2.12.2021, 9:00 - 11:00: Due to updates GitLab may be unavailable for some minutes between 09:00 and 11:00.

parsers.py 45.4 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()
331
        self._assign_parser_names()  # prepare class first
332
        self.root__ = root if root else copy.deepcopy(self.__class__.root__)
333
        if self.wspL__:
Eckhart Arnold's avatar
Eckhart Arnold committed
334
            self.wsp_left_parser__ = Whitespace(self.wspL__)  # type: ParserBase
335
            self.wsp_left_parser__.grammar = self
336
            self.all_parsers.add(self.wsp_left_parser__)  # don't you forget about me...
337
338
339
        else:
            self.wsp_left_parser__ = ZOMBIE_PARSER
        if self.wspR__:
Eckhart Arnold's avatar
Eckhart Arnold committed
340
            self.wsp_right_parser__ = Whitespace(self.wspR__)  # type: ParserBase
341
            self.wsp_right_parser__.grammar = self
342
            self.all_parsers.add(self.wsp_right_parser__)  # don't you forget about me...
343
344
345
346
        else:
            self.wsp_right_parser__ = ZOMBIE_PARSER
        self.root__.apply(self._add_parser)

347
    def __getitem__(self, key):
348
349
350
351
352
353
354
355
        try:
            return self.__dict__[key]
        except KeyError:
            parser = getattr(self, key, None)
            if parser:
                raise KeyError(('Parser "%s" inaccesible, because it is not connected '
                                'to the root parser "%s" !') % (key,  self.root__.name))
            raise KeyError('Unknown parser "%s" !' % key)
356

357
    def _reset(self):
358
359
360
361
362
363
364
365
366
367
368
369
        # 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

370
    # 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
371
    def _add_parser(self, parser: Parser) -> None:
372
        """Adds the copy of the classes parser object to this
373
        particular instance of Grammar.
374
        """
375
376
        if parser.name:
            setattr(self, parser.name, parser)
377
378
379
        self.all_parsers.add(parser)
        parser.grammar = self

Eckhart Arnold's avatar
Eckhart Arnold committed
380
    def __call__(self, document: str, start_parser="root__") -> Node:
381
382
383
384
        """Parses a document with with parser-combinators.

        Args:
            document (str): The source text to be parsed.
385
386
387
            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.)
388
389
390
        Returns:
            Node: The root node ot the parse tree.
        """
391
        # assert isinstance(document, str), type(document)
392
393
394
395
396
397
398
399
        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
400
        self.history_tracking = is_logging()
401
        self.document = document
Eckhart Arnold's avatar
Eckhart Arnold committed
402
        parser = self[start_parser] if isinstance(start_parser, str) else start_parser
403
404
        assert parser.grammar == self, "Cannot run parsers from a different grammar object!" \
                                       " %s vs. %s" % (str(self), str(parser.grammar))
405
        stitches = []  # type: List[Node]
406
        rest = document
407
408
        if not rest:
            result, ignore = parser(rest)
409
410
411
412
413
414
415
416
417
418
        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" + \
419
420
421
                                (("! trying to recover" +
                                  (" but stopping history recording at this point."
                                   if self.history_tracking else "..."))
422
423
424
425
                                 if len(stitches) < MAX_DROPOUTS
                                 else " too often! Terminating parser.")
                stitches.append(Node(None, skip))
                stitches[-1].add_error(error_msg)
426
                if self.history_tracking:
427
428
429
430
431
432
433
                    # 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
434
435
436
                    record = HistoryRecord(self.call_stack.copy(), stitches[-1], len(rest))
                    self.history.append(record)
                    self.history_tracking = False
437
438
439
440
441
442
443
        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
444
    def log_parsing_history(self, log_file_name: str='') -> None:
445
446
447
448
449
450
        """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
451
            return record.stack, record.status, excerpt
452
453

        def write_log(history, log_name):
Eckhart Arnold's avatar
Eckhart Arnold committed
454
            path = os.path.join(log_dir(), log_name + "_parser.log")
455
456
457
458
459
460
            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
461
462
463
464
465
466
467
        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)
468
            if record.node and record.node.parser.ptype != WHITESPACE_PTYPE:
Eckhart Arnold's avatar
Eckhart Arnold committed
469
470
471
472
473
474
                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')
475
476


Eckhart Arnold's avatar
Eckhart Arnold committed
477
478
def dsl_error_msg(parser: Parser, error_str: str) -> str:
    """Returns an error message for errors in the parser configuration,
479
480
481
    e.g. errors that result in infinite loops.

