ebnf.py 43.3 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__)
133
    wsp__ = Whitespace(WSP_RE__)
di68kap's avatar
di68kap committed
134
    EOF = NegativeLookahead(RegExp('.'))
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
    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__))),
151
152
                         Series(Option(flowmarker), literal), Series(Option(flowmarker), plaintext),
                         Series(Option(flowmarker), regexp), Series(Option(flowmarker), whitespace),
153
                         Series(Option(flowmarker), oneormore), Series(Option(flowmarker), group),
154
                         Series(Option(flowmarker), unordered), repetition, option)
155
156
157
158
159
160
161
    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
162
163
164
    root__ = syntax


165
def grammar_changed(grammar_class, grammar_source: str) -> bool:
166
167
    """
    Returns ``True`` if ``grammar_class`` does not reflect the latest
Eckhart Arnold's avatar
Eckhart Arnold committed
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
    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()
186
        m = re.search('class \w*\(Grammar\)', pycode)
Eckhart Arnold's avatar
Eckhart Arnold committed
187
188
189
190
191
192
193
194
195
196
        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__


197
def get_ebnf_grammar() -> EBNFGrammar:
Eckhart Arnold's avatar
Eckhart Arnold committed
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
    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
#
########################################################################


214
EBNF_AST_transformation_table = {
215
    # AST Transformations for EBNF-grammar
216
    "+":
217
        remove_expendables,
218
    "syntax":
219
        [],  # otherwise '"*": replace_by_single_child' would be applied
220
    "directive, definition":
221
        remove_tokens('@', '='),
Eckhart Arnold's avatar
Eckhart Arnold committed
222
    "expression":
223
        [replace_by_single_child, flatten, remove_tokens('|')],  # remove_infix_operator],
224
    "term":
225
226
        [replace_by_single_child, flatten],  # supports both idioms:
                                             # "{ factor }+" and "factor { factor }"
227
    "factor, flowmarker, retrieveop":
228
        replace_by_single_child,
229
    "group":
230
        [remove_brackets, replace_by_single_child],
231
232
    "unordered":
        remove_brackets,
233
    "oneormore, repetition, option":
Eckhart Arnold's avatar
Eckhart Arnold committed
234
        [reduce_single_child, remove_brackets,
235
         forbid('repetition', 'option', 'oneormore'), assert_content(r'(?!§)(?:.|\n)*')],
236
    "symbol, literal, regexp":
Eckhart Arnold's avatar
Eckhart Arnold committed
237
        reduce_single_child,
238
    (TOKEN_PTYPE, WHITESPACE_PTYPE):
Eckhart Arnold's avatar
Eckhart Arnold committed
239
        reduce_single_child,
240
    "list_":
241
        [flatten, remove_infix_operator],
242
    "*":
243
        replace_by_single_child
244
245
}

246

Eckhart Arnold's avatar
Eckhart Arnold committed
247
def EBNFTransform() -> TransformationFunc:
248
    return partial(traverse, processing_table=EBNF_AST_transformation_table.copy())
di68kap's avatar
di68kap committed
249

250
def get_ebnf_transformer() -> TransformationFunc:
251
252
253
254
255
256
257
    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
258
259
260
261
262
263
264
265


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

266

267
PreprocessorFactoryFunc = Callable[[], PreprocessorFunc]
268
ParserFactoryFunc = Callable[[], Grammar]
269
TransformerFactoryFunc = Callable[[], TransformationFunc]
270
271
CompilerFactoryFunc = Callable[[], Compiler]

272
273
274
PREPROCESSOR_FACTORY = '''
def get_preprocessor() -> PreprocessorFunc:
    return {NAME}Preprocessor
275
276
277
278
'''


GRAMMAR_FACTORY = '''
279
def get_grammar() -> {NAME}Grammar:
280
281
282
283
284
    global thread_local_{NAME}_grammar_singleton
    try:
        grammar = thread_local_{NAME}_grammar_singleton
    except NameError:
        thread_local_{NAME}_grammar_singleton = {NAME}Grammar()
285
286
        grammar = thread_local_{NAME}_grammar_singleton
    return grammar
287
288
289
290
'''


