parsers.py 45 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
        return getattr(self, key)
349

350
    def _reset(self):
351
352
353
354
355
356
357
358
359
360
361
362
        # 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

363
    # 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
364
    def _add_parser(self, parser: Parser) -> None:
365
        """Adds the copy of the classes parser object to this
366
        particular instance of Grammar.
367
        """
368
369
        if parser.name:
            setattr(self, parser.name, parser)
370
371
372
        self.all_parsers.add(parser)
        parser.grammar = self

Eckhart Arnold's avatar
Eckhart Arnold committed
373
    def __call__(self, document: str, start_parser="root__") -> Node:
374
375
376
377
        """Parses a document with with parser-combinators.

        Args:
            document (str): The source text to be parsed.
378
379
380
            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.)
381
382
383
        Returns:
            Node: The root node ot the parse tree.
        """
384
        # assert isinstance(document, str), type(document)
385
386
387
388
389
390
391
392
        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
393
        self.history_tracking = is_logging()
394
        self.document = document
Eckhart Arnold's avatar
Eckhart Arnold committed
395
        parser = self[start_parser] if isinstance(start_parser, str) else start_parser
396
397
        assert parser.grammar == self, "Cannot run parsers from a different grammar object!" \
                                       " %s vs. %s" % (str(self), str(parser.grammar))
398
        stitches = []  # type: List[Node]
399
        rest = document
400
401
        if not rest:
            result, ignore = parser(rest)
402
403
404
405
406
407
408
409
410
411
        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" + \
412
413
414
                                (("! trying to recover" +
                                  (" but stopping history recording at this point."
                                   if self.history_tracking else "..."))
415
416
417
418
                                 if len(stitches) < MAX_DROPOUTS
                                 else " too often! Terminating parser.")
                stitches.append(Node(None, skip))
                stitches[-1].add_error(error_msg)
419
                if self.history_tracking:
420
421
422
423
424
425
426
                    # 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
427
428
429
                    record = HistoryRecord(self.call_stack.copy(), stitches[-1], len(rest))
                    self.history.append(record)
                    self.history_tracking = False
430
431
432
433
434
435
436
        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
437
    def log_parsing_history(self, log_file_name: str='') -> None:
438
439
440
441
442
443
        """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
444
            return record.stack, record.status, excerpt
445
446

        def write_log(history, log_name):
Eckhart Arnold's avatar
Eckhart Arnold committed
447
            path = os.path.join(log_dir(), log_name + "_parser.log")
448
449
450
451
452
453
            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
454
455
456
457
458
459
460
        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)
461
            if record.node and record.node.parser.ptype != WHITESPACE_PTYPE:
Eckhart Arnold's avatar
Eckhart Arnold committed
462
463
464
465
466
467
                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')
468
469


Eckhart Arnold's avatar
Eckhart Arnold committed
470
471
def dsl_error_msg(parser: Parser, error_str: str) -> str:
    """Returns an error message for errors in the parser configuration,
472
473
474
    e.g. errors that result in infinite loops.

    Args:
Eckhart Arnold's avatar
Eckhart Arnold committed
475
        parser (Parser):  The parser where the error was noticed. Note
476
            that this is not necessarily the parser that caused the
Eckhart Arnold's avatar
Eckhart Arnold committed
477
            error but only where the error became aparent.
478
479
480
481
482
483
484
485
486
487
488
489
490
        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.
    """
    msg = ["DSL parser specification error:", error_str, "caught by parser", str(parser)]
    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)


491
492
493
494
495
496
497
498
499
500
501
502
########################################################################
#
# Token and Regular Expression parser classes (i.e. leaf classes)
#
########################################################################


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


503
def make_token(token: str, argument: str = '') -> str:
504
505
506
507
508
509
510
511
512
513
514
515
516
    """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


517
def nil_scanner(text: str) -> str:
Eckhart Arnold's avatar
Eckhart Arnold committed
518
    return text
519
520
521


class ScannerToken(Parser):
522
523
524
525
526
527
528
529
530
531
    """
    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.
    """
532
533
534

    def __init__(self, scanner_token: str) -> None:
        assert scanner_token and scanner_token.isupper()
535
        assert RX_SCANNER_TOKEN.match(scanner_token)
536
        super(ScannerToken, self).__init__(scanner_token)
537

