ebnf.py 55.8 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
from collections import OrderedDict
28
from functools import partial
eckhart's avatar
eckhart committed
29
30
import keyword
import os
31

eckhart's avatar
eckhart committed
32
from DHParser.compile import CompilerError, Compiler, compile_source, visitor_name
33
from DHParser.error import Error
34
35
from DHParser.parse import Grammar, mixin_comment, Forward, RegExp, Whitespace, \
    NegativeLookahead, Alternative, Series, Option, OneOrMore, ZeroOrMore, Token
36
from DHParser.preprocess import nil_preprocessor, PreprocessorFunc
eckhart's avatar
eckhart committed
37
from DHParser.syntaxtree import Node, WHITESPACE_PTYPE, TOKEN_PTYPE
eckhart's avatar
eckhart committed
38
from DHParser.toolkit import load_if_file, escape_re, md5, sane_parser_name, re, expand_table, \
39
    GLOBALS, CONFIG_PRESET, get_config_value, unrepr, typing
eckhart's avatar
eckhart committed
40
from DHParser.transform import TransformationFunc, traverse, remove_brackets, \
Eckhart Arnold's avatar
Eckhart Arnold committed
41
    reduce_single_child, replace_by_single_child, remove_expendables, \
eckhart's avatar
eckhart committed
42
    remove_tokens, flatten, forbid, assert_content
43
from DHParser.versionnumber import __version__
44
from typing import Callable, Dict, List, Set, Tuple, Sequence, Union, Optional, Any
eckhart's avatar
eckhart committed
45

46

47
__all__ = ('get_ebnf_preprocessor',
48
49
50
51
           'get_ebnf_grammar',
           'get_ebnf_transformer',
           'get_ebnf_compiler',
           'EBNFGrammar',
52
           'EBNFTransform',
Eckhart Arnold's avatar
Eckhart Arnold committed
53
           'EBNFCompilerError',
54
           'EBNFCompiler',
55
           'grammar_changed',
eckhart's avatar
eckhart committed
56
           'compile_ebnf',
57
           'PreprocessorFactoryFunc',
58
59
           'ParserFactoryFunc',
           'TransformerFactoryFunc',
60
           'CompilerFactoryFunc')
61
62


Eckhart Arnold's avatar
Eckhart Arnold committed
63
64
65
66
67
68
69
########################################################################
#
# EBNF scanning
#
########################################################################


70
71
def get_ebnf_preprocessor() -> PreprocessorFunc:
    return nil_preprocessor
Eckhart Arnold's avatar
Eckhart Arnold committed
72
73
74
75
76
77
78
79


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

80

di68kap's avatar
di68kap committed
81
class EBNFGrammar(Grammar):
eckhart's avatar
eckhart committed
82
    r"""
eckhart's avatar
eckhart committed
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
122
123
    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 = !/./
124
    """
di68kap's avatar
di68kap committed
125
    expression = Forward()
eckhart's avatar
eckhart committed
126
    source_hash__ = "82a7c668f86b83f86515078e6c9093ed"
eckhart's avatar
eckhart committed
127
    parser_initialization__ = ["upon instantiation"]
di68kap's avatar
di68kap committed
128
129
    COMMENT__ = r'#.*(?:\n|$)'
    WHITESPACE__ = r'\s*'
130
    WSP_RE__ = mixin_comment(whitespace=WHITESPACE__, comment=COMMENT__)
131
    wsp__ = Whitespace(WSP_RE__)
di68kap's avatar
di68kap committed
132
    EOF = NegativeLookahead(RegExp('.'))
133
134
135
136
137
138
139
140
141
142
143
144
145
146
    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
147
148
149
150
                                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)
151
152
153
    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
154
155
                       Alternative(regexp, literal, symbol),
                       ZeroOrMore(Series(Series(Token(","), wsp__), Alternative(regexp, literal, symbol))), mandatory=1)
156
    definition = Series(symbol, Series(Token("="), wsp__), expression, mandatory=1)
eckhart's avatar
eckhart committed
157
    syntax = Series(Option(Series(wsp__, RegExp(''))), ZeroOrMore(Alternative(definition, directive)), EOF, mandatory=2)
di68kap's avatar
di68kap committed
158
159
160
    root__ = syntax


161
def grammar_changed(grammar_class, grammar_source: str) -> bool:
162
163
    """
    Returns ``True`` if ``grammar_class`` does not reflect the latest
Eckhart Arnold's avatar
Eckhart Arnold committed
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
    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
182
        m = re.search(r'class \w*\(Grammar\)', pycode)
Eckhart Arnold's avatar
Eckhart Arnold committed
183
184
185
186
187
188
189
190
191
192
        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__


