1086

Eckhart Arnold's avatar
Eckhart Arnold committed
1087
1088
1089
1090
1091
1092
1093
    def __deepcopy__(self, memo):
        duplicate = self.__class__()
        memo[id(self)] = duplicate
        parser = copy.deepcopy(self.parser, memo)
        duplicate.set(parser)
        return duplicate

1094
    def __call__(self, text: str) -> Tuple[Node, str]:
1095
1096
        return self.parser(text)

1097
1098
1099
1100
1101
1102
1103
1104
    def __str__(self):
        if self.cycle_reached:
            return "..."
        else:
            self.cycle_reached = True
            s = str(self.parser)
            self.cycle_reached = False
            return s
1105

1106
1107
    def set(self, parser: Parser):
        # assert isinstance(parser, Parser)
1108
        self.name = parser.name  # redundant, see Grammar-constructor
1109
1110
        self.parser = parser

1111
    def apply(self, func: Parser.ApplyFunc):
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
        if super(Forward, self).apply(func):
            assert not self.visited
            self.parser.apply(func)


#######################################################################
#
# Syntax driven compilation support
#
#######################################################################


1124
class Compiler:
Eckhart Arnold's avatar
Eckhart Arnold committed
1125
    def __init__(self, grammar_name="", grammar_source=""):
1126
        self.dirty_flag = False
Eckhart Arnold's avatar
Eckhart Arnold committed
1127
        self.set_grammar_name(grammar_name, grammar_source)
1128
1129
1130
1131

    def _reset(self):
        pass

1132
    def __call__(self, node: Node) -> Any:
1133
        """Compiles the abstract syntax tree with the root ``node``.
1134
        
Eckhart Arnold's avatar
Eckhart Arnold committed
1135
        It's called `compile_ast`` to avoid confusion with the 
1136
1137
        ``_compile`` that is called from within the local node 
        compiler methods.
1138
1139
1140
1141
1142
1143
1144
        """
        if self.dirty_flag:
            self._reset()
        else:
            self.dirty_flag = True
        return self._compile(node)

Eckhart Arnold's avatar
Eckhart Arnold committed
1145
1146
1147
1148
1149
1150
1151
    def set_grammar_name(self, grammar_name, grammar_source):
        assert grammar_name == "" or re.match('\w+\Z', grammar_name)
        if not grammar_name and re.fullmatch(r'[\w/:\\]+', grammar_source):
            grammar_name = os.path.splitext(os.path.basename(grammar_source))[0]
        self.grammar_name = grammar_name
        self.grammar_source = load_if_file(grammar_source)

1152
    @staticmethod
1153
    def derive_method_name(node_name: str) -> str:
1154
        """Returns the method name for ``node_name``, e.g.
1155
        >>> Compiler.method_name('expression')
1156
1157
1158
1159
        'on_expression'
        """
        return 'on_' + node_name

1160
    def _compile(self, node: Node) -> Any:
1161
1162
1163
1164
1165
        """Calls the compilation method for the given node and returns
         the result of the compilation.
        
        The method's name is dreived from either the node's parser 
        name or, if the parser is anonymous, the node's parser's class