ebnf.py 38.9 KB
Newer Older
1
"""ebnf.py - EBNF -> Python-Parser compilation for DHParser
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18

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.
"""

19
import keyword
20
from collections import OrderedDict
21
from functools import partial
22

23
from DHParser.error import Error
24
25
from DHParser.parser import Grammar, mixin_comment, nil_preprocessor, Forward, RegExp, RE, \
    NegativeLookahead, Alternative, Series, Option, OneOrMore, ZeroOrMore, Token, \
26
    Compiler, PreprocessorFunc
27
from DHParser.syntaxtree import Node, TransformationFunc, WHITESPACE_PTYPE, TOKEN_PTYPE
di68kap's avatar
di68kap committed
28
from DHParser.toolkit import load_if_file, escape_re, md5, sane_parser_name, re, typing
29
from DHParser.transform import traverse, remove_brackets, \
30
    reduce_single_child, replace_by_single_child, remove_expendables, \
31
    remove_tokens, flatten, forbid, assert_content, remove_infix_operator
32
from DHParser.versionnumber import __version__
di68kap's avatar
di68kap committed
33
from typing import Callable, Dict, List, Set, Tuple
34

35
__all__ = ('get_ebnf_preprocessor',
36
37
38
39
           'get_ebnf_grammar',
           'get_ebnf_transformer',
           'get_ebnf_compiler',
           'EBNFGrammar',
40
           'EBNFTransform',
Eckhart Arnold's avatar
Eckhart Arnold committed
41
           'EBNFCompilerError',
42
           'EBNFCompiler',
43
           'grammar_changed',
44
           'PreprocessorFactoryFunc',
45
46
           'ParserFactoryFunc',
           'TransformerFactoryFunc',
47
           'CompilerFactoryFunc')
48
49


Eckhart Arnold's avatar
Eckhart Arnold committed
50
51
52
53
54
55
56
########################################################################
#
# EBNF scanning
#
########################################################################


57
58
def get_ebnf_preprocessor() -> PreprocessorFunc:
    return nil_preprocessor
Eckhart Arnold's avatar
Eckhart Arnold committed
59
60
61
62
63
64
65
66


########################################################################
#
# EBNF parsing
#
########################################################################

67

di68kap's avatar
di68kap committed
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
class EBNFGrammar(Grammar):
    r"""Parser for an EBNF source file, with this grammar:

    # EBNF-Grammar in EBNF

    @ comment    =  /#.*(?:\n|$)/                    # comments start with '#' and eat all chars up to and including '\n'
    @ whitespace =  /\s*/                            # whitespace includes linefeed
    @ literalws  =  right                            # trailing whitespace of literals will be ignored tacitly

    syntax     =  [~//] { definition | directive } §EOF
    definition =  symbol §"=" expression
    directive  =  "@" §symbol "=" ( regexp | literal | list_ )

    expression =  term { "|" term }
    term       =  { ["§"] factor }+                       # "§" means all following factors mandatory
    factor     =  [flowmarker] [retrieveop] symbol !"="   # negative lookahead to be sure it's not a definition
                | [flowmarker] literal
                | [flowmarker] regexp
                | [flowmarker] oneormore
87
88
                | [flowmarker] group
                | [flowmarker] unordered
di68kap's avatar
di68kap committed
89
90
91
92
93
94
95
                | repetition
                | option

    flowmarker =  "!"  | "&"                         # '!' negative lookahead, '&' positive lookahead
                | "-!" | "-&"                        # '-' negative lookbehind, '-&' positive lookbehind
    retrieveop =  "::" | ":"                         # '::' pop, ':' retrieve

96
97
    group      =  "(" §expression ")"
    unordered  =  "<" §expression ">"                # elements of expression in arbitrary order
di68kap's avatar
di68kap committed
98
    oneormore  =  "{" expression "}+"
99
100
    repetition =  "{" §expression "}"
    option     =  "[" §expression "]"
di68kap's avatar
di68kap committed
101
102
103
104
105
106
107
108
109
110
111
112

