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

ebnf.py 43.4 KB
Newer Older
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
# ebnf.py - EBNF -> Python-Parser compilation for DHParser
#
# 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.
17
18


19
20
21
22
23
"""
Module ``ebnf`` provides a self-hosting parser for EBNF-Grammars as
well as an EBNF-compiler that compiles an EBNF-Grammar into a
DHParser based Grammar class that can be executed to parse source text
conforming to this grammar into contrete syntax trees.
24
25
"""

26

27
import keyword
28
from collections import OrderedDict
29
from functools import partial
30

eckhart's avatar
eckhart committed
31
from DHParser.compile import CompilerError, Compiler
32
from DHParser.error import Error
33
34
from DHParser.parse import Grammar, mixin_comment, Forward, RegExp, Whitespace, \
    NegativeLookahead, Alternative, Series, Option, OneOrMore, ZeroOrMore, Token
35
from DHParser.preprocess import nil_preprocessor, PreprocessorFunc
36
from DHParser.syntaxtree import Node, WHITESPACE_PTYPE, TOKEN_PTYPE
eckhart's avatar
eckhart committed
37
38
39
from DHParser.toolkit import load_if_file, escape_re, md5, sane_parser_name, re, expand_table, \
    typing
from DHParser.transform import TransformationFunc, traverse, remove_brackets, \
Eckhart Arnold's avatar
Eckhart Arnold committed
40
    reduce_single_child, replace_by_single_child, remove_expendables, \
41
    remove_tokens, flatten, forbid, assert_content, remove_infix_operator
42
from DHParser.versionnumber import __version__
eckhart's avatar
eckhart committed
43
44
from typing import Callable, Dict, List, Set, Tuple

45

46
__all__ = ('get_ebnf_preprocessor',
47
48
49
50
           'get_ebnf_grammar',
           'get_ebnf_transformer',
           'get_ebnf_compiler',
           'EBNFGrammar',
51
           'EBNFTransform',
Eckhart Arnold's avatar
Eckhart Arnold committed
52
           'EBNFCompilerError',
53
           'EBNFCompiler',
54
           'grammar_changed',
55
           'PreprocessorFactoryFunc',
56
57
           'ParserFactoryFunc',
           'TransformerFactoryFunc',
58
           'CompilerFactoryFunc')
59
60


Eckhart Arnold's avatar
Eckhart Arnold committed
61
62
63
64
65
66
67
########################################################################
#
# EBNF scanning
#
########################################################################


68
69
def get_ebnf_preprocessor() -> PreprocessorFunc:
    return nil_preprocessor
Eckhart Arnold's avatar
Eckhart Arnold committed
70
71
72
73
74
75
76
77


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

78

di68kap's avatar
di68kap committed
79
class EBNFGrammar(Grammar):
eckhart's avatar
eckhart committed
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
    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] plaintext
                   | [flowmarker] regexp
                   | [flowmarker] whitespace
                   | [flowmarker] oneormore
                   | [flowmarker] group
                   | [flowmarker] unordered
                   | repetition
                   | option

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

        group      = "(" §expression ")"
        unordered  = "<" §expression ">"                # elements of expression in arbitrary order
        oneormore  = "{" expression "}+"
        repetition = "{" §expression "}"
        option     = "[" §expression "]"