TRANSFORMER_FACTORY = '''
291
292
293
def {NAME}Transform() -> TransformationDict:
    return partial(traverse, processing_table={NAME}_AST_transformation_table.copy())

294
def get_transformer() -> TransformationFunc:
295
296
297
298
299
300
301
    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
302
303
304
305
'''


COMPILER_FACTORY = '''
306
def get_compiler(grammar_name="{NAME}", grammar_source="") -> {NAME}Compiler:
307
308
309
310
311
312
313
    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)
314
315
        compiler = thread_local_{NAME}_compiler_singleton
    return compiler
316
317
'''

Eckhart Arnold's avatar
Eckhart Arnold committed
318

eckhart's avatar
eckhart committed
319
class EBNFCompilerError(CompilerError):
320
    """Error raised by `EBNFCompiler` class. (Not compilation errors
321
    in the strict sense, see `CompilationError` in module ``dsl.py``)"""
322
323
324
    pass


325
class EBNFCompiler(Compiler):
326
327
    """
    Generates a Parser from an abstract syntax tree of a grammar specified
328
    in EBNF-Notation.
329
330
331
332
333
334
335
336
337
338
339
340
341

    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:
342
        current_symbols:  During compilation, a list containing the root
343
344
345
346
                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.

347
        rules:  Dictionary that maps rule names to a list of Nodes that
348
349
350
351
352
353
                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`

354
                Now `[node.content for node in self.rules['alternative']]`
355
356
                yields `['alternative = a | b', 'a', 'b']`

357
        symbols:  A mapping of symbol names to their first usage (not
358
359
                their definition!) in the EBNF source.

360
        variables:  A set of symbols names that are used with the
361
362
363
364
                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.

365
        recursive:  A set of symbols that are used recursively and
366
367
                therefore require a `Forward`-operator.

368
        definitions:  A dictionary of definitions. Other than `rules`
369
370
                this maps the symbols to their compiled definienda.

371
        deferred_taks:  A list of callables that is filled during
372
373
374
375
376
                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.

377
        root:   The name of the root symbol.
378

379
        directives:  A dictionary of all directives and their default
380
                values.
381
382
383

        re_flags:  A set of regular expression flags to be added to all
                regular expressions found in the current parsing process
384
385
    """
    COMMENT_KEYWORD = "COMMENT__"
386
    WHITESPACE_KEYWORD = "WSP_RE__"
Eckhart Arnold's avatar
Eckhart Arnold committed
387
    RAW_WS_KEYWORD = "WHITESPACE__"
388
    WHITESPACE_PARSER_KEYWORD = "wsp__"
Eckhart Arnold's avatar
Eckhart Arnold committed
389
    RESERVED_SYMBOLS = {WHITESPACE_KEYWORD, RAW_WS_KEYWORD, COMMENT_KEYWORD}
390
    AST_ERROR = "Badly structured syntax tree. " \
Eckhart Arnold's avatar
Eckhart Arnold committed
391
                "Potentially due to erroneous AST transformation."
392
393
394
395
    PREFIX_TABLE = {'§': 'Required',
                    '&': 'Lookahead', '!': 'NegativeLookahead',
                    '-&': 'Lookbehind', '-!': 'NegativeLookbehind',
                    '::': 'Pop', ':': 'Retrieve'}
396
397
398
    WHITESPACE = {'horizontal': r'[\t ]*',  # default: horizontal
                  'linefeed': r'[ \t]*\n?(?!\s*\n)[ \t]*',
                  'vertical': r'\s*'}
399
    REPEATABLE_DIRECTIVES = {'tokens'}
400

401

402
    def __init__(self, grammar_name="", grammar_source=""):
Eckhart Arnold's avatar
Eckhart Arnold committed
403
        super(EBNFCompiler, self).__init__(grammar_name, grammar_source)
404
405
        self._reset()

