ebnf.py 44.3 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, RootNode, WHITESPACE_PTYPE, TOKEN_PTYPE
eckhart's avatar
eckhart committed
37
from DHParser.toolkit import load_if_file, escape_re, md5, sane_parser_name, re, expand_table, \
38
    GLOBALS, CONFIG_PRESET, get_config_value, typing
eckhart's avatar
eckhart committed
39
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__
43
from typing import Callable, Dict, List, Set, Tuple, Any
eckhart's avatar
eckhart committed
44

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
    r"""
eckhart's avatar
eckhart committed
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
120
121
    Parser for an EBNF source file, with this grammar:

    @ 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 | symbol) { "," (regexp | literal | symbol) }

    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
    regexp     = /\/(?:\\\/|[^\/])*?\//~            # e.g. /\w+/, ~/#.*(?:\n|$)/~
    whitespace = /~/~                               # insignificant whitespace

    EOF = !/./
122
    """
di68kap's avatar
di68kap committed
123
    expression = Forward()
eckhart's avatar
eckhart committed
124
    source_hash__ = "82a7c668f86b83f86515078e6c9093ed"
125
    parser_initialization__ = "upon instantiation"
di68kap's avatar
di68kap committed
126
127
    COMMENT__ = r'#.*(?:\n|$)'
    WHITESPACE__ = r'\s*'
128
    WSP_RE__ = mixin_comment(whitespace=WHITESPACE__, comment=COMMENT__)
129
    wsp__ = Whitespace(WSP_RE__)
di68kap's avatar
di68kap committed
130
    EOF = NegativeLookahead(RegExp('.'))
131
132
133
134
135
136
137
138
139
140
141
142
143
144
    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,
eckhart's avatar
eckhart committed
145
146
147
148
                                NegativeLookahead(Series(Token("="), wsp__))), Series(Option(flowmarker), literal),
                         Series(Option(flowmarker), plaintext), Series(Option(flowmarker), regexp),
                         Series(Option(flowmarker), whitespace), Series(Option(flowmarker), oneormore),
                         Series(Option(flowmarker), group), Series(Option(flowmarker), unordered), repetition, option)
149
150
151
    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__),
eckhart's avatar
eckhart committed
152
153
                       Alternative(regexp, literal, symbol),
                       ZeroOrMore(Series(Series(Token(","), wsp__), Alternative(regexp, literal, symbol))), mandatory=1)
154
    definition = Series(symbol, Series(Token("="), wsp__), expression, mandatory=1)
eckhart's avatar
eckhart committed
155
    syntax = Series(Option(Series(wsp__, RegExp(''))), ZeroOrMore(Alternative(definition, directive)), EOF, mandatory=2)
di68kap's avatar
di68kap committed
156
157
158
    root__ = syntax


159
def grammar_changed(grammar_class, grammar_source: str) -> bool:
160
161
    """
    Returns ``True`` if ``grammar_class`` does not reflect the latest
Eckhart Arnold's avatar
Eckhart Arnold committed
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
    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()
eckhart's avatar
eckhart committed
180
        m = re.search(r'class \w*\(Grammar\)', pycode)
Eckhart Arnold's avatar
Eckhart Arnold committed
181
182
183
184
185
186
187
188
189
190
        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__


191
def get_ebnf_grammar() -> EBNFGrammar:
Eckhart Arnold's avatar
Eckhart Arnold committed
192
    try:
193
        grammar = GLOBALS.ebnf_grammar_singleton
Eckhart Arnold's avatar
Eckhart Arnold committed
194
        return grammar
195
    except AttributeError:
196
197
        GLOBALS.ebnf_grammar_singleton = EBNFGrammar()
        return GLOBALS.ebnf_grammar_singleton
Eckhart Arnold's avatar
Eckhart Arnold committed
198
199
200
201
202
203
204
205
206