        symbol     = /(?!\d)\w+/~                       # e.g. expression, factor, parameter_list
        literal    = /"(?:[^"]|\\")*?"/~                # e.g. "(", '+', 'while'
                   | /'(?:[^']|\\')*?'/~                # whitespace following literals will be ignored tacitly.
        plaintext  = /`(?:[^"]|\\")*?`/~                # like literal but does not eat whitespace
120
        regexp     = /\/(?:\\\/|[^\/])*?\//~            # e.g. /\w+/, ~/#.*(?:\n|$)/~
eckhart's avatar
eckhart committed
121
122
123
124
125
126
                                                        # '~' is a whitespace-marker, if present leading or trailing
                                                        # whitespace of a regular expression will be ignored tacitly.
        whitespace = /~/~                               # implicit or default whitespace
        list_      = /\w+/~ { "," /\w+/~ }              # comma separated list of symbols, e.g. BEGIN_LIST, END_LIST,
                                                        # BEGIN_QUOTE, END_QUOTE ; see CommonMark/markdown.py for an exmaple
        EOF = !/./
127
    """
di68kap's avatar
di68kap committed
128
    expression = Forward()
129
    parser_initialization__ = "upon instantiation"
di68kap's avatar
di68kap committed
130
131
    COMMENT__ = r'#.*(?:\n|$)'
    WHITESPACE__ = r'\s*'
132
    WSP_RE__ = mixin_comment(whitespace=WHITESPACE__, comment=COMMENT__)
di68kap's avatar
di68kap committed
133
    wspL__ = ''
134
    wspR__ = WSP_RE__
135
    wsp__ = Whitespace(WSP_RE__)
di68kap's avatar
di68kap committed
136
    EOF = NegativeLookahead(RegExp('.'))
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
    list_ = Series(RegExp('\\w+'), wsp__, ZeroOrMore(Series(Series(Token(","), wsp__), RegExp('\\w+'), wsp__)))
    whitespace = Series(RegExp('~'), wsp__)
    regexp = Series(RegExp('/(?:\\\\/|[^/])*?/'), wsp__)
    plaintext = Series(RegExp('`(?:[^"]|\\\\")*?`'), wsp__)
    literal = Alternative(Series(RegExp('"(?:[^"]|\\\\")*?"'), wsp__), Series(RegExp("'(?:[^']|\\\\')*?'"), wsp__))
    symbol = Series(RegExp('(?!\\d)\\w+'), wsp__)
    option = Series(Series(Token("["), wsp__), expression, Series(Token("]"), wsp__), mandatory=1)
    repetition = Series(Series(Token("{"), wsp__), expression, Series(Token("}"), wsp__), mandatory=1)
    oneormore = Series(Series(Token("{"), wsp__), expression, Series(Token("}+"), wsp__))
    unordered = Series(Series(Token("<"), wsp__), expression, Series(Token(">"), wsp__), mandatory=1)
    group = Series(Series(Token("("), wsp__), expression, Series(Token(")"), wsp__), mandatory=1)
    retrieveop = Alternative(Series(Token("::"), wsp__), Series(Token(":"), wsp__))
    flowmarker = Alternative(Series(Token("!"), wsp__), Series(Token("&"), wsp__),
                             Series(Token("-!"), wsp__), Series(Token("-&"), wsp__))
    factor = Alternative(Series(Option(flowmarker), Option(retrieveop), symbol,
                                NegativeLookahead(Series(Token("="), wsp__))),
153
154
                         Series(Option(flowmarker), literal), Series(Option(flowmarker), plaintext),
                         Series(Option(flowmarker), regexp), Series(Option(flowmarker), whitespace),
155
                         Series(Option(flowmarker), oneormore), Series(Option(flowmarker), group),
156
                         Series(Option(flowmarker), unordered), repetition, option)
157
158
159
160
161
162
163
    term = OneOrMore(Series(Option(Series(Token("§"), wsp__)), factor))
    expression.set(Series(term, ZeroOrMore(Series(Series(Token("|"), wsp__), term))))
    directive = Series(Series(Token("@"), wsp__), symbol, Series(Token("="), wsp__),
                       Alternative(regexp, literal, list_), mandatory=1)
    definition = Series(symbol, Series(Token("="), wsp__), expression, mandatory=1)
    syntax = Series(Option(Series(wsp__, RegExp(''))),
                    ZeroOrMore(Alternative(definition, directive)), EOF, mandatory=2)
di68kap's avatar
di68kap committed
164
165
166
    root__ = syntax


167
def grammar_changed(grammar_class, grammar_source: str) -> bool:
168
169
    """
    Returns ``True`` if ``grammar_class`` does not reflect the latest
Eckhart Arnold's avatar
Eckhart Arnold committed
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
    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()
188
        m = re.search('class \w*\(Grammar\)', pycode)
Eckhart Arnold's avatar
Eckhart Arnold committed
189
190
191
192
193
194
195
196
197
198
        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__


199
def get_ebnf_grammar() -> EBNFGrammar:
Eckhart Arnold's avatar
Eckhart Arnold committed
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
    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
#
########################################################################


216
EBNF_AST_transformation_table = {
217
    # AST Transformations for EBNF-grammar
218
    "+":
219
        remove_expendables,
220
    "syntax":
221
        [],  # otherwise '"*": replace_by_single_child' would be applied
222
    "directive, definition":
223
        remove_tokens('@', '='),
Eckhart Arnold's avatar
Eckhart Arnold committed
224
    "expression":
225
        [replace_by_single_child, flatten, remove_tokens('|')],  # remove_infix_operator],
226
    "term":
227
228
        [replace_by_single_child, flatten],  # supports both idioms:
                                             # "{ factor }+" and "factor { factor }"
229
    "factor, flowmarker, retrieveop":
230
        replace_by_single_child,
231
    "group":
232
        [remove_brackets, replace_by_single_child],
233
234
    "unordered":
        remove_brackets,
235
    "oneormore, repetition, option":
Eckhart Arnold's avatar
Eckhart Arnold committed
236
        [reduce_single_child, remove_brackets,
237
         forbid('repetition', 'option', 'oneormore'), assert_content(r'(?!§)(?:.|\n)*')],
238
    "symbol, literal, regexp":
Eckhart Arnold's avatar
Eckhart Arnold committed
239
        reduce_single_child,
240
    (TOKEN_PTYPE, WHITESPACE_PTYPE):
Eckhart Arnold's avatar
Eckhart Arnold committed
241
        reduce_single_child,
242
    "list_":
243
        [flatten, remove_infix_operator],
244
    "*":
245
        replace_by_single_child
246
247
}

