parsers.py 46.1 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
Eckhart Arnold's avatar
Eckhart Arnold committed
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
        try:
            location = len(text)
164
165
166
167
168
169
            grammar = parser.grammar  # grammar may be 'None' for unconnected parsers!

            if grammar.history_tracking:
                grammar.call_stack.append(parser)
                grammar.moving_forward = True

170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
            # if location has already been visited by the current parser,
            # return saved result
            if location in parser.visited:
                return parser.visited[location]
            # break left recursion at the maximum allowed depth
            if parser.recursion_counter.setdefault(location, 0) > LEFT_RECURSION_DEPTH:
                return None, text

            parser.recursion_counter[location] += 1

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

            if node is not None:
                # in case of a recursive call saves the result of the first
                # (or left-most) call that matches
                parser.visited[location] = (node, rest)
187
                grammar.last_node = node   # store last node for Lookbehind operator
188
189
190
191
192
193
194
            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

195
196
197
198
199
200
201
202
203
            if grammar.history_tracking:
                # don't track returning parsers except in case an error has occurred
                if grammar.moving_forward or (node and node._errors):
                    grammar.moving_forward = False
                    record = HistoryRecord(grammar.call_stack.copy(), node, len(rest))
                    grammar.history.append(record)
                    # print(record.stack, record.status, rest[:20].replace('\n', '|'))
                grammar.call_stack.pop()

204
205
206
207
208
209
210
211
212
213
214
        except RecursionError:
            node = Node(None, text[:min(10, max(1, text.find("\n")))] + " ...")
            node.add_error("maximum recursion depth of parser reached; "
                           "potentially due to too many errors!")
            rest = ''

        return node, rest

    return guarded_call


Eckhart Arnold's avatar
Eckhart Arnold committed
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
                # add parser to grammar object on the fly...
                setattr(self, key, copy.deepcopy(parser))
360
                self[key].apply(self._add_parser)
Eckhart Arnold's avatar
Eckhart Arnold committed
361
                return self[key]
362
            raise KeyError('Unknown parser "%s" !' % key)
363

364
    def _reset(self):
365
366
367
368
369
370
371
372
373
374
375
376
        # 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

377
    # 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
378
    def _add_parser(self, parser: Parser) -> None:
Eckhart Arnold's avatar
Eckhart Arnold committed
379
        """Adds the particular copy of the parser object to this
380
        particular instance of Grammar.
381
        """
382
383
        if parser.name:
            setattr(self, parser.name, parser)
384
385
386
        self.all_parsers.add(parser)
        parser.grammar = self

Eckhart Arnold's avatar
Eckhart Arnold committed
387
    def __call__(self, document: str, start_parser="root__") -> Node:
388
389
390
391
        """Parses a document with with parser-combinators.

        Args:
            document (str): The source text to be parsed.
392
393
394
            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.)
395
396
397
        Returns:
            Node: The root node ot the parse tree.
        """
398
        # assert isinstance(document, str), type(document)
399
400
401
402
403
404
405
406
        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
407
        self.history_tracking = is_logging()
408
        self.document = document
Eckhart Arnold's avatar
Eckhart Arnold committed
409
        parser = self[start_parser] if isinstance(start_parser, str) else start_parser
410
411
        assert parser.grammar == self, "Cannot run parsers from a different grammar object!" \
                                       " %s vs. %s" % (str(self), str(parser.grammar))
412
        stitches = []  # type: List[Node]
413
        rest = document
414
415
        if not rest:
            result, ignore = parser(rest)
416
417
418
419
420
421
422
423
424
425
        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" + \
426
427
428
                                (("! trying to recover" +
                                  (" but stopping history recording at this point."
                                   if self.history_tracking else "..."))
429
430
431
432
                                 if len(stitches) < MAX_DROPOUTS
                                 else " too often! Terminating parser.")
                stitches.append(Node(None, skip))
                stitches[-1].add_error(error_msg)
433
                if self.history_tracking:
434
435
436
437
438
439
440
                    # 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
441
442
443
                    record = HistoryRecord(self.call_stack.copy(), stitches[-1], len(rest))
                    self.history.append(record)
                    self.history_tracking = False
444
445
446
447
        if stitches:
            if rest:
                stitches.append(Node(None, rest))
            result = Node(None, tuple(stitches))
448
449
450
        if any(self.variables.values()):
            result.add_error("Capture-retrieve-stack not empty after end of parsing: "
                             + str(self.variables))
451
452
453
        result.pos = 0  # calculate all positions
        return result

