ebnf.py 52.8 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
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__
eckhart's avatar
eckhart committed
44
from typing import Callable, Dict, List, Set, Tuple, 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"
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}_grammar_singleton
278
    except AttributeError:
279
        GLOBALS.{NAME}_{ID}_grammar_singleton = {NAME}Grammar()
eckhart's avatar
eckhart committed
280
281
        if hasattr(get_grammar, 'python_src__'):
            GLOBALS.{NAME}_{ID}_grammar_singleton.python_src__ = get_grammar.python_src__
282
        grammar = GLOBALS.{NAME}_{ID}_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}_transformer_singleton
294
    except AttributeError:
295
296
        GLOBALS.{NAME}_{ID}_transformer_singleton = {NAME}Transform()
        transformer = GLOBALS.{NAME}_{ID}_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}_compiler_singleton
305
    except AttributeError:
eckhart's avatar
eckhart committed
306
        GLOBALS.{NAME}_{ID}_compiler_singleton = {NAME}Compiler()
307
        compiler = GLOBALS.{NAME}_{ID}_compiler_singleton
308
    return compiler
309
310
'''

Eckhart Arnold's avatar
Eckhart Arnold committed
311

312
313
314
315
316
WHITESPACE_TYPES = {'horizontal': r'[\t ]*',  # default: horizontal
                    'linefeed': r'[ \t]*\n?(?!\s*\n)[ \t]*',
                    'vertical': r'\s*'}


eckhart's avatar
eckhart committed
317
318
# Representation of Python code or, rather, something that will be output as Python code
ReprType = Union[str, unrepr]
eckhart's avatar
eckhart committed
319
320


321
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
351
352
353
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.

        resume: mapping of symbols to a list of search conditions. A
                search condition can be either a string ot a regular
                expression. The closest match from all search conditions
                is the point of reentry for the parser after a parser
                has error occurred.
    """
eckhart's avatar
eckhart committed
354
355
    __slots__ = ['whitespace', 'comment', 'literalws', 'tokens', 'filter', 'error', 'resume']

356
357
358
359
    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
360
        self.tokens = set()   # type: Collection[str]
361
        self.filter = dict()  # type: Dict[str, str]
eckhart's avatar
eckhart committed
362
        self.error = dict()   # type: Dict[str, List[Tuple[ReprType, ReprType]]]
363
364
365
366
367
368
369
370
371
372
373
374
375
        self.resume = dict()  # type: Dict[str, List[Union[unrepr, str]]]

    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
376
class EBNFCompilerError(CompilerError):
377
    """Error raised by `EBNFCompiler` class. (Not compilation errors
378
    in the strict sense, see `CompilationError` in module ``dsl.py``)"""
379
380
381
    pass


382
class EBNFCompiler(Compiler):
383
384
    """
    Generates a Parser from an abstract syntax tree of a grammar specified
385
    in EBNF-Notation.
386
387
388
389
390
391
392
393
394
395
396
397
398

    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:
399
        current_symbols:  During compilation, a list containing the root
400
401
402
403
                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.

404
        rules:  Dictionary that maps rule names to a list of Nodes that
405
406
407
408
409
410
                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`

411
                Now `[node.content for node in self.rules['alternative']]`
412
413
                yields `['alternative = a | b', 'a', 'b']`

414
        symbols:  A mapping of symbol names to their first usage (not
415
416
                their definition!) in the EBNF source.

417
        variables:  A set of symbols names that are used with the
418
419
420
421
                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.

422
        recursive:  A set of symbols that are used recursively and
423
424
                therefore require a `Forward`-operator.

425
        definitions:  A dictionary of definitions. Other than `rules`
426
427
                this maps the symbols to their compiled definienda.

428
        deferred_tasks:  A list of callables that is filled during
429
430
431
432
433
                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.