406

407
    def _reset(self):
408
        super(EBNFCompiler, self)._reset()
409
        self._result = ''           # type: str
410
        self.re_flags = set()       # type: Set[str]
411
412
413
        self.rules = OrderedDict()  # type: OrderedDict[str, List[Node]]
        self.current_symbols = []   # type: List[Node]
        self.symbols = {}           # type: Dict[str, Node]
414
415
        self.variables = set()      # type: Set[str]
        self.recursive = set()      # type: Set[str]
416
        self.definitions = {}       # type: Dict[str, str]
417
        self.deferred_tasks = []    # type: List[Callable]
418
        self.root_symbol = ""       # type: str
eckhart's avatar
eckhart committed
419
        self.directives = {'whitespace': self.WHITESPACE['vertical'],
420
                           'comment': '',
421
                           'literalws': {'right'},
422
                           'tokens': set(),  # alt. 'preprocessor_tokens'
423
424
425
                           'filter': dict()}  # alt. 'filter'
        # self.directives['ignorecase']: False
        self.defined_directives = set()  # type: Set[str]
426

Eckhart Arnold's avatar
Eckhart Arnold committed
427
    @property
428
    def result(self) -> str:
Eckhart Arnold's avatar
Eckhart Arnold committed
429
430
        return self._result

431
    # methods for generating skeleton code for preprocessor, transformer, and compiler
432

433
    def gen_preprocessor_skeleton(self) -> str:
434
435
436
437
        """
        Returns Python-skeleton-code for a preprocessor-function for
        the previously compiled formal language.
        """
438
        name = self.grammar_name + "Preprocessor"
439
        return "def %s(text):\n    return text, lambda i: i\n" % name \
440
               + PREPROCESSOR_FACTORY.format(NAME=self.grammar_name)
441

442

443
    def gen_transformer_skeleton(self) -> str:
444
445
446
447
        """
        Returns Python-skeleton-code for the AST-transformation for the
        previously compiled formal language.
        """
448
        if not self.rules:
Eckhart Arnold's avatar
Eckhart Arnold committed
449
450
            raise EBNFCompilerError('Compiler must be run before calling '
                                    '"gen_transformer_Skeleton()"!')
451
        tt_name = self.grammar_name + '_AST_transformation_table'
di68kap's avatar
di68kap committed
452
        transtable = [tt_name + ' = {',
Eckhart Arnold's avatar
Eckhart Arnold committed
453
                      '    # AST Transformations for the ' + self.grammar_name + '-grammar']
Eckhart Arnold's avatar
Eckhart Arnold committed
454
        transtable.append('    "+": remove_empty,')
455
        for name in self.rules:
eckhart's avatar
eckhart committed
456
            transformations = '[]'
457
458
            rule = self.definitions[name]
            if rule.startswith('Alternative'):
eckhart's avatar
eckhart committed
459
                transformations = '[replace_or_reduce]'
460
            elif rule.startswith('Synonym'):
Eckhart Arnold's avatar
Eckhart Arnold committed
461
                transformations = '[reduce_single_child]'
eckhart's avatar
eckhart committed
462
            transtable.append('    "' + name + '": %s,' % transformations)
463
        transtable.append('    ":Token": reduce_single_child,')
464
        transtable += ['    "*": replace_by_single_child', '}', '']
465
        transtable += [TRANSFORMER_FACTORY.format(NAME=self.grammar_name)]
466
467
        return '\n'.join(transtable)

468

469
    def gen_compiler_skeleton(self) -> str:
470
471
472
473
        """
        Returns Python-skeleton-code for a Compiler-class for the
        previously compiled formal language.
        """
474
        if not self.rules:
475
476
            raise EBNFCompilerError('Compiler has not been run before calling '
                                    '"gen_Compiler_Skeleton()"!')
477
        compiler = ['class ' + self.grammar_name + 'Compiler(Compiler):',
478
479
480
481
                    '    """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
482
                    self.grammar_name + '", grammar_source=""):',
483
                    '        super(' + self.grammar_name +
Eckhart Arnold's avatar
Eckhart Arnold committed
484
                    'Compiler, self).__init__(grammar_name, grammar_source)',
485
486
487
488
                    r"        assert re.match('\w+\Z', grammar_name)", '',
                    '    def _reset(self):',
                    '        super()._reset()',
                    '        # initialize your variables here, not in the constructor!']
489
        for name in self.rules:
490
            method_name = Compiler.method_name(name)
491
            if name == self.root_symbol:
492
                compiler += ['    def ' + method_name + '(self, node):',
493
                             '        return self.fallback_compiler(node)', '']
494
            else:
di68kap's avatar
di68kap committed
495
                compiler += ['    # def ' + method_name + '(self, node):',
eckhart's avatar
eckhart committed
496
                             '    #     return node', '']
497
        compiler += [COMPILER_FACTORY.format(NAME=self.grammar_name)]
498
        return '\n'.join(compiler)
499