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


207
EBNF_AST_transformation_table = {
208
    # AST Transformations for EBNF-grammar
209
    "<":
210
        remove_expendables,
211
    "syntax":
212
        [],  # otherwise '"*": replace_by_single_child' would be applied
213
    "directive, definition":
eckhart's avatar
eckhart committed
214
        [flatten, remove_tokens('@', '=', ',')],
Eckhart Arnold's avatar
Eckhart Arnold committed
215
    "expression":
216
        [replace_by_single_child, flatten, remove_tokens('|')],  # remove_infix_operator],
217
    "term":
218
219
        [replace_by_single_child, flatten],  # supports both idioms:
                                             # "{ factor }+" and "factor { factor }"
220
    "factor, flowmarker, retrieveop":
221
        replace_by_single_child,
222
    "group":
223
        [remove_brackets, replace_by_single_child],
224
225
    "unordered":
        remove_brackets,
226
    "oneormore, repetition, option":
Eckhart Arnold's avatar
Eckhart Arnold committed
227
        [reduce_single_child, remove_brackets,
228
         forbid('repetition', 'option', 'oneormore'), assert_content(r'(?!§)(?:.|\n)*')],
229
    "symbol, literal, regexp":
Eckhart Arnold's avatar
Eckhart Arnold committed
230
        reduce_single_child,
231
    (TOKEN_PTYPE, WHITESPACE_PTYPE):
Eckhart Arnold's avatar
Eckhart Arnold committed
232
        reduce_single_child,
eckhart's avatar
eckhart committed
233
234
    # "list_":
    #     [flatten, remove_infix_operator],
235
    "*":
236
        replace_by_single_child
237
238
}

239

Eckhart Arnold's avatar
Eckhart Arnold committed
240
def EBNFTransform() -> TransformationFunc:
241
    return partial(traverse, processing_table=EBNF_AST_transformation_table.copy())
di68kap's avatar
di68kap committed
242

eckhart's avatar
eckhart committed
243

244
def get_ebnf_transformer() -> TransformationFunc:
245
    try:
246
        transformer = GLOBALS.EBNF_transformer_singleton
247
    except AttributeError:
248
249
        GLOBALS.EBNF_transformer_singleton = EBNFTransform()
        transformer = GLOBALS.EBNF_transformer_singleton
250
    return transformer
Eckhart Arnold's avatar
Eckhart Arnold committed
251
252
253
254
255
256
257
258


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

259

260
PreprocessorFactoryFunc = Callable[[], PreprocessorFunc]
261
ParserFactoryFunc = Callable[[], Grammar]
262
TransformerFactoryFunc = Callable[[], TransformationFunc]
263
264
CompilerFactoryFunc = Callable[[], Compiler]

265
266
267
PREPROCESSOR_FACTORY = '''
def get_preprocessor() -> PreprocessorFunc:
    return {NAME}Preprocessor
268
269
270
271
'''


GRAMMAR_FACTORY = '''
272
def get_grammar() -> {NAME}Grammar:
273
    try:
274
        grammar = GLOBALS.{NAME}_{ID}_grammar_singleton
275
    except AttributeError:
276
277
        GLOBALS.{NAME}_{ID}_grammar_singleton = {NAME}Grammar()
        grammar = GLOBALS.{NAME}_{ID}_grammar_singleton
278
    return grammar
279
280
281
282
'''


TRANSFORMER_FACTORY = '''
283
284
285
def {NAME}Transform() -> TransformationDict:
    return partial(traverse, processing_table={NAME}_AST_transformation_table.copy())

286
def get_transformer() -> TransformationFunc:
287
    try:
288
        transformer = GLOBALS.{NAME}_{ID}_transformer_singleton
289
    except AttributeError:
290
291
        GLOBALS.{NAME}_{ID}_transformer_singleton = {NAME}Transform()
        transformer = GLOBALS.{NAME}_{ID}_transformer_singleton
292
    return transformer
293
294
295
296
'''


COMPILER_FACTORY = '''
eckhart's avatar
eckhart committed
297
def get_compiler() -> {NAME}Compiler:
298
    try:
299
        compiler = GLOBALS.{NAME}_{ID}_compiler_singleton
300
    except AttributeError:
eckhart's avatar
eckhart committed
301
        GLOBALS.{NAME}_{ID}_compiler_singleton = {NAME}Compiler()
302
        compiler = GLOBALS.{NAME}_{ID}_compiler_singleton
303
    return compiler
304
305
'''

