ebnf.py 44.2 KB
Newer Older
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
# ebnf.py - EBNF -> Python-Parser compilation for DHParser
#
# Copyright 2016  by Eckhart Arnold (arnold@badw.de)
#                 Bavarian Academy of Sciences an Humanities (badw.de)
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
# implied.  See the License for the specific language governing
# permissions and limitations under the License.
17
18


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

26

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

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

45

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


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


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


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

78

di68kap's avatar
di68kap committed
79
class EBNFGrammar(Grammar):
eckhart's avatar
eckhart committed
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
    r"""
    Parser for an EBNF source file, with this grammar::

        # EBNF-Grammar in EBNF

        @ comment    = /#.*(?:\n|$)/                    # comments start with '#' and eat all chars up to and including '\n'
        @ whitespace = /\s*/                            # whitespace includes linefeed
        @ literalws  = right                            # trailing whitespace of literals will be ignored tacitly

        syntax     = [~//] { definition | directive } §EOF
        definition = symbol §"=" expression
        directive  = "@" §symbol "=" ( regexp | literal | list_ )

        expression = term { "|" term }
        term       = { ["§"] factor }+                       # "§" means all following factors mandatory
        factor     = [flowmarker] [retrieveop] symbol !"="   # negative lookahead to be sure it's not a definition
                   | [flowmarker] literal
                   | [flowmarker] plaintext
                   | [flowmarker] regexp
                   | [flowmarker] whitespace
                   | [flowmarker] oneormore
                   | [flowmarker] group
                   | [flowmarker] unordered
                   | repetition
                   | option

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

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

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


165
def grammar_changed(grammar_class, grammar_source: str) -> bool:
166
167
    """
    Returns ``True`` if ``grammar_class`` does not reflect the latest
Eckhart Arnold's avatar
Eckhart Arnold committed
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
    changes of ``grammar_source``

    Parameters:
        grammar_class:  the parser class representing the grammar
            or the file name of a compiler suite containing the grammar
        grammar_source:  File name or string representation of the
            EBNF code of the grammar

    Returns (bool):
        True, if the source text of the grammar is different from the
        source from which the grammar class was generated
    """
    grammar = load_if_file(grammar_source)
    chksum = md5(grammar, __version__)
    if isinstance(grammar_class, str):
        # grammar_class = load_compiler_suite(grammar_class)[1]
        with open(grammar_class, 'r', encoding='utf8') as f:
            pycode = f.read()
eckhart's avatar
eckhart committed
186
        m = re.search(r'class \w*\(Grammar\)', pycode)
Eckhart Arnold's avatar
Eckhart Arnold committed
187
188
189
190
191
192
193
194
195
196
        if m:
            m = re.search('    source_hash__ *= *"([a-z0-9]*)"',
                          pycode[m.span()[1]:])
            return not (m and m.groups() and m.groups()[-1] == chksum)
        else:
            return True
    else:
        return chksum != grammar_class.source_hash__


197
def get_ebnf_grammar() -> EBNFGrammar:
Eckhart Arnold's avatar
Eckhart Arnold committed
198
    try:
199
        grammar = GLOBALS.ebnf_grammar_singleton
Eckhart Arnold's avatar
Eckhart Arnold committed
200
        return grammar
201
    except AttributeError:
202
203
        GLOBALS.ebnf_grammar_singleton = EBNFGrammar()
        return GLOBALS.ebnf_grammar_singleton
Eckhart Arnold's avatar
Eckhart Arnold committed
204
205
206
207
208
209
210
211
212