    symbol     =  /(?!\d)\w+/~                       # e.g. expression, factor, parameter_list
    literal    =  /"(?:[^"]|\\")*?"/~                # e.g. "(", '+', 'while'
                | /'(?:[^']|\\')*?'/~                # whitespace following literals will be ignored tacitly.
    regexp     =  /~?\/(?:\\\/|[^\/])*?\/~?/~        # e.g. /\w+/, ~/#.*(?:\n|$)/~
                                                     # '~' is a whitespace-marker, if present leading or trailing
                                                     # whitespace of a regular expression will be ignored tacitly.
    list_      =  /\w+/~ { "," /\w+/~ }              # comma separated list of symbols, e.g. BEGIN_LIST, END_LIST,
                                                     # BEGIN_QUOTE, END_QUOTE ; see CommonMark/markdown.py for an exmaple
    EOF =  !/./
    """
    expression = Forward()
113
114
    source_hash__ = "3fc9f5a340f560e847d9af0b61a68743"
    parser_initialization__ = "upon instantiation"
di68kap's avatar
di68kap committed
115
116
117
118
119
120
    COMMENT__ = r'#.*(?:\n|$)'
    WHITESPACE__ = r'\s*'
    WSP__ = mixin_comment(whitespace=WHITESPACE__, comment=COMMENT__)
    wspL__ = ''
    wspR__ = WSP__
    EOF = NegativeLookahead(RegExp('.'))
Eckhart Arnold's avatar
Eckhart Arnold committed
121
    list_ = Series(RE('\\w+'), ZeroOrMore(Series(Token(","), RE('\\w+'))))
di68kap's avatar
di68kap committed
122
123
124
    regexp = RE('~?/(?:\\\\/|[^/])*?/~?')
    literal = Alternative(RE('"(?:[^"]|\\\\")*?"'), RE("'(?:[^']|\\\\')*?'"))
    symbol = RE('(?!\\d)\\w+')
125
126
    option = Series(Token("["), expression, Token("]"), mandatory=1)
    repetition = Series(Token("{"), expression, Token("}"), mandatory=1)
Eckhart Arnold's avatar
Eckhart Arnold committed
127
    oneormore = Series(Token("{"), expression, Token("}+"))
128
    unordered = Series(Token("<"), expression, Token(">"), mandatory=1)
129
    group = Series(Token("("), expression, Token(")"), mandatory=1)
di68kap's avatar
di68kap committed
130
131
    retrieveop = Alternative(Token("::"), Token(":"))
    flowmarker = Alternative(Token("!"), Token("&"), Token("-!"), Token("-&"))
132
133
134
135
136
    factor = Alternative(
        Series(Option(flowmarker), Option(retrieveop), symbol, NegativeLookahead(Token("="))),
        Series(Option(flowmarker), literal), Series(Option(flowmarker), regexp),
        Series(Option(flowmarker), oneormore), Series(Option(flowmarker), group),
        Series(Option(flowmarker), unordered), repetition, option)
Eckhart Arnold's avatar
Eckhart Arnold committed
137
138
    term = OneOrMore(Series(Option(Token("§")), factor))
    expression.set(Series(term, ZeroOrMore(Series(Token("|"), term))))
139
140
    directive = Series(Token("@"), symbol, Token("="), Alternative(regexp, literal, list_),
                       mandatory=1)
di68kap's avatar
di68kap committed
141
    definition = Series(symbol, Token("="), expression, mandatory=1)
142
143
    syntax = Series(Option(RE('', wR='', wL=WSP__)), ZeroOrMore(Alternative(definition, directive)),
                    EOF, mandatory=2)
di68kap's avatar
di68kap committed
144
145
146
    root__ = syntax


147
def grammar_changed(grammar_class, grammar_source: str) -> bool:
Eckhart Arnold's avatar
Eckhart Arnold committed
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
    """Returns ``True`` if ``grammar_class`` does not reflect the latest
    changes of ``grammar_source``

    Parameters:
        grammar_class:  the parser class representing the grammar
            or the file name of a compiler suite containing the grammar
        grammar_source:  File name or string representation of the
            EBNF code of the grammar

    Returns (bool):
        True, if the source text of the grammar is different from the
        source from which the grammar class was generated
    """
    grammar = load_if_file(grammar_source)
    chksum = md5(grammar, __version__)
    if isinstance(grammar_class, str):
        # grammar_class = load_compiler_suite(grammar_class)[1]
        with open(grammar_class, 'r', encoding='utf8') as f:
            pycode = f.read()
167
        m = re.search('class \w*\(Grammar\)', pycode)
Eckhart Arnold's avatar
Eckhart Arnold committed
168
169
170
171
172
173
174
175
176
177
        if m:
            m = re.search('    source_hash__ *= *"([a-z0-9]*)"',
                          pycode[m.span()[1]:])
            return not (m and m.groups() and m.groups()[-1] == chksum)
        else:
            return True
    else:
        return chksum != grammar_class.source_hash__


178
def get_ebnf_grammar() -> EBNFGrammar:
Eckhart Arnold's avatar
Eckhart Arnold committed
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
    global thread_local_ebnf_grammar_singleton
    try:
        grammar = thread_local_ebnf_grammar_singleton
        return grammar
    except NameError:
        thread_local_ebnf_grammar_singleton = EBNFGrammar()
        return thread_local_ebnf_grammar_singleton


########################################################################
#
# EBNF concrete to abstract syntax tree transformation and validation
#
########################################################################


195
EBNF_AST_transformation_table = {
196
    # AST Transformations for EBNF-grammar
197
    "+":
198
        remove_expendables,
199
    "syntax":
200
        [],  # otherwise '"*": replace_by_single_child' would be applied
201
    "directive, definition":
202
        remove_tokens('@', '='),
Eckhart Arnold's avatar
Eckhart Arnold committed
203
    "expression":
204
        [replace_by_single_child, flatten, remove_tokens('|')],  # remove_infix_operator],
205
206
207
208
209
    "term":
        [replace_by_single_child, flatten],  # supports both idioms:  "{ factor }+" and "factor { factor }"
    "factor, flowmarker, retrieveop":
        replace_by_single_child,
    "group":
210
        [remove_brackets, replace_by_single_child],
211
212
    "unordered":
        remove_brackets,
213
    "oneormore, repetition, option":
214
215
        [reduce_single_child, remove_brackets,
         forbid('repetition', 'option', 'oneormore'), assert_content(r'(?!§)')],
216
    "symbol, literal, regexp":
217
        reduce_single_child,
218
    (TOKEN_PTYPE, WHITESPACE_PTYPE):
219
        reduce_single_child,
220
    "list_":
221
        [flatten, remove_infix_operator],
222
    "*":
223
        replace_by_single_child
224
225
}