248

Eckhart Arnold's avatar
Eckhart Arnold committed
249
def EBNFTransform() -> TransformationFunc:
250
    return partial(traverse, processing_table=EBNF_AST_transformation_table.copy())
di68kap's avatar
di68kap committed
251

252
def get_ebnf_transformer() -> TransformationFunc:
253
254
255
256
257
258
259
    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
260
261
262
263
264
265
266
267


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

268

269
PreprocessorFactoryFunc = Callable[[], PreprocessorFunc]
270
ParserFactoryFunc = Callable[[], Grammar]
271
TransformerFactoryFunc = Callable[[], TransformationFunc]
272
273
CompilerFactoryFunc = Callable[[], Compiler]

274
275
276
PREPROCESSOR_FACTORY = '''
def get_preprocessor() -> PreprocessorFunc:
    return {NAME}Preprocessor
277
278
279
280
'''


GRAMMAR_FACTORY = '''
281
def get_grammar() -> {NAME}Grammar:
282
283
284
285
286
    global thread_local_{NAME}_grammar_singleton
    try:
        grammar = thread_local_{NAME}_grammar_singleton
    except NameError:
        thread_local_{NAME}_grammar_singleton = {NAME}Grammar()
287
288
        grammar = thread_local_{NAME}_grammar_singleton
    return grammar
289
290
291
292
'''


TRANSFORMER_FACTORY = '''
293
294
295
def {NAME}Transform() -> TransformationDict:
    return partial(traverse, processing_table={NAME}_AST_transformation_table.copy())

296
def get_transformer() -> TransformationFunc:
297
298
299
300
301
302
303
    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
304
305
306
307
'''


COMPILER_FACTORY = '''
308
def get_compiler(grammar_name="{NAME}", grammar_source="") -> {NAME}Compiler:
309
310
311
312
313
314
315
    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)
316
317
        compiler = thread_local_{NAME}_compiler_singleton
    return compiler
318
319
'''

Eckhart Arnold's avatar
Eckhart Arnold committed
320

eckhart's avatar
eckhart committed
321
class EBNFCompilerError(CompilerError):
322
    """Error raised by `EBNFCompiler` class. (Not compilation errors
323
    in the strict sense, see `CompilationError` in module ``dsl.py``)"""
324
325
326
    pass


327
class EBNFCompiler(Compiler):
328
329
    """
    Generates a Parser from an abstract syntax tree of a grammar specified
330
    in EBNF-Notation.
331
332
333
334
335
336
337
338
339
340
341
342
343

    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:
344
        current_symbols:  During compilation, a list containing the root
345
346
347
348
                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.

349
        rules:  Dictionary that maps rule names to a list of Nodes that
350
351
352
353
354
355
                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`

356
                Now `[node.content for node in self.rules['alternative']]`
357
358
                yields `['alternative = a | b', 'a', 'b']`

359
        symbols:  A mapping of symbol names to their first usage (not
360
361
                their definition!) in the EBNF source.

362
        variables:  A set of symbols names that are used with the
363
364
365
366
                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.

367
        recursive:  A set of symbols that are used recursively and
368
369
                therefore require a `Forward`-operator.

370
        definitions:  A dictionary of definitions. Other than `rules`
371
372
                this maps the symbols to their compiled definienda.

373
        deferred_taks:  A list of callables that is filled during
374
375
376
377
378
                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.

379
        root:   The name of the root symbol.
380

381
        directives:  A dictionary of all directives and their default
382
                values.
383
384
385