193
def get_ebnf_grammar() -> EBNFGrammar:
Eckhart Arnold's avatar
Eckhart Arnold committed
194
    try:
195
        grammar = GLOBALS.ebnf_grammar_singleton
Eckhart Arnold's avatar
Eckhart Arnold committed
196
        return grammar
197
    except AttributeError:
198
199
        GLOBALS.ebnf_grammar_singleton = EBNFGrammar()
        return GLOBALS.ebnf_grammar_singleton
Eckhart Arnold's avatar
Eckhart Arnold committed
200
201
202
203
204
205
206
207
208


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


209
EBNF_AST_transformation_table = {
210
    # AST Transformations for EBNF-grammar
211
    "<":
212
        remove_expendables,
213
    "syntax":
214
        [],  # otherwise '"*": replace_by_single_child' would be applied
215
    "directive, definition":
eckhart's avatar
eckhart committed
216
        [flatten, remove_tokens('@', '=', ',')],
Eckhart Arnold's avatar
Eckhart Arnold committed
217
    "expression":
218
        [replace_by_single_child, flatten, remove_tokens('|')],  # remove_infix_operator],
219
    "term":
220
221
        [replace_by_single_child, flatten],  # supports both idioms:
                                             # "{ factor }+" and "factor { factor }"
222
    "factor, flowmarker, retrieveop":
223
        replace_by_single_child,
224
    "group":
225
        [remove_brackets, replace_by_single_child],
226
227
    "unordered":
        remove_brackets,
228
    "oneormore, repetition, option":
Eckhart Arnold's avatar
Eckhart Arnold committed
229
        [reduce_single_child, remove_brackets,
230
         forbid('repetition', 'option', 'oneormore'), assert_content(r'(?!§)(?:.|\n)*')],
231
    "symbol, literal, regexp":
Eckhart Arnold's avatar
Eckhart Arnold committed
232
        reduce_single_child,
233
    (TOKEN_PTYPE, WHITESPACE_PTYPE):
Eckhart Arnold's avatar
Eckhart Arnold committed
234
        reduce_single_child,
eckhart's avatar
eckhart committed
235
236
    # "list_":
    #     [flatten, remove_infix_operator],
237
    "*":
238
        replace_by_single_child
239
240
}

241

Eckhart Arnold's avatar
Eckhart Arnold committed
242
def EBNFTransform() -> TransformationFunc:
243
    return partial(traverse, processing_table=EBNF_AST_transformation_table.copy())
di68kap's avatar
di68kap committed
244

eckhart's avatar
eckhart committed
245

246
def get_ebnf_transformer() -> TransformationFunc:
247
    try:
248
        transformer = GLOBALS.EBNF_transformer_singleton
249
    except AttributeError:
250
251
        GLOBALS.EBNF_transformer_singleton = EBNFTransform()
        transformer = GLOBALS.EBNF_transformer_singleton
252
    return transformer
Eckhart Arnold's avatar
Eckhart Arnold committed
253
254
255
256
257
258
259
260


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

261

262
PreprocessorFactoryFunc = Callable[[], PreprocessorFunc]
263
ParserFactoryFunc = Callable[[], Grammar]
264
TransformerFactoryFunc = Callable[[], TransformationFunc]
265
266
CompilerFactoryFunc = Callable[[], Compiler]

267
268
269
PREPROCESSOR_FACTORY = '''
def get_preprocessor() -> PreprocessorFunc:
    return {NAME}Preprocessor
270
271
272
273
'''


GRAMMAR_FACTORY = '''
274
def get_grammar() -> {NAME}Grammar:
eckhart's avatar
eckhart committed
275
    global GLOBALS
276
    try:
277
        grammar = GLOBALS.{NAME}_{ID:08d}_grammar_singleton
278
    except AttributeError:
279
        GLOBALS.{NAME}_{ID:08d}_grammar_singleton = {NAME}Grammar()
eckhart's avatar
eckhart committed
280
        if hasattr(get_grammar, 'python_src__'):
281
282
            GLOBALS.{NAME}_{ID:08d}_grammar_singleton.python_src__ = get_grammar.python_src__
        grammar = GLOBALS.{NAME}_{ID:08d}_grammar_singleton
283
    return grammar
284
285
286
287
'''


TRANSFORMER_FACTORY = '''
288
289
290
def {NAME}Transform() -> TransformationDict:
    return partial(traverse, processing_table={NAME}_AST_transformation_table.copy())