Eckhart Arnold's avatar
Eckhart Arnold committed
454
    def log_parsing_history(self, log_file_name: str='') -> None:
455
456
457
458
459
460
        """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
461
            return record.stack, record.status, excerpt
462
463

        def write_log(history, log_name):
Eckhart Arnold's avatar
Eckhart Arnold committed
464
            path = os.path.join(log_dir(), log_name + "_parser.log")
465
466
467
468
469
470
            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
471
472
473
474
475
476
477
        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)
478
            if record.node and record.node.parser.ptype != WHITESPACE_PTYPE:
Eckhart Arnold's avatar
Eckhart Arnold committed
479
                match_history.append(line)
480
                if record.node.error_flag:
Eckhart Arnold's avatar
Eckhart Arnold committed
481
482
483
484
                    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')
485
486


Eckhart Arnold's avatar
Eckhart Arnold committed
487
488
def dsl_error_msg(parser: Parser, error_str: str) -> str:
    """Returns an error message for errors in the parser configuration,
489
490
491
    e.g. errors that result in infinite loops.

    Args:
Eckhart Arnold's avatar
Eckhart Arnold committed
492
        parser (Parser):  The parser where the error was noticed. Note
493
            that this is not necessarily the parser that caused the
Eckhart Arnold's avatar
Eckhart Arnold committed
494
            error but only where the error became aparent.
495
496
497
498
499
        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.
    """
500
    msg = ["DSL parser specification error:", error_str, 'Caught by parser "%s".' % str(parser)]
501
502
503
504
505
506
507
    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)


508
509
510
511
512
513
514
515
516
517
518
519
########################################################################
#
# Token and Regular Expression parser classes (i.e. leaf classes)
#
########################################################################


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


520
def make_token(token: str, argument: str = '') -> str:
521
522
523
524
525
526
527
528
529
530
531
532
533
    """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


534
def nil_scanner(text: str) -> str:
Eckhart Arnold's avatar
Eckhart Arnold committed
535
    return text
536
537
538


class ScannerToken(Parser):
539
540
541
542
543
544
545
546
547
548
    """
    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.
    """
549
550
551

    def __init__(self, scanner_token: str) -> None:
        assert scanner_token and scanner_token.isupper()
552
        assert RX_SCANNER_TOKEN.match(scanner_token)
553
        super(ScannerToken, self).__init__(scanner_token)
554

555
    def __call__(self, text: str) -> Tuple[Node, str]:
556
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
        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):
582
583
    """
    Regular expression parser.
584
585
586
587
588
589
    
    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.
    """
590
591

    def __init__(self, regexp, name: str = '') -> None:
592
        super(RegExp, self).__init__(name)
593
594
595
        self.regexp = re.compile(regexp) if isinstance(regexp, str) else regexp

    def __deepcopy__(self, memo):
Eckhart Arnold's avatar
Eckhart Arnold committed
596
        # `regex` supports deep copies, but not `re`
597
        try:
Eckhart Arnold's avatar
Eckhart Arnold committed
598
            regexp = copy.deepcopy(self.regexp, memo)
599
600
        except TypeError:
            regexp = self.regexp.pattern
601
        return RegExp(regexp, self.name)
602

603
    def __call__(self, text: str) -> Tuple[Node, str]:
604
605
606
607
608
609
        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

610
611
612
    def __str__(self):
        return self.name or self.ptype + ' /%s/' % self.regexp.pattern

613

614
615
class Whitespace(RegExp):
    assert WHITESPACE_PTYPE == ":Whitespace"
616
617


618
619
class RE(Parser):
    """Regular Expressions with optional leading or trailing whitespace.
620
621
622
623
624
625
626
627
628
629
    
    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.
630
    """
631
    def __init__(self, regexp, wL=None, wR=None, name=''):
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
        """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.
        """
647
        super(RE, self).__init__(name)
648
649
        self.wL = wL
        self.wR = wR
650
651
        self.wspLeft = Whitespace(wL) if wL else ZOMBIE_PARSER
        self.wspRight = Whitespace(wR) if wR else ZOMBIE_PARSER
652
653
        self.main = RegExp(regexp)

Eckhart Arnold's avatar
Eckhart Arnold committed
654
655
656
657
658
    def __deepcopy__(self, memo={}):
        try:
            regexp = copy.deepcopy(self.main.regexp, memo)
        except TypeError:
            regexp = self.main.regexp.pattern
659
        return self.__class__(regexp, self.wL, self.wR, self.name)
Eckhart Arnold's avatar
Eckhart Arnold committed
660

661
    def __call__(self, text: str) -> Tuple[Node, str]:
662
663
664
665
666
667
668
669
670
671
672
        # 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