    Args:
Eckhart Arnold's avatar
Eckhart Arnold committed
482
        parser (Parser):  The parser where the error was noticed. Note
483
            that this is not necessarily the parser that caused the
Eckhart Arnold's avatar
Eckhart Arnold committed
484
            error but only where the error became aparent.
485
486
487
488
489
        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.
    """
490
    msg = ["DSL parser specification error:", error_str, 'Caught by parser "%s".' % str(parser)]
491
492
493
494
495
496
497
    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)


498
499
500
501
502
503
504
505
506
507
508
509
########################################################################
#
# Token and Regular Expression parser classes (i.e. leaf classes)
#
########################################################################


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


510
def make_token(token: str, argument: str = '') -> str:
511
512
513
514
515
516
517
518
519
520
521
522
523
    """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


524
def nil_scanner(text: str) -> str:
Eckhart Arnold's avatar
Eckhart Arnold committed
525
    return text
526
527
528


class ScannerToken(Parser):
529
530
531
532
533
534
535
536
537
538
    """
    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.
    """
539
540
541

    def __init__(self, scanner_token: str) -> None:
        assert scanner_token and scanner_token.isupper()
542
        assert RX_SCANNER_TOKEN.match(scanner_token)
543
        super(ScannerToken, self).__init__(scanner_token)
544

545
    def __call__(self, text: str) -> Tuple[Node, str]:
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
        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):
572
573
    """
    Regular expression parser.
574
575
576
577
578
579
    
    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.
    """
580
581

    def __init__(self, regexp, name: str = '') -> None:
582
        super(RegExp, self).__init__(name)
583
584
585
        self.regexp = re.compile(regexp) if isinstance(regexp, str) else regexp

    def __deepcopy__(self, memo):
Eckhart Arnold's avatar
Eckhart Arnold committed
586
        # `regex` supports deep copies, but not `re`
587
        try:
Eckhart Arnold's avatar
Eckhart Arnold committed
588
            regexp = copy.deepcopy(self.regexp, memo)
589
590
        except TypeError:
            regexp = self.regexp.pattern
591
        return RegExp(regexp, self.name)
592

593
    def __call__(self, text: str) -> Tuple[Node, str]:
594
595
596
597
598
599
        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

600
601
602
    def __str__(self):
        return self.name or self.ptype + ' /%s/' % self.regexp.pattern

603

604
605
class Whitespace(RegExp):
    assert WHITESPACE_PTYPE == ":Whitespace"
606
607


608
609
class RE(Parser):
    """Regular Expressions with optional leading or trailing whitespace.
610
611
612
613
614
615
616
617
618
619
    
    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.
620
    """
621
    def __init__(self, regexp, wL=None, wR=None, name=''):
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
        """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.
        """
637
        super(RE, self).__init__(name)
638
639
        self.wL = wL
        self.wR = wR
640
641
        self.wspLeft = Whitespace(wL) if wL else ZOMBIE_PARSER
        self.wspRight = Whitespace(wR) if wR else ZOMBIE_PARSER
642
643
        self.main = RegExp(regexp)

Eckhart Arnold's avatar
Eckhart Arnold committed
644
645
646
647
648
    def __deepcopy__(self, memo={}):
        try:
            regexp = copy.deepcopy(self.main.regexp, memo)
        except TypeError:
            regexp = self.main.regexp.pattern
649
        return self.__class__(regexp, self.wL, self.wR, self.name)
Eckhart Arnold's avatar
Eckhart Arnold committed
650

651
    def __call__(self, text: str) -> Tuple[Node, str]:
652
653
654
655
656
657
658
659
660
661
662
        # 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

663
    def __str__(self):
664
        if self.ptype == TOKEN_PTYPE:
665
666
667
            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 ''))
668
669
670

    def _grammar_assigned_notifier(self):
        if self.grammar:
671
            # use default whitespace parsers if not otherwise specified
672
673
674
675
676
            if self.wL is None:
                self.wspLeft = self.grammar.wsp_left_parser__
            if self.wR is None:
                self.wspRight = self.grammar.wsp_right_parser__

677
    def apply(self, func: Parser.ApplyFunc):
678
679
680
681
682
683
684
685
        if super(RE, self).apply(func):
            if self.wL:
                self.wspLeft.apply(func)
            if self.wR:
                self.wspRight.apply(func)
            self.main.apply(func)