291
def get_transformer() -> TransformationFunc:
292
    try:
293
        transformer = GLOBALS.{NAME}_{ID:08d}_transformer_singleton
294
    except AttributeError:
295
296
        GLOBALS.{NAME}_{ID:08d}_transformer_singleton = {NAME}Transform()
        transformer = GLOBALS.{NAME}_{ID:08d}_transformer_singleton
297
    return transformer
298
299
300
301
'''


COMPILER_FACTORY = '''
eckhart's avatar
eckhart committed
302
def get_compiler() -> {NAME}Compiler:
303
    try:
304
        compiler = GLOBALS.{NAME}_{ID:08d}_compiler_singleton
305
    except AttributeError:
306
307
        GLOBALS.{NAME}_{ID:08d}_compiler_singleton = {NAME}Compiler()
        compiler = GLOBALS.{NAME}_{ID:08d}_compiler_singleton
308
    return compiler
309
310
'''

Eckhart Arnold's avatar
Eckhart Arnold committed
311

312
313
314
315
WHITESPACE_TYPES = {'horizontal': r'[\t ]*',  # default: horizontal
                    'linefeed': r'[ \t]*\n?(?!\s*\n)[ \t]*',
                    'vertical': r'\s*'}

316
317
318
DROP_TOKEN  = 'token'
DROP_WSPC   = 'whitespace'
DROP_VALUES = {DROP_TOKEN, DROP_WSPC}
319

eckhart's avatar
eckhart committed
320
321
# Representation of Python code or, rather, something that will be output as Python code
ReprType = Union[str, unrepr]
eckhart's avatar
eckhart committed
322
323


324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
class EBNFDirectives:
    """
    A Record that keeps information about compiler directives
    during the compilation process.

    Attributes:
        whitespace:  the regular expression string for (insignificant)
                whitespace

        comment:  the regular expression string for comments

        literalws:  automatic whitespace eating next to literals. Can
                be either 'left', 'right', 'none', 'both'

        tokens:  set of the names of preprocessor tokens
        filter:  mapping of symbols to python filter functions that
                will be called on any retrieve / pop - operations on
                these symbols

        error:  mapping of symbols to tuples of match conditions and
                customized error messages. A match condition can be
                either a string or a regular expression. The first
                error message where the search condition matches will
                be displayed. An empty string '' as search condition
                always matches, so in case of multiple error messages,
                this condition should be placed at the end.

351
352
353
354
355
356
357
358
359
360
361
362
        skip:  mapping of symbols to a list of search expressions. A
                search expressions can be either a string ot a regular
                expression. The closest match is the point of reentry
                for the series-parser when a mandatory item failed to
                match the following text.

        resume:  mapping of symbols to a list of search expressions. A
                search expressions can be either a string ot a regular
                expression. The closest match is the point of reentry
                for after a parsing error has error occurred. Other
                than the skip field, this configures resuming after
                the failing parser has returned.
363
    """
364
    __slots__ = ['whitespace', 'comment', 'literalws', 'tokens', 'filter', 'error', 'skip',
365
                 'resume', 'drop']
eckhart's avatar
eckhart committed
366

367
368
369
370
    def __init__(self):
        self.whitespace = WHITESPACE_TYPES['vertical']  # type: str
        self.comment = ''     # type: str
        self.literalws = {'right'}  # type: Collection[str]
eckhart's avatar
eckhart committed
371
        self.tokens = set()   # type: Collection[str]
372
        self.filter = dict()  # type: Dict[str, str]
eckhart's avatar
eckhart committed
373
        self.error = dict()   # type: Dict[str, List[Tuple[ReprType, ReprType]]]
374
        self.skip = dict()    # type: Dict[str, List[Union[unrepr, str]]]
375
        self.resume = dict()  # type: Dict[str, List[Union[unrepr, str]]]
376
        self.drop = set()     # type: Set[str]
377
378
379
380
381
382
383
384
385
386
387
388

    def __getitem__(self, key):
        return getattr(self, key)

    def __setitem__(self, key, value):
        assert hasattr(self, key)
        setattr(self, key, value)

    def keys(self):
        return self.__dict__.keys()


eckhart's avatar
eckhart committed
389
class EBNFCompilerError(CompilerError):
390
    """Error raised by `EBNFCompiler` class. (Not compilation errors
391
    in the strict sense, see `CompilationError` in module ``dsl.py``)"""
392
393
394
    pass


395
class EBNFCompiler(Compiler):
396
397
    """
    Generates a Parser from an abstract syntax tree of a grammar specified
398
    in EBNF-Notation.
399
400
401
402
403
404
405

    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.