673
    def __str__(self):
674
        if self.ptype == TOKEN_PTYPE:
675
676
677
            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 ''))
678
679
680

    def _grammar_assigned_notifier(self):
        if self.grammar:
681
            # use default whitespace parsers if not otherwise specified
682
683
684
685
686
            if self.wL is None:
                self.wspLeft = self.grammar.wsp_left_parser__
            if self.wR is None:
                self.wspRight = self.grammar.wsp_right_parser__

687
    def apply(self, func: Parser.ApplyFunc):
688
689
690
691
692
693
694
695
        if super(RE, self).apply(func):
            if self.wL:
                self.wspLeft.apply(func)
            if self.wR:
                self.wspRight.apply(func)
            self.main.apply(func)


696
697
698
class Token(RE):
    assert TOKEN_PTYPE == ":Token"

699
    def __init__(self, token: str, wL=None, wR=None, name: str = '') -> None:
700
701
702
703
704
        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)
705
706


707
def mixin_comment(whitespace: str, comment: str) -> str:
708
709
710
711
712
713
714
    """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).
715
716
717
718
719
720
721
722
723
724
725
726
727
    """
    wspc = '(?:' + whitespace + '(?:' + comment + whitespace + ')*)'
    return wspc


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


class UnaryOperator(Parser):
728
    def __init__(self, parser: Parser, name: str = '') -> None:
729
        super(UnaryOperator, self).__init__(name)
730
        # assert isinstance(parser, Parser)
731
        self.parser = parser  # type: Parser
732

Eckhart Arnold's avatar
Eckhart Arnold committed
733
734
    def __deepcopy__(self, memo):
        parser = copy.deepcopy(self.parser, memo)
735
        return self.__class__(parser, self.name)
Eckhart Arnold's avatar
Eckhart Arnold committed
736

737
    def apply(self, func: Parser.ApplyFunc):
738
739
740
741
742
        if super(UnaryOperator, self).apply(func):
            self.parser.apply(func)


class NaryOperator(Parser):
743
    def __init__(self, *parsers: Parser, name: str = '') -> None:
744
        super(NaryOperator, self).__init__(name)
745
        # assert all([isinstance(parser, Parser) for parser in parsers]), str(parsers)
746
        self.parsers = parsers  # type: Container  ## [Parser]
747

Eckhart Arnold's avatar
Eckhart Arnold committed
748
749
    def __deepcopy__(self, memo):
        parsers = copy.deepcopy(self.parsers, memo)
750
        return self.__class__(*parsers, name=self.name)
Eckhart Arnold's avatar
Eckhart Arnold committed
751

752
    def apply(self, func: Parser.ApplyFunc):
753
754
755
756
757
        if super(NaryOperator, self).apply(func):
            for parser in self.parsers:
                parser.apply(func)


758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
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


776
class Optional(UnaryOperator):
777
    def __init__(self, parser: Parser, name: str = '') -> None:
778
        super(Optional, self).__init__(parser, name)
779
        # assert isinstance(parser, Parser)
780
        assert not isinstance(parser, Optional), \
781
            "Redundant nesting of options: %s(%s)" % \
782
783
            (str(name), str(parser.name))
        assert not isinstance(parser, Required), \
784
            "Nesting options with required elements is contradictory: " \
785
786
            "%s(%s)" % (str(name), str(parser.name))

787
    def __call__(self, text: str) -> Tuple[Node, str]:
788
789
790
791
792
793
794
        node, text = self.parser(text)
        if node:
            return Node(self, node), text
        return Node(self, ()), text


class ZeroOrMore(Optional):
795
796
    def __call__(self, text: str) -> Tuple[Node, str]:
        results = ()  # type: Tuple[Node, ...]
797
798
799
        n = len(text) + 1
        while text and len(text) < n:
            n = len(text)
800
801
802
            node, text = self.parser(text)
            if not node:
                break
803
            if len(text) == n:
di68kap's avatar
di68kap committed
804
                node.add_error(dsl_error_msg(self, 'Infinite Loop detected.'))
805
806
807
808
809
            results += (node,)
        return Node(self, results), text


class OneOrMore(UnaryOperator):
810
    def __init__(self, parser: Parser, name: str = '') -> None:
811
        super(OneOrMore, self).__init__(parser, name)
812
813
814
815
        assert not isinstance(parser, Optional), \
            "Use ZeroOrMore instead of nesting OneOrMore and Optional: " \
            "%s(%s)" % (str(name), str(parser.name))