        re_flags:  A set of regular expression flags to be added to all
                regular expressions found in the current parsing process
386
387
    """
    COMMENT_KEYWORD = "COMMENT__"
388
    WHITESPACE_KEYWORD = "WSP_RE__"
Eckhart Arnold's avatar
Eckhart Arnold committed
389
    RAW_WS_KEYWORD = "WHITESPACE__"
390
    WHITESPACE_PARSER_KEYWORD = "wsp__"
Eckhart Arnold's avatar
Eckhart Arnold committed
391
    RESERVED_SYMBOLS = {WHITESPACE_KEYWORD, RAW_WS_KEYWORD, COMMENT_KEYWORD}
392
    AST_ERROR = "Badly structured syntax tree. " \
Eckhart Arnold's avatar
Eckhart Arnold committed
393
                "Potentially due to erroneous AST transformation."
394
395
396
397
    PREFIX_TABLE = {'§': 'Required',
                    '&': 'Lookahead', '!': 'NegativeLookahead',
                    '-&': 'Lookbehind', '-!': 'NegativeLookbehind',
                    '::': 'Pop', ':': 'Retrieve'}
398
399
400
    WHITESPACE = {'horizontal': r'[\t ]*',  # default: horizontal
                  'linefeed': r'[ \t]*\n?(?!\s*\n)[ \t]*',
                  'vertical': r'\s*'}
401
    REPEATABLE_DIRECTIVES = {'tokens'}
402

403

404
    def __init__(self, grammar_name="", grammar_source=""):
Eckhart Arnold's avatar
Eckhart Arnold committed
405
        super(EBNFCompiler, self).__init__(grammar_name, grammar_source)
406
407
        self._reset()

408

409
    def _reset(self):
410
        super(EBNFCompiler, self)._reset()
411
        self._result = ''           # type: str
412
        self.re_flags = set()       # type: Set[str]
413
414
415
        self.rules = OrderedDict()  # type: OrderedDict[str, List[Node]]
        self.current_symbols = []   # type: List[Node]
        self.symbols = {}           # type: Dict[str, Node]
416
417
        self.variables = set()      # type: Set[str]
        self.recursive = set()      # type: Set[str]
418
        self.definitions = {}       # type: Dict[str, str]
419
        self.deferred_tasks = []    # type: List[Callable]
420
        self.root_symbol = ""       # type: str
eckhart's avatar
eckhart committed
421
        self.directives = {'whitespace': self.WHITESPACE['vertical'],
422
                           'comment': '',
423
                           'literalws': {'right'},
424
                           'tokens': set(),  # alt. 'preprocessor_tokens'
425
426
427
                           'filter': dict()}  # alt. 'filter'
        # self.directives['ignorecase']: False
        self.defined_directives = set()  # type: Set[str]
428

Eckhart Arnold's avatar
Eckhart Arnold committed
429
    @property
430
    def result(self) -> str:
Eckhart Arnold's avatar
Eckhart Arnold committed
431
432
        return self._result

433
    # methods for generating skeleton code for preprocessor, transformer, and compiler
434

435
    def gen_preprocessor_skeleton(self) -> str:
436
437
438
439
        """
        Returns Python-skeleton-code for a preprocessor-function for
        the previously compiled formal language.
        """
440
        name = self.grammar_name + "Preprocessor"
441
        return "def %s(text):\n    return text, lambda i: i\n" % name \
442
               + PREPROCESSOR_FACTORY.format(NAME=self.grammar_name)
443

444

445
    def gen_transformer_skeleton(self) -> str:
446
447
448
449
        """
        Returns Python-skeleton-code for the AST-transformation for the
        previously compiled formal language.
        """
450
        if not self.rules:
Eckhart Arnold's avatar
Eckhart Arnold committed
451
452
            raise EBNFCompilerError('Compiler must be run before calling '
                                    '"gen_transformer_Skeleton()"!')
453
        tt_name = self.grammar_name + '_AST_transformation_table'
di68kap's avatar
di68kap committed
454
        transtable = [tt_name + ' = {',
Eckhart Arnold's avatar
Eckhart Arnold committed
455
                      '    # AST Transformations for the ' + self.grammar_name + '-grammar']
Eckhart Arnold's avatar
Eckhart Arnold committed
456
        transtable.append('    "+": remove_empty,')
457
        for name in self.rules:
eckhart's avatar
eckhart committed
458
            transformations = '[]'
459
460
            rule = self.definitions[name]
            if rule.startswith('Alternative'):
eckhart's avatar
eckhart committed
461
                transformations = '[replace_or_reduce]'
462
            elif rule.startswith('Synonym'):
Eckhart Arnold's avatar
Eckhart Arnold committed
463
                transformations = '[reduce_single_child]'
eckhart's avatar
eckhart committed
464
            transtable.append('    "' + name + '": %s,' % transformations)
465
        transtable.append('    ":Token": reduce_single_child,')
466
        transtable += ['    "*": replace_by_single_child', '}', '']
467
        transtable += [TRANSFORMER_FACTORY.format(NAME=self.grammar_name)]
468
469
        return '\n'.join(transtable)