500
    def verify_transformation_table(self, transtable):
501
502
503
504
505
506
        """
        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.
        """
507
508
509
510
511
512
513
514
        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
515
                                      0, Error.UNDEFINED_SYMBOL_IN_TRANSFORMATION_TABLE))
516
517
        return messages

518

519
520
521
522
523
    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
        """
524
525
526
527
528
529
530
531
532
533

        # 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

534
535
536
        if self.variables:
            for i in range(len(definitions)):
                if definitions[i][0] in self.variables:
537
                    definitions[i] = (definitions[i][0], 'Capture(%s)' % definitions[i][1])
538

539
540
        # add special fields for Grammar class

541
542
        definitions.append((self.WHITESPACE_PARSER_KEYWORD,
                            'Whitespace(%s)' % self.WHITESPACE_KEYWORD))
543
544
545
546
        # definitions.append(('wspR__', self.WHITESPACE_KEYWORD
        #                     if 'right' in self.directives['literalws'] else "''"))
        # definitions.append(('wspL__', self.WHITESPACE_KEYWORD
        #                     if 'left' in self.directives['literalws'] else "''"))
547
        definitions.append((self.WHITESPACE_KEYWORD,
Eckhart Arnold's avatar
Eckhart Arnold committed
548
549
550
                            ("mixin_comment(whitespace=" + self.RAW_WS_KEYWORD +
                             ", comment=" + self.COMMENT_KEYWORD + ")")))
        definitions.append((self.RAW_WS_KEYWORD, "r'{whitespace}'".format(**self.directives)))
551
552
553
554
        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
555

556
        article = 'an ' if self.grammar_name[0:1] in "AaEeIiOoUu" else 'a '  # what about 'hour', 'universe' etc.?
557
        declarations = ['class ' + self.grammar_name +
558
                        'Grammar(Grammar):',
559
560
                        'r"""Parser for ' + article + self.grammar_name +
                        ' source file' +
561
                        (', with this grammar:' if self.grammar_source else '.')]
562
        definitions.append(('parser_initialization__', '"upon instantiation"'))
563
        if self.grammar_source:
564
            definitions.append(('source_hash__',
565
                                '"%s"' % md5(self.grammar_source, __version__)))
566
            declarations.append('')
567
            declarations += [line for line in self.grammar_source.split('\n')]
568
569
570
571
572
            while declarations[-1].strip() == '':
                declarations = declarations[:-1]
        declarations.append('"""')

        # turn definitions into declarations in reverse order
573

574
        self.root_symbol = definitions[0][0] if definitions else ""
575
576
577
578
579
580
581
582
        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]
583
584
585
586
587
588

        # 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
589
                self.tree.new_error(self.symbols[symbol],
590
                               "Missing definition for symbol '%s'" % symbol)
591
                # root_node.error_flag = True
592
593
594

        # check for unconnected rules

Eckhart Arnold's avatar
Eckhart Arnold committed
595
596
597
        defined_symbols.difference_update(self.RESERVED_SYMBOLS)

        def remove_connections(symbol):
598
            """Recursively removes all symbols which appear in the
eckhart's avatar
eckhart committed
599
            definiens of a particular symbol."""
Eckhart Arnold's avatar
Eckhart Arnold committed
600
601
602
603
604
605
606
            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
607
608
609
            self.tree.new_error(self.rules[leftover][0],
                                ('Rule "%s" is not connected to parser root "%s" !') %
                                (leftover, self.root_symbol), Error.WARNING)
610