406
    Additionally, class EBNFCompiler provides helper methods to generate
407
408
409
410
411
    code-skeletons for a preprocessor, AST-transformation and full
    compilation of the formal language. These method's names start with
    the prefix `gen_`.

    Attributes:
412
        current_symbols:  During compilation, a list containing the root
413
414
415
416
                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.

417
        rules:  Dictionary that maps rule names to a list of Nodes that
418
419
420
421
422
423
                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`

424
                Now `[node.content for node in self.rules['alternative']]`
425
426
                yields `['alternative = a | b', 'a', 'b']`

427
        symbols:  A mapping of symbol names to their first usage (not
428
429
                their definition!) in the EBNF source.

430
        variables:  A set of symbols names that are used with the
431
432
433
434
                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.

435
        recursive:  A set of symbols that are used recursively and
436
437
                therefore require a `Forward`-operator.

438
        definitions:  A dictionary of definitions. Other than `rules`
439
440
                this maps the symbols to their compiled definienda.

441
        deferred_tasks:  A list of callables that is filled during
442
443
444
445
446
                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.

447
        root_symbol: The name of the root symbol.
448

449
450
451
452
453
454
455
456
457
458
459
460
        directives:  A record of all directives and their default values.

        defined_directives:  A set of all directives that have already been
                defined. With the exception of those directives contained
                in EBNFCompiler.REPEATABLE_DIRECTIVES, directives must only
                be defined once.

        consumed_custom_errors:  A set of symbols for which a custom error
                has been defined and(!) consumed during compilation. This
                allows to add a compiler error in those cases where (i) an
                error message has been defined but will never used or (ii)
                an error message is accidently used twice. For examples, see
461
462
463
464
                `test_ebnf.TestErrorCustomization`.

        consumed_skip_rules: The same as `consumed_custom_errors` only for
                in-series-resume-rules (aka 'skip-rules') for Series-parsers.
465
466
467

        re_flags:  A set of regular expression flags to be added to all
                regular expressions found in the current parsing process
468

eckhart's avatar
eckhart committed
469
470
471
472
        grammar_name:  The name of the grammar to be compiled

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

473
474
475
        grammar_id: a unique id for every compiled grammar. (Required for
                disambiguation of of thread local variables storing
                compiled texts.)
476
477
    """
    COMMENT_KEYWORD = "COMMENT__"
478
    WHITESPACE_KEYWORD = "WSP_RE__"
Eckhart Arnold's avatar
Eckhart Arnold committed
479
    RAW_WS_KEYWORD = "WHITESPACE__"
480
    WHITESPACE_PARSER_KEYWORD = "wsp__"
481
    DROP_WHITESPACE_PARSER_KEYWORD = "dwsp__"
482
    RESUME_RULES_KEYWORD = "resume_rules__"
483
    SKIP_RULES_SUFFIX = '_skip__'
eckhart's avatar
eckhart committed
484
485
486
    ERR_MSG_SUFFIX = '_err_msg__'
    RESERVED_SYMBOLS = {WHITESPACE_KEYWORD, RAW_WS_KEYWORD, COMMENT_KEYWORD,
                        RESUME_RULES_KEYWORD, ERR_MSG_SUFFIX}
487
    AST_ERROR = "Badly structured syntax tree. " \
Eckhart Arnold's avatar
Eckhart Arnold committed
488
                "Potentially due to erroneous AST transformation."
489
490
491
492
    PREFIX_TABLE = {'§': 'Required',
                    '&': 'Lookahead', '!': 'NegativeLookahead',
                    '-&': 'Lookbehind', '-!': 'NegativeLookbehind',
                    '::': 'Pop', ':': 'Retrieve'}
493
    REPEATABLE_DIRECTIVES = {'tokens'}
494

495

eckhart's avatar
eckhart committed
496
    def __init__(self, grammar_name="DSL", grammar_source=""):
497
        self.grammar_id = 0
eckhart's avatar
eckhart committed
498
499
        super(EBNFCompiler, self).__init__()  # calls the _reset()-method
        self.set_grammar_name(grammar_name, grammar_source)
500

501

502
    def _reset(self):
503
        super(EBNFCompiler, self)._reset()
504
        self._result = ''           # type: str
505
        self.re_flags = set()       # type: Set[str]
506
507
508
        self.rules = OrderedDict()  # type: OrderedDict[str, List[Node]]
        self.current_symbols = []   # type: List[Node]
        self.symbols = {}           # type: Dict[str, Node]
509
510
        self.variables = set()      # type: Set[str]
        self.recursive = set()      # type: Set[str]
511
        self.definitions = {}       # type: Dict[str, str]
512
        self.deferred_tasks = []    # type: List[Callable]
513
        self.root_symbol = ""       # type: str
514
515
516
        self.directives = EBNFDirectives()   # type: EBNFDirectives
        self.defined_directives = set()      # type: Set[str]
        self.consumed_custom_errors = set()  # type: Set[str]
517
        self.consumed_skip_rules = set()     # type: Set[str]
518
519
        self.grammar_id += 1