686
687
688
class Token(RE):
    assert TOKEN_PTYPE == ":Token"

689
    def __init__(self, token: str, wL=None, wR=None, name: str = '') -> None:
690
691
692
693
694
        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)
695
696


697
def mixin_comment(whitespace: str, comment: str) -> str:
698
699
700
701
702
703
704
    """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).
705
706
707
708
709
710
711
712
713
714
715
716
717
    """
    wspc = '(?:' + whitespace + '(?:' + comment + whitespace + ')*)'
    return wspc


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


class UnaryOperator(Parser):
718
    def __init__(self, parser: Parser, name: str = '') -> None:
719
        super(UnaryOperator, self).__init__(name)
720
        # assert isinstance(parser, Parser)
721
        self.parser = parser  # type: Parser
722

Eckhart Arnold's avatar
Eckhart Arnold committed
723
724
    def __deepcopy__(self, memo):
        parser = copy.deepcopy(self.parser, memo)
725
        return self.__class__(parser, self.name)
Eckhart Arnold's avatar
Eckhart Arnold committed
726

727
    def apply(self, func: Parser.ApplyFunc):
728
729
730
731
732
        if super(UnaryOperator, self).apply(func):
            self.parser.apply(func)


class NaryOperator(Parser):
733
    def __init__(self, *parsers: Parser, name: str = '') -> None:
734
        super(NaryOperator, self).__init__(name)
735
        # assert all([isinstance(parser, Parser) for parser in parsers]), str(parsers)
736
        self.parsers = parsers  # type: Container  ## [Parser]
737

Eckhart Arnold's avatar
Eckhart Arnold committed
738
739
    def __deepcopy__(self, memo):
        parsers = copy.deepcopy(self.parsers, memo)
740
        return self.__class__(*parsers, name=self.name)
Eckhart Arnold's avatar
Eckhart Arnold committed
741

742
    def apply(self, func: Parser.ApplyFunc):
743
744
745
746
747
        if super(NaryOperator, self).apply(func):
            for parser in self.parsers:
                parser.apply(func)


748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
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


766
class Optional(UnaryOperator):
767
    def __init__(self, parser: Parser, name: str = '') -> None:
768
        super(Optional, self).__init__(parser, name)
769
        # assert isinstance(parser, Parser)
770
        assert not isinstance(parser, Optional), \
771
            "Redundant nesting of options: %s(%s)" % \
772
773
            (str(name), str(parser.name))
        assert not isinstance(parser, Required), \
774
            "Nesting options with required elements is contradictory: " \
775
776
            "%s(%s)" % (str(name), str(parser.name))

777
    def __call__(self, text: str) -> Tuple[Node, str]:
778
779
780
781
782
783
784
        node, text = self.parser(text)
        if node:
            return Node(self, node), text
        return Node(self, ()), text


class ZeroOrMore(Optional):
785
786
    def __call__(self, text: str) -> Tuple[Node, str]:
        results = ()  # type: Tuple[Node, ...]
787
788
789
        n = len(text) + 1
        while text and len(text) < n:
            n = len(text)
790
791
792
            node, text = self.parser(text)
            if not node:
                break
793
            if len(text) == n:
di68kap's avatar
di68kap committed
794
                node.add_error(dsl_error_msg(self, 'Infinite Loop detected.'))
795
796
797
798
799
            results += (node,)
        return Node(self, results), text


class OneOrMore(UnaryOperator):
800
    def __init__(self, parser: Parser, name: str = '') -> None:
801
        super(OneOrMore, self).__init__(parser, name)
802
803
804
805
        assert not isinstance(parser, Optional), \
            "Use ZeroOrMore instead of nesting OneOrMore and Optional: " \
            "%s(%s)" % (str(name), str(parser.name))

806
807
808
    def __call__(self, text: str) -> Tuple[Node, str]:
        results = ()  # type: Tuple[Node, ...]
        text_ = text  # type: str
809
810
811
        n = len(text) + 1
        while text_ and len(text_) < n:
            n = len(text_)
812
813
814
            node, text_ = self.parser(text_)
            if not node:
                break
815
            if len(text_) == n:
816
                node.add_error(dsl_error_msg(self, 'Infinite Loop detected.'))
817
818
819
820
821
822
823
            results += (node,)
        if results == ():
            return None, text
        return Node(self, results), text_


class Sequence(NaryOperator):
824
    def __init__(self, *parsers: Parser, name: str = '') -> None:
825
        super(Sequence, self).__init__(*parsers, name=name)
826
827
        assert len(self.parsers) >= 1