538
    def __call__(self, text: str) -> Tuple[Node, str]:
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
        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):
565
566
    """
    Regular expression parser.
567
568
569
570
571
572
    
    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.
    """
573
574

    def __init__(self, regexp, name: str = '') -> None:
575
        super(RegExp, self).__init__(name)
576
577
578
        self.regexp = re.compile(regexp) if isinstance(regexp, str) else regexp

    def __deepcopy__(self, memo):
Eckhart Arnold's avatar
Eckhart Arnold committed
579
        # `regex` supports deep copies, but not `re`
580
        try:
Eckhart Arnold's avatar
Eckhart Arnold committed
581
            regexp = copy.deepcopy(self.regexp, memo)
582
583
        except TypeError:
            regexp = self.regexp.pattern
584
        return RegExp(regexp, self.name)
585

586
    def __call__(self, text: str) -> Tuple[Node, str]:
587
588
589
590
591
592
        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

593
594
595
    def __str__(self):
        return self.name or self.ptype + ' /%s/' % self.regexp.pattern

596

597
598
class Whitespace(RegExp):
    assert WHITESPACE_PTYPE == ":Whitespace"
599
600


601
602
class RE(Parser):
    """Regular Expressions with optional leading or trailing whitespace.
603
604
605
606
607
608
609
610
611
612
    
    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.
613
    """
614
    def __init__(self, regexp, wL=None, wR=None, name=''):
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
        """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.
        """
630
        super(RE, self).__init__(name)
631
632
        self.wL = wL
        self.wR = wR
633
634
        self.wspLeft = Whitespace(wL) if wL else ZOMBIE_PARSER
        self.wspRight = Whitespace(wR) if wR else ZOMBIE_PARSER
635
636
        self.main = RegExp(regexp)

Eckhart Arnold's avatar
Eckhart Arnold committed
637
638
639
640
641
    def __deepcopy__(self, memo={}):
        try:
            regexp = copy.deepcopy(self.main.regexp, memo)
        except TypeError:
            regexp = self.main.regexp.pattern
642
        return self.__class__(regexp, self.wL, self.wR, self.name)
Eckhart Arnold's avatar
Eckhart Arnold committed
643

644
    def __call__(self, text: str) -> Tuple[Node, str]:
645
646
647
648
649
650
651
652
653
654
655
        # 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

656
    def __str__(self):
657
        if self.ptype == TOKEN_PTYPE:
658
659
660
            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 ''))
661
662
663

    def _grammar_assigned_notifier(self):
        if self.grammar:
664
            # use default whitespace parsers if not otherwise specified
665
666
667
668
669
            if self.wL is None:
                self.wspLeft = self.grammar.wsp_left_parser__
            if self.wR is None:
                self.wspRight = self.grammar.wsp_right_parser__

670
    def apply(self, func: Parser.ApplyFunc):
671
672
673
674
675
676
677
678
        if super(RE, self).apply(func):
            if self.wL:
                self.wspLeft.apply(func)
            if self.wR:
                self.wspRight.apply(func)
            self.main.apply(func)


679
680
681
class Token(RE):
    assert TOKEN_PTYPE == ":Token"

682
    def __init__(self, token: str, wL=None, wR=None, name: str = '') -> None:
683
684
685
686
687
        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)
688
689


690
def mixin_comment(whitespace: str, comment: str) -> str:
691
692
693
694
695
696
697
    """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).
698
699
700
701
702
703
704
705
706
707
708
709
710
    """
    wspc = '(?:' + whitespace + '(?:' + comment + whitespace + ')*)'
    return wspc


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


class UnaryOperator(Parser):
711
    def __init__(self, parser: Parser, name: str = '') -> None:
712
        super(UnaryOperator, self).__init__(name)
713
        # assert isinstance(parser, Parser)
714
        self.parser = parser  # type: Parser
715