Eckhart Arnold's avatar
Eckhart Arnold committed
306

eckhart's avatar
eckhart committed
307
class EBNFCompilerError(CompilerError):
308
    """Error raised by `EBNFCompiler` class. (Not compilation errors
309
    in the strict sense, see `CompilationError` in module ``dsl.py``)"""
310
311
312
    pass


313
class EBNFCompiler(Compiler):
314
315
    """
    Generates a Parser from an abstract syntax tree of a grammar specified
316
    in EBNF-Notation.
317
318
319
320
321
322
323
324
325
326
327
328
329

    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:
330
        current_symbols:  During compilation, a list containing the root
331
332
333
334
                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.

335
        rules:  Dictionary that maps rule names to a list of Nodes that
336
337
338
339
340
341
                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`

342
                Now `[node.content for node in self.rules['alternative']]`
343
344
                yields `['alternative = a | b', 'a', 'b']`

345
        symbols:  A mapping of symbol names to their first usage (not
346
347
                their definition!) in the EBNF source.

348
        variables:  A set of symbols names that are used with the
349
350
351
352
                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.

353
        recursive:  A set of symbols that are used recursively and
354
355
                therefore require a `Forward`-operator.

356
        definitions:  A dictionary of definitions. Other than `rules`
357
358
                this maps the symbols to their compiled definienda.

359
        deferred_tasks:  A list of callables that is filled during
360
361
362
363
364
                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.

365
        root_symbol: The name of the root symbol.
366

367
        directives:  A dictionary of all directives and their default
368
                values.
369
370
371

        re_flags:  A set of regular expression flags to be added to all
                regular expressions found in the current parsing process
372

eckhart's avatar
eckhart committed
373
374
375
376
        grammar_name:  The name of the grammar to be compiled

        grammar_source:  The source code of the grammar to be compiled.

377
378
379
        grammar_id: a unique id for every compiled grammar. (Required for
                disambiguation of of thread local variables storing
                compiled texts.)
380
381
    """
    COMMENT_KEYWORD = "COMMENT__"
382
    WHITESPACE_KEYWORD = "WSP_RE__"
Eckhart Arnold's avatar
Eckhart Arnold committed
383
    RAW_WS_KEYWORD = "WHITESPACE__"
384
    WHITESPACE_PARSER_KEYWORD = "wsp__"
Eckhart Arnold's avatar
Eckhart Arnold committed
385
    RESERVED_SYMBOLS = {WHITESPACE_KEYWORD, RAW_WS_KEYWORD, COMMENT_KEYWORD}
386
    AST_ERROR = "Badly structured syntax tree. " \
Eckhart Arnold's avatar
Eckhart Arnold committed
387
                "Potentially due to erroneous AST transformation."
388
389
390
391
    PREFIX_TABLE = {'§': 'Required',
                    '&': 'Lookahead', '!': 'NegativeLookahead',
                    '-&': 'Lookbehind', '-!': 'NegativeLookbehind',
                    '::': 'Pop', ':': 'Retrieve'}
392
393
394
    WHITESPACE = {'horizontal': r'[\t ]*',  # default: horizontal
                  'linefeed': r'[ \t]*\n?(?!\s*\n)[ \t]*',
                  'vertical': r'\s*'}
395
    REPEATABLE_DIRECTIVES = {'tokens'}
396

397

eckhart's avatar
eckhart committed
398
    def __init__(self, grammar_name="DSL", grammar_source=""):
399
        self.grammar_id = 0
eckhart's avatar
eckhart committed
400
401
        super(EBNFCompiler, self).__init__()  # calls the _reset()-method
        self.set_grammar_name(grammar_name, grammar_source)
402