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


213
EBNF_AST_transformation_table = {
214
    # AST Transformations for EBNF-grammar
215
    "<":
216
        remove_expendables,
217
    "syntax":
218
        [],  # otherwise '"*": replace_by_single_child' would be applied
219
    "directive, definition":
220
        remove_tokens('@', '='),
Eckhart Arnold's avatar
Eckhart Arnold committed
221
    "expression":
222
        [replace_by_single_child, flatten, remove_tokens('|')],  # remove_infix_operator],
223
    "term":
224
225
        [replace_by_single_child, flatten],  # supports both idioms:
                                             # "{ factor }+" and "factor { factor }"
226
    "factor, flowmarker, retrieveop":
227
        replace_by_single_child,
228
    "group":
229
        [remove_brackets, replace_by_single_child],
230
231
    "unordered":
        remove_brackets,
232
    "oneormore, repetition, option":
Eckhart Arnold's avatar
Eckhart Arnold committed
233
        [reduce_single_child, remove_brackets,
234
         forbid('repetition', 'option', 'oneormore'), assert_content(r'(?!§)(?:.|\n)*')],
235
    "symbol, literal, regexp":
Eckhart Arnold's avatar
Eckhart Arnold committed
236
        reduce_single_child,
237
    (TOKEN_PTYPE, WHITESPACE_PTYPE):
Eckhart Arnold's avatar
Eckhart Arnold committed
238
        reduce_single_child,
239
    "list_":
240
        [flatten, remove_infix_operator],
241
    "*":
242
        replace_by_single_child
243
244
}

245

Eckhart Arnold's avatar
Eckhart Arnold committed
246
def EBNFTransform() -> TransformationFunc:
247
    return partial(traverse, processing_table=EBNF_AST_transformation_table.copy())
di68kap's avatar
di68kap committed
248

eckhart's avatar
eckhart committed
249

250
def get_ebnf_transformer() -> TransformationFunc:
251
    try:
252
        transformer = GLOBALS.EBNF_transformer_singleton
253
    except AttributeError:
254
255
        GLOBALS.EBNF_transformer_singleton = EBNFTransform()
        transformer = GLOBALS.EBNF_transformer_singleton
256
    return transformer
Eckhart Arnold's avatar
Eckhart Arnold committed
257
258
259
260
261
262
263
264


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

265

266
PreprocessorFactoryFunc = Callable[[], PreprocessorFunc]
267
ParserFactoryFunc = Callable[[], Grammar]
268
TransformerFactoryFunc = Callable[[], TransformationFunc]
269
270
CompilerFactoryFunc = Callable[[], Compiler]

271
272
273
PREPROCESSOR_FACTORY = '''
def get_preprocessor() -> PreprocessorFunc:
    return {NAME}Preprocessor
274
275
276
277
'''


GRAMMAR_FACTORY = '''
278
def get_grammar() -> {NAME}Grammar:
279
    try:
280
        grammar = GLOBALS.{NAME}_{ID}_grammar_singleton
281
    except AttributeError:
282
283
        GLOBALS.{NAME}_{ID}_grammar_singleton = {NAME}Grammar()
        grammar = GLOBALS.{NAME}_{ID}_grammar_singleton
284
    return grammar
285
286
287
288
'''


TRANSFORMER_FACTORY = '''
289
290
291
def {NAME}Transform() -> TransformationDict:
    return partial(traverse, processing_table={NAME}_AST_transformation_table.copy())

292
def get_transformer() -> TransformationFunc:
293
    try:
294
        transformer = GLOBALS.{NAME}_{ID}_transformer_singleton
295
    except AttributeError:
296
297
        GLOBALS.{NAME}_{ID}_transformer_singleton = {NAME}Transform()
        transformer = GLOBALS.{NAME}_{ID}_transformer_singleton
298
    return transformer
299
300
301
302
'''


COMPILER_FACTORY = '''
eckhart's avatar
eckhart committed
303
def get_compiler() -> {NAME}Compiler:
304
    try:
305
        compiler = GLOBALS.{NAME}_{ID}_compiler_singleton
306
    except AttributeError:
eckhart's avatar
eckhart committed
307
        GLOBALS.{NAME}_{ID}_compiler_singleton = {NAME}Compiler()
308
        compiler = GLOBALS.{NAME}_{ID}_compiler_singleton
309
    return compiler
310
311
'''

Eckhart Arnold's avatar
Eckhart Arnold committed
312

eckhart's avatar
eckhart committed
313
class EBNFCompilerError(CompilerError):
314
    """Error raised by `EBNFCompiler` class. (Not compilation errors
315
    in the strict sense, see `CompilationError` in module ``dsl.py``)"""
316
317
318
    pass


319
class EBNFCompiler(Compiler):
320
321
    """
    Generates a Parser from an abstract syntax tree of a grammar specified
322
    in EBNF-Notation.
323
324
325
326
327
328
329
330
331
332
333
334
335

    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:
336
        current_symbols:  During compilation, a list containing the root
337
338
339
340
                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.

341
        rules:  Dictionary that maps rule names to a list of Nodes that
342
343
344
345
346
347
                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`

348
                Now `[node.content for node in self.rules['alternative']]`
349
350
                yields `['alternative = a | b', 'a', 'b']`

351
        symbols:  A mapping of symbol names to their first usage (not
352
353
                their definition!) in the EBNF source.

354
        variables:  A set of symbols names that are used with the
355
356
357
358
                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.

359
        recursive:  A set of symbols that are used recursively and
360
361
                therefore require a `Forward`-operator.

362
        definitions:  A dictionary of definitions. Other than `rules`
363
364
                this maps the symbols to their compiled definienda.

365
        deferred_tasks:  A list of callables that is filled during
366
367
368
369
370
                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.

371
        root_symbol: The name of the root symbol.
372

373
        directives:  A dictionary of all directives and their default
374
                values.
375
376
377

        re_flags:  A set of regular expression flags to be added to all
                regular expressions found in the current parsing process
378

eckhart's avatar
eckhart committed
379
380
381
382
        grammar_name:  The name of the grammar to be compiled

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

383
384
385
        grammar_id: a unique id for every compiled grammar. (Required for
                disambiguation of of thread local variables storing
                compiled texts.)
386
387
    """
    COMMENT_KEYWORD = "COMMENT__"
388
    WHITESPACE_KEYWORD = "WSP_RE__"
Eckhart Arnold's avatar
Eckhart Arnold committed
389
    RAW_WS_KEYWORD = "WHITESPACE__"
390
    WHITESPACE_PARSER_KEYWORD = "wsp__"
Eckhart Arnold's avatar
Eckhart Arnold committed
391
    RESERVED_SYMBOLS = {WHITESPACE_KEYWORD, RAW_WS_KEYWORD, COMMENT_KEYWORD}
392
    AST_ERROR = "Badly structured syntax tree. " \
Eckhart Arnold's avatar
Eckhart Arnold committed
393
                "Potentially due to erroneous AST transformation."
394
395
396
397
    PREFIX_TABLE = {'§': 'Required',
                    '&': 'Lookahead', '!': 'NegativeLookahead',
                    '-&': 'Lookbehind', '-!': 'NegativeLookbehind',
                    '::': 'Pop', ':': 'Retrieve'}
398
399
400
    WHITESPACE = {'horizontal': r'[\t ]*',  # default: horizontal
                  'linefeed': r'[ \t]*\n?(?!\s*\n)[ \t]*',
                  'vertical': r'\s*'}
401
    REPEATABLE_DIRECTIVES = {'tokens'}
402

403

eckhart's avatar
eckhart committed
404
    def __init__(self, grammar_name="DSL", grammar_source=""):
405
        self.grammar_id = 0
eckhart's avatar
eckhart committed
406
407
        super(EBNFCompiler, self).__init__()  # calls the _reset()-method
        self.set_grammar_name(grammar_name, grammar_source)
408

409

410
    def _reset(self):
411
        super(EBNFCompiler, self)._reset()
412
        self._result = ''           # type: str
413
        self.re_flags = set()       # type: Set[str]
414
415
416
        self.rules = OrderedDict()  # type: OrderedDict[str, List[Node]]
        self.current_symbols = []   # type: List[Node]
        self.symbols = {}           # type: Dict[str, Node]
417
418
        self.variables = set()      # type: Set[str]
        self.recursive = set()      # type: Set[str]
419
        self.definitions = {}       # type: Dict[str, str]
420
        self.deferred_tasks = []    # type: List[Callable]
421
        self.root_symbol = ""       # type: str
eckhart's avatar
eckhart committed
422
        self.directives = {'whitespace': self.WHITESPACE['vertical'],
423
                           'comment': '',
424
                           'literalws': {'right'},
425
                           'tokens': set(),  # alt. 'preprocessor_tokens'
426
427
428
                           'filter': dict()}  # alt. 'filter'
        # self.directives['ignorecase']: False
        self.defined_directives = set()  # type: Set[str]
429
430
        self.grammar_id += 1