Eckhart Arnold's avatar
Eckhart Arnold committed
716
717
    def __deepcopy__(self, memo):
        parser = copy.deepcopy(self.parser, memo)
718
        return self.__class__(parser, self.name)
Eckhart Arnold's avatar
Eckhart Arnold committed
719

720
    def apply(self, func: Parser.ApplyFunc):
721
722
723
724
725
        if super(UnaryOperator, self).apply(func):
            self.parser.apply(func)


class NaryOperator(Parser):
726
    def __init__(self, *parsers: Parser, name: str = '') -> None:
727
        super(NaryOperator, self).__init__(name)
728
        # assert all([isinstance(parser, Parser) for parser in parsers]), str(parsers)
729
        self.parsers = parsers  # type: Container  ## [Parser]
730

Eckhart Arnold's avatar
Eckhart Arnold committed
731
732
    def __deepcopy__(self, memo):
        parsers = copy.deepcopy(self.parsers, memo)
733
        return self.__class__(*parsers, name=self.name)
Eckhart Arnold's avatar
Eckhart Arnold committed
734

735
    def apply(self, func: Parser.ApplyFunc):
736
737
738
739
740
        if super(NaryOperator, self).apply(func):
            for parser in self.parsers:
                parser.apply(func)


741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
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


759
class Optional(UnaryOperator):
760
    def __init__(self, parser: Parser, name: str = '') -> None:
761
        super(Optional, self).__init__(parser, name)
762
        # assert isinstance(parser, Parser)
763
764
765
766
        assert not isinstance(parser, Optional), \
            "Nesting options would be redundant: %s(%s)" % \
            (str(name), str(parser.name))
        assert not isinstance(parser, Required), \
767
            "Nesting options with required elements is contradictory: " \
768
769
            "%s(%s)" % (str(name), str(parser.name))

770
    def __call__(self, text: str) -> Tuple[Node, str]:
771
772
773
774
775
776
777
        node, text = self.parser(text)
        if node:
            return Node(self, node), text
        return Node(self, ()), text


class ZeroOrMore(Optional):
778
779
    def __call__(self, text: str) -> Tuple[Node, str]:
        results = ()  # type: Tuple[Node, ...]
780
781
782
        n = len(text) + 1
        while text and len(text) < n:
            n = len(text)
783
784
785
            node, text = self.parser(text)
            if not node:
                break
786
            if len(text) == n:
di68kap's avatar
di68kap committed
787
                node.add_error(dsl_error_msg(self, 'Infinite Loop detected.'))
788
789
790
791
792
            results += (node,)
        return Node(self, results), text


class OneOrMore(UnaryOperator):
793
    def __init__(self, parser: Parser, name: str = '') -> None:
794
        super(OneOrMore, self).__init__(parser, name)
795
796
797
798
        assert not isinstance(parser, Optional), \
            "Use ZeroOrMore instead of nesting OneOrMore and Optional: " \
            "%s(%s)" % (str(name), str(parser.name))

799
800
801
    def __call__(self, text: str) -> Tuple[Node, str]:
        results = ()  # type: Tuple[Node, ...]
        text_ = text  # type: str
802
803
804
        n = len(text) + 1
        while text_ and len(text_) < n:
            n = len(text_)
805
806
807
            node, text_ = self.parser(text_)
            if not node:
                break
808
            if len(text_) == n:
809
                node.add_error(dsl_error_msg(self, 'Infinite Loop detected.'))
810
811
812
813
814
815
816
            results += (node,)
        if results == ():
            return None, text
        return Node(self, results), text_


class Sequence(NaryOperator):
817
    def __init__(self, *parsers: Parser, name: str = '') -> None:
818
        super(Sequence, self).__init__(*parsers, name=name)
819
820
        assert len(self.parsers) >= 1

821
822
823
    def __call__(self, text: str) -> Tuple[Node, str]:
        results = ()  # type: Tuple[Node, ...]
        text_ = text  # type: str
824
825
826
827
828
829
830
831
832
833
834
        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_