434
        root_symbol: The name of the root symbol.
435

436
437
438
439
440
441
442
443
444
445
446
447
448
        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
                `test_ebnf.TestErrorCustomization`
449
450
451

        re_flags:  A set of regular expression flags to be added to all
                regular expressions found in the current parsing process
452

eckhart's avatar
eckhart committed
453
454
455
456
        grammar_name:  The name of the grammar to be compiled

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

457
458
459
        grammar_id: a unique id for every compiled grammar. (Required for
                disambiguation of of thread local variables storing
                compiled texts.)
460
461
    """
    COMMENT_KEYWORD = "COMMENT__"
462
    WHITESPACE_KEYWORD = "WSP_RE__"
Eckhart Arnold's avatar
Eckhart Arnold committed
463
    RAW_WS_KEYWORD = "WHITESPACE__"
464
    WHITESPACE_PARSER_KEYWORD = "wsp__"
465
    RESUME_RULES_KEYWORD = "resume_rules__"
eckhart's avatar
eckhart committed
466
467
468
    ERR_MSG_SUFFIX = '_err_msg__'
    RESERVED_SYMBOLS = {WHITESPACE_KEYWORD, RAW_WS_KEYWORD, COMMENT_KEYWORD,
                        RESUME_RULES_KEYWORD, ERR_MSG_SUFFIX}
469
    AST_ERROR = "Badly structured syntax tree. " \
Eckhart Arnold's avatar
Eckhart Arnold committed
470
                "Potentially due to erroneous AST transformation."
471
472
473
474
    PREFIX_TABLE = {'§': 'Required',
                    '&': 'Lookahead', '!': 'NegativeLookahead',
                    '-&': 'Lookbehind', '-!': 'NegativeLookbehind',
                    '::': 'Pop', ':': 'Retrieve'}
475
    REPEATABLE_DIRECTIVES = {'tokens'}
476

477

eckhart's avatar
eckhart committed
478
    def __init__(self, grammar_name="DSL", grammar_source=""):
479
        self.grammar_id = 0
eckhart's avatar
eckhart committed
480
481
        super(EBNFCompiler, self).__init__()  # calls the _reset()-method
        self.set_grammar_name(grammar_name, grammar_source)
482

483

484
    def _reset(self):
485
        super(EBNFCompiler, self)._reset()
486
        self._result = ''           # type: str
487
        self.re_flags = set()       # type: Set[str]
488
489
490
        self.rules = OrderedDict()  # type: OrderedDict[str, List[Node]]
        self.current_symbols = []   # type: List[Node]
        self.symbols = {}           # type: Dict[str, Node]
491
492
        self.variables = set()      # type: Set[str]
        self.recursive = set()      # type: Set[str]
493
        self.definitions = {}       # type: Dict[str, str]
494
        self.deferred_tasks = []    # type: List[Callable]
495
        self.root_symbol = ""       # type: str
496
497
498
        self.directives = EBNFDirectives()   # type: EBNFDirectives
        self.defined_directives = set()      # type: Set[str]
        self.consumed_custom_errors = set()  # type: Set[str]
499
500
        self.grammar_id += 1

501

Eckhart Arnold's avatar
Eckhart Arnold committed
502
    @property
503
    def result(self) -> str:
Eckhart Arnold's avatar
Eckhart Arnold committed
504
505
        return self._result

eckhart's avatar
eckhart committed
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522

    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


523
    # methods for generating skeleton code for preprocessor, transformer, and compiler
524

525
    def gen_preprocessor_skeleton(self) -> str:
526
527
528
529
        """
        Returns Python-skeleton-code for a preprocessor-function for
        the previously compiled formal language.
        """
530
        name = self.grammar_name + "Preprocessor"
531
        return "def %s(text):\n    return text, lambda i: i\n" % name \
532
               + PREPROCESSOR_FACTORY.format(NAME=self.grammar_name)
533