611
        # set root_symbol parser and assemble python grammar definition
612

613
614
        if self.root_symbol and 'root__' not in self.rules:
            declarations.append('root__ = ' + self.root_symbol)
615
        declarations.append('')
Eckhart Arnold's avatar
Eckhart Arnold committed
616
617
618
        self._result = '\n    '.join(declarations) \
                       + GRAMMAR_FACTORY.format(NAME=self.grammar_name)
        return self._result
619

620
621
622

    ## compilation methods

623
    def on_syntax(self, node: Node) -> str:
624
        definitions = []  # type: List[Tuple[str, str]]
625
626

        # drop the wrapping sequence node
627
628
        if len(node.children) == 1 and not node.children[0].parser.name:
            node = node.children[0]
629
630

        # compile definitions and directives and collect definitions
631
        for nd in node.children:
632
            if nd.parser.name == "definition":
633
                definitions.append(self.compile(nd))
634
            else:
635
                assert nd.parser.name == "directive", nd.as_sxpr()
636
                self.compile(nd)
637
            # node.error_flag = max(node.error_flag, nd.error_flag)
638
        self.definitions.update(definitions)
639

640
        return self.assemble_parser(definitions, node)
641

642

643
    def on_definition(self, node: Node) -> Tuple[str, str]:
644
        rule = node.children[0].content
645
        if rule in self.rules:
Eckhart Arnold's avatar
Eckhart Arnold committed
646
            first = self.rules[rule][0]
eckhart's avatar
eckhart committed
647
            if not first.errors:
eckhart's avatar
eckhart committed
648
                self.tree.new_error(first, 'First definition of rule "%s" '
649
                               'followed by illegal redefinitions.' % rule)
eckhart's avatar
eckhart committed
650
            self.tree.new_error(node, 'A rule "%s" has already been defined earlier.' % rule)
651
        elif rule in EBNFCompiler.RESERVED_SYMBOLS:
eckhart's avatar
eckhart committed
652
            self.tree.new_error(node, 'Symbol "%s" is a reserved symbol.' % rule)
653
        elif not sane_parser_name(rule):
eckhart's avatar
eckhart committed
654
            self.tree.new_error(node, 'Illegal symbol "%s". Symbols must not start or '
eckhart's avatar
eckhart committed
655
                                ' end with a doube underscore "__".' % rule)
656
        elif rule in self.directives['tokens']:
eckhart's avatar
eckhart committed
657
            self.tree.new_error(node, 'Symbol "%s" has already been defined as '
eckhart's avatar
eckhart committed
658
                                'a preprocessor token.' % rule)
659
        elif keyword.iskeyword(rule):
eckhart's avatar
eckhart committed
660
            self.tree.new_error(node, 'Python keyword "%s" may not be used as a symbol. '
eckhart's avatar
eckhart committed
661
                                % rule + '(This may change in the future.)')
662
        try:
663
664
            self.current_symbols = [node]
            self.rules[rule] = self.current_symbols
665
            defn = self.compile(node.children[1])
666
            if rule in self.variables:
667
                defn = 'Capture(%s)' % defn
668
                self.variables.remove(rule)
669
670
671
            elif defn.find("(") < 0:
                # assume it's a synonym, like 'page = REGEX_PAGE_NR'
                defn = 'Synonym(%s)' % defn
672
        except TypeError as error:
673
674
675
676
            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
677
            self.tree.new_error(node, errmsg)
678
            rule, defn = rule + ':error', '"' + errmsg + '"'
Eckhart Arnold's avatar
Eckhart Arnold committed
679
        return rule, defn
680

681

682
    def _check_rx(self, node: Node, rx: str) -> str:
683
684
        """
        Checks whether the string `rx` represents a valid regular
685
686
687
        expression. Makes sure that multiline regular expressions are
        prepended by the multiline-flag. Returns the regular expression string.
        """
688
        flags = self.re_flags | {'x'} if rx.find('\n') >= 0 else self.re_flags
689
        if flags:  rx = "(?%s)%s" % ("".join(flags), rx)
690
691
692
        try:
            re.compile(rx)
        except Exception as re_error:
eckhart's avatar
eckhart committed
693
            self.tree.new_error(node, "malformed regular expression %s: %s" %
eckhart's avatar
eckhart committed
694
                                (repr(rx), str(re_error)))
695
696
        return rx