431

Eckhart Arnold's avatar
Eckhart Arnold committed
432
    @property
433
    def result(self) -> str:
Eckhart Arnold's avatar
Eckhart Arnold committed
434
435
        return self._result

eckhart's avatar
eckhart committed
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452

    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


453
    # methods for generating skeleton code for preprocessor, transformer, and compiler
454

455
    def gen_preprocessor_skeleton(self) -> str:
456
457
458
459
        """
        Returns Python-skeleton-code for a preprocessor-function for
        the previously compiled formal language.
        """
460
        name = self.grammar_name + "Preprocessor"
461
        return "def %s(text):\n    return text, lambda i: i\n" % name \
462
               + PREPROCESSOR_FACTORY.format(NAME=self.grammar_name)
463

464

465
    def gen_transformer_skeleton(self) -> str:
466
467
468
469
        """
        Returns Python-skeleton-code for the AST-transformation for the
        previously compiled formal language.
        """
470
        if not self.rules:
Eckhart Arnold's avatar
Eckhart Arnold committed
471
472
            raise EBNFCompilerError('Compiler must be run before calling '
                                    '"gen_transformer_Skeleton()"!')
473
        tt_name = self.grammar_name + '_AST_transformation_table'
di68kap's avatar
di68kap committed
474
        transtable = [tt_name + ' = {',
Eckhart Arnold's avatar
Eckhart Arnold committed
475
                      '    # AST Transformations for the ' + self.grammar_name + '-grammar']
476
        transtable.append('    "<": remove_empty,')
477
        for name in self.rules:
eckhart's avatar
eckhart committed
478
            transformations = '[]'
479
480
            rule = self.definitions[name]
            if rule.startswith('Alternative'):
eckhart's avatar
eckhart committed
481
                transformations = '[replace_or_reduce]'
482
            elif rule.startswith('Synonym'):
Eckhart Arnold's avatar
Eckhart Arnold committed
483
                transformations = '[reduce_single_child]'
eckhart's avatar
eckhart committed
484
            transtable.append('    "' + name + '": %s,' % transformations)
485
        transtable.append('    ":Token": reduce_single_child,')
486
        transtable += ['    "*": replace_by_single_child', '}', '']
487
        transtable += [TRANSFORMER_FACTORY.format(NAME=self.grammar_name, ID=self.grammar_id)]
488
489
        return '\n'.join(transtable)

490

491
    def gen_compiler_skeleton(self) -> str:
492
493
494
495
        """
        Returns Python-skeleton-code for a Compiler-class for the
        previously compiled formal language.
        """
496
        if not self.rules:
497
498
            raise EBNFCompilerError('Compiler has not been run before calling '
                                    '"gen_Compiler_Skeleton()"!')
499
        compiler = ['class ' + self.grammar_name + 'Compiler(Compiler):',
eckhart's avatar
eckhart committed
500
501
                    '    """Compiler for the abstract-syntax-tree of a '
                    + self.grammar_name + ' source file.',
502
                    '    """', '',
eckhart's avatar
eckhart committed
503
504
505
                    '    def __init__(self):',
                    '        super(' + self.grammar_name + 'Compiler, self).__init__()',
                    '',
506
507
508
                    '    def _reset(self):',
                    '        super()._reset()',
                    '        # initialize your variables here, not in the constructor!']
509
        for name in self.rules:
510
            method_name = Compiler.method_name(name)
511
            if name == self.root_symbol:
512
                compiler += ['    def ' + method_name + '(self, node):',
513
                             '        return self.fallback_compiler(node)', '']
514
            else:
di68kap's avatar
di68kap committed
515
                compiler += ['    # def ' + method_name + '(self, node):',
eckhart's avatar
eckhart committed
516
                             '    #     return node', '']
517
        compiler += [COMPILER_FACTORY.format(NAME=self.grammar_name, ID=self.grammar_id)]
518
        return '\n'.join(compiler)
519

520
    def verify_transformation_table(self, transtable):
521
522
523
524
525
526
        """
        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.
        """
527
        assert self._dirty_flag
528
        table_entries = set(expand_table(transtable).keys()) - {'*', '<', '>', '~'}
529
530
531
532
533
534
        symbols = self.rules.keys()
        messages = []
        for entry in table_entries:
            if entry not in symbols and not entry.startswith(":"):
                messages.append(Error(('Symbol "%s" is not defined in grammar %s but appears in '
                                       'the transformation table!') % (entry, self.grammar_name),
Eckhart Arnold's avatar
Eckhart Arnold committed
535
                                      0, Error.UNDEFINED_SYMBOL_IN_TRANSFORMATION_TABLE))
536
537
        return messages

538
539
540
541
542
543
544
    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
545

546
547
548
549
550
    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
        """
551
552
553
554
555
556
557
558
559
560

        # 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

561
562
563
        if self.variables:
            for i in range(len(definitions)):
                if definitions[i][0] in self.variables:
564
                    definitions[i] = (definitions[i][0], 'Capture(%s)' % definitions[i][1])
565

566
567
        # add special fields for Grammar class

568
569
        definitions.append((self.WHITESPACE_PARSER_KEYWORD,
                            'Whitespace(%s)' % self.WHITESPACE_KEYWORD))
570
        definitions.append((self.WHITESPACE_KEYWORD,
eckhart's avatar
eckhart committed
571
572
                            ("mixin_comment(whitespace=" + self.RAW_WS_KEYWORD
                             + ", comment=" + self.COMMENT_KEYWORD + ")")))
Eckhart Arnold's avatar
Eckhart Arnold committed
573
        definitions.append((self.RAW_WS_KEYWORD, "r'{whitespace}'".format(**self.directives)))
574
575
576
577
        definitions.append((self.COMMENT_KEYWORD, "r'{comment}'".format(**self.directives)))

        # prepare parser class header and docstring and
        # add EBNF grammar to the doc string of the parser class
578

579
        article = 'an ' if self.grammar_name[0:1] in "AaEeIiOoUu" else 'a '  # what about 'hour', 'universe' etc.?
580
        show_source = get_config_value('add_grammar_source_to_parser_docstring')
eckhart's avatar
eckhart committed
581
582
583
584
        declarations = ['class ' + self.grammar_name
                        + 'Grammar(Grammar):',
                        'r"""Parser for ' + article + self.grammar_name
                        + ' source file'
585
                        + ('. Grammar:' if self.grammar_source and show_source else '.')]
586
        definitions.append(('parser_initialization__', '"upon instantiation"'))
587
        if self.grammar_source:
588
            definitions.append(('source_hash__',
589
                                '"%s"' % md5(self.grammar_source, __version__)))
590
            declarations.append('')
591
592
            if show_source:
                declarations += [line for line in self.grammar_source.split('\n')]
593
594
595
596
597
            while declarations[-1].strip() == '':
                declarations = declarations[:-1]
        declarations.append('"""')

        # turn definitions into declarations in reverse order
598

599
        self.root_symbol = definitions[0][0] if definitions else ""
600
601
602
603
604
605
606
607
        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]
608
609
610
611
612
613

        # 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
614
                self.tree.new_error(self.symbols[symbol],
eckhart's avatar
eckhart committed
615
                                    "Missing definition for symbol '%s'" % symbol)
616
                # root_node.error_flag = True
617
618
619

        # check for unconnected rules

Eckhart Arnold's avatar
Eckhart Arnold committed
620
621
622
        defined_symbols.difference_update(self.RESERVED_SYMBOLS)

        def remove_connections(symbol):
623
            """Recursively removes all symbols which appear in the
eckhart's avatar
eckhart committed
624
            definiens of a particular symbol."""
Eckhart Arnold's avatar
Eckhart Arnold committed
625
626
627
628
629
630
631
            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
632
633
634
            self.tree.new_error(self.rules[leftover][0],
                                ('Rule "%s" is not connected to parser root "%s" !') %
                                (leftover, self.root_symbol), Error.WARNING)
635

636
        # set root_symbol parser and assemble python grammar definition
637

638
639
        if self.root_symbol and 'root__' not in self.rules:
            declarations.append('root__ = ' + self.root_symbol)
640
        declarations.append('')
Eckhart Arnold's avatar
Eckhart Arnold committed
641
        self._result = '\n    '.join(declarations) \
642
                       + GRAMMAR_FACTORY.format(NAME=self.grammar_name, ID=self.grammar_id)
Eckhart Arnold's avatar
Eckhart Arnold committed
643
        return self._result
644

645
646
647

    ## compilation methods

648
    def on_syntax(self, node: Node) -> str:
649
        definitions = []  # type: List[Tuple[str, str]]
650
651

        # drop the wrapping sequence node
652
653
        if len(node.children) == 1 and not node.children[0].parser.name:
            node = node.children[0]
654
655

        # compile definitions and directives and collect definitions
656
        for nd in node.children:
657
            if nd.parser.name == "definition":
658
                definitions.append(self.compile(nd))
659
            else:
660
                assert nd.parser.name == "directive", nd.as_sxpr()
661
                self.compile(nd)
662
            # node.error_flag = max(node.error_flag, nd.error_flag)
663
        self.definitions.update(definitions)
664

665
        return self.assemble_parser(definitions, node)
666

667

668
    def on_definition(self, node: Node) -> Tuple[str, str]:
669
        rule = node.children[0].content
670
        if rule in self.rules:
Eckhart Arnold's avatar
Eckhart Arnold committed
671
            first = self.rules[rule][0]
eckhart's avatar
eckhart committed
672
            if not first.errors:
eckhart's avatar
eckhart committed
673
                self.tree.new_error(first, 'First definition of rule "%s" '
eckhart's avatar
eckhart committed
674
                                    'followed by illegal redefinitions.' % rule)
eckhart's avatar
eckhart committed
675
            self.tree.new_error(node, 'A rule "%s" has already been defined earlier.' % rule)
676
        elif rule in EBNFCompiler.RESERVED_SYMBOLS:
eckhart's avatar
eckhart committed
677
            self.tree.new_error(node, 'Symbol "%s" is a reserved symbol.' % rule)
678
        elif not sane_parser_name(rule):
eckhart's avatar
eckhart committed
679
            self.tree.new_error(node, 'Illegal symbol "%s". Symbols must not start or '
eckhart's avatar
eckhart committed
680
                                ' end with a doube underscore "__".' % rule)
681
        elif rule in self.directives['tokens']:
eckhart's avatar
eckhart committed
682
            self.tree.new_error(node, 'Symbol "%s" has already been defined as '
eckhart's avatar
eckhart committed
683
                                'a preprocessor token.' % rule)
684
        elif keyword.iskeyword(rule):
eckhart's avatar
eckhart committed
685
            self.tree.new_error(node, 'Python keyword "%s" may not be used as a symbol. '
eckhart's avatar
eckhart committed
686
                                % rule + '(This may change in the future.)')
687
        try:
688
689
            self.current_symbols = [node]
            self.rules[rule] = self.current_symbols
690
            defn = self.compile(node.children[1])
691
            if rule in self.variables:
692
                defn = 'Capture(%s)' % defn
693
                self.variables.remove(rule)
694
695
696
            elif defn.find("(") < 0:
                # assume it's a synonym, like 'page = REGEX_PAGE_NR'
                defn = 'Synonym(%s)' % defn
697
        except TypeError as error:
698
699
700
701
            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
702
            self.tree.new_error(node, errmsg)
703
            rule, defn = rule + ':error', '"' + errmsg + '"'
Eckhart Arnold's avatar
Eckhart Arnold committed
704
        return rule, defn
705

706

707
    def _check_rx(self, node: Node, rx: str) -> str:
708
709
        """
        Checks whether the string `rx` represents a valid regular
710
711
712
        expression. Makes sure that multiline regular expressions are
        prepended by the multiline-flag. Returns the regular expression string.
        """
713
        flags = self.re_flags | {'x'} if rx.find('\n') >= 0 else self.re_flags
eckhart's avatar
eckhart committed
714
715
        if flags:
            rx = "(?%s)%s" % ("".join(flags), rx)
716
717
718
        try:
            re.compile(rx)
        except Exception as re_error:
eckhart's avatar
eckhart committed
719
            self.tree.new_error(node, "malformed regular expression %s: %s" %
eckhart's avatar
eckhart committed
720
                                (repr(rx), str(re_error)))
721
722
        return rx

723

724
    def on_directive(self, node: Node) -> str:
725
        key = node.children[0].content.lower()
726
        assert key not in self.directives['tokens']
727

728
729
        if key not in self.REPEATABLE_DIRECTIVES:
            if key in self.defined_directives:
eckhart's avatar
eckhart committed
730
                self.tree.new_error(node, 'Directive "%s" has already been defined earlier. '
eckhart's avatar
eckhart committed
731
732
                                    % key + 'Later definition will be ignored!',
                                    code=Error.REDEFINED_DIRECTIVE_WARNING)
733
734
735
                return ""
            self.defined_directives.add(key)

736
        if key in {'comment', 'whitespace'}:
737
738
            if node.children[1].parser.name == "list_":
                if len(node.children[1].result) != 1:
eckhart's avatar
eckhart committed
739
                    self.tree.new_error(node, 'Directive "%s" must have one, but not %i values.'
eckhart's avatar
eckhart committed
740
                                        % (key, len(node.children[1].result)))
741
                value = self.compile(node.children[1]).pop()
742
743
                if key == 'whitespace' and value in EBNFCompiler.WHITESPACE:
                    value = EBNFCompiler.WHITESPACE[value]  # replace whitespace-name by regex
744
                else:
eckhart's avatar
eckhart committed
745
                    self.tree.new_error(node, 'Value "%s" not allowed for directive "%s".'
eckhart's avatar
eckhart committed
746
                                        % (value, key))
747
            else:
748
749
750
                value = node.children[1].content.strip("~")  # cast(str, node.children[
                # 1].result).strip("~")
                if value != node.children[1].content:  # cast(str, node.children[1].result):
eckhart's avatar
eckhart committed
751
                    self.tree.new_error(node, "Whitespace marker '~' not allowed in definition "
eckhart's avatar
eckhart committed
752
                                        "of %s regular expression." % key)
753
754
755
756
                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])
757
                if key == 'whitespace' and not re.match(value, ''):
eckhart's avatar
eckhart committed
758
                    self.tree.new_error(node, "Implicit whitespace should always "
eckhart's avatar
eckhart committed
759
                                        "match the empty string, /%s/ does not." % value)
760
            self.directives[key] = value
761

762
        elif key == 'ignorecase':
763
            if node.children[1].content.lower() not in {"off", "false", "no"}:
764
765
                self.re_flags.add('i')

Eckhart Arnold's avatar
Eckhart Arnold committed
766
        # elif key == 'testing':
767
        #     value = node.children[1].content
Eckhart Arnold's avatar
Eckhart Arnold committed
768
        #     self.directives['testing'] = value.lower() not in {"off", "false", "no"}
769

770
        elif key == 'literalws':
771
            value = {item.lower() for item in self.compile(node.children[1])}
eckhart's avatar
eckhart committed
772
            if ((value - {'left', 'right', 'both', 'none'})
Eckhart Arnold's avatar
Eckhart Arnold committed
773
                    or ('none' in value and len(value) > 1)):
eckhart's avatar
eckhart committed
774
                self.tree.new_error(node, 'Directive "literalws" allows only `left`, `right`, '
eckhart's avatar
eckhart committed
775
                                    '`both` or `none`, not `%s`' % ", ".join(value))
eckhart's avatar
eckhart committed
776
            wsp = {'left', 'right'} if 'both' in value \
777
                else {} if 'none' in value else value
eckhart's avatar
eckhart committed
778
            self.directives[key] = list(wsp)
779

780
        elif key in {'tokens', 'preprocessor_tokens'}:
781
            tokens = self.compile(node.children[1])
782
            redeclared = self.directives['tokens'] & tokens
783
            if redeclared:
eckhart's avatar
eckhart committed
784
                self.tree.new_error(node, 'Tokens %s have already been declared earlier. '
eckhart's avatar
eckhart committed
785
786
                                    % str(redeclared) + 'Later declaration will be ignored',
                                    code=Error.REDECLARED_TOKEN_WARNING)
787
            self.directives['tokens'] |= tokens - redeclared
788

789
        elif key.endswith('_filter'):
790
            filter_set = self.compile(node.children[1])
791
            if not isinstance(filter_set, set) or len(filter_set) != 1:
eckhart's avatar
eckhart committed
792
                self.tree.new_error(node, 'Directive "%s" accepts exactly on symbol, not %s'
eckhart's avatar
eckhart committed
793
                                    % (key, str(filter_set)))
794
            self.directives['filter'][key[:-7]] = filter_set.pop()
795

796
        else:
eckhart's avatar
eckhart committed
797
            self.tree.new_error(node, 'Unknown directive %s ! (Known ones are %s .)' %
eckhart's avatar
eckhart committed
798
                                (key, ', '.join(list(self.directives.keys()))))
799
800
        return ""

801

eckhart's avatar
eckhart committed
802
    def non_terminal(self, node: Node, parser_class: str, custom_args: List[str] = []) -> str:
803
804
        """
        Compiles any non-terminal, where `parser_class` indicates the Parser class
805
806
        name for the particular non-terminal.
        """
807
        arguments = [self.compile(r) for r in node.children] + custom_args
808
809
        return parser_class + '(' + ', '.join(arguments) + ')'

810

811
    def on_expression(self, node) -> str:
di68kap's avatar
di68kap committed
812
        # TODO: Add check for errors like "a" | "ab" (which will always yield a, even for ab)
813
814
        return self.non_terminal(node, 'Alternative')

815

816
    def on_term(self, node) -> str:
di68kap's avatar
di68kap committed
817
818
819
820
821
        # Basically, the following code does only this:
        #       return self.non_terminal(node, 'Series')
        # What makes it (look) more complicated is the handling of the
        # mandatory §-operator
        mandatory_marker = []
eckhart's avatar
eckhart committed
822
        filtered_children = []  # type: List[Node]
di68kap's avatar
di68kap committed
823
        for nd in node.children:
824
            if nd.parser.ptype == TOKEN_PTYPE and nd.content == "§":
eckhart's avatar
eckhart committed
825
826
                mandatory_marker.append(len(filtered_children))
                # if len(filtered_children) == 0:
eckhart's avatar
eckhart committed
827
                #     self.tree.new_error(nd.pos, 'First item of a series should not be mandatory.',
eckhart's avatar
eckhart committed
828
                #                         Error.WARNING)
eckhart's avatar
eckhart committed
829
                if len(mandatory_marker) > 1:
eckhart's avatar
eckhart committed
830
                    self.tree.new_error(nd, 'One mandatory marker (§) sufficient to declare '
eckhart's avatar
eckhart committed
831
                                        'the rest of the series as mandatory.', Error.WARNING)
di68kap's avatar
di68kap committed
832
833
834
835
            else:
                filtered_children.append(nd)
        saved_result = node.result
        node.result = tuple(filtered_children)
eckhart's avatar
eckhart committed
836
837
838
839
840
        if len(filtered_children) == 1:
            compiled = self.non_terminal(node, 'Required')
        else:
            custom_args = ['mandatory=%i' % mandatory_marker[0]] if mandatory_marker else []
            compiled = self.non_terminal(node, 'Series', custom_args)
di68kap's avatar
di68kap committed
841
842
        node.result = saved_result
        return compiled
Eckhart Arnold's avatar