534

535
    def gen_transformer_skeleton(self) -> str:
536
537
538
539
        """
        Returns Python-skeleton-code for the AST-transformation for the
        previously compiled formal language.
        """
540
        if not self.rules:
Eckhart Arnold's avatar
Eckhart Arnold committed
541
542
            raise EBNFCompilerError('Compiler must be run before calling '
                                    '"gen_transformer_Skeleton()"!')
543
        tt_name = self.grammar_name + '_AST_transformation_table'
di68kap's avatar
di68kap committed
544
        transtable = [tt_name + ' = {',
Eckhart Arnold's avatar
Eckhart Arnold committed
545
                      '    # AST Transformations for the ' + self.grammar_name + '-grammar']
546
        transtable.append('    "<": remove_empty,')
547
        for name in self.rules:
eckhart's avatar
eckhart committed
548
            transformations = '[]'
549
550
            rule = self.definitions[name]
            if rule.startswith('Alternative'):
eckhart's avatar
eckhart committed
551
                transformations = '[replace_or_reduce]'
552
            elif rule.startswith('Synonym'):
Eckhart Arnold's avatar
Eckhart Arnold committed
553
                transformations = '[reduce_single_child]'
eckhart's avatar
eckhart committed
554
            transtable.append('    "' + name + '": %s,' % transformations)
555
        transtable.append('    ":Token": reduce_single_child,')
556
        transtable += ['    "*": replace_by_single_child', '}', '']
557
        transtable += [TRANSFORMER_FACTORY.format(NAME=self.grammar_name, ID=self.grammar_id)]
558
559
        return '\n'.join(transtable)

560

561
    def gen_compiler_skeleton(self) -> str:
562
563
564
565
        """
        Returns Python-skeleton-code for a Compiler-class for the
        previously compiled formal language.
        """
566
        if not self.rules:
567
568
            raise EBNFCompilerError('Compiler has not been run before calling '
                                    '"gen_Compiler_Skeleton()"!')
569
        compiler = ['class ' + self.grammar_name + 'Compiler(Compiler):',
eckhart's avatar
eckhart committed
570
571
                    '    """Compiler for the abstract-syntax-tree of a '
                    + self.grammar_name + ' source file.',
572
                    '    """', '',
eckhart's avatar
eckhart committed
573
574
575
                    '    def __init__(self):',
                    '        super(' + self.grammar_name + 'Compiler, self).__init__()',
                    '',
576
577
578
                    '    def _reset(self):',
                    '        super()._reset()',
                    '        # initialize your variables here, not in the constructor!']
579
        for name in self.rules:
580
            method_name = Compiler.method_name(name)
581
            if name == self.root_symbol:
582
                compiler += ['    def ' + method_name + '(self, node):',
583
                             '        return self.fallback_compiler(node)', '']
584
            else:
di68kap's avatar
di68kap committed
585
                compiler += ['    # def ' + method_name + '(self, node):',
eckhart's avatar
eckhart committed
586
                             '    #     return node', '']
587
        compiler += [COMPILER_FACTORY.format(NAME=self.grammar_name, ID=self.grammar_id)]
588
        return '\n'.join(compiler)
589

590
    def verify_transformation_table(self, transtable):
591
592
593
594
595
596
        """
        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.
        """
597
        assert self._dirty_flag
598
        table_entries = set(expand_table(transtable).keys()) - {'*', '<', '>', '~'}
599
600
601
602
603
604
        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),
605
                                      0, Error.UNDEFINED_SYMBOL_IN_TRANSTABLE_WARNING))
606
607
        return messages

608
609
610
611
612
613
614
    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
615

616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643

    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
644
    def _gen_search_rule(self, nd: Node) -> ReprType:
645
646
647
        """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.
648
649
650
651
652
653
654
655
656
        """
        if nd.parser.name == 'regexp':
            return unrepr("re.compile(r'%s')" % self._extract_regex(nd))
        elif nd.parser.name == 'literal':
            s = nd.content.strip()
            return s.strip('"') if s[0] == '"' else s.strip("'")
        return ''


657
658
659
660
661
    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
        """
662
663
664
665
666
667
668
669
670
671

        # 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

672
673
674
        if self.variables:
            for i in range(len(definitions)):
                if definitions[i][0] in self.variables:
675
                    definitions[i] = (definitions[i][0], 'Capture(%s)' % definitions[i][1])
676

677
678
        # add special fields for Grammar class

679
680
        definitions.append((self.WHITESPACE_PARSER_KEYWORD,
                            'Whitespace(%s)' % self.WHITESPACE_KEYWORD))
681
        definitions.append((self.WHITESPACE_KEYWORD,
eckhart's avatar
eckhart committed
682
683
                            ("mixin_comment(whitespace=" + self.RAW_WS_KEYWORD
                             + ", comment=" + self.COMMENT_KEYWORD + ")")))
684
685
        definitions.append((self.RAW_WS_KEYWORD, "r'{}'".format(self.directives.whitespace)))
        definitions.append((self.COMMENT_KEYWORD, "r'{}'".format(self.directives.comment)))
686
687
688

        # prepare and add resume-rules

eckhart's avatar
eckhart committed
689
        resume_rules = dict()  # type: Dict[str, List[ReprType]]
690
        for symbol, raw_rules in self.directives.resume.items():
691
692
693
694
695
            refined_rules = []
            for rule in raw_rules:
                if isinstance(rule, unrepr) and rule.s.isidentifier():
                    try:
                        nd = self.rules[rule.s][0].children[1]
696
                        refined = self._gen_search_rule(nd)
697
698
699
700
701
702
703
704
705
706
707
                    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)))
708

eckhart's avatar
eckhart committed
709
710
        # prepare and add customized error-messages

711
        for symbol, err_msgs in self.directives.error.items():
eckhart's avatar
eckhart committed
712
            custom_errors = []  # type: List[Tuple[ReprType, ReprType]]
eckhart's avatar
eckhart committed
713
714
715
716
717
718
719
720
721
722
            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)))

723
724
725
726
727
728
729
730
        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!'
                    .format(symbol), Error.UNUSED_ERROR_MSG_WARNING)

731
732
        # prepare parser class header and docstring and
        # add EBNF grammar to the doc string of the parser class
733

734
        article = 'an ' if self.grammar_name[0:1] in "AaEeIiOoUu" else 'a '  # what about 'hour', 'universe' etc.?
735
        show_source = get_config_value('add_grammar_source_to_parser_docstring')
eckhart's avatar
eckhart committed
736
737
738
739
        declarations = ['class ' + self.grammar_name
                        + 'Grammar(Grammar):',
                        'r"""Parser for ' + article + self.grammar_name
                        + ' source file'
740
                        + ('. Grammar:' if self.grammar_source and show_source else '.')]
741
        definitions.append(('parser_initialization__', '"upon instantiation"'))
742
        if self.grammar_source:
743
            definitions.append(('source_hash__',
744
                                '"%s"' % md5(self.grammar_source, __version__)))
745
            declarations.append('')
746
747
            if show_source:
                declarations += [line for line in self.grammar_source.split('\n')]
748
749
750
751
752
            while declarations[-1].strip() == '':
                declarations = declarations[:-1]
        declarations.append('"""')

        # turn definitions into declarations in reverse order
753

754
        self.root_symbol = definitions[0][0] if definitions else ""
755
756
757
758
759
760
761
762
        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]
763
764
765
766
767
768

        # 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
769
                self.tree.new_error(self.symbols[symbol],
eckhart's avatar
eckhart committed
770
                                    "Missing definition for symbol '%s'" % symbol)
771
                # root_node.error_flag = True
772
773
774

        # check for unconnected rules

Eckhart Arnold's avatar
Eckhart Arnold committed
775
776
777
        defined_symbols.difference_update(self.RESERVED_SYMBOLS)

        def remove_connections(symbol):
778
            """Recursively removes all symbols which appear in the
eckhart's avatar
eckhart committed
779
            definiens of a particular symbol."""
Eckhart Arnold's avatar
Eckhart Arnold committed
780
781
782
783
784
785
786
            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
787
788
789
            self.tree.new_error(self.rules[leftover][0],
                                ('Rule "%s" is not connected to parser root "%s" !') %
                                (leftover, self.root_symbol), Error.WARNING)
790

791
        # set root_symbol parser and assemble python grammar definition
792

793
794
        if self.root_symbol and 'root__' not in self.rules:
            declarations.append('root__ = ' + self.root_symbol)
795
        declarations.append('')
Eckhart Arnold's avatar
Eckhart Arnold committed
796
        self._result = '\n    '.join(declarations) \
797
                       + GRAMMAR_FACTORY.format(NAME=self.grammar_name, ID=self.grammar_id)
Eckhart Arnold's avatar
Eckhart Arnold committed
798
        return self._result
799

800
801
802

    ## compilation methods

803
    def on_syntax(self, node: Node) -> str:
804
        definitions = []  # type: List[Tuple[str, str]]
805
806

        # drop the wrapping sequence node
807
808
        if len(node.children) == 1 and not node.children[0].parser.name:
            node = node.children[0]
809
810

        # compile definitions and directives and collect definitions
811
        for nd in node.children:
812
            if nd.parser.name == "definition":
813
                definitions.append(self.compile(nd))
814
            else:
815
                assert nd.parser.name == "directive", nd.as_sxpr()
816
                self.compile(nd)
817
            # node.error_flag = max(node.error_flag, nd.error_flag)
818
        self.definitions.update(definitions)
819

820
        return self.assemble_parser(definitions, node)
821

822

823
    def on_definition(self, node: Node) -> Tuple[str, str]:
824
        rule = node.children[0].content
825
        if rule in self.rules:
Eckhart Arnold's avatar
Eckhart Arnold committed
826
            first = self.rules[rule][0]
eckhart's avatar
eckhart committed
827
            if not first.errors:
eckhart's avatar
eckhart committed
828
                self.tree.new_error(first, 'First definition of rule "%s" '
eckhart's avatar
eckhart committed
829
                                    'followed by illegal redefinitions.' % rule)
eckhart's avatar
eckhart committed
830
            self.tree.new_error(node, 'A rule "%s" has already been defined earlier.' % rule)
831
        elif rule in EBNFCompiler.RESERVED_SYMBOLS:
eckhart's avatar
eckhart committed
832
            self.tree.new_error(node, 'Symbol "%s" is a reserved symbol.' % rule)
833
        elif not sane_parser_name(rule):
eckhart's avatar
eckhart committed
834
            self.tree.new_error(node, 'Illegal symbol "%s". Symbols must not start or '
eckhart's avatar
eckhart committed
835
                                ' end with a doube underscore "__".' % rule)
836
        elif rule in self.directives.tokens:
eckhart's avatar
eckhart committed
837
            self.tree.new_error(node, 'Symbol "%s" has already been defined as '
eckhart's avatar
eckhart committed
838
                                'a preprocessor token.' % rule)
839
        elif keyword.iskeyword(rule):
eckhart's avatar
eckhart committed
840
            self.tree.new_error(node, 'Python keyword "%s" may not be used as a symbol. '
eckhart's avatar
eckhart committed
841
                                % rule + '(This may change in the future.)')
842
        try:
843
844
            self.current_symbols = [node]
            self.rules[rule] = self.current_symbols
845
            defn = self.compile(node.children[1])
846
            if rule in self.variables:
847
                defn = 'Capture(%s)' % defn
848
                self.variables.remove(rule)
849
850
851
            elif defn.find("(") < 0:
                # assume it's a synonym, like 'page = REGEX_PAGE_NR'
                defn = 'Synonym(%s)' % defn
852
        except TypeError as error:
853
854
855
856
            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
857
            self.tree.new_error(node, errmsg)
858
            rule, defn = rule + ':error', '"' + errmsg + '"'
Eckhart Arnold's avatar
Eckhart Arnold committed
859
        return rule, defn
860

861

862
    def on_directive(self, node: Node) -> str:
863
        key = node.children[0].content
864
        assert key not in self.directives.tokens
865

866
867
        if key not in self.REPEATABLE_DIRECTIVES:
            if key in self.defined_directives:
eckhart's avatar
eckhart committed
868
                self.tree.new_error(node, 'Directive "%s" has already been defined earlier. '
eckhart's avatar
eckhart committed
869
870
                                    % key + 'Later definition will be ignored!',
                                    code=Error.REDEFINED_DIRECTIVE_WARNING)
871
872
873
                return ""
            self.defined_directives.add(key)

eckhart's avatar
eckhart committed
874
875
        def check_argnum(n: int = 1):
            if len(node.children) > n + 1:
876
                self.tree.new_error(node, 'Directive "%s" can have at most %i values.' % (key, n))
eckhart's avatar
eckhart committed
877

878
        if key in {'comment', 'whitespace'}:
eckhart's avatar
eckhart committed
879
880
881
            check_argnum()
            if node.children[1].parser.name == "symbol":
                value = node.children[1].content
882
883
                if key == 'whitespace' and value in WHITESPACE_TYPES:
                    value = WHITESPACE_TYPES[value]  # replace whitespace-name by regex
884
                else:
eckhart's avatar
eckhart committed
885
                    self.tree.new_error(node, 'Value "%s" not allowed for directive "%s".'
eckhart's avatar
eckhart committed
886
                                        % (value, key))
887
            else:
888
                value = self._extract_regex(node.children[1])
889
                if key == 'whitespace' and not re.match(value, ''):
eckhart's avatar
eckhart committed
890
                    self.tree.new_error(node, "Implicit whitespace should always "
eckhart's avatar
eckhart committed
891
                                        "match the empty string, /%s/ does not." % value)
892
            self.directives[key] = value
893

894
        elif key == 'ignorecase':
eckhart's avatar
eckhart committed
895
            check_argnum()
896
            if node.children[1].content.lower() not in {"off", "false", "no"}:
897
898
                self.re_flags.add('i')

899
        elif key == 'literalws':
eckhart's avatar
eckhart committed
900
901
902
            values = {child.content.strip().lower() for child in node.children[1:]}
            if ((values - {'left', 'right', 'both', 'none'})
                    or ('none' in values and len(values) > 1)):
eckhart's avatar
eckhart committed
903
                self.tree.new_error(node, 'Directive "literalws" allows only `left`, `right`, '
eckhart's avatar
eckhart committed
904
905
906
                                    '`both` or `none`, not `%s`' % ", ".join(values))
            wsp = {'left', 'right'} if 'both' in values \
                else {} if 'none' in values else values
907
            self.directives.literalws = wsp
908

909
        elif key in {'tokens', 'preprocessor_tokens'}:
eckhart's avatar
eckhart committed
910
            tokens = {child.content.strip() for child in node.children[1:]}
911
            redeclared = self.directives.tokens & tokens
912
            if redeclared:
eckhart's avatar
eckhart committed
913
                self.tree.new_error(node, 'Tokens %s have already been declared earlier. '
eckhart's avatar
eckhart committed
914
915
                                    % str(redeclared) + 'Later declaration will be ignored',
                                    code=Error.REDECLARED_TOKEN_WARNING)
916
            self.directives.tokens |= tokens - redeclared
917

918
        elif key.endswith('_filter'):
eckhart's avatar
eckhart committed
919
            check_argnum()
920
            symbol = key[:-7]
921
            self.directives.filter[symbol] = node.children[1].content.strip()
922

923
        elif key.endswith('_error'):
924
            check_argnum(2)
925
            symbol = key[:-6]
926
            error_msgs = self.directives.error.get(symbol, [])
eckhart's avatar
eckhart committed
927
928
929
            if symbol in self.rules:
                self.tree.new_error(node, 'Custom error message for symbol "%s"' % symbol