697

698
    def on_directive(self, node: Node) -> str:
699
        key = node.children[0].content.lower()
700
        assert key not in self.directives['tokens']
701

702
703
        if key not in self.REPEATABLE_DIRECTIVES:
            if key in self.defined_directives:
eckhart's avatar
eckhart committed
704
                self.tree.new_error(node, 'Directive "%s" has already been defined earlier. '
eckhart's avatar
eckhart committed
705
706
                                    % key + 'Later definition will be ignored!',
                                    code=Error.REDEFINED_DIRECTIVE_WARNING)
707
708
709
                return ""
            self.defined_directives.add(key)

710
        if key in {'comment', 'whitespace'}:
711
712
            if node.children[1].parser.name == "list_":
                if len(node.children[1].result) != 1:
eckhart's avatar
eckhart committed
713
                    self.tree.new_error(node, 'Directive "%s" must have one, but not %i values.'
eckhart's avatar
eckhart committed
714
                                        % (key, len(node.children[1].result)))
715
                value = self.compile(node.children[1]).pop()
716
717
                if key == 'whitespace' and value in EBNFCompiler.WHITESPACE:
                    value = EBNFCompiler.WHITESPACE[value]  # replace whitespace-name by regex
718
                else:
eckhart's avatar
eckhart committed
719
                    self.tree.new_error(node, 'Value "%s" not allowed for directive "%s".'
eckhart's avatar
eckhart committed
720
                                        % (value, key))
721
            else:
722
723
724
                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
725
                    self.tree.new_error(node, "Whitespace marker '~' not allowed in definition "
eckhart's avatar
eckhart committed
726
                                        "of %s regular expression." % key)
727
728
729
730
                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])
731
                if key == 'whitespace' and not re.match(value, ''):
eckhart's avatar
eckhart committed
732
                    self.tree.new_error(node, "Implicit whitespace should always "
eckhart's avatar
eckhart committed
733
                                        "match the empty string, /%s/ does not." % value)
734
            self.directives[key] = value
735

736
        elif key == 'ignorecase':
737
            if node.children[1].content.lower() not in {"off", "false", "no"}:
738
739
                self.re_flags.add('i')

Eckhart Arnold's avatar
Eckhart Arnold committed
740
        # elif key == 'testing':
741
        #     value = node.children[1].content
Eckhart Arnold's avatar
Eckhart Arnold committed
742
        #     self.directives['testing'] = value.lower() not in {"off", "false", "no"}
743

744
        elif key == 'literalws':
745
            value = {item.lower() for item in self.compile(node.children[1])}
eckhart's avatar
eckhart committed
746
            if ((value - {'left', 'right', 'both', 'none'})
Eckhart Arnold's avatar
Eckhart Arnold committed
747
                    or ('none' in value and len(value) > 1)):
eckhart's avatar
eckhart committed
748
                self.tree.new_error(node, 'Directive "literalws" allows only `left`, `right`, '
eckhart's avatar
eckhart committed
749
                                    '`both` or `none`, not `%s`' % ", ".join(value))
eckhart's avatar
eckhart committed
750
            wsp = {'left', 'right'} if 'both' in value \
751
                else {} if 'none' in value else value
eckhart's avatar
eckhart committed
752
            self.directives[key] = list(wsp)
753

754
        elif key in {'tokens', 'preprocessor_tokens'}:
755
            tokens = self.compile(node.children[1])
756
            redeclared = self.directives['tokens'] & tokens
757
            if redeclared:
eckhart's avatar
eckhart committed
758
                self.tree.new_error(node, 'Tokens %s have already been declared earlier. '
eckhart's avatar
eckhart committed
759
760
                                    % str(redeclared) + 'Later declaration will be ignored',
                                    code=Error.REDECLARED_TOKEN_WARNING)
761
            self.directives['tokens'] |= tokens - redeclared
762