403

404
    def _reset(self):
405
        super(EBNFCompiler, self)._reset()
406
        self._result = ''           # type: str
407
        self.re_flags = set()       # type: Set[str]
408
409
410
        self.rules = OrderedDict()  # type: OrderedDict[str, List[Node]]
        self.current_symbols = []   # type: List[Node]
        self.symbols = {}           # type: Dict[str, Node]
411
412
        self.variables = set()      # type: Set[str]
        self.recursive = set()      # type: Set[str]
413
        self.definitions = {}       # type: Dict[str, str]
414
        self.deferred_tasks = []    # type: List[Callable]
415
        self.root_symbol = ""       # type: str
eckhart's avatar
eckhart committed
416
        self.directives = {'whitespace': self.WHITESPACE['vertical'],
417
                           'comment': '',
418
                           'literalws': {'right'},
419
                           'tokens': set(),  # alt. 'preprocessor_tokens'
420
421
                           'filter': dict(),  # alt. 'filter'
                           'error': dict()}  # customized error messages
422
423
        # self.directives['ignorecase']: False
        self.defined_directives = set()  # type: Set[str]
424
425
        self.grammar_id += 1

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

eckhart's avatar
eckhart committed
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447

    def set_grammar_name(self, grammar_name: str = "", grammar_source: str = ""):
        """
        Changes the grammar name and source.

        The grammar name and the source text are metadata that do not affect the
        compilation process. It is used to name and annotate the output.
        Returns `self`.
        """
        assert grammar_name == "" or re.match(r'\w+\Z', grammar_name)
        if not grammar_name and re.fullmatch(r'[\w/:\\]+', grammar_source):
            grammar_name = os.path.splitext(os.path.basename(grammar_source))[0]
        self.grammar_name = grammar_name
        self.grammar_source = load_if_file(grammar_source)
        return self


448
    # methods for generating skeleton code for preprocessor, transformer, and compiler
449

450
    def gen_preprocessor_skeleton(self) -> str:
451
452
453
454
        """
        Returns Python-skeleton-code for a preprocessor-function for
        the previously compiled formal language.
        """
455
        name = self.grammar_name + "Preprocessor"
456
        return "def %s(text):\n    return text, lambda i: i\n" % name \
457
               + PREPROCESSOR_FACTORY.format(NAME=self.grammar_name)
458

459

460
    def gen_transformer_skeleton(self) -> str:
461
462
463
464
        """
        Returns Python-skeleton-code for the AST-transformation for the
        previously compiled formal language.
        """
465
        if not self.rules:
Eckhart Arnold's avatar
Eckhart Arnold committed
466
467
            raise EBNFCompilerError('Compiler must be run before calling '
                                    '"gen_transformer_Skeleton()"!')
468
        tt_name = self.grammar_name + '_AST_transformation_table'
di68kap's avatar
di68kap committed
469
        transtable = [tt_name + ' = {',
Eckhart Arnold's avatar
Eckhart Arnold committed
470
                      '    # AST Transformations for the ' + self.grammar_name + '-grammar']
471
        transtable.append('    "<": remove_empty,')
472
        for name in self.rules:
eckhart's avatar
eckhart committed
473
            transformations = '[]'
474
475
            rule = self.definitions[name]
            if rule.startswith('Alternative'):
eckhart's avatar
eckhart committed
476
                transformations = '[replace_or_reduce]'
477
            elif rule.startswith('Synonym'):
Eckhart Arnold's avatar
Eckhart Arnold committed
478
                transformations = '[reduce_single_child]'
eckhart's avatar
eckhart committed
479
            transtable.append('    "' + name + '": %s,' % transformations)
480
        transtable.append('    ":Token": reduce_single_child,')
481
        transtable += ['    "*": replace_by_single_child', '}', '']
482
        transtable += [TRANSFORMER_FACTORY.format(NAME=self.grammar_name, ID=self.grammar_id)]
483
484
        return '\n'.join(transtable)