226

Eckhart Arnold's avatar
Eckhart Arnold committed
227
def EBNFTransform() -> TransformationFunc:
228
    return partial(traverse, processing_table=EBNF_AST_transformation_table.copy())
di68kap's avatar
di68kap committed
229

230
def get_ebnf_transformer() -> TransformationFunc:
231
232
233
234
235
236
237
    global thread_local_EBNF_transformer_singleton
    try:
        transformer = thread_local_EBNF_transformer_singleton
    except NameError:
        thread_local_EBNF_transformer_singleton = EBNFTransform()
        transformer = thread_local_EBNF_transformer_singleton
    return transformer
Eckhart Arnold's avatar
Eckhart Arnold committed
238
239
240
241
242
243
244
245


########################################################################
#
# EBNF abstract syntax tree to Python parser compilation
#
########################################################################

246

247
PreprocessorFactoryFunc = Callable[[], PreprocessorFunc]
248
ParserFactoryFunc = Callable[[], Grammar]
249
TransformerFactoryFunc = Callable[[], TransformationFunc]
250
251
CompilerFactoryFunc = Callable[[], Compiler]

252
253
254
PREPROCESSOR_FACTORY = '''
def get_preprocessor() -> PreprocessorFunc:
    return {NAME}Preprocessor
255
256
257
258
'''


GRAMMAR_FACTORY = '''
259
def get_grammar() -> {NAME}Grammar:
260
261
262
263
264
    global thread_local_{NAME}_grammar_singleton
    try:
        grammar = thread_local_{NAME}_grammar_singleton
    except NameError:
        thread_local_{NAME}_grammar_singleton = {NAME}Grammar()
265
266
        grammar = thread_local_{NAME}_grammar_singleton
    return grammar
267
268
269
270
'''


TRANSFORMER_FACTORY = '''
271
272
273
def {NAME}Transform() -> TransformationDict:
    return partial(traverse, processing_table={NAME}_AST_transformation_table.copy())

274
def get_transformer() -> TransformationFunc:
275
276
277
278
279
280
281
    global thread_local_{NAME}_transformer_singleton
    try:
        transformer = thread_local_{NAME}_transformer_singleton
    except NameError:
        thread_local_{NAME}_transformer_singleton = {NAME}Transform()
        transformer = thread_local_{NAME}_transformer_singleton
    return transformer
282
283
284
285
'''


COMPILER_FACTORY = '''
286
def get_compiler(grammar_name="{NAME}", grammar_source="") -> {NAME}Compiler:
287
288
289
290
291
292
293
    global thread_local_{NAME}_compiler_singleton
    try:
        compiler = thread_local_{NAME}_compiler_singleton
        compiler.set_grammar_name(grammar_name, grammar_source)
    except NameError:
        thread_local_{NAME}_compiler_singleton = \\
            {NAME}Compiler(grammar_name, grammar_source)
294
295
        compiler = thread_local_{NAME}_compiler_singleton
    return compiler
296
297
'''

Eckhart Arnold's avatar
Eckhart Arnold committed
298

299
300
class EBNFCompilerError(Exception):
    """Error raised by `EBNFCompiler` class. (Not compilation errors
301
    in the strict sense, see `CompilationError` in module ``dsl.py``)"""
302
303
304
    pass


305
class EBNFCompiler(Compiler):
306
307
    """
    Generates a Parser from an abstract syntax tree of a grammar specified
308
    in EBNF-Notation.
309
310
311
312
313
314
315
316
317
318
319
320
321

    Instances of this class must be called with the root-node of the
    abstract syntax tree from an EBNF-specification of a formal language.
    The returned value is the Python-source-code of a Grammar class for
    this language that can be used to parse texts in this language.
    See classes `parser.Compiler` and `parser.Grammar` for more information.

    Addionally, class EBNFCompiler provides helper methods to generate
    code-skeletons for a preprocessor, AST-transformation and full
    compilation of the formal language. These method's names start with
    the prefix `gen_`.

    Attributes:
322
        current_symbols:  During compilation, a list containing the root
323
324
325
326
                node of the currently compiled definition as first element
                and then the nodes of the symbols that are referred to in
                the currently compiled definition.

327
        rules:  Dictionary that maps rule names to a list of Nodes that
328
329
330
331
332
333
334
335
336
                contain symbol-references in the definition of the rule.
                The first item in the list is the node of the rule-
                definition itself. Example:

                           `alternative = a | b`

                Now `[str(node) for node in self.rules['alternative']]`
                yields `['alternative = a | b', 'a', 'b']`

337
        symbols:  A mapping of symbol names to their first usage (not
338
339
                their definition!) in the EBNF source.

340
        variables:  A set of symbols names that are used with the
341
342
343
344
                Pop or Retrieve operator. Because the values of these
                symbols need to be captured they are called variables.
                See `test_parser.TestPopRetrieve` for an example.

345
        recursive:  A set of symbols that are used recursively and
346
347
                therefore require a `Forward`-operator.

348
        definitions:  A dictionary of definitions. Other than `rules`
349
350
                this maps the symbols to their compiled definienda.

351
        deferred_taks:  A list of callables that is filled during
352
353
354
355
356
                compilatation, but that will be executed only after
                compilation has finished. Typically, it contains
                sementatic checks that require information that
                is only available upon completion of compilation.