520

Eckhart Arnold's avatar
Eckhart Arnold committed
521
    @property
522
    def result(self) -> str:
Eckhart Arnold's avatar
Eckhart Arnold committed
523
524
        return self._result

eckhart's avatar
eckhart committed
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541

    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


542
    # methods for generating skeleton code for preprocessor, transformer, and compiler
543

544
    def gen_preprocessor_skeleton(self) -> str:
545
546
547
548
        """
        Returns Python-skeleton-code for a preprocessor-function for
        the previously compiled formal language.
        """
549
        name = self.grammar_name + "Preprocessor"
550
        return "def %s(text):\n    return text, lambda i: i\n" % name \
551
               + PREPROCESSOR_FACTORY.format(NAME=self.grammar_name)
552

553

554
    def gen_transformer_skeleton(self) -> str:
555
556
557
558
        """
        Returns Python-skeleton-code for the AST-transformation for the
        previously compiled formal language.
        """
559
        if not self.rules:
Eckhart Arnold's avatar
Eckhart Arnold committed
560
561
            raise EBNFCompilerError('Compiler must be run before calling '
                                    '"gen_transformer_Skeleton()"!')
562
        tt_name = self.grammar_name + '_AST_transformation_table'
di68kap's avatar
di68kap committed
563
        transtable = [tt_name + ' = {',
Eckhart Arnold's avatar
Eckhart Arnold committed
564
                      '    # AST Transformations for the ' + self.grammar_name + '-grammar']
565
        transtable.append('    "<": remove_empty,')
566
        for name in self.rules:
eckhart's avatar
eckhart committed
567
            transformations = '[]'
568
569
            rule = self.definitions[name]
            if rule.startswith('Alternative'):
eckhart's avatar
eckhart committed
570
                transformations = '[replace_or_reduce]'
571
            elif rule.startswith('Synonym'):
Eckhart Arnold's avatar
Eckhart Arnold committed
572
                transformations = '[reduce_single_child]'
eckhart's avatar
eckhart committed
573
            transtable.append('    "' + name + '": %s,' % transformations)
574
        transtable.append('    ":Token": reduce_single_child,')
575
        transtable += ['    "*": replace_by_single_child', '}', '']
576
        transtable += [TRANSFORMER_FACTORY.format(NAME=self.grammar_name, ID=self.grammar_id)]
577
578
        return '\n'.join(transtable)

579

580
    def gen_compiler_skeleton(self) -> str:
581
582
583
584
        """
        Returns Python-skeleton-code for a Compiler-class for the
        previously compiled formal language.
        """
585
        if not self.rules:
586
587
            raise EBNFCompilerError('Compiler has not been run before calling '
                                    '"gen_Compiler_Skeleton()"!')
588
        compiler = ['class ' + self.grammar_name + 'Compiler(Compiler):',
eckhart's avatar
eckhart committed
589
590
                    '    """Compiler for the abstract-syntax-tree of a '
                    + self.grammar_name + ' source file.',
591
                    '    """', '',
eckhart's avatar
eckhart committed
592
593
594
                    '    def __init__(self):',
                    '        super(' + self.grammar_name + 'Compiler, self).__init__()',
                    '',
595
596
597
                    '    def _reset(self):',
                    '        super()._reset()',
                    '        # initialize your variables here, not in the constructor!']
598
        for name in self.rules:
eckhart's avatar
eckhart committed
599
            method_name = visitor_name(name)
600
            if name == self.root_symbol:
601
                compiler += ['    def ' + method_name + '(self, node):',
602
                             '        return self.fallback_compiler(node)', '']
603
            else:
di68kap's avatar
di68kap committed
604
                compiler += ['    # def ' + method_name + '(self, node):',
eckhart's avatar
eckhart committed
605
                             '    #     return node', '']
606
        compiler += [COMPILER_FACTORY.format(NAME=self.grammar_name, ID=self.grammar_id)]
607
        return '\n'.join(compiler)
608

609
    def verify_transformation_table(self, transtable):
610
611
612
613
614
615
        """
        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.
        """
616
        assert self._dirty_flag
617
        table_entries = set(expand_table(transtable).keys()) - {'*', '<', '>', '~'}
618
619
620
621
622
623
        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),
624
                                      0, Error.UNDEFINED_SYMBOL_IN_TRANSTABLE_WARNING))
625
626
        return messages