485

486
    def gen_compiler_skeleton(self) -> str:
487
488
489
490
        """
        Returns Python-skeleton-code for a Compiler-class for the
        previously compiled formal language.
        """
491
        if not self.rules:
492
493
            raise EBNFCompilerError('Compiler has not been run before calling '
                                    '"gen_Compiler_Skeleton()"!')
494
        compiler = ['class ' + self.grammar_name + 'Compiler(Compiler):',
eckhart's avatar
eckhart committed
495
496
                    '    """Compiler for the abstract-syntax-tree of a '
                    + self.grammar_name + ' source file.',
497
                    '    """', '',
eckhart's avatar
eckhart committed
498
499
500
                    '    def __init__(self):',
                    '        super(' + self.grammar_name + 'Compiler, self).__init__()',
                    '',
501
502
503
                    '    def _reset(self):',
                    '        super()._reset()',
                    '        # initialize your variables here, not in the constructor!']
504
        for name in self.rules:
505
            method_name = Compiler.method_name(name)
506
            if name == self.root_symbol:
507
                compiler += ['    def ' + method_name + '(self, node):',
508
                             '        return self.fallback_compiler(node)', '']
509
            else:
di68kap's avatar
di68kap committed
510
                compiler += ['    # def ' + method_name + '(self, node):',
eckhart's avatar
eckhart committed
511
                             '    #     return node', '']
512
        compiler += [COMPILER_FACTORY.format(NAME=self.grammar_name, ID=self.grammar_id)]
513
        return '\n'.join(compiler)
514

515
    def verify_transformation_table(self, transtable):
516
517
518
519
520
521
        """
        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.
        """
522
        assert self._dirty_flag
523
        table_entries = set(expand_table(transtable).keys()) - {'*', '<', '>', '~'}
524
525
526
527
528
529
        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
530
                                      0, Error.UNDEFINED_SYMBOL_IN_TRANSFORMATION_TABLE))
531
532
        return messages

533
534
535
536
537
538
539
    def verify_compiler(self, compiler):
        """
        Checks for on_XXXX()-methods that occur in the compiler, although XXXX
        has never been defined in the grammar. Usually, this kind of
        inconsistency results from an error like a typo in the compiler-code.
        """
        pass  # TODO: add verification code here
540

541
542
543
544
545
    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
        """
546
547
548
549
550
551
552
553
554
555

        # 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

556
557
558
        if self.variables:
            for i in range(len(definitions)):
                if definitions[i][0] in self.variables:
559
                    definitions[i] = (definitions[i][0], 'Capture(%s)' % definitions[i][1])
560

561
562
        # add special fields for Grammar class

563
564
        definitions.append((self.WHITESPACE_PARSER_KEYWORD,
                            'Whitespace(%s)' % self.WHITESPACE_KEYWORD))
565
        definitions.append((self.WHITESPACE_KEYWORD,
eckhart's avatar
eckhart committed
566
567
                            ("mixin_comment(whitespace=" + self.RAW_WS_KEYWORD
                             + ", comment=" + self.COMMENT_KEYWORD + ")")))
Eckhart Arnold's avatar
Eckhart Arnold committed
568
        definitions.append((self.RAW_WS_KEYWORD, "r'{whitespace}'".format(**self.directives)))
569
570
571
572
        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
573

574
        article = 'an ' if self.grammar_name[0:1] in "AaEeIiOoUu" else 'a '  # what about 'hour', 'universe' etc.?
575
        show_source = get_config_value('add_grammar_source_to_parser_docstring')
eckhart's avatar
eckhart committed
576
577
578
579
        declarations = ['class ' + self.grammar_name
                        + 'Grammar(Grammar):',
                        'r"""Parser for ' + article + self.grammar_name
                        + ' source file'
580
                        + ('. Grammar:' if self.grammar_source and show_source else '.')]
581
        definitions.append(('parser_initialization__', '"upon instantiation"'))
582
        if self.grammar_source:
583
            definitions.append(('source_hash__',
584
                                '"%s"' % md5(self.grammar_source, __version__)))
585
            declarations.append('')
586
587
            if show_source:
                declarations += [line for line in self.grammar_source.split('\n')]
588
589
590
591
592
            while declarations[-1].strip() == '':
                declarations = declarations[:-1]
        declarations.append('"""')

        # turn definitions into declarations in reverse order
593

594
        self.root_symbol = definitions[0][0] if definitions else ""
595
596
597
598
599
600
601
602
        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]
603
604
605
606
607
608

        # 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
609
                self.tree.new_error(self.symbols[symbol],
eckhart's avatar
eckhart committed
610
                                    "Missing definition for symbol '%s'" % symbol)
611
                # root_node.error_flag = True
612
613
614