357
        root:   The name of the root symbol.
358

359
        directives:  A dictionary of all directives and their default
360
                values.
361
362
363

        re_flags:  A set of regular expression flags to be added to all
                regular expressions found in the current parsing process
364
365
    """
    COMMENT_KEYWORD = "COMMENT__"
366
    WHITESPACE_KEYWORD = "WSP__"
Eckhart Arnold's avatar
Eckhart Arnold committed
367
368
    RAW_WS_KEYWORD = "WHITESPACE__"
    RESERVED_SYMBOLS = {WHITESPACE_KEYWORD, RAW_WS_KEYWORD, COMMENT_KEYWORD}
369
    AST_ERROR = "Badly structured syntax tree. " \
Eckhart Arnold's avatar
Eckhart Arnold committed
370
                "Potentially due to erroneous AST transformation."
371
372
373
374
    PREFIX_TABLE = {'§': 'Required',
                    '&': 'Lookahead', '!': 'NegativeLookahead',
                    '-&': 'Lookbehind', '-!': 'NegativeLookbehind',
                    '::': 'Pop', ':': 'Retrieve'}
375
376
377
    WHITESPACE = {'horizontal': r'[\t ]*',  # default: horizontal
                  'linefeed': r'[ \t]*\n?(?!\s*\n)[ \t]*',
                  'vertical': r'\s*'}
378

379

380
    def __init__(self, grammar_name="", grammar_source=""):
Eckhart Arnold's avatar
Eckhart Arnold committed
381
        super(EBNFCompiler, self).__init__(grammar_name, grammar_source)
382
383
        self._reset()

384

385
    def _reset(self):
386
        super(EBNFCompiler, self)._reset()
387
        self._result = ''           # type: str
388
        self.re_flags = set()       # type: Set[str]
389
390
391
        self.rules = OrderedDict()  # type: OrderedDict[str, List[Node]]
        self.current_symbols = []   # type: List[Node]
        self.symbols = {}           # type: Dict[str, Node]
392
393
        self.variables = set()      # type: Set[str]
        self.recursive = set()      # type: Set[str]
394
        self.definitions = {}       # type: Dict[str, str]
395
        self.deferred_tasks = []    # type: List[Callable]
396
        self.root_symbol = ""  # type: str
397
        self.directives = {'whitespace': self.WHITESPACE['horizontal'],
398
                           'comment': '',
399
                           'literalws': ['right'],
400
401
                           'tokens': set(),  # alt. 'preprocessor_tokens'
                           'filter': dict(),  # alt. 'filter'
Eckhart Arnold's avatar
Eckhart Arnold committed
402
                           'ignorecase': False}
403

Eckhart Arnold's avatar
Eckhart Arnold committed
404
    @property
405
    def result(self) -> str:
Eckhart Arnold's avatar
Eckhart Arnold committed
406
407
        return self._result

408
    # methods for generating skeleton code for preprocessor, transformer, and compiler
409

410
    def gen_preprocessor_skeleton(self) -> str:
411
412
413
414
        """
        Returns Python-skeleton-code for a preprocessor-function for
        the previously compiled formal language.
        """
415
        name = self.grammar_name + "Preprocessor"
416
        return "def %s(text):\n    return text\n" % name \
417
               + PREPROCESSOR_FACTORY.format(NAME=self.grammar_name)
418

419

420
    def gen_transformer_skeleton(self) -> str:
421
422
423
424
        """
        Returns Python-skeleton-code for the AST-transformation for the
        previously compiled formal language.
        """
425
        if not self.rules:
Eckhart Arnold's avatar
Eckhart Arnold committed
426
427
            raise EBNFCompilerError('Compiler must be run before calling '
                                    '"gen_transformer_Skeleton()"!')
428
        tt_name = self.grammar_name + '_AST_transformation_table'
di68kap's avatar
di68kap committed
429
        transtable = [tt_name + ' = {',
Eckhart Arnold's avatar
Eckhart Arnold committed
430
                      '    # AST Transformations for the ' + self.grammar_name + '-grammar']
Eckhart Arnold's avatar
Eckhart Arnold committed
431
        transtable.append('    "+": remove_empty,')
432
        for name in self.rules:
eckhart's avatar
eckhart committed
433
            transformations = '[]'
434
435
            rule = self.definitions[name]
            if rule.startswith('Alternative'):
eckhart's avatar
eckhart committed
436
                transformations = '[replace_or_reduce]'
437
            elif rule.startswith('Synonym'):
eckhart's avatar
eckhart committed
438
439
                transformations = '[replace_by_single_child]'
            transtable.append('    "' + name + '": %s,' % transformations)
440
        transtable.append('    ":Token, :RE": reduce_single_child,')
441
        transtable += ['    "*": replace_by_single_child', '}', '']
442
        transtable += [TRANSFORMER_FACTORY.format(NAME=self.grammar_name)]
443
444
        return '\n'.join(transtable)