816
817
818
    def __call__(self, text: str) -> Tuple[Node, str]:
        results = ()  # type: Tuple[Node, ...]
        text_ = text  # type: str
819
820
821
        n = len(text) + 1
        while text_ and len(text_) < n:
            n = len(text_)
822
823
824
            node, text_ = self.parser(text_)
            if not node:
                break
825
            if len(text_) == n:
826
                node.add_error(dsl_error_msg(self, 'Infinite Loop detected.'))
827
828
829
830
831
832
833
            results += (node,)
        if results == ():
            return None, text
        return Node(self, results), text_


class Sequence(NaryOperator):
834
    def __init__(self, *parsers: Parser, name: str = '') -> None:
835
        super(Sequence, self).__init__(*parsers, name=name)
836
837
        assert len(self.parsers) >= 1

838
839
840
    def __call__(self, text: str) -> Tuple[Node, str]:
        results = ()  # type: Tuple[Node, ...]
        text_ = text  # type: str
841
842
843
        for parser in self.parsers:
            node, text_ = parser(text_)
            if not node:
844
845
                return None, text
            results += (node,)
846
847
848
849
850
            if node.error_flag:
                break
        assert len(results) <= len(self.parsers)
        return Node(self, results), text_

851
    def __add__(self, other: 'Sequence') -> 'Sequence':
Eckhart Arnold's avatar
Eckhart Arnold committed
852
        return Sequence(*(self.parsers + (other,)))
853

854
    def __radd__(self, other: 'Sequence') -> 'Sequence':
855
856
857
858
859
860
861
862
863
        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

864
865

class Alternative(NaryOperator):
866
867
868
869
870
871
872
    """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
873
874
    # the order of the sub-expression matters!
    >>> number = RE('\d+') | RE('\d+') + RE('\.') + RE('\d+')
875
876
    >>> Grammar(number)("3.1416").show()
    '3 <<< Error on ".1416" | Parser stopped before end! trying to recover... >>> '
877
878

    # the most selective expression should be put first:
Eckhart Arnold's avatar
Eckhart Arnold committed
879
    >>> number = RE('\d+') + RE('\.') + RE('\d+') | RE('\d+')
880
    >>> Grammar(number)("3.1416").show()
Eckhart Arnold's avatar
Eckhart Arnold committed
881
    '3.1416'
882
    """
883
884

    def __init__(self, *parsers: Parser, name: str = '') -> None:
885
        super(Alternative, self).__init__(*parsers, name=name)
886
887
888
        assert len(self.parsers) >= 1
        assert all(not isinstance(p, Optional) for p in self.parsers)

889
    def __call__(self, text: str) -> Tuple[Node, str]:
890
891
892
893
894
895
        for parser in self.parsers:
            node, text_ = parser(text)
            if node:
                return Node(self, node), text_
        return None, text

896
    def __or__(self, other):
Eckhart Arnold's avatar
Eckhart Arnold committed
897
        return Alternative(*(self.parsers + (other,)))
898
899
900
901
902
903
904
905
906
907
908

    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

909
910
911
912
913
914
915
916
917

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


class FlowOperator(UnaryOperator):
918
    def __init__(self, parser: Parser, name: str = '') -> None:
919
        super(FlowOperator, self).__init__(parser, name)
920
921
922


class Required(FlowOperator):
923
    # Add constructor that checks for logical errors, like `Required(Optional(...))` constructs ?
924
    def __call__(self, text: str) -> Tuple[Node, str]:
925
926
927
928
929
930
931
932
933
934
935
936
937
        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):
938
    def __init__(self, parser: Parser, name: str = '') -> None:
939
        super(Lookahead, self).__init__(parser, name)
940

941
    def __call__(self, text: str) -> Tuple[Node, str]:
942
943
944
945
946
947
        node, text_ = self.parser(text)
        if self.sign(node is not None):
            return Node(self, ''), text
        else:
            return None, text

948
    def sign(self, bool_value) -> bool:
949
950
951
952
        return bool_value


class NegativeLookahead(Lookahead):
953
    def sign(self, bool_value) -> bool:
954
955
956
        return not bool_value


957
def iter_right_branch(node) -> Iterator[Node]:
958
959
960
961
962
963
964
965
966
967
968
969
    """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):
970
    def __init__(self, parser: Parser, name: str = '') -> None:
971
        super(Lookbehind, self).__init__(parser, name)
972
973
        print("WARNING: Lookbehind Operator is experimental!")

974
    def __call__(self, text: str) -> Tuple[Node, str]:
975
976
977
978
979
980
981
982
        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

983
    def sign(self, bool_value) -> bool:
984
985
986
987
988
989
990
991
992
993
994
995
996
997
        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):
998
    def sign(self, bool_value) -> bool:
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
        return not bool_value


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


class Capture(UnaryOperator):
1010
1011
    """STILL EXPERIMENTAL!"""