        # check for unconnected rules

Eckhart Arnold's avatar
Eckhart Arnold committed
615
616
617
        defined_symbols.difference_update(self.RESERVED_SYMBOLS)

        def remove_connections(symbol):
618
            """Recursively removes all symbols which appear in the
eckhart's avatar
eckhart committed
619
            definiens of a particular symbol."""
Eckhart Arnold's avatar
Eckhart Arnold committed
620
621
622
623
624
625
626
            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
627
628
629
            self.tree.new_error(self.rules[leftover][0],
                                ('Rule "%s" is not connected to parser root "%s" !') %
                                (leftover, self.root_symbol), Error.WARNING)
630

631
        # set root_symbol parser and assemble python grammar definition
632

633
634
        if self.root_symbol and 'root__' not in self.rules:
            declarations.append('root__ = ' + self.root_symbol)
635
        declarations.append('')
Eckhart Arnold's avatar
Eckhart Arnold committed
636
        self._result = '\n    '.join(declarations) \
637
                       + GRAMMAR_FACTORY.format(NAME=self.grammar_name, ID=self.grammar_id)
Eckhart Arnold's avatar
Eckhart Arnold committed
638
        return self._result
639

640
641
642

    ## compilation methods

643
    def on_syntax(self, node: Node) -> str:
644
        definitions = []  # type: List[Tuple[str, str]]
645
646

        # drop the wrapping sequence node
647
648
        if len(node.children) == 1 and not node.children[0].parser.name:
            node = node.children[0]
649
650

        # compile definitions and directives and collect definitions
651
        for nd in node.children:
652
            if nd.parser.name == "definition":
653
                definitions.append(self.compile(nd))
654
            else:
655
                assert nd.parser.name == "directive", nd.as_sxpr()
656
                self.compile(nd)
657
            # node.error_flag = max(node.error_flag, nd.error_flag)
658
        self.definitions.update(definitions)
659

660
        return self.assemble_parser(definitions, node)
661

662

663
    def on_definition(self, node: Node) -> Tuple[str, str]:
664
        rule = node.children[0].content
665
        if rule in self.rules:
Eckhart Arnold's avatar
Eckhart Arnold committed
666
            first = self.rules[rule][0]
eckhart's avatar
eckhart committed
667
            if not first.errors:
eckhart's avatar
eckhart committed
668
                self.tree.new_error(first, 'First definition of rule "%s" '
eckhart's avatar
eckhart committed
669
                                    'followed by illegal redefinitions.' % rule)
eckhart's avatar
eckhart committed
670
            self.tree.new_error(node, 'A rule "%s" has already been defined earlier.' % rule)
671
        elif rule in EBNFCompiler.RESERVED_SYMBOLS:
eckhart's avatar
eckhart committed
672
            self.tree.new_error(node, 'Symbol "%s" is a reserved symbol.' % rule)
673
        elif not sane_parser_name(rule):
eckhart's avatar
eckhart committed
674
            self.tree.new_error(node, 'Illegal symbol "%s". Symbols must not start or '
eckhart's avatar
eckhart committed
675
                                ' end with a doube underscore "__".' % rule)
676
        elif rule in self.directives['tokens']:
eckhart's avatar
eckhart committed
677
            self.tree.new_error(node, 'Symbol "%s" has already been defined as '
eckhart's avatar
eckhart committed
678
                                'a preprocessor token.' % rule)
679
        elif keyword.iskeyword(rule):
eckhart's avatar
eckhart committed
680
            self.tree.new_error(node, 'Python keyword "%s" may not be used as a symbol. '
eckhart's avatar
eckhart committed
681
                                % rule + '(This may change in the future.)')
682
        try:
683
684
            self.current_symbols = [node]
            self.rules[rule] = self.current_symbols
685
            defn = self.compile(node.children[1])
686
            if rule in self.variables:
687
                defn = 'Capture(%s)' % defn
688
                self.variables.remove(rule)
689
690
691
            elif defn.find("(") < 0:
                # assume it's a synonym, like 'page = REGEX_PAGE_NR'
                defn = 'Synonym(%s)' % defn
692
        except TypeError as error:
693
694
695
696
            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
697
            self.tree.new_error(node, errmsg)
698
            rule, defn = rule + ':error', '"' + errmsg + '"'
Eckhart Arnold's avatar
Eckhart Arnold committed
699
        return rule, defn
700