445

446
    def gen_compiler_skeleton(self) -> str:
447
448
449
450
        """
        Returns Python-skeleton-code for a Compiler-class for the
        previously compiled formal language.
        """
451
        if not self.rules:
452
453
            raise EBNFCompilerError('Compiler has not been run before calling '
                                    '"gen_Compiler_Skeleton()"!')
454
        compiler = ['class ' + self.grammar_name + 'Compiler(Compiler):',
455
456
457
458
                    '    """Compiler for the abstract-syntax-tree of a ' +
                    self.grammar_name + ' source file.',
                    '    """', '',
                    '    def __init__(self, grammar_name="' +
Eckhart Arnold's avatar
Eckhart Arnold committed
459
                    self.grammar_name + '", grammar_source=""):',
460
                    '        super(' + self.grammar_name +
Eckhart Arnold's avatar
Eckhart Arnold committed
461
                    'Compiler, self).__init__(grammar_name, grammar_source)',
eckhart's avatar
eckhart committed
462
                    r"        assert re.match('\w+\Z', grammar_name)", '']
463
        for name in self.rules:
464
            method_name = Compiler.method_name(name)
465
            if name == self.root_symbol:
466
                compiler += ['    def ' + method_name + '(self, node):',
467
468
                             '        return node', '']
            else:
469
                compiler += ['    def ' + method_name + '(self, node):',
470
                             '        pass', '']
471
        compiler += [COMPILER_FACTORY.format(NAME=self.grammar_name)]
472
        return '\n'.join(compiler)
473

474

475
476
477
478
479
    def assemble_parser(self, definitions: List[Tuple[str, str]], root_node: Node) -> str:
        """
        Creates the Python code for the parser after compilation of
        the EBNF-Grammar
        """
480
481
482
483
484
485
486
487
488
489

        # execute deferred tasks, for example semantic checks that cannot
        # be done before the symbol table is complete

        for task in self.deferred_tasks:
            task()

        # provide for capturing of symbols that are variables, i.e. the
        # value of will be retrieved at some point during the parsing process

490
491
492
        if self.variables:
            for i in range(len(definitions)):
                if definitions[i][0] in self.variables:
493
                    definitions[i] = (definitions[i][0], 'Capture(%s)' % definitions[i][1])
494

495
496
        # add special fields for Grammar class

497
        definitions.append(('wspR__', self.WHITESPACE_KEYWORD
Eckhart Arnold's avatar
Eckhart Arnold committed
498
                            if 'right' in self.directives['literalws'] else "''"))
499
        definitions.append(('wspL__', self.WHITESPACE_KEYWORD
Eckhart Arnold's avatar
Eckhart Arnold committed
500
                            if 'left' in self.directives['literalws'] else "''"))
501
        definitions.append((self.WHITESPACE_KEYWORD,
Eckhart Arnold's avatar
Eckhart Arnold committed
502
503
504
                            ("mixin_comment(whitespace=" + self.RAW_WS_KEYWORD +
                             ", comment=" + self.COMMENT_KEYWORD + ")")))
        definitions.append((self.RAW_WS_KEYWORD, "r'{whitespace}'".format(**self.directives)))
505
506
507
508
        definitions.append((self.COMMENT_KEYWORD, "r'{comment}'".format(**self.directives)))

        # prepare parser class header and docstring and
        # add EBNF grammar to the doc string of the parser class
509

510
        article = 'an ' if self.grammar_name[0:1] in "AaEeIiOoUu" else 'a '  # what about 'hour', 'universe' etc.?
511
        declarations = ['class ' + self.grammar_name +
512
                        'Grammar(Grammar):',
513
514
                        'r"""Parser for ' + article + self.grammar_name +
                        ' source file' +
515
                        (', with this grammar:' if self.grammar_source else '.')]
516
        definitions.append(('parser_initialization__', '"upon instantiation"'))
517
        if self.grammar_source:
518
            definitions.append(('source_hash__',
519
                                '"%s"' % md5(self.grammar_source, __version__)))
520
            declarations.append('')
521
            declarations += [line for line in self.grammar_source.split('\n')]
522
523
524
525
526
            while declarations[-1].strip() == '':
                declarations = declarations[:-1]
        declarations.append('"""')

        # turn definitions into declarations in reverse order
527

528
        self.root_symbol = definitions[0][0] if definitions else ""
529
530
531
532
533
534
535
536
        definitions.reverse()
        declarations += [symbol + ' = Forward()'
                         for symbol in sorted(list(self.recursive))]
        for symbol, statement in definitions:
            if symbol in self.recursive:
                declarations += [symbol + '.set(' + statement + ')']
            else:
                declarations += [symbol + ' = ' + statement]
537
538
539
540
541
542
543

        # check for symbols used but never defined

        defined_symbols = set(self.rules.keys()) | self.RESERVED_SYMBOLS
        for symbol in self.symbols:
            if symbol not in defined_symbols:
                self.symbols[symbol].add_error("Missing definition for symbol '%s'" % symbol)
544
                # root_node.error_flag = True
545
546
547

        # check for unconnected rules

Eckhart Arnold's avatar
Eckhart Arnold committed
548
549
550
        defined_symbols.difference_update(self.RESERVED_SYMBOLS)

        def remove_connections(symbol):
eckhart's avatar
eckhart committed
551
552
            """Recursiviely removes all symbols which appear in the
            definiens of a particular symbol."""
Eckhart Arnold's avatar
Eckhart Arnold committed
553
554
555
556
557
558
559
            if symbol in defined_symbols:
                defined_symbols.remove(symbol)
                for related in self.rules[symbol][1:]:
                    remove_connections(str(related))

        remove_connections(self.root_symbol)
        for leftover in defined_symbols:
eckhart's avatar
eckhart committed
560
561
562
            self.rules[leftover][0].add_error(
                ('Rule "%s" is not connected to parser root "%s" !') % 
                (leftover, self.root_symbol), Error.WARNING)
563