763
        elif key.endswith('_filter'):
764
            filter_set = self.compile(node.children[1])
765
            if not isinstance(filter_set, set) or len(filter_set) != 1:
eckhart's avatar
eckhart committed
766
                self.tree.new_error(node, 'Directive "%s" accepts exactly on symbol, not %s'
eckhart's avatar
eckhart committed
767
                                    % (key, str(filter_set)))
768
            self.directives['filter'][key[:-7]] = filter_set.pop()
769

770
        else:
eckhart's avatar
eckhart committed
771
            self.tree.new_error(node, 'Unknown directive %s ! (Known ones are %s .)' %
eckhart's avatar
eckhart committed
772
                                (key, ', '.join(list(self.directives.keys()))))
773
774
        return ""

775

776
    def non_terminal(self, node: Node, parser_class: str, custom_args: List[str]=[]) -> str:
777
778
        """
        Compiles any non-terminal, where `parser_class` indicates the Parser class
779
780
        name for the particular non-terminal.
        """
781
        arguments = [self.compile(r) for r in node.children] + custom_args
782
783
        return parser_class + '(' + ', '.join(arguments) + ')'

784

785
    def on_expression(self, node) -> str:
di68kap's avatar
di68kap committed
786
        # TODO: Add check for errors like "a" | "ab" (which will always yield a, even for ab)
787
788
        return self.non_terminal(node, 'Alternative')

789

790
    def on_term(self, node) -> str:
di68kap's avatar
di68kap committed
791
792
793
794
795
        # 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
796
        filtered_children = []  # type: List[Node]
di68kap's avatar
di68kap committed
797
        for nd in node.children:
798
            if nd.parser.ptype == TOKEN_PTYPE and nd.content == "§":
eckhart's avatar
eckhart committed
799
800
                mandatory_marker.append(len(filtered_children))
                # if len(filtered_children) == 0:
eckhart's avatar
eckhart committed
801
                #     self.tree.new_error(nd.pos, 'First item of a series should not be mandatory.',
eckhart's avatar
eckhart committed
802
                #                         Error.WARNING)
eckhart's avatar
eckhart committed
803
                if len(mandatory_marker) > 1:
eckhart's avatar
eckhart committed
804
                    self.tree.new_error(nd, 'One mandatory marker (§) sufficient to declare '
eckhart's avatar
eckhart committed
805
                                        'the rest of the series as mandatory.', Error.WARNING)
di68kap's avatar
di68kap committed
806
807
808
809
            else:
                filtered_children.append(nd)
        saved_result = node.result
        node.result = tuple(filtered_children)
eckhart's avatar
eckhart committed
810
811
812
813
814
        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
815
816
        node.result = saved_result
        return compiled
817

818

819
    def on_factor(self, node: Node) -> str:
820
        assert node.children
821
        assert len(node.children) >= 2, node.as_sxpr()
822
        prefix = node.children[0].content
823
        custom_args = []  # type: List[str]
824
825

        if prefix in {'::', ':'}:
826
827
            assert len(node.children) == 2
            arg = node.children[-1]
828
            if arg.parser.name != 'symbol':
eckhart's avatar
eckhart committed
829
                self.tree.new_error(node, ('Retrieve Operator "%s" requires a symbol, '
eckhart's avatar
eckhart committed
830
                                    'and not a %s.') % (prefix, str(arg.parser)))
831
                return str(arg.result)
832
            if str(arg) in self.directives['filter']:
833
                custom_args = ['rfilter=%s' % self.directives['filter'][str(arg)]]
834
            self.variables.add(str(arg))  # cast(str, arg.result)
835

836
        elif len(node.children) > 2:
837
838
            # shift = (Node(node.parser, node.result[1].result),)
            # node.result[1].result = shift + node.result[2:]
839
840
841
842
            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])
843

844
        node.result = node.children[1:]
845
846
        try:
            parser_class = self.PREFIX_TABLE[prefix]
847
848
849
850
            result = self.non_terminal(node, parser_class, custom_args)
            if prefix[:1] == '-':
                def check(node):
                    nd = node
851
852
853
                    if len(nd.children) >= 1:
                        nd = nd.children[0]
                    while nd.parser.name == "symbol":
854
                        symlist = self.rules.get(nd.content, [])
855
856
857
858
859
860
                        if <