701

702
    def _check_rx(self, node: Node, rx: str) -> str:
703
704
        """
        Checks whether the string `rx` represents a valid regular
705
706
707
        expression. Makes sure that multiline regular expressions are
        prepended by the multiline-flag. Returns the regular expression string.
        """
708
        flags = self.re_flags | {'x'} if rx.find('\n') >= 0 else self.re_flags
eckhart's avatar
eckhart committed
709
710
        if flags:
            rx = "(?%s)%s" % ("".join(flags), rx)
711
712
713
        try:
            re.compile(rx)
        except Exception as re_error:
eckhart's avatar
eckhart committed
714
            self.tree.new_error(node, "malformed regular expression %s: %s" %
eckhart's avatar
eckhart committed
715
                                (repr(rx), str(re_error)))
716
717
        return rx

718

719
    def on_directive(self, node: Node) -> str:
720
        key = node.children[0].content.lower()
721
        assert key not in self.directives['tokens']
722

723
724
        if key not in self.REPEATABLE_DIRECTIVES:
            if key in self.defined_directives:
eckhart's avatar
eckhart committed
725
                self.tree.new_error(node, 'Directive "%s" has already been defined earlier. '
eckhart's avatar
eckhart committed
726
727
                                    % key + 'Later definition will be ignored!',
                                    code=Error.REDEFINED_DIRECTIVE_WARNING)
728
729
730
                return ""
            self.defined_directives.add(key)

eckhart's avatar
eckhart committed
731
732
733
734
735
        def check_argnum(n: int = 1):
            if len(node.children) > n + 1:
                self.tree.new_error(node, 'Directive "%s" must have one, but not %i values.'
                                    % (key, len(node.children) - 1))

736
        if key in {'comment', 'whitespace'}:
eckhart's avatar
eckhart committed
737
738
739
            check_argnum()
            if node.children[1].parser.name == "symbol":
                value = node.children[1].content
740
741
                if key == 'whitespace' and value in EBNFCompiler.WHITESPACE:
                    value = EBNFCompiler.WHITESPACE[value]  # replace whitespace-name by regex
742
                else:
eckhart's avatar
eckhart committed
743
                    self.tree.new_error(node, 'Value "%s" not allowed for directive "%s".'
eckhart's avatar
eckhart committed
744
                                        % (value, key))
745
            else:
eckhart's avatar
eckhart committed
746
747
748
                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
749
                    self.tree.new_error(node, "Whitespace marker '~' not allowed in definition "
eckhart's avatar
eckhart committed
750
                                        "of %s regular expression." % key)
751
752
753
754
                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])
755
                if key == 'whitespace' and not re.match(value, ''):
eckhart's avatar
eckhart committed
756
                    self.tree.new_error(node, "Implicit whitespace should always "
eckhart's avatar
eckhart committed
757
                                        "match the empty string, /%s/ does not." % value)
758
            self.directives[key] = value
759

760
        elif key == 'ignorecase':
eckhart's avatar
eckhart committed
761
            check_argnum()
762
            if node.children[1].content.lower() not in {"off", "false", "no"}:
763
764
                self.re_flags.add('i')

Eckhart Arnold's avatar
Eckhart Arnold committed
765
        # elif key == 'testing':
766
        #     value = node.children[1].content
Eckhart Arnold's avatar
Eckhart Arnold committed
767
        #     self.directives['testing'] = value.lower() not in {"off", "false", "no"}
768

769
        elif key == 'literalws':
eckhart's avatar
eckhart committed
770
            value = {child.content.strip().lower() for child in node.children[1:]}
eckhart's avatar
eckhart committed
771
            if ((value - {'left', 'right', 'both', 'none'})
Eckhart Arnold's avatar
Eckhart Arnold committed
772
                    or ('none' in value and len(value) > 1)):
eckhart's avatar
eckhart committed
773
                self.tree.new_error(node, 'Directive "literalws" allows only `left`, `right`, '
eckhart's avatar
eckhart committed
774
                                    '`both` or `none`, not `%s`' % ", ".join(value))
eckhart's avatar
eckhart committed
775
            wsp = {'left', 'right'} if 'both' in value \
776
                else {} if 'none' in value else value
eckhart's avatar
eckhart committed
777
            self.directives[key] = list(wsp)
778