470

471
    def gen_compiler_skeleton(self) -> str:
472
473
474
475
        """
        Returns Python-skeleton-code for a Compiler-class for the
        previously compiled formal language.
        """
476
        if not self.rules:
477
478
            raise EBNFCompilerError('Compiler has not been run before calling '
                                    '"gen_Compiler_Skeleton()"!')
479
        compiler = ['class ' + self.grammar_name + 'Compiler(Compiler):',
480
481
482
483
                    '    """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
484
                    self.grammar_name + '", grammar_source=""):',
485
                    '        super(' + self.grammar_name +
Eckhart Arnold's avatar
Eckhart Arnold committed
486
                    'Compiler, self).__init__(grammar_name, grammar_source)',
487
488
489
490
                    r"        assert re.match('\w+\Z', grammar_name)", '',
                    '    def _reset(self):',
                    '        super()._reset()',
                    '        # initialize your variables here, not in the constructor!']
491
        for name in self.rules:
492
            method_name = Compiler.method_name(name)
493
            if name == self.root_symbol:
494
                compiler += ['    def ' + method_name + '(self, node):',
495
                             '        return self.fallback_compiler(node)', '']
496
            else:
di68kap's avatar
di68kap committed
497
                compiler += ['    # def ' + method_name + '(self, node):',
eckhart's avatar
eckhart committed
498
                             '    #     return node', '']
499
        compiler += [COMPILER_FACTORY.format(NAME=self.grammar_name)]
500
        return '\n'.join(compiler)
501

502
    def verify_transformation_table(self, transtable):
503
504
505
506
507
508
        """
        Checks for symbols that occur in the transformation-table but have
        never been defined in the grammar. Usually, this kind of
        inconsistency results from an error like a typo in the transformation
        table.
        """
509
510
511
512
513
514
515
516
        assert self._dirty_flag
        table_entries = set(expand_table(transtable).keys()) - {'*', '+', '~'}
        symbols = self.rules.keys()
        messages = []
        for entry in table_entries:
            if entry not in symbols and not entry.startswith(":"):
                messages.append(Error(('Symbol "%s" is not defined in grammar %s but appears in '
                                       'the transformation table!') % (entry, self.grammar_name),
Eckhart Arnold's avatar
Eckhart Arnold committed
517
                                      0, Error.UNDEFINED_SYMBOL_IN_TRANSFORMATION_TABLE))
518
519
        return messages

520

521
522
523
524
525
    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
        """
526
527
528
529
530
531
532
533
534
535

        # 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

536
537
538
        if self.variables:
            for i in range(len(definitions)):
                if definitions[i][0] in self.variables:
539
                    definitions[i] = (definitions[i][0], 'Capture(%s)' % definitions[i][1])
540

541
542
        # add special fields for Grammar class

543
544
        definitions.append((self.WHITESPACE_PARSER_KEYWORD,
                            'Whitespace(%s)' % self.WHITESPACE_KEYWORD))
545
        definitions.append(('wspR__', self.WHITESPACE_KEYWORD
Eckhart Arnold's avatar
Eckhart Arnold committed
546
                            if 'right' in self.directives['literalws'] else "''"))
547
        definitions.append(('wspL__', self.WHITESPACE_KEYWORD
Eckhart Arnold's avatar
Eckhart Arnold committed
548
                            if 'left' in self.directives['literalws'] else "''"))
549
        definitions.append((self.WHITESPACE_KEYWORD,
Eckhart Arnold's avatar
Eckhart Arnold committed
550
551
552
                            ("mixin_comment(whitespace=" + self.RAW_WS_KEYWORD +
                             ", comment=" + self.COMMENT_KEYWORD + ")")))
        definitions.append((self.RAW_WS_KEYWORD, "r'{whitespace}'".format(**self.directives)))
553
554
555
556
        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
557

558
        article = 'an ' if self.grammar_name[0:1] in "AaEeIiOoUu" else 'a '  # what about 'hour', 'universe' etc.?
559
        declarations = ['class ' + self.grammar_name +
560
                        'Grammar(Grammar):',
561
562
                        'r"""Parser for ' + article + self.grammar_name +
                        ' source file' +
563
                        (', with this grammar:' if self.grammar_source else '.')]
564
        definitions.append(('parser_initialization__', '"upon instantiation"'))
565
        if self.grammar_source:
566
            definitions.append(('source_hash__',
567
                                '"%s"' % md5(self.grammar_source, __version__)))
568
            declarations.append('')
569
            declarations += [line for line in self.grammar_source.split('\n')]
570
571
572
573
574
            while declarations[-1].strip() == '':
                declarations = declarations[:-1]
        declarations.append('"""')

        # turn definitions into declarations in reverse order
575

576
        self.root_symbol = definitions[0][0] if definitions else ""
577
578
579
580
581
582
583
584
        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]
585
586
587
588
589
590