564
        # set root_symbol parser and assemble python grammar definition
565

566
567
        if self.root_symbol and 'root__' not in self.rules:
            declarations.append('root__ = ' + self.root_symbol)
568
        declarations.append('')
Eckhart Arnold's avatar
Eckhart Arnold committed
569
570
571
        self._result = '\n    '.join(declarations) \
                       + GRAMMAR_FACTORY.format(NAME=self.grammar_name)
        return self._result
572

573
574
575

    ## compilation methods

576
    def on_syntax(self, node: Node) -> str:
577
        definitions = []  # type: List[Tuple[str, str]]
578

eckhart's avatar
eckhart committed
579
580
581
582
583
584
        # transfer directives to re_flags where needed
        if self.directives['ignorecase']:
            self.re_flags.add('i')
        elif 'i' in self.re_flags:
            self.re_flags.remove('i')

585
        # drop the wrapping sequence node
586
587
        if len(node.children) == 1 and not node.children[0].parser.name:
            node = node.children[0]
588
589

        # compile definitions and directives and collect definitions
590
        for nd in node.children:
591
            if nd.parser.name == "definition":
592
                definitions.append(self.compile(nd))
593
            else:
594
                assert nd.parser.name == "directive", nd.as_sxpr()
595
                self.compile(nd)
596
            node.error_flag = max(node.error_flag, nd.error_flag)
597
        self.definitions.update(definitions)
598

599
        return self.assemble_parser(definitions, node)
600

601

602
    def on_definition(self, node: Node) -> Tuple[str, str]:
603
        rule = str(node.children[0])
604
        if rule in self.rules:
Eckhart Arnold's avatar
Eckhart Arnold committed
605
606
607
608
609
            first = self.rules[rule][0]
            if not first._errors:
                first.add_error('First definition of rule "%s" '
                                'followed by illegal redefinitions.' % rule)
            node.add_error('A rule with name "%s" has already been defined earlier.' % rule)
610
        elif rule in EBNFCompiler.RESERVED_SYMBOLS:
611
612
613
614
            node.add_error('Symbol "%s" is a reserved symbol.' % rule)
        elif not sane_parser_name(rule):
            node.add_error('Illegal symbol "%s". Symbols must not start or '
                           ' end with a doube underscore "__".' % rule)
615
        elif rule in self.directives['tokens']:
616
            node.add_error('Symbol "%s" has already been defined as '
617
                           'a preprocessor token.' % rule)
618
619
        elif keyword.iskeyword(rule):
            node.add_error('Python keyword "%s" may not be used as a symbol. '
620
                           % rule + '(This may change in the future.)')
621
        try:
622
623
            self.current_symbols = [node]
            self.rules[rule] = self.current_symbols
624
            defn = self.compile(node.children[1])
625
            if rule in self.variables:
626
                defn = 'Capture(%s)' % defn
627
                self.variables.remove(rule)
628
629
630
            elif defn.find("(") < 0:
                # assume it's a synonym, like 'page = REGEX_PAGE_NR'
                defn = 'Synonym(%s)' % defn
631
        except TypeError as error:
632
633
634
635
            from traceback import extract_tb
            trace = str(extract_tb(error.__traceback__)[-1])
            errmsg = "%s (TypeError: %s; %s)\n%s" \
                     % (EBNFCompiler.AST_ERROR, str(error), trace, node.as_sxpr())
636
637
            node.add_error(errmsg)
            rule, defn = rule + ':error', '"' + errmsg + '"'
Eckhart Arnold's avatar
Eckhart Arnold committed
638
        return rule, defn
639

640

641
    def _check_rx(self, node: Node, rx: str) -> str:
642
643
        """
        Checks whether the string `rx` represents a valid regular
644
645
646
        expression. Makes sure that multiline regular expressions are
        prepended by the multiline-flag. Returns the regular expression string.
        """
647
        flags = self.re_flags | {'x'} if rx.find('\n') >= 0 else self.re_flags
648
        if flags:  rx = "(?%s)%s" % ("".join(flags), rx)
649
650
651
652
653
654
655
        try:
            re.compile(rx)
        except Exception as re_error:
            node.add_error("malformed regular expression %s: %s" %
                           (repr(rx), str(re_error)))
        return rx

656

657
    def on_directive(self, node: Node) -> str:
658
        key = str(node.children[0]).lower()
659
        assert key not in self.directives['tokens']
660