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
RetrieveFilter = Callable[[List[str]], str]


1028
def last_value(stack: List[str]) -> str:
1029
1030
1031
    return stack[-1]


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


1037
1038
def accumulate(stack: List[str]) -> str:
    return "".join(stack) if len(stack) > 1 else stack[-1]  # provoke IndexError if stack empty
1039
1040


1041
class Retrieve(Parser):
1042
    """STILL EXPERIMENTAL!"""
1043
    def __init__(self, symbol: Parser, filter: RetrieveFilter = None, name: str = '') -> None:
1044
        super(Retrieve, self).__init__(name)
1045
        self.symbol = symbol
1046
        self.filter = filter if filter else last_value
1047

Eckhart Arnold's avatar
Eckhart Arnold committed
1048
    def __deepcopy__(self, memo):
1049
        return self.__class__(self.symbol, self.filter, self.name)
Eckhart Arnold's avatar
Eckhart Arnold committed
1050

1051
    def __call__(self, text: str) -> Tuple[Node, str]:
1052
1053
1054
        return self.call(text)  # allow call method to be called from subclass circumventing the parser guard

    def call(self, text: str) -> Tuple[Node, str]:
1055
1056
        try:
            stack = self.grammar.variables[self.symbol.name]
1057
            value = self.filter(stack)
1058
1059
        except (KeyError, IndexError):
            return Node(self, '').add_error(dsl_error_msg(self,
1060
                                                          "'%s' undefined or exhausted." % self.symbol.name)), text
1061
1062
1063
1064
1065
1066
1067
        if text.startswith(value):
            return Node(self, value), text[len(value):]
        else:
            return None, text


class Pop(Retrieve):
1068
1069
1070
1071
1072
1073
1074
1075
    """STILL EXPERIMENTAL!!!"""

    def __call__(self, text: str) -> Tuple[Node, str]:
        nd, txt = super(Pop, self).call(text)  # call() instead of __call__() to avoid parser guard
        if nd and not nd.error_flag:
            stack = self.grammar.variables[self.symbol.name]
            stack.pop()
        return nd, txt
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088


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


class Forward(Parser):
    def __init__(self):
        Parser.__init__(self)
        self.parser = None
1089
        self.cycle_reached = False
1090

Eckhart Arnold's avatar
Eckhart Arnold committed
1091
1092
1093
1094
1095
1096
1097
    def __deepcopy__(self, memo):
        duplicate = self.__class__()
        memo[id(self)] = duplicate
        parser = copy.deepcopy(self.parser, memo)
        duplicate.set(parser)
        return duplicate

1098
    def __call__(self, text: str) -> Tuple[Node, str]:
1099
1100
        return self.parser(text)

1101
1102
1103
1104
1105
1106
1107
1108
    def __str__(self):
        if self.cycle_reached:
            return "..."
        else:
            self.cycle_reached = True
            s = str(self.parser)
            self.cycle_reached = False
            return s
1109

1110
1111
    def set(self, parser: Parser):
        # assert isinstance(parser, Parser)
1112
        # self.name = parser.name  # redundant, see Grammar-constructor
1113
1114
        self.parser = parser

1115
    def apply(self, func: Parser.ApplyFunc):
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
        if super(Forward, self).apply(func):
            assert not self.visited
            self.parser.apply(func)


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


1128
class Compiler:
Eckhart Arnold's avatar
Eckhart Arnold committed
1129
    def __init__(self, grammar_name="", grammar_source=""):
1130
        self.dirty_flag = False
Eckhart Arnold's avatar
Eckhart Arnold committed
1131
        self.set_grammar_name(grammar_name, grammar_source)
1132
1133
1134
1135

    def _reset(self):
        pass

1136
    def __call__(self, node: Node) -> Any:
1137
        """Compiles the abstract syntax tree with the root ``node``.
1138
        
Eckhart Arnold's avatar
Eckhart Arnold committed
1139
        It's called `compile_ast`` to avoid confusion with the 
1140
1141
        ``_compile`` that is called from within the local node 
        compiler methods.
1142
1143
1144
1145
1146
1147
1148
        """
        if self.dirty_flag:
            self._reset()
        else:
            self.dirty_flag = True
        return self._compile(node)

Eckhart Arnold's avatar
Eckhart Arnold committed
1149
1150
1151
1152
1153
1154
1155
    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)

1156
    @staticmethod
1157
    def derive_method_name(node_name: str) -> str:
1158
        """Returns the method name for ``node_name``, e.g.
1159
        >>> Compiler.method_name('expression')
Eckhart Arnold's avatar