        # 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:
eckhart's avatar
eckhart committed
591
                self.tree.new_error(self.symbols[symbol],
592
                               "Missing definition for symbol '%s'" % symbol)
593
                # root_node.error_flag = True
594
595
596

        # check for unconnected rules

Eckhart Arnold's avatar
Eckhart Arnold committed
597
598
599
        defined_symbols.difference_update(self.RESERVED_SYMBOLS)

        def remove_connections(symbol):
600
            """Recursively removes all symbols which appear in the
eckhart's avatar
eckhart committed
601
            definiens of a particular symbol."""
Eckhart Arnold's avatar
Eckhart Arnold committed
602
603
604
605
606
607
608
            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
609
610
611
            self.tree.new_error(self.rules[leftover][0],
                                ('Rule "%s" is not connected to parser root "%s" !') %
                                (leftover, self.root_symbol), Error.WARNING)
612

613
        # set root_symbol parser and assemble python grammar definition
614

615
616
        if self.root_symbol and 'root__' not in self.rules:
            declarations.append('root__ = ' + self.root_symbol)
617
        declarations.append('')
Eckhart Arnold's avatar
Eckhart Arnold committed
618
619
620
        self._result = '\n    '.join(declarations) \
                       + GRAMMAR_FACTORY.format(NAME=self.grammar_name)
        return self._result
621

622
623
624

    ## compilation methods

625
    def on_syntax(self, node: Node) -> str:
626
        definitions = []  # type: List[Tuple[str, str]]
627
628

        # drop the wrapping sequence node
629
630
        if len(node.children) == 1 and not node.children[0].parser.name:
            node = node.children[0]
631
632

        # compile definitions and directives and collect definitions
633
        for nd in node.children:
634
            if nd.parser.name == "definition":
635
                definitions.append(self.compile(nd))
636
            else:
637
                assert nd.parser.name == "directive", nd.as_sxpr()
638
                self.compile(nd)
639
            # node.error_flag = max(node.error_flag, nd.error_flag)
640
        self.definitions.update(definitions)
641

642
        return self.assemble_parser(definitions, node)
643

644

645
    def on_definition(self, node: Node) -> Tuple[str, str]:
646
        rule = node.children[0].content
647
        if rule in self.rules:
Eckhart Arnold's avatar
Eckhart Arnold committed
648
            first = self.rules[rule][0]
eckhart's avatar
eckhart committed
649
            if not first.errors:
eckhart's avatar
eckhart committed
650
                self.tree.new_error(first, 'First definition of rule "%s" '
651
                               'followed by illegal redefinitions.' % rule)
eckhart's avatar
eckhart committed
652
            self.tree.new_error(node, 'A rule "%s" has already been defined earlier.' % rule)
653
        elif rule in EBNFCompiler.RESERVED_SYMBOLS:
eckhart's avatar
eckhart committed
654
            self.tree.new_error(node, 'Symbol "%s" is a reserved symbol.' % rule)
655
        elif not sane_parser_name(rule):
eckhart's avatar
eckhart committed
656
            self.tree.new_error(node, 'Illegal symbol "%s". Symbols must not start or '
eckhart's avatar
eckhart committed
657
                                ' end with a doube underscore "__".' % rule)
658
        elif rule in self.directives['tokens']:
eckhart's avatar
eckhart committed
659
            self.tree.new_error(node, 'Symbol "%s" has already been defined as '
eckhart's avatar
eckhart committed
660
                                'a preprocessor token.' % rule)
661
        elif keyword.iskeyword(rule):
eckhart's avatar
eckhart committed
662
            self.tree.new_error(node, 'Python keyword "%s" may not be used as a symbol. '
eckhart's avatar
eckhart committed
663
                                % rule + '(This may change in the future.)')
664
        try:
665
666
            self.current_symbols = [node]
            self.rules[rule] = self.current_symbols
667
            defn = self.compile(node.children[1])
668
            if rule in self.variables:
669
                defn = 'Capture(%s)' % defn
670
                self.variables.remove(rule)
671
672
673
            elif defn.find("(") < 0:
                # assume it's a synonym, like 'page = REGEX_PAGE_NR'
                defn = 'Synonym(%s)' % defn
674
        except TypeError as error:
675
676
677
678
            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())
eckhart's avatar
eckhart committed
679
            self.tree.new_error(node, errmsg)
680
            rule, defn = rule + ':error', '"' + errmsg + '"'
Eckhart Arnold's avatar
Eckhart Arnold committed
681
        return rule, defn
682

683

684
    def _check_rx(self, node: Node, rx: str) -> str:
685
686
        """
        Checks whether the string `rx` represents a valid regular
687
688
689
        expression. Makes sure that multiline regular expressions are
        prepended by the multiline-flag. Returns the regular expression string.
        """
690
        flags = self.re_flags | {'x'} if rx.find('\n') >= 0 else self.re_flags
691
        if flags:  rx = "(?%s)%s" % ("".join(flags), rx)
692
693
694
        try:
            re.compile(rx)
        except Exception as re_error:
eckhart's avatar
eckhart committed
695
            self.tree.new_error(node, "malformed regular expression %s: %s" %
eckhart's avatar
eckhart committed
696
                                (repr(rx), str(re_error)))
697
698
        return rx

699

700
    def on_directive(self, node: Node) -> str:
701
        key = node.children[0].content.lower()
702
        assert key not in self.directives['tokens']
703

704
705
        if key not in self.REPEATABLE_DIRECTIVES:
            if key in self.defined_directives:
eckhart's avatar
eckhart committed
706
                self.tree.new_error(node, 'Directive "%s" has already been defined earlier. '
eckhart's avatar
eckhart committed
707
708
                                    % key + 'Later definition will be ignored!',
                                    code=Error.REDEFINED_DIRECTIVE_WARNING)
709
710
711
                return ""
            self.defined_directives.add(key)

712
        if key in {'comment', 'whitespace'}:
713
714
            if node.children[1].parser.name == "list_":
                if len(node.children[1].result) != 1:
eckhart's avatar
eckhart committed
715
                    self.tree.new_error(node, 'Directive "%s" must have one, but not %i values.'
eckhart's avatar
eckhart committed
716
                                        % (key, len(node.children[1].result)))
717
                value = self.compile(node.children[1]).pop()
718
719
                if key == 'whitespace' and value in EBNFCompiler.WHITESPACE:
                    value = EBNFCompiler.WHITESPACE[value]  # replace whitespace-name by regex
720
                else:
eckhart's avatar
eckhart committed
721
                    self.tree.new_error(node, 'Value "%s" not allowed for directive "%s".'
eckhart's avatar
eckhart committed
722
                                        % (value, key))
723
            else:
724
725
726
                value = node.children[1].content.strip("~")  # cast(str, node.children[
                # 1].result).strip("~")
                if value != node.children[1].content:  # cast(str, node.children[1].result):
eckhart's avatar
eckhart committed
727
                    self.tree.new_error(node, "Whitespace marker '~' not allowed in definition "
eckhart's avatar
eckhart committed
728
                                        "of %s regular expression." % key)
729
730
731
732
                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])
733
                if key == 'whitespace' and not re.match(value, ''):
eckhart's avatar
eckhart committed
734
                    self.tree.new_error(node, "Implicit whitespace should always "
eckhart's avatar
eckhart committed
735
                                        "match the empty string, /%s/ does not." % value)
736
            self.directives[key] = value
737

738
        elif key == 'ignorecase':
739
            if node.children[1].content.lower() not in {"off", "false", "no"}:
740
741
                self.re_flags.add('i')

Eckhart Arnold's avatar
Eckhart Arnold committed
742
        # elif key == 'testing':
743
        #     value = node.children[1].content
Eckhart Arnold's avatar
Eckhart Arnold committed
744
        #     self.directives['testing'] = value.lower() not in {"off", "false", "no"}
745

746
        elif key == 'literalws':
747
            value = {item.lower() for item in self.compile(node.children[1])}
eckhart's avatar
eckhart committed
748
            if ((value - {'left', 'right', 'both', 'none'})
Eckhart Arnold's avatar
Eckhart Arnold committed
749
                    or ('none' in value and len(value) > 1)):
eckhart's avatar
eckhart committed
750
                self.tree.new_error(node, 'Directive "literalws" allows only `left`, `right`, '
eckhart's avatar
eckhart committed
751
                                    '`both` or `none`, not `%s`' % ", ".join(value))
eckhart's avatar
eckhart committed
752
            wsp = {'left', 'right'} if 'both' in value \
753
                else {} if 'none' in value else value
eckhart's avatar
eckhart committed
754
            self.directives[key] = list(wsp)
755

756
        elif key in {'tokens', 'preprocessor_tokens'}:
757
            tokens = self.compile(node.children[1])
758
            redeclared = self.directives['tokens'] & tokens
759
            if redeclared:
eckhart's avatar
eckhart committed
760
                self.tree.new_error(node, 'Tokens %s have already been declared earlier. '
eckhart's avatar
eckhart committed
761
762
                                    % str(redeclared) + 'Later declaration will be ignored',
                                    code=Error.REDECLARED_TOKEN_WARNING)
763
            self.directives['tokens'] |= tokens - redeclared
764

765
        elif key.endswith('_filter'):
766
            filter_set = self.compile(node.children[1])
767
            if not isinstance(filter_set, set) or len(filter_set) != 1:
eckhart's avatar
eckhart committed
768
                self.tree.new_error(node, 'Directive "%s" accepts exactly on symbol, not %s'
eckhart's avatar
eckhart committed
769
                                    % (key, str(filter_set)))
770
            self.directives['filter'][key[:-7]] = filter_set.pop()
771

772
        else:
eckhart's avatar
eckhart committed
773
            self.tree.new_error(node, 'Unknown directive %s ! (Known ones are %s .)' %
eckhart's avatar
eckhart committed
774
                                (key, ', '.join(list(self.directives.keys()))))
775
776
        return ""

777

778
    def non_terminal(self, node: Node, parser_class: str, custom_args: List[str]=[]) -> str:
779
780
        """
        Compiles any non-terminal, where `parser_class` indicates the Parser class
781
782
        name for the particular non-terminal.
        """
783
        arguments = [self.compile(r) for r in node.children] + custom_args
784
785
        return parser_class + '(' + ', '.join(arguments) + ')'

786

787
    def on_expression(self, node) -> str:
di68kap's avatar
di68kap committed
788
        # TODO: Add check for errors like "a" | "ab" (which will always yield a, even for ab)
789
790
        return self.non_terminal(node, 'Alternative')

791

792
    def on_term(self, node) -> str:
di68kap's avatar
di68kap committed
793
794
795
796
797
        # 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 = []
eckhart's avatar
eckhart committed
798
        filtered_children = []  # type: List[Node]
di68kap's avatar
di68kap committed
799
        for nd in node.children:
800
            if nd.parser.ptype == TOKEN_PTYPE and nd.content == "§":
eckhart's avatar
eckhart committed
801
802
                mandatory_marker.append(len(filtered_children))
                # if len(filtered_children) == 0:
eckhart's avatar
eckhart committed
803
                #     self.tree.new_error(nd.pos, 'First item of a series should not be mandatory.',
eckhart's avatar
eckhart committed
804
                #                         Error.WARNING)
eckhart's avatar
eckhart committed
805
                if len(mandatory_marker) > 1:
eckhart's avatar
eckhart committed
806
                    self.tree.new_error(nd, 'One mandatory marker (§) sufficient to declare '
eckhart's avatar
eckhart committed
807
                                        'the rest of the series as mandatory.', Error.WARNING)
di68kap's avatar
di68kap committed
808
809
810
811
            else:
                filtered_children.append(nd)
        saved_result = node.result
        node.result = tuple(filtered_children)
eckhart's avatar
eckhart committed
812
813
814
815
816
        if len(filtered_children) == 1:
            compiled = self.non_terminal(node, 'Required')
        else:
            custom_args = ['mandatory=%i' % mandatory_marker[0]] if mandatory_marker else []
            compiled = self.non_terminal(node, 'Series', custom_args)
di68kap's avatar
di68kap committed
817
818
        node.result = saved_result
        return compiled
819

820

821
    def on_factor(self, node: Node) -> str:
822
        assert node.children
823
        assert len(node.children) >= 2, node.as_sxpr()
824
        prefix = node.children[0].content
825
        custom_args = []  # type: List[str]
826
827

        if prefix in {'::', ':'}:
828
829
            assert len(node.children) == 2
            arg = node.children[-1]
830
            if arg.parser.name != 'symbol':
eckhart's avatar
eckhart committed
831
                self.tree.new_error(node, ('Retrieve Operator "%s" requires a symbol, '
eckhart's avatar
eckhart committed
832
                                    'and not a %s.') % (prefix, str(arg.parser)))
833
                return str(arg.result)
834
            if str(arg) in self.directives['filter']:
835
                custom_args = ['rfilter=%s' % self.directives['filter'][str(arg)]]
836
            self.variables.add(str(arg))  # cast(str, arg.result)
837

838
        elif len(node.children) > 2:
839
840
            # shift = (Node(node.parser, node.result[1].result),)
            # node.result[1].result = shift + node.result[2:]
841
842
843
844
            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])
845

846
        node.result = node.children[1:]
847
848
        try:
            parser_class = self.PREFIX_TABLE[prefix]
849
850
851
852
            result = self.non_terminal(node, parser_class, custom_args)
            if prefix[:1] == '-':
                def check(node):
                    nd = node
853
854
855
                    if len(nd.children) >= 1:
                        nd = nd.children[0]
                    while nd.parser.name == "symbol":
856
                        symlist = self.rules.get(nd.content, [])
857
858
859
860
861
862
                        if len(symlist) == 2:
                            nd = symlist[1]
                        else:
                            if len(symlist) == 1:
                                nd = symlist[0].children[1]
                            break
863
864
                    if (nd.parser.name !=