661
        if key in {'comment', 'whitespace'}:
662
663
            if node.children[1].parser.name == "list_":
                if len(node.children[1].result) != 1:
Eckhart Arnold's avatar
Eckhart Arnold committed
664
                    node.add_error('Directive "%s" must have one, but not %i values.' %
665
                                   (key, len(node.children[1].result)))
666
                value = self.compile(node.children[1]).pop()
667
668
                if key == 'whitespace' and value in EBNFCompiler.WHITESPACE:
                    value = EBNFCompiler.WHITESPACE[value]  # replace whitespace-name by regex
669
                else:
670
                    node.add_error('Value "%s" not allowed for directive "%s".' % (value, key))
671
            else:
672
673
                value = str(node.children[1]).strip("~")  # cast(str, node.children[1].result).strip("~")
                if value != str(node.children[1]):  # cast(str, node.children[1].result):
674
675
676
677
678
679
                    node.add_error("Whitespace marker '~' not allowed in definition of "
                                   "%s regular expression." % key)
                if value[0] + value[-1] in {'""', "''"}:
                    value = escape_re(value[1:-1])
                elif value[0] + value[-1] == '//':
                    value = self._check_rx(node, value[1:-1])
680
681
682
                if key == 'whitespace' and not re.match(value, ''):
                    node.add_error("Implicit whitespace should always match the empty string, "
                                   "/%s/ does not." % value)
683
            self.directives[key] = value
684

685
686
687
688
689
        elif key == 'ignorecase':
            value = str(node.children[1]).lower() not in {"off", "false", "no"}
            self.directives['ignorecase'] == value
            if value:
                self.re_flags.add('i')
eckhart's avatar
eckhart committed
690
691
            elif 'i' in self.re_flags:
                self.re_flags.remove('i')
692

Eckhart Arnold's avatar
Eckhart Arnold committed
693
694
695
        # elif key == 'testing':
        #     value = str(node.children[1])
        #     self.directives['testing'] = value.lower() not in {"off", "false", "no"}
696

697
        elif key == 'literalws':
698
            value = {item.lower() for item in self.compile(node.children[1])}
eckhart's avatar
eckhart committed
699
            if ((value - {'left', 'right', 'both', 'none'})
Eckhart Arnold's avatar
Eckhart Arnold committed
700
                    or ('none' in value and len(value) > 1)):
701
702
703
                node.add_error('Directive "literalws" allows the values '
                               '`left`, `right`, `both` or `none`, '
                               'but not `%s`' % ", ".join(value))
eckhart's avatar
eckhart committed
704
            wsp = {'left', 'right'} if 'both' in value \
705
                else {} if 'none' in value else value
eckhart's avatar
eckhart committed
706
            self.directives[key] = list(wsp)
707

708
        elif key in {'tokens', 'preprocessor_tokens'}:
709
            self.directives['tokens'] |= self.compile(node.children[1])
710

711
        elif key.endswith('_filter'):
712
            filter_set = self.compile(node.children[1])
713
714
715
716
            if not isinstance(filter_set, set) or len(filter_set) != 1:
                node.add_error('Directive "%s" accepts exactly on symbol, not %s'
                               % (key, str(filter_set)))
            self.directives['filter'][key[:-7]] = filter_set.pop()
717

718
719
        else:
            node.add_error('Unknown directive %s ! (Known ones are %s .)' %
720
                           (key, ', '.join(list(self.directives.keys()))))
721
722
        return ""

723

724
    def non_terminal(self, node: Node, parser_class: str, custom_args: List[str]=[]) -> str:
725
726
        """
        Compiles any non-terminal, where `parser_class` indicates the Parser class
727
728
        name for the particular non-terminal.
        """
729
        arguments = [self.compile(r) for r in node.children] + custom_args
730
        node.error_flag = max(node.error_flag, max(t.error_flag for t in node.children))
731
732
        return parser_class + '(' + ', '.join(arguments) + ')'

733

734
    def on_expression(self, node) -> str:
di68kap's avatar
di68kap committed
735
        # TODO: Add check for errors like "a" | "ab" (which will always yield a, even for ab)
736
737
        return self.non_terminal(node, 'Alternative')

738

739
    def on_term(self, node) -> str:
di68kap's avatar
di68kap committed
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
        # Basically, the following code does only this:
        #       return self.non_terminal(node, 'Series')
        # What makes it (look) more complicated is the handling of the
        # mandatory §-operator
        mandatory_marker = []
        filtered_children = []
        i = 0
        for nd in node.children:
            if nd.parser.ptype == TOKEN_PTYPE and str(nd) == "§":
                mandatory_marker.append(i)
                if i == 0:
                    nd.add_error('First item of a series should not be mandatory.',
                                 Error.WARNING)
                elif len(mandatory_marker) > 1:
                    nd.add_error('One mandatory marker (§) sufficient to declare the '
                                 'rest of the series as mandatory.', Error.WARNING)
            else:
                filtered_children.append(nd)
                i += 1
        saved_result = node.result
        node.result = tuple(filtered_children)
eckhart's avatar
eckhart committed
761
        custom_args = ['mandatory=%i' % mandatory_marker[0]] if mandatory_marker else []
di68kap's avatar
di68kap committed
762
763
764
        compiled = self.non_terminal(node, 'Series', custom_args)
        node.result = saved_result
        return compiled
765

766

767
    def on_factor(self, node: Node) -> str:
768
        assert node.children
769
        assert len(node.children) >= 2, node.as_sxpr()
770
        prefix = str(node.children[0])  # cast(str, node.children[0].result)
771
        custom_args = []  # type: List[str]
772
773

        if prefix in {'::', ':'}:
774
775
            assert len(node.children) == 2
            arg = node.children[-1]
776
            if arg.parser.name != 'symbol':
Eckhart Arnold's avatar
Eckhart Arnold committed
777
                node.add_error(('Retrieve Operator "%s" requires a symbol, '
778
779
                                'and not a %s.') % (prefix, str(arg.parser)))
                return str(arg.result)
780
            if str(arg) in self.directives['filter']:
781
                custom_args = ['rfilter=%s' % self.directives['filter'][str(arg)]]
782
            self.variables.add(str(arg))  # cast(str, arg.result)
783