835
    def __add__(self, other: 'Sequence') -> 'Sequence':
836
837
        return Sequence(*self.parsers, other)

838
    def __radd__(self, other: 'Sequence') -> 'Sequence':
839
840
841
842
843
844
845
846
847
        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

848
849

class Alternative(NaryOperator):
850
851
852
853
854
855
856
    """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
857
858
    # the order of the sub-expression matters!
    >>> number = RE('\d+') | RE('\d+') + RE('\.') + RE('\d+')
859
860
    >>> Grammar(number)("3.1416").show()
    '3 <<< Error on ".1416" | Parser stopped before end! trying to recover... >>> '
861
862

    # the most selective expression should be put first:
Eckhart Arnold's avatar
Eckhart Arnold committed
863
    >>> number = RE('\d+') + RE('\.') + RE('\d+') | RE('\d+')
864
    >>> Grammar(number)("3.1416").show()
Eckhart Arnold's avatar
Eckhart Arnold committed
865
    '3.1416'
866
    """
867
868

    def __init__(self, *parsers: Parser, name: str = '') -> None:
869
        super(Alternative, self).__init__(*parsers, name=name)
870
871
872
        assert len(self.parsers) >= 1
        assert all(not isinstance(p, Optional) for p in self.parsers)

873
    def __call__(self, text: str) -> Tuple[Node, str]:
874
875
876
877
878
879
        for parser in self.parsers:
            node, text_ = parser(text)
            if node:
                return Node(self, node), text_
        return None, text

880
881
882
883
884
885
886
887
888
889
890
891
892
    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

893
894
895
896
897
898
899
900
901

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


class FlowOperator(UnaryOperator):
902
    def __init__(self, parser: Parser, name: str = '') -> None:
903
        super(FlowOperator, self).__init__(parser, name)
904
905
906


class Required(FlowOperator):
907
    # Add constructor that checks for logical errors, like `Required(Optional(...))` constructs ?
908
    def __call__(self, text: str) -> Tuple[Node, str]:
909
910
911
912
913
914
915
916
917
918
919
920
921
        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):
922
    def __init__(self, parser: Parser, name: str = '') -> None:
923
        super(Lookahead, self).__init__(parser, name)
924

925
    def __call__(self, text: str) -> Tuple[Node, str]:
926
927
928
929
930
931
        node, text_ = self.parser(text)
        if self.sign(node is not None):
            return Node(self, ''), text
        else:
            return None, text

932
    def sign(self, bool_value) -> bool:
933
934
935
936
        return bool_value


class NegativeLookahead(Lookahead):
937
    def sign(self, bool_value) -> bool:
938
939
940
        return not bool_value


941
def iter_right_branch(node) -> Iterator[Node]:
942
943
944
945
946
947
948
949
950
951
952
953
    """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):
954
    def __init__(self, parser: Parser, name: str = '') -> None:
955
        super(Lookbehind, self).__init__(parser, name)
956
957
        print("WARNING: Lookbehind Operator is experimental!")

958
    def __call__(self, text: str) -> Tuple[Node, str]:
959
960
961
962
963
964
965
966
        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

967
    def sign(self, bool_value) -> bool:
968
969
970
971
972
973
974
975
976
977
978
979
980
981
        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):
982
    def sign(self, bool_value) -> bool:
983
984
985
986
987
988
989
990
991
992
993
        return not bool_value


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


class Capture(UnaryOperator):
994
    def __init__(self, parser: Parser, name: str = '') -> None:
995
        super(Capture, self).__init__(parser, name)
996

997
    def __call__(self, text: str) -> Tuple[Node, str]:
998
999
1000
1001
        node, text = self.parser(text)
        if node:
            stack = self.grammar.variables.setdefault(self.name, [])
            stack.append(str(node))
1002
1003
1004
            return Node(self, node), text
        else:
            return None, text
1005
1006


1007
1008
1009
1010
RetrieveFilter = Callable[[List[str]], str]


def nop_filter(stack: List[str]) -> str:
1011
1012
1013
    return stack[-1]