779
        elif key in {'tokens', 'preprocessor_tokens'}:
eckhart's avatar
eckhart committed
780
            tokens = {child.content.strip() for child in node.children[1:]}
781
            redeclared = self.directives['tokens'] & tokens
782
            if redeclared:
eckhart's avatar
eckhart committed
783
                self.tree.new_error(node, 'Tokens %s have already been declared earlier. '
eckhart's avatar
eckhart committed
784
785
                                    % str(redeclared) + 'Later declaration will be ignored',
                                    code=Error.REDECLARED_TOKEN_WARNING)
786
            self.directives['tokens'] |= tokens - redeclared
787

788
        elif key.endswith('_filter'):
eckhart's avatar
eckhart committed
789
790
            check_argnum()
            self.directives['filter'][key[:-7]] = node.children[1].content.strip()
791

792
        elif key.endswith('_error'):
eckhart's avatar
eckhart committed
793
794
795
            check_argnum()
            if not node.children[1].parser.name == "literal":
                self.tree.new_error(node, 'Directive "%s" requires message string as argument')
796
797
798
799
800
801
802
            error_msg = node.children[1].content
            symbol = key[:-6]
            if symbol in self.rules:
                self.tree.new_error(node, 'Custom error message for symbol "%s"' % symbol
                                    + 'must be defined before the symbol!')
            self.directives['error'][symbol] = error_msg

803
        else:
eckhart's avatar
eckhart committed
804
            self.tree.new_error(node, 'Unknown directive %s ! (Known ones are %s .)' %
eckhart's avatar
eckhart committed
805
                                (key, ', '.join(list(self.directives.keys()))))
806
807
        return ""

808

eckhart's avatar
eckhart committed
809
    def non_terminal(self, node: Node, parser_class: str, custom_args: List[str] = []) -> str:
810
811
        """
        Compiles any non-terminal, where `parser_class` indicates the Parser class
812
813
        name for the particular non-terminal.
        """
814
        arguments = [self.compile(r) for r in node.children] + custom_args
815
816
        return parser_class + '(' + ', '.join(arguments) + ')'

817

818
    def on_expression(self, node) -> str:
di68kap's avatar
di68kap committed
819
        # TODO: Add check for errors like "a" | "ab" (which will always yield a, even for ab)
820
821
        return self.non_terminal(node, 'Alternative')

822

823
    def on_term(self, node) -> str:
di68kap's avatar
di68kap committed
824
825
826
827
828
        # 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
829
        filtered_children = []  # type: List[Node]
di68kap's avatar
di68kap committed
830
        for nd in node.children:
831
            if nd.parser.ptype == TOKEN_PTYPE and nd.content == "§":
eckhart's avatar
eckhart committed
832
833
                mandatory_marker.append(len(filtered_children))
                # if len(filtered_children) == 0:
eckhart's avatar
eckhart committed
834
                #     self.tree.new_error(nd.pos, 'First item of a series should not be mandatory.',
eckhart's avatar
eckhart committed
835
                #                         Error.WARNING)
eckhart's avatar
eckhart committed
836
                if len(mandatory_marker) > 1:
eckhart's avatar
eckhart committed
837
                    self.tree.new_error(nd, 'One mandatory marker (§) sufficient to declare '
eckhart's avatar
eckhart committed
838
                                        'the rest of the series as mandatory.', Error.WARNING)
di68kap's avatar
di68kap committed
839
840
841
842
            else:
                filtered_children.append(nd)
        saved_result = node.result
        node.result = tuple(filtered_children)
eckhart's avatar
eckhart committed
843
844
845
846
        if len(filtered_children) == 1:
            compiled = self.non_terminal(node, 'Required')
        else:
            custom_args = ['mandatory=%i' % mandatory_marker[0]] if mandatory_marker else []
847
848
849
850
851
852
            # add custom error message if it has been declared for the currend definition
            if custom_args:
                current_symbol = next(reversed(self.rules.keys()))
                msg = self