828
829
830
    def __call__(self, text: str) -> Tuple[Node, str]:
        results = ()  # type: Tuple[Node, ...]
        text_ = text  # type: str
831
832
833
834
835
836
837
838
839
840
841
        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_

842
    def __add__(self, other: 'Sequence') -> 'Sequence':
843
844
        return Sequence(*self.parsers, other)

845
    def __radd__(self, other: 'Sequence') -> 'Sequence':
846
847
848
849
850
851
852
853
854
        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

855
856

class Alternative(NaryOperator):
857
858
859
860
861
862
863
    """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
864
865
    # the order of the sub-expression matters!
    >>> number = RE('\d+') | RE('\d+') + RE('\.') + RE('\d+')
866
867
    >>> Grammar(number)("3.1416").show()
    '3 <<< Error on ".1416" | Parser stopped before end! trying to recover... >>> '
868
869

    # the most selective expression should be put first:
Eckhart Arnold's avatar
Eckhart Arnold committed
870
    >>> number = RE('\d+') + RE('\.') + RE('\d+') | RE('\d+')
871
    >>> Grammar(number)("3.1416").show()
Eckhart Arnold's avatar
Eckhart Arnold committed
872
    '3.1416'
873
    """
874
875

    def __init__(self, *parsers: Parser, name: str = '') -> None:
876
        super(Alternative, self).__init__(*parsers, name=name)
877
878
879
        assert len(self.parsers) >= 1
        assert all(not isinstance(p, Optional) for p in self.parsers)

880
    def __call__(self, text: str) -> Tuple[Node, str]:
881
882
883
884
885
886
        for parser in self.parsers:
            node, text_ = parser(text)
            if node:
                return Node(self, node), text_
        return None, text

887
888
889
890
891
892
893
894
895
896
897
898
899
    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

900
901
902
903
904
905
906
907
908

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


class FlowOperator(UnaryOperator):
909
    def __init__(self, parser: Parser, name: str = '') -> None:
910
        super(FlowOperator, self).__init__(parser, name)
911
912
913


class Required(FlowOperator):
914
    # Add constructor that checks for logical errors, like `Required(Optional(...))` constructs ?
915
    def __call__(self, text: str) -> Tuple[Node, str]:
916
917
918
919
920
921
922
923
924
925
926
927
928
        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):
929
    def __init__(self, parser: Parser, name: str = '') -> None:
930
        super(Lookahead, self).__init__(parser, name)
931

932
    def __call__(self, text: str) -> Tuple[Node, str]:
933
934
935
936
937
938
        node, text_ = self.parser(text)
        if self.sign(node is not None):
            return Node(self, ''), text
        else:
            return None, text

939
    def sign(self, bool_value) -> bool:
940
941
942
943
        return bool_value


class NegativeLookahead(Lookahead):
944
    def sign(self, bool_value) -> bool:
945
946
947
        return not bool_value


948
def iter_right_branch(node) -> Iterator[Node]:
949
950
951
952
953
954
955
956
957
958
959
960
    """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):
961
    def __init__(self, parser: Parser, name: str = '') -> None:
962
        super(Lookbehind, self).__init__(parser, name)
963
964
        print("WARNING: Lookbehind Operator is experimental!")

965
    def __call__(self, text: str) -> Tuple[Node, str]:
966
967
968
969
970
971
972
973
        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

974
    def sign(self, bool_value) -> bool:
975
976
977
978
979
980
981
982
983
984
985
986
987
988
        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):
989
    def sign(self, bool_value) -> bool:
990
991
992
993
994
995
996
997
998
999
1000
        return not bool_value


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


class Capture(UnaryOperator):
1001
    def __init__(self, parser: Parser, name: str = '') -> None:
1002
        super(Capture, self).__init__(parser, name)
1003

1004
    def __call__(self, text: str) -> Tuple[Node, str]:
1005
1006
1007
1008
        node, text = self.parser(text)
        if node:
            stack = self.grammar.variables.setdefault(self.name, [])
            stack.append(str(node))
1009
1010
1011
            return Node(self, node), text
        else:
            return None, text
1012
1013


1014
1015
1016
1017
RetrieveFilter = Callable[[List[str]], str]


def nop_filter(stack: List[str]) -> str:
1018
1019
1020
    return stack[-1]