1014
def counterpart_filter(stack: List[str]) -> str:
1015
1016
1017
1018
    value = stack[-1]
    return value.replace("(", ")").replace("[", "]").replace("{", "}").replace(">", "<")


1019
def accumulating_filter(stack: List[str]) -> str:
1020
1021
1022
    return "".join(stack)


1023
class Retrieve(Parser):
1024
    def __init__(self, symbol: Parser, filter: RetrieveFilter = None, name: str = '') -> None:
1025
1026
        if not name:
            name = symbol.name
1027
        super(Retrieve, self).__init__(name)
1028
        self.symbol = symbol
1029
        self.filter = filter if filter else nop_filter
1030

Eckhart Arnold's avatar
Eckhart Arnold committed
1031
    def __deepcopy__(self, memo):
1032
        return self.__class__(self.symbol, self.filter, self.name)
Eckhart Arnold's avatar
Eckhart Arnold committed
1033

1034
    def __call__(self, text: str) -> Tuple[Node, str]:
1035
1036
        try:
            stack = self.grammar.variables[self.symbol.name]
1037
            value = self.filter(stack)
1038
1039
1040
1041
            self.pick_value(stack)
        except (KeyError, IndexError):
            return Node(self, '').add_error(dsl_error_msg(self,
                "%s undefined or exhausted" % self.symbol.name)), text
1042
1043
1044
1045
1046
        if text.startswith(value):
            return Node(self, value), text[len(value):]
        else:
            return None, text

1047
    def pick_value(self, stack: List[str]) -> str:
1048
1049
1050
1051
        return stack[-1]


class Pop(Retrieve):
1052
    def pick_value(self, stack: List[str]) -> str:
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
        return stack.pop()


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


class Forward(Parser):
    def __init__(self):
        Parser.__init__(self)
        self.parser = None
1067
        self.cycle_reached = False
1068

Eckhart Arnold's avatar
Eckhart Arnold committed
1069
1070
1071
1072
1073
1074
1075
    def __deepcopy__(self, memo):
        duplicate = self.__class__()
        memo[id(self)] = duplicate
        parser = copy.deepcopy(self.parser, memo)
        duplicate.set(parser)
        return duplicate

1076
    def __call__(self, text: str) -> Tuple[Node, str]:
1077
1078
        return self.parser(text)

1079
1080
1081
1082
1083
1084
1085
1086
    def __str__(self):
        if self.cycle_reached:
            return "..."
        else:
            self.cycle_reached = True
            s = str(self.parser)
            self.cycle_reached = False
            return s
1087

1088
1089
    def set(self, parser: Parser):
        # assert isinstance(parser, Parser)
1090
        self.name = parser.name  # redundant, see Grammar-constructor
1091
1092
        self.parser = parser

1093
    def apply(self, func: Parser.ApplyFunc):
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
        if super(Forward, self).apply(func):
            assert not self.visited
            self.parser.apply(func)


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


1106
class Compiler:
Eckhart Arnold's avatar
Eckhart Arnold committed
1107
    def __init__(self, grammar_name="", grammar_source=""):
1108
        self.dirty_flag = False
Eckhart Arnold's avatar
Eckhart Arnold committed
1109
        self.set_grammar_name(grammar_name, grammar_source)
1110
1111
1112
1113

    def _reset(self):
        pass

1114
    def __call__(self, node: Node) -> Any:
1115
        """Compiles the abstract syntax tree with the root ``node``.
1116
        
Eckhart Arnold's avatar
Eckhart Arnold committed
1117
        It's called `compile_ast`` to avoid confusion with the 
1118
1119
        ``_compile`` that is called from within the local node 
        compiler methods.
1120
1121
1122
1123
1124
1125
1126
        """
        if self.dirty_flag:
            self._reset()
        else:
            self.dirty_flag = True
        return self._compile(node)

Eckhart Arnold's avatar
Eckhart Arnold committed
1127
1128
1129
1130
1131
1132
1133
    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)

1134
    @staticmethod
1135
    def derive_method_name(node_name: str) -&