627
628
629
630
631
632
633
    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
634

635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662

    def _check_rx(self, node: Node, rx: str) -> str:
        """
        Checks whether the string `rx` represents a valid regular
        expression. Makes sure that multiline regular expressions are
        prepended by the multiline-flag. Returns the regular expression string.
        """
        flags = self.re_flags | {'x'} if rx.find('\n') >= 0 else self.re_flags
        if flags:
            rx = "(?%s)%s" % ("".join(flags), rx)
        try:
            re.compile(rx)
        except Exception as re_error:
            self.tree.new_error(node, "malformed regular expression %s: %s" %
                                (repr(rx), str(re_error)))
        return rx


    def _extract_regex(self, node: Node) -> str:
        """Extracts regular expression string from regexp-Node."""
        value = node.content.strip("~")
        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])
        return value


eckhart's avatar
eckhart committed
663
    def _gen_search_rule(self, nd: Node) -> ReprType:
664
665
666
        """Generates a search rule, which can be either a string for simple
        string search or a regular expression from the nodes content. Returns
        an empty string in case the node is neither regexp nor literal.
667
        """
668
        if nd.tag_name == 'regexp':
669
            return unrepr("re.compile(r'%s')" % self._extract_regex(nd))
670
        elif nd.tag_name == 'literal':
671
672
673
674
            s = nd.content.strip()
            return s.strip('"') if s[0] == '"' else s.strip("'")
        return ''

675
676
677
678
679
680
681
    def _gen_search_list(self, nodes: Sequence[Node]) -> List[Union[unrepr, str]]:
        search_list = []  # type: List[Union[unrepr, str]]
        for child in nodes:
            rule = self._gen_search_rule(child)
            search_list.append(rule if rule else unrepr(child.content.strip()))
        return search_list

682

683
684
685
686
687
    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
        """
688
689
690
691
692
693
694
695
696
697

        # 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

698
699
700
        if self.variables:
            for i in range(len(definitions)):
                if definitions[i][0] in self.variables:
701
                    definitions[i] = (definitions[i][0], 'Capture(%s)' % definitions[i][1])
702

703
704
        # add special fields for Grammar class

705
706
        definitions.append((self.WHITESPACE_PARSER_KEYWORD,
                            'Whitespace(%s)' % self.WHITESPACE_KEYWORD))
707
708
709
        if DROP_WSPC in self.directives.drop:
            definitions.append((self.DROP_WHITESPACE_PARSER_KEYWORD,
                                'DropWhitespace(%s)' % self.WHITESPACE_KEYWORD))
710
        definitions.append((self.WHITESPACE_KEYWORD,
eckhart's avatar
eckhart committed
711
712
                            ("mixin_comment(whitespace=" + self.RAW_WS_KEYWORD
                             + ", comment=" + self.COMMENT_KEYWORD + ")")))
713
714
        definitions.append((self.RAW_WS_KEYWORD, "r'{}'".format(self.directives.whitespace)))
        definitions.append((self.COMMENT_KEYWORD, "r'{}'".format(self.directives.comment)))
715
716
717

        # prepare and add resume-rules

eckhart's avatar
eckhart committed
718
        resume_rules = dict()  # type: Dict[str, List[ReprType]]
719
        for symbol, raw_rules in self.directives.resume.items():
720
721
722
723
724
            refined_rules = []
            for rule in raw_rules:
                if isinstance(rule, unrepr) and rule.s.isidentifier():
                    try:
                        nd = self.rules[rule.s][0].children[1]
725
                        refined = self._gen_search_rule(nd)
726
727
728
729
730
731
732
733
734
735
736
                    except IndexError:
                        refined = ""
                    if refined:
                        refined_rules.append(refined)
                    else:
                        self.tree.new_error(nd, 'Symbol "%s" cannot be used in resume rule, since'
                                                ' it represents neither literal nor regexp!')
                else:
                    refined_rules.append(rule)
            resume_rules[symbol] = refined_rules
        definitions.append((self.RESUME_RULES_KEYWORD, repr(resume_rules)))
737

eckhart's avatar
eckhart committed
738
739
        # prepare and add customized error-messages

740
        for symbol, err_msgs in self.directives.error.items():
eckhart's avatar
eckhart committed
741
            custom_errors = []  # type: List[Tuple[ReprType, ReprType]]
eckhart's avatar
eckhart committed
742
743
744
745
746
747
748
749
750
751
            for search, message in err_msgs:
                if isinstance(search, unrepr) and search.s.isidentifier():
                    try:
                        nd = self.rules[search.s][0].children[1]
                        search = self._gen_search_rule(nd)
                    except IndexError:
                        search = ''
                custom_errors.append((search, message))
            definitions.append((symbol + self.ERR_MSG_SUFFIX, repr(custom_errors)))

752
753
754
755
756
757
        for symbol in self.directives.error.keys():
            if symbol not in self.consumed_custom_errors:
                def_node = self.rules[symbol][0]
                self.tree.new_error(
                    def_node, 'Customized error message for symbol "{}" will never be used, '
                    'because the mandatory marker "§" appears nowhere in its definiendum!'
758
759
760
761
762
                    .format(symbol), Error.UNUSED_ERROR_HANDLING_WARNING)

        # prepare and add skip-rules

        for symbol, skip in self.directives.skip.items():
Eckhart Arnold's avatar
Eckhart Arnold committed
763
            skip_rules = []  # type: List[ReprType]
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
            for search in skip:
                if isinstance(search, unrepr) and search.s.isidentifier():
                    try:
                        nd = self.rules[search.s][0].children[1]
                        search = self._gen_search_rule(nd)
                    except IndexError:
                        search = ''
                skip_rules.append(search)
            definitions.append((symbol + self.SKIP_RULES_SUFFIX, repr(skip_rules)))

        for symbol in self.directives.error.keys():
            if symbol not in self.consumed_skip_rules:
                def_node = self.rules[symbol][0]
                self.tree.new_error(
                    def_node, '"Skip-rules" for symbol "{}" will never be used, '
                    'because the mandatory marker "§" appears nowhere in its definiendum!'
                    .format(symbol), Error.UNUSED_ERROR_HANDLING_WARNING)
781

782
783
        # prepare parser class header and docstring and
        # add EBNF grammar to the doc string of the parser class
784

785
        article = 'an ' if self.grammar_name[0:1] in "AaEeIiOoUu" else 'a '  # what about 'hour', 'universe' etc.?
786
        show_source = get_config_value('add_grammar_source_to_parser_docstring')
eckhart's avatar
eckhart committed
787
788
789
790
        declarations = ['class ' + self.grammar_name
                        + 'Grammar(Grammar):',
                        'r"""Parser for ' + article + self.grammar_name
                        + ' source file'
791
                        + ('. Grammar:' if self.grammar_source and show_source else '.')]
eckhart's avatar
eckhart committed
792
        definitions.append(('parser_initialization__', '["upon instantiation"]'))
793
        if self.grammar_source:
794
            definitions.append(('source_hash__',
795
                                '"%s"' % md5(self.grammar_source, __version__)))
796
            declarations.append('')
797
798
            if show_source:
                declarations += [line for line in self.grammar_source.split('\n')]
799
800
801
802
803
            while declarations[-1].strip() == '':
                declarations = declarations[:-1]
        declarations.append('"""')

        # turn definitions into declarations in reverse order
804

805
        self.root_symbol = definitions[0][0] if definitions else ""
806
807
808
809
810
811
812
813
        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]
814
815
816
817
818
819

        # 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
820
                self.tree.new_error(self.symbols[symbol],
eckhart's avatar
eckhart committed
821
                                    "Missing definition for symbol '%s'" % symbol)
822
                # root_node.error_flag = True
823
824
825

        # check for unconnected rules

Eckhart Arnold's avatar
Eckhart Arnold committed
826
827
828
        defined_symbols.difference_update(self.RESERVED_SYMBOLS)

        def remove_connections(symbol):
829
            """Recursively removes all symbols which appear in the
eckhart's avatar
eckhart committed
830
            definiens of a particular symbol."""
Eckhart Arnold's avatar
Eckhart Arnold committed
831
832
833
834
835
836
837
            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
838
839
840
            self.tree.new_error(self.rules[leftover][0],
                                ('Rule "%s" is not connected to parser root "%s" !') %
                                (leftover, self.root_symbol), Error.WARNING)
841

842
        # set root_symbol parser and assemble python grammar definition
843

844
845
        if self.root_symbol and 'root__' not in self.rules:
            declarations.append('root__ = ' + self.root_symbol)
846
        declarations.append('')
Eckhart Arnold's avatar
Eckhart Arnold committed
847
        self._result = '\n    '.join(declarations) \
848
                       + GRAMMAR_FACTORY.format(NAME=self.grammar_name, ID=self.grammar_id)
Eckhart Arnold's avatar
Eckhart Arnold committed
849
        return self._result
850

851
852
853

    ## compilation methods

854
    def on_syntax(self, node: Node) -> str:
855
        definitions = []  # type: List[Tuple[str, str]]
856
857

        # drop the wrapping sequence node
858
        if len(node.children) == 1 and node.children[0].is_anonymous():
859
            node = node.children[0]
860
861

        # compile definitions and directives and collect definitions
862
        for nd in node.children:
863
            if nd.tag_name == "definition":
864
                definitions.append(self.compile(nd))
865
            else:
866
                assert nd.tag_name == "directive", nd.as_sxpr()
867
                self.compile(nd)
868
            # node.error_flag = max(node.error_flag, nd.error_flag)
869
        self.definitions.update(definitions)
870

871
        return self.assemble_parser(definitions, node)
872

873

874
    def on_definition(self, node: Node) -> Tuple[str, str]:
875
        rule = node.children[0].content
876
        if rule in self.rules:
Eckhart Arnold's avatar
Eckhart Arnold committed
877
            first = self.rules[rule][0]
eckhart's avatar
eckhart committed
878
            if not first.errors:
eckhart's avatar
eckhart committed
879
                self.tree.new_error(first, 'First definition of rule "%s" '
eckhart's avatar
eckhart committed
880
                                    'followed by illegal redefinitions.' % rule)
eckhart's avatar
eckhart committed
881
            self.tree.new_error(node, 'A rule "%s" has already been defined earlier.' % rule)
882
        elif rule in EBNFCompiler.RESERVED_SYMBOLS:
eckhart's avatar
eckhart committed
883
            self.tree.new_error(node, 'Symbol "%s" is a reserved symbol.' % rule)
884
        elif not sane_parser_name(rule):
eckhart's avatar
eckhart committed
885
            self.tree.new_error(node, 'Illegal symbol "%s". Symbols must not start or '
eckhart's avatar
eckhart committed
886
                                ' end with a doube underscore "__".' % rule)
887
        elif rule in self.directives.tokens:
eckhart's avatar
eckhart committed
888
            self.tree.new_error(node, 'Symbol "%s" has already been defined as '
eckhart's avatar
eckhart committed
889
                                'a preprocessor token.' % rule)
890
        elif keyword.iskeyword(rule):
eckhart's avatar
eckhart committed
891
            self.tree.new_error(node, 'Python keyword "%s" may not be used as a symbol. '
eckhart's avatar
eckhart committed
892
                                % rule + '(This may change in the future.)')
893
        try:
894
895
            self.current_symbols = [node]
            self.rules[rule] = self.current_symbols
896
            defn = self.compile(node.children[1])
897
            if rule in self.variables:
898
                defn = 'Capture(%s)' % defn
899
                self.variables.remove(rule)
900
901
902
            elif defn.find("(") < 0:
                # assume it's a synonym, like 'page = REGEX_PAGE_NR'
                defn = 'Synonym(%s)' % defn
903
        except TypeError as error:
904
905
906
907
            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
908
            self.tree.new_error(node, errmsg)
909
            rule, defn = rule + ':error', '"' + errmsg + '"'
Eckhart Arnold's avatar
Eckhart Arnold committed
910
        return rule, defn
911

912

913
    def on_directive(self, node: Node) -> str:
914
        key = node.children[0].content
915
        assert key not in self.directives.tokens
916

917
        if key not in self.REPEATABLE_DIRECTIVES and not key.endswith('_error'):
918
            if key in self.defined_directives:
eckhart's avatar
eckhart committed
919
                self.tree.new_error(node, 'Directive "%s" has already been defined earlier. '
eckhart's avatar
eckhart committed
920
                                    % key + 'Later definition will be ignored!',
921
                                    code=Error.REDEFINED_DIRECTIVE)
922
923
924
                return ""
            self.defined_directives.add(key)

eckhart's avatar
eckhart committed
925
926
        def check_argnum(n: int = 1):
            if len(node.children) > n + 1:
927
                self.tree.new_error(node, 'Directive "%s" can have at most %i values.' % (key, n))
eckhart's avatar
eckhart committed
928

929
        if key in {'comment', 'whitespace'}:
eckhart's avatar
eckhart committed
930
            check_argnum()
931
            if node.children[1].tag_name == "symbol":
eckhart's avatar
eckhart committed
932
                value = node.children[1].content
933
934
                if key == 'whitespace' and value in WHITESPACE_TYPES:
                    value = WHITESPACE_TYPES[value]  # replace whitespace-name by regex
935
                else:
eckhart's avatar
eckhart committed
936
                    self.tree.new_error(node, 'Value "%s" not allowed for directive "%s".'
eckhart's avatar
eckhart committed
937
                                        % (value, key))
938
            else:
939
                value = self._extract_regex(node.children[1])
940