1021
def counterpart_filter(stack: List[str]) -> str:
1022
1023
1024
1025
    value = stack[-1]
    return value.replace("(", ")").replace("[", "]").replace("{", "}").replace(">", "<")


1026
def accumulating_filter(stack: List[str]) -> str:
1027
1028
1029
    return "".join(stack)


1030
class Retrieve(Parser):
1031
    def __init__(self, symbol: Parser, filter: RetrieveFilter = None, name: str = '') -> None:
1032
1033
        if not name:
            name = symbol.name
1034
        super(Retrieve, self).__init__(name)
1035
        self.symbol = symbol
1036
        self.filter = filter if filter else nop_filter
1037

Eckhart Arnold's avatar
Eckhart Arnold committed
1038
    def __deepcopy__(self, memo):
1039
        return self.__class__(self.symbol, self.filter, self.name)
Eckhart Arnold's avatar
Eckhart Arnold committed
1040

1041
    def __call__(self, text: str) -> Tuple[Node, str]:
1042
1043
        try:
            stack = self.grammar.variables[self.symbol.name]
1044
            value = self.filter(stack)
1045
1046
1047
1048
            self.pick_value(stack)
        except (KeyError, IndexError):
            return Node(self, '').add_error(dsl_error_msg(self,
                "%s undefined or exhausted" % self.symbol.name)), text
1049
1050
1051
1052
1053
        if text.startswith(value):
            return Node(self, value), text[len(value):]
        else:
            return None, text

1054
    def pick_value(self, stack: List[str]) -> str:
1055
1056
1057
1058
        return stack[-1]


class Pop(Retrieve):
1059
    def pick_value(self, stack: List[str]) -> str:
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
        return stack.pop()


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


class Forward(Parser):
    def __init__(self):
        Parser.__init__(self)
        self.parser = None
1074
        self.cycle_reached = False
1075

Eckhart Arnold's avatar
Eckhart Arnold committed
1076
1077
1078
1079
1080
1081
1082
    def __deepcopy__(self, memo):
        duplicate = self.__class__()
        memo[id(self)] = duplicate
        parser = copy.deepcopy(self.parser, memo)
        duplicate.set(parser)
        return duplicate

1083
    def __call__(self, text: str) -> Tuple[Node, str]:
1084
1085
        return self.parser(text)

1086
1087
1088
1089
1090
1091
1092
1093
    def __str__(self):
        if self.cycle_reached:
            return "..."
        else:
            self.cycle_reached = True
            s = str(self.parser)
            self.cycle_reached = False
            return s
1094

1095
1096
    def set(self, parser: Parser):
        # assert isinstance(parser, Parser)
1097
        self.name = parser.name  # redundant, see Grammar-constructor
1098
1099
        self.parser = parser

1100
    def apply(self, func: Parser.ApplyFunc):
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
        if super(Forward, self).apply(func):
            assert not self.visited
            self.parser.apply(func)


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


1113
class Compiler:
Eckhart Arnold's avatar
Eckhart Arnold committed
1114
    def __init__(self, grammar_name="", grammar_source=""):
1115
        self.dirty_flag = False
Eckhart Arnold's avatar
Eckhart Arnold committed
1116
        self.set_grammar_name(grammar_name, grammar_source)
1117
1118
1119
1120

    def _reset(self):
        pass

1121
    def __call__(self, node: Node) -> Any:
1122
        """Compiles the abstract syntax tree with the root ``node``.
1123
        
Eckhart Arnold's avatar
Eckhart Arnold committed
1124
        It's called `compile_ast`` to avoid confusion with the 
1125
1126
        ``_compile`` that is called from within the local node 
        compiler methods.
1127
1128
1129
1130
1131
1132
1133
        """
        if self.dirty_flag:
            self._reset()
        else:
            self.dirty_flag = True
        return self._compile(node)

Eckhart Arnold's avatar
Eckhart Arnold committed
1134
1135
1136
1137
1138
1139
1140
    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)

1141
    @staticmethod
1142
    def derive_method_name(node_name: str) -> str:
1143
        """Returns the method name for ``node_name``, e.g.
1144
        >>> Compiler.method_name('expression')
1145
1146
1147
1148
        'on_expression'
        """
        return 'on_' + node_name

1149
    def _compile(self, node: Node) -> Any:
1150
1151
1152
1153
1154
        """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
1155
        name by adding the prefix 'on_'.
1156
1157
<