784
        elif len(node.children) > 2:
785
786
            # shift = (Node(node.parser, node.result[1].result),)
            # node.result[1].result = shift + node.result[2:]
787
788
789
790
            node.children[1].result = (Node(node.children[1].parser, node.children[1].result),) \
                                    + node.children[2:]
            node.children[1].parser = node.parser
            node.result = (node.children[0], node.children[1])
791

792
        node.result = node.children[1:]
793
794
        try:
            parser_class = self.PREFIX_TABLE[prefix]
795
796
797
798
            result = self.non_terminal(node, parser_class, custom_args)
            if prefix[:1] == '-':
                def check(node):
                    nd = node
799
800
801
802
803
804
805
806
807
808
                    if len(nd.children) >= 1:
                        nd = nd.children[0]
                    while nd.parser.name == "symbol":
                        symlist = self.rules.get(str(nd), [])
                        if len(symlist) == 2:
                            nd = symlist[1]
                        else:
                            if len(symlist) == 1:
                                nd = symlist[0].children[1]
                            break
809
810
811
                    if (nd.parser.name != "regexp" or str(nd)[:1] != '/'
                        or str(nd)[-1:] != '/'):
                        node.add_error("Lookbehind-parser can only be used with plain RegExp-"
812
                                       "parsers, not with: " + nd.parser.name + nd.parser.ptype)
813
814
815
816

                if not result.startswith('RegExp('):
                    self.deferred_tasks.append(lambda: check(node))
            return result
817
818
        except KeyError:
            node.add_error('Unknown prefix "%s".' % prefix)
819
        return ""
820

821

822
    def on_option(self, node) -> str:
823
        return self.non_terminal(node, 'Option')
824

825

826
    def on_repetition(self, node) -> str:
827
828
        return self.non_terminal(node, 'ZeroOrMore')

829

830
    def on_oneormore(self, node) -> str:
831
832
        return self.non_terminal(node, 'OneOrMore')

833

834
    def on_regexchain(self, node) -> str:
835
836
        raise EBNFCompilerError("Not yet implemented!")

837

838
    def on_group(self, node) -> str:
839
840
841
        raise EBNFCompilerError("Group nodes should have been eliminated by "
                                "AST transformation!")

842
    def on_unordered(self, node) -> str:
843
844
845
        # return self.non_terminal(node, 'Unordered')
        assert len(node.children) == 1
        nd = node.children[0]
eckhart's avatar
eckhart committed
846
847
        for child in nd.children:
            if child.parser.ptype == TOKEN_PTYPE and str(nd) == "§":
848
                node.add_error("Unordered parser lists cannot contain mandatory (§) items.")
eckhart's avatar
eckhart committed
849
        args = ', '.join(self.compile(child) for child in nd.children)
850
851
852
853
854
855
856
        if nd.parser.name == "term":
            return "AllOf(" + args + ")"
        elif nd.parser.name == "expression":
            return "SomeOf(" + args + ")"
        else:
            node.add_error("Unordered sequence or alternative requires at least two elements.")
            return ""
857

858
859
860
    def on_symbol(self, node: Node) -> str:     # called only for symbols on the right hand side!
        symbol = str(node)  # ; assert result == cast(str, node.result)
        if symbol in self.directives['tokens']:
861
            return 'PreprocessorToken("' + symbol + '")'
862
        else:
863
864
            self.current_symbols.append(node)
            if symbol not in self.symbols:
865
                self.symbols[symbol] = node  # remember first use of symbol
866
867
            if symbol in self.rules:
                self.recursive.add(symbol)
Eckhart Arnold's avatar
Eckhart Arnold committed
868
            if symbol in EBNFCompiler.RESERVED_SYMBOLS:  # (EBNFCompiler.WHITESPACE_KEYWORD, EBNFCompiler.COMMENT_KEYWORD):
Eckhart Arnold's avatar
Eckhart Arnold committed
869
                return "RegExp(%s)" % symbol
870
            return symbol
871

872

873
    def on_literal(self, node) -> str:
874
        return 'Token(' + str(node).replace('\\', r'\\') + ')'  # return 'Token(' + ', '.merge_children([node.result]) + ')' ?
875

876

877
    def on_regexp(self, node: Node) -> str:
878
        rx = str(node)
879
        name = []   # type: List[str]
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
        if rx[0] == '/' and rx[-1] == '/':
            parser = 'RegExp('
        else:
            parser = 'RE('
            if rx[:2] == '~/':
                if not 'left' in self.directives['literalws']:
                    name = ['wL=' + self.WHITESPACE_KEYWORD] + name
                rx = rx[1:]
            elif 'left' in self.directives['literalws']:
                name = ["wL=''"] + name
            if rx[-2:] == '/~':
                if 'right' not in self.directives['literalws']:
                    name = ['wR=' + self.WHITESPACE_KEYWORD] + name
                rx = rx[:-1]
            elif 'right' in self.directives['literalws']:
                name = ["wR=''"] + name
896
897
898
        try:
            arg = repr(self._check_rx(node, rx[1:-1].replace(r'\/', '/')))
        except AttributeError as error: