ebnf.py 40.3 KB
Newer Older
1
"""ebnf.py - EBNF -> Python-Parser compilation for DHParser
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18

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.
"""

19
import keyword
20
from collections import OrderedDict
21
from functools import partial
22
from typing import Callable, Dict, List, Set, Tuple
23

24
from DHParser.error import Error
25
from DHParser.parse import Grammar, mixin_comment, Forward, RegExp, RE, \
26
    NegativeLookahead, Alternative, Series, Option, OneOrMore, ZeroOrMore, Token, \
27
28
    Compiler
from DHParser.preprocess import nil_preprocessor, PreprocessorFunc
29
from DHParser.syntaxtree import Node, TransformationFunc, WHITESPACE_PTYPE, TOKEN_PTYPE
30
from DHParser.toolkit import load_if_file, escape_re, md5, sane_parser_name, re, expand_table
31
from DHParser.transform import traverse, remove_brackets, \
Eckhart Arnold's avatar
Eckhart Arnold committed
32
    reduce_single_child, replace_by_single_child, remove_expendables, \
33
    remove_tokens, flatten, forbid, assert_content, remove_infix_operator
34
from DHParser.versionnumber import __version__
35

36
__all__ = ('get_ebnf_preprocessor',
37
38
39
40
           'get_ebnf_grammar',
           'get_ebnf_transformer',
           'get_ebnf_compiler',
           'EBNFGrammar',
41
           'EBNFTransform',
Eckhart Arnold's avatar
Eckhart Arnold committed
42
           'EBNFCompilerError',
43
           'EBNFCompiler',
44
           'grammar_changed',
45
           'PreprocessorFactoryFunc',
46
47
           'ParserFactoryFunc',
           'TransformerFactoryFunc',
48
           'CompilerFactoryFunc')
49
50


Eckhart Arnold's avatar
Eckhart Arnold committed
51
52
53
54
55
56
57
########################################################################
#
# EBNF scanning
#
########################################################################


58
59
def get_ebnf_preprocessor() -> PreprocessorFunc:
    return nil_preprocessor
Eckhart Arnold's avatar
Eckhart Arnold committed
60
61
62
63
64
65
66
67


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

68

di68kap's avatar
di68kap committed
69
70
71
72
73
class EBNFGrammar(Grammar):
    r"""Parser for an EBNF source file, with this grammar:

    # EBNF-Grammar in EBNF

74
75
76
77
78
    @ 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
di68kap's avatar
di68kap committed
79
80
81
82
83
84

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

    expression =  term { "|" term }
85
86
87
    term       =  { ["§"] factor }+               # "§" means all following factors mandatory
    factor     =  [flowmarker] [retrieveop] symbol !"="   # negative lookahead to be sure
                                                          # it's not a definition
di68kap's avatar
di68kap committed
88
89
90
                | [flowmarker] literal
                | [flowmarker] regexp
                | [flowmarker] oneormore
91
92
                | [flowmarker] group
                | [flowmarker] unordered
di68kap's avatar
di68kap committed
93
94
95
                | repetition
                | option

96
97
98
    flowmarker =  "!"  | "&"                     # '!' negative lookahead, '&' positive lookahead
                | "-!" | "-&"                    # '-' negative lookbehind, '-&' positive lookbehind
    retrieveop =  "::" | ":"                     # '::' pop, ':' retrieve
di68kap's avatar
di68kap committed
99

100
    group      =  "(" §expression ")"
101
    unordered  =  "<" §expression ">"            # elements of expression in arbitrary order
di68kap's avatar
di68kap committed
102
    oneormore  =  "{" expression "}+"
103
104
    repetition =  "{" §expression "}"
    option     =  "[" §expression "]"
di68kap's avatar
di68kap committed
105

106
107
108
109
110
111
112
113
114
115
116
    symbol     =  /(?!\d)\w+/~                   # e.g. expression, factor, parameter_list
    literal    =  /"(?:[^"]|\\")*?"/~            # e.g. "(", '+', 'while'
                | /'(?:[^']|\\')*?'/~            # whitespace following literals will be ignored
    regexp     =  /~?\/(?:\\\/|[^\/])*?\/~?/~    # e.g. /\w+/, ~/#.*(?:\n|$)/~
                                                 # '~' is a whitespace-marker, if present leading
                                                 # or trailing whitespace of a regular expression
                                                 # will be ignored tacitly.
    list_      =  /\w+/~ { "," /\w+/~ }          # comma separated list of symbols,
                                                 # e.g. BEGIN_LIST, END_LIST,
                                                 # BEGIN_QUOTE, END_QUOTE
                                                 # see CommonMark/markdown.py for an exmaple
di68kap's avatar
di68kap committed
117
118
119
    EOF =  !/./
    """
    expression = Forward()
120
121
    source_hash__ = "3fc9f5a340f560e847d9af0b61a68743"
    parser_initialization__ = "upon instantiation"
di68kap's avatar
di68kap committed
122
123
124
125
126
127
    COMMENT__ = r'#.*(?:\n|$)'
    WHITESPACE__ = r'\s*'
    WSP__ = mixin_comment(whitespace=WHITESPACE__, comment=COMMENT__)
    wspL__ = ''
    wspR__ = WSP__
    EOF = NegativeLookahead(RegExp('.'))
Eckhart Arnold's avatar
Eckhart Arnold committed
128
    list_ = Series(RE('\\w+'), ZeroOrMore(Series(Token(","), RE('\\w+'))))
di68kap's avatar
di68kap committed
129
130
131
    regexp = RE('~?/(?:\\\\/|[^/])*?/~?')
    literal = Alternative(RE('"(?:[^"]|\\\\")*?"'), RE("'(?:[^']|\\\\')*?'"))
    symbol = RE('(?!\\d)\\w+')
132
133
    option = Series(Token("["), expression, Token("]"), mandatory=1)
    repetition = Series(Token("{"), expression, Token("}"), mandatory=1)
Eckhart Arnold's avatar
Eckhart Arnold committed
134
    oneormore = Series(Token("{"), expression, Token("}+"))
135
    unordered = Series(Token("<"), expression, Token(">"), mandatory=1)
136
    group = Series(Token("("), expression, Token(")"), mandatory=1)
di68kap's avatar
di68kap committed
137
138
    retrieveop = Alternative(Token("::"), Token(":"))
    flowmarker = Alternative(Token("!"), Token("&"), Token("-!"), Token("-&"))
139
140
141
142
143
    factor = Alternative(
        Series(Option(flowmarker), Option(retrieveop), symbol, NegativeLookahead(Token("="))),
        Series(Option(flowmarker), literal), Series(Option(flowmarker), regexp),
        Series(Option(flowmarker), oneormore), Series(Option(flowmarker), group),
        Series(Option(flowmarker), unordered), repetition, option)
Eckhart Arnold's avatar
Eckhart Arnold committed
144
145
    term = OneOrMore(Series(Option(Token("§")), factor))
    expression.set(Series(term, ZeroOrMore(Series(Token("|"), term))))
146
147
    directive = Series(Token("@"), symbol, Token("="), Alternative(regexp, literal, list_),
                       mandatory=1)
di68kap's avatar
di68kap committed
148
    definition = Series(symbol, Token("="), expression, mandatory=1)
149
150
    syntax = Series(Option(RE('', wR='', wL=WSP__)), ZeroOrMore(Alternative(definition, directive)),
                    EOF, mandatory=2)
di68kap's avatar
di68kap committed
151
152
153
    root__ = syntax


154
def grammar_changed(grammar_class, grammar_source: str) -> bool:
155
156
    """
    Returns ``True`` if ``grammar_class`` does not reflect the latest
Eckhart Arnold's avatar
Eckhart Arnold committed
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
    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()
175
        m = re.search('class \w*\(Grammar\)', pycode)
Eckhart Arnold's avatar
Eckhart Arnold committed
176
177
178
179
180
181
182
183
184
185
        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__


186
def get_ebnf_grammar() -> EBNFGrammar:
Eckhart Arnold's avatar
Eckhart Arnold committed
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
    global thread_local_ebnf_grammar_singleton
    try:
        grammar = thread_local_ebnf_grammar_singleton
        return grammar
    except NameError:
        thread_local_ebnf_grammar_singleton = EBNFGrammar()
        return thread_local_ebnf_grammar_singleton


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


203
EBNF_AST_transformation_table = {
204
    # AST Transformations for EBNF-grammar
205
    "+":
206
        remove_expendables,
207
    "syntax":
208
        [],  # otherwise '"*": replace_by_single_child' would be applied
209
    "directive, definition":
210
        remove_tokens('@', '='),
Eckhart Arnold's avatar
Eckhart Arnold committed
211
    "expression":
212
        [replace_by_single_child, flatten, remove_tokens('|')],  # remove_infix_operator],
213
    "term":
214
215
        [replace_by_single_child, flatten],  # supports both idioms:
                                             # "{ factor }+" and "factor { factor }"
216
    "factor, flowmarker, retrieveop":
217
        replace_by_single_child,
218
    "group":
219
        [remove_brackets, replace_by_single_child],
220
221
    "unordered":
        remove_brackets,
222
    "oneormore, repetition, option":
Eckhart Arnold's avatar
Eckhart Arnold committed
223
        [reduce_single_child, remove_brackets,
224
         forbid('repetition', 'option', 'oneormore'), assert_content(r'(?!§)')],
225
    "symbol, literal, regexp":
Eckhart Arnold's avatar
Eckhart Arnold committed
226
        reduce_single_child,
227
    (TOKEN_PTYPE, WHITESPACE_PTYPE):
Eckhart Arnold's avatar
Eckhart Arnold committed
228
        reduce_single_child,
229
    "list_":
230
        [flatten, remove_infix_operator],
231
    "*":
232
        replace_by_single_child
233
234
}

235

Eckhart Arnold's avatar
Eckhart Arnold committed
236
def EBNFTransform() -> TransformationFunc:
237
    return partial(traverse, processing_table=EBNF_AST_transformation_table.copy())
di68kap's avatar
di68kap committed
238

239
def get_ebnf_transformer() -> TransformationFunc:
240
241
242
243
244
245
246
    global thread_local_EBNF_transformer_singleton
    try:
        transformer = thread_local_EBNF_transformer_singleton
    except NameError:
        thread_local_EBNF_transformer_singleton = EBNFTransform()
        transformer = thread_local_EBNF_transformer_singleton
    return transformer
Eckhart Arnold's avatar
Eckhart Arnold committed
247
248
249
250
251
252
253
254


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

255

256
PreprocessorFactoryFunc = Callable[[], PreprocessorFunc]
257
ParserFactoryFunc = Callable[[], Grammar]
258
TransformerFactoryFunc = Callable[[], TransformationFunc]
259
260
CompilerFactoryFunc = Callable[[], Compiler]

261
262
263
PREPROCESSOR_FACTORY = '''
def get_preprocessor() -> PreprocessorFunc:
    return {NAME}Preprocessor
264
265
266
267
'''


GRAMMAR_FACTORY = '''
268
def get_grammar() -> {NAME}Grammar:
269
270
271
272
273
    global thread_local_{NAME}_grammar_singleton
    try:
        grammar = thread_local_{NAME}_grammar_singleton
    except NameError:
        thread_local_{NAME}_grammar_singleton = {NAME}Grammar()
274
275
        grammar = thread_local_{NAME}_grammar_singleton
    return grammar
276
277
278
279
'''


TRANSFORMER_FACTORY = '''
280
281
282
def {NAME}Transform() -> TransformationDict:
    return partial(traverse, processing_table={NAME}_AST_transformation_table.copy())

283
def get_transformer() -> TransformationFunc:
284
285
286
287
288
289
290
    global thread_local_{NAME}_transformer_singleton
    try:
        transformer = thread_local_{NAME}_transformer_singleton
    except NameError:
        thread_local_{NAME}_transformer_singleton = {NAME}Transform()
        transformer = thread_local_{NAME}_transformer_singleton
    return transformer
291
292
293
294
'''


COMPILER_FACTORY = '''
295
def get_compiler(grammar_name="{NAME}", grammar_source="") -> {NAME}Compiler:
296
297
298
299
300
301
302
    global thread_local_{NAME}_compiler_singleton
    try:
        compiler = thread_local_{NAME}_compiler_singleton
        compiler.set_grammar_name(grammar_name, grammar_source)
    except NameError:
        thread_local_{NAME}_compiler_singleton = \\
            {NAME}Compiler(grammar_name, grammar_source)
303
304
        compiler = thread_local_{NAME}_compiler_singleton
    return compiler
305
306
'''

Eckhart Arnold's avatar
Eckhart Arnold committed
307

308
309
class EBNFCompilerError(Exception):
    """Error raised by `EBNFCompiler` class. (Not compilation errors
310
    in the strict sense, see `CompilationError` in module ``dsl.py``)"""
311
312
313
    pass


314
class EBNFCompiler(Compiler):
315
316
    """
    Generates a Parser from an abstract syntax tree of a grammar specified
317
    in EBNF-Notation.
318
319
320
321
322
323
324
325
326
327
328
329
330

    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:
331
        current_symbols:  During compilation, a list containing the root
332
333
334
335
                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.

336
        rules:  Dictionary that maps rule names to a list of Nodes that
337
338
339
340
341
342
                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`

343
                Now `[node.content for node in self.rules['alternative']]`
344
345
                yields `['alternative = a | b', 'a', 'b']`

346
        symbols:  A mapping of symbol names to their first usage (not
347
348
                their definition!) in the EBNF source.

349
        variables:  A set of symbols names that are used with the
350
351
352
353
                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.

354
        recursive:  A set of symbols that are used recursively and
355
356
                therefore require a `Forward`-operator.

357
        definitions:  A dictionary of definitions. Other than `rules`
358
359
                this maps the symbols to their compiled definienda.

360
        deferred_taks:  A list of callables that is filled during
361
362
363
364
365
                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.

366
        root:   The name of the root symbol.
367

368
        directives:  A dictionary of all directives and their default
369
                values.
370
371
372

        re_flags:  A set of regular expression flags to be added to all
                regular expressions found in the current parsing process
373
374
    """
    COMMENT_KEYWORD = "COMMENT__"
375
    WHITESPACE_KEYWORD = "WSP__"
Eckhart Arnold's avatar
Eckhart Arnold committed
376
377
    RAW_WS_KEYWORD = "WHITESPACE__"
    RESERVED_SYMBOLS = {WHITESPACE_KEYWORD, RAW_WS_KEYWORD, COMMENT_KEYWORD}
378
    AST_ERROR = "Badly structured syntax tree. " \
Eckhart Arnold's avatar
Eckhart Arnold committed
379
                "Potentially due to erroneous AST transformation."
380
381
382
383
    PREFIX_TABLE = {'§': 'Required',
                    '&': 'Lookahead', '!': 'NegativeLookahead',
                    '-&': 'Lookbehind', '-!': 'NegativeLookbehind',
                    '::': 'Pop', ':': 'Retrieve'}
384
385
386
    WHITESPACE = {'horizontal': r'[\t ]*',  # default: horizontal
                  'linefeed': r'[ \t]*\n?(?!\s*\n)[ \t]*',
                  'vertical': r'\s*'}
387
    REPEATABLE_DIRECTIVES = {'tokens'}
388

389

390
    def __init__(self, grammar_name="", grammar_source=""):
Eckhart Arnold's avatar
Eckhart Arnold committed
391
        super(EBNFCompiler, self).__init__(grammar_name, grammar_source)
392
393
        self._reset()

394

395
    def _reset(self):
396
        super(EBNFCompiler, self)._reset()
397
        self._result = ''           # type: str
398
        self.re_flags = set()       # type: Set[str]
399
400
401
        self.rules = OrderedDict()  # type: OrderedDict[str, List[Node]]
        self.current_symbols = []   # type: List[Node]
        self.symbols = {}           # type: Dict[str, Node]
402
403
        self.variables = set()      # type: Set[str]
        self.recursive = set()      # type: Set[str]
404
        self.definitions = {}       # type: Dict[str, str]
405
        self.deferred_tasks = []    # type: List[Callable]
406
        self.root_symbol = ""       # type: str
407
        self.directives = {'whitespace': self.WHITESPACE['horizontal'],
408
                           'comment': '',
409
                           'literalws': {'right'},
410
                           'tokens': set(),  # alt. 'preprocessor_tokens'
411
412
413
                           'filter': dict()}  # alt. 'filter'
        # self.directives['ignorecase']: False
        self.defined_directives = set()  # type: Set[str]
414

Eckhart Arnold's avatar
Eckhart Arnold committed
415
    @property
416
    def result(self) -> str:
Eckhart Arnold's avatar
Eckhart Arnold committed
417
418
        return self._result

419
    # methods for generating skeleton code for preprocessor, transformer, and compiler
420

421
    def gen_preprocessor_skeleton(self) -> str:
422
423
424
425
        """
        Returns Python-skeleton-code for a preprocessor-function for
        the previously compiled formal language.
        """
426
        name = self.grammar_name + "Preprocessor"
427
        return "def %s(text):\n    return text, lambda i: i\n" % name \
428
               + PREPROCESSOR_FACTORY.format(NAME=self.grammar_name)
429

430

431
    def gen_transformer_skeleton(self) -> str:
432
433
434
435
        """
        Returns Python-skeleton-code for the AST-transformation for the
        previously compiled formal language.
        """
436
        if not self.rules:
Eckhart Arnold's avatar
Eckhart Arnold committed
437
438
            raise EBNFCompilerError('Compiler must be run before calling '
                                    '"gen_transformer_Skeleton()"!')
439
        tt_name = self.grammar_name + '_AST_transformation_table'
di68kap's avatar
di68kap committed
440
        transtable = [tt_name + ' = {',
Eckhart Arnold's avatar
Eckhart Arnold committed
441
                      '    # AST Transformations for the ' + self.grammar_name + '-grammar']
Eckhart Arnold's avatar
Eckhart Arnold committed
442
        transtable.append('    "+": remove_empty,')
443
        for name in self.rules:
eckhart's avatar
eckhart committed
444
            transformations = '[]'
445
446
            rule = self.definitions[name]
            if rule.startswith('Alternative'):
eckhart's avatar
eckhart committed
447
                transformations = '[replace_or_reduce]'
448
            elif rule.startswith('Synonym'):
Eckhart Arnold's avatar
Eckhart Arnold committed
449
                transformations = '[reduce_single_child]'
eckhart's avatar
eckhart committed
450
            transtable.append('    "' + name + '": %s,' % transformations)
Eckhart Arnold's avatar
Eckhart Arnold committed
451
        transtable.append('    ":Token, :RE": reduce_single_child,')
452
        transtable += ['    "*": replace_by_single_child', '}', '']
453
        transtable += [TRANSFORMER_FACTORY.format(NAME=self.grammar_name)]
454
455
        return '\n'.join(transtable)

456

457
    def gen_compiler_skeleton(self) -> str:
458
459
460
461
        """
        Returns Python-skeleton-code for a Compiler-class for the
        previously compiled formal language.
        """
462
        if not self.rules:
463
464
            raise EBNFCompilerError('Compiler has not been run before calling '
                                    '"gen_Compiler_Skeleton()"!')
465
        compiler = ['class ' + self.grammar_name + 'Compiler(Compiler):',
466
467
468
469
                    '    """Compiler for the abstract-syntax-tree of a ' +
                    self.grammar_name + ' source file.',
                    '    """', '',
                    '    def __init__(self, grammar_name="' +
Eckhart Arnold's avatar
Eckhart Arnold committed
470
                    self.grammar_name + '", grammar_source=""):',
471
                    '        super(' + self.grammar_name +
Eckhart Arnold's avatar
Eckhart Arnold committed
472
                    'Compiler, self).__init__(grammar_name, grammar_source)',
eckhart's avatar
eckhart committed
473
                    r"        assert re.match('\w+\Z', grammar_name)", '']
474
        for name in self.rules:
475
            method_name = Compiler.method_name(name)
476
            if name == self.root_symbol:
477
                compiler += ['    def ' + method_name + '(self, node):',
478
479
                             '        return node', '']
            else:
di68kap's avatar
di68kap committed
480
481
                compiler += ['    # def ' + method_name + '(self, node):',
                             '    #     return node', '']
482
        compiler += [COMPILER_FACTORY.format(NAME=self.grammar_name)]
483
        return '\n'.join(compiler)
484

485
486
487
488
489
490
491
492
493
494
495
496
    def verify_transformation_table(self, transtable):
        assert self._dirty_flag
        table_entries = set(expand_table(transtable).keys()) - {'*', '+', '~'}
        symbols = self.rules.keys()
        messages = []
        for entry in table_entries:
            if entry not in symbols and not entry.startswith(":"):
                messages.append(Error(('Symbol "%s" is not defined in grammar %s but appears in '
                                       'the transformation table!') % (entry, self.grammar_name),
                                      Error.UNDEFINED_SYMBOL_IN_TRANSFORMATION_TABLE))
        return messages

497

498
499
500
501
502
    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
        """
503
504
505
506
507
508
509
510
511
512

        # 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

513
514
515
        if self.variables:
            for i in range(len(definitions)):
                if definitions[i][0] in self.variables:
516
                    definitions[i] = (definitions[i][0], 'Capture(%s)' % definitions[i][1])
517

518
519
        # add special fields for Grammar class

520
        definitions.append(('wspR__', self.WHITESPACE_KEYWORD
Eckhart Arnold's avatar
Eckhart Arnold committed
521
                            if 'right' in self.directives['literalws'] else "''"))
522
        definitions.append(('wspL__', self.WHITESPACE_KEYWORD
Eckhart Arnold's avatar
Eckhart Arnold committed
523
                            if 'left' in self.directives['literalws'] else "''"))
524
        definitions.append((self.WHITESPACE_KEYWORD,
Eckhart Arnold's avatar
Eckhart Arnold committed
525
526
527
                            ("mixin_comment(whitespace=" + self.RAW_WS_KEYWORD +
                             ", comment=" + self.COMMENT_KEYWORD + ")")))
        definitions.append((self.RAW_WS_KEYWORD, "r'{whitespace}'".format(**self.directives)))
528
529
530
531
        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
532

533
        article = 'an ' if self.grammar_name[0:1] in "AaEeIiOoUu" else 'a '  # what about 'hour', 'universe' etc.?
534
        declarations = ['class ' + self.grammar_name +
535
                        'Grammar(Grammar):',
536
537
                        'r"""Parser for ' + article + self.grammar_name +
                        ' source file' +
538
                        (', with this grammar:' if self.grammar_source else '.')]
539
        definitions.append(('parser_initialization__', '"upon instantiation"'))
540
        if self.grammar_source:
541
            definitions.append(('source_hash__',
542
                                '"%s"' % md5(self.grammar_source, __version__)))
543
            declarations.append('')
544
            declarations += [line for line in self.grammar_source.split('\n')]
545
546
547
548
549
            while declarations[-1].strip() == '':
                declarations = declarations[:-1]
        declarations.append('"""')

        # turn definitions into declarations in reverse order
550

551
        self.root_symbol = definitions[0][0] if definitions else ""
552
553
554
555
556
557
558
559
        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]
560
561
562
563
564
565
566

        # 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:
                self.symbols[symbol].add_error("Missing definition for symbol '%s'" % symbol)
567
                # root_node.error_flag = True
568
569
570

        # check for unconnected rules

Eckhart Arnold's avatar
Eckhart Arnold committed
571
572
573
        defined_symbols.difference_update(self.RESERVED_SYMBOLS)

        def remove_connections(symbol):
574
            """Recursively removes all symbols which appear in the
eckhart's avatar
eckhart committed
575
            definiens of a particular symbol."""
Eckhart Arnold's avatar
Eckhart Arnold committed
576
577
578
579
580
581
582
            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
583
            self.rules[leftover][0].add_error(
584
                ('Rule "%s" is not connected to parser root "%s" !') %
eckhart's avatar
eckhart committed
585
                (leftover, self.root_symbol), Error.WARNING)
586

587
        # set root_symbol parser and assemble python grammar definition
588

589
590
        if self.root_symbol and 'root__' not in self.rules:
            declarations.append('root__ = ' + self.root_symbol)
591
        declarations.append('')
Eckhart Arnold's avatar
Eckhart Arnold committed
592
593
594
        self._result = '\n    '.join(declarations) \
                       + GRAMMAR_FACTORY.format(NAME=self.grammar_name)
        return self._result
595

596
597
598

    ## compilation methods

599
    def on_syntax(self, node: Node) -> str:
600
        definitions = []  # type: List[Tuple[str, str]]
601
602

        # drop the wrapping sequence node
603
604
        if len(node.children) == 1 and not node.children[0].parser.name:
            node = node.children[0]
605
606

        # compile definitions and directives and collect definitions
607
        for nd in node.children:
608
            if nd.parser.name == "definition":
609
                definitions.append(self.compile(nd))
610
            else:
611
                assert nd.parser.name == "directive", nd.as_sxpr()
612
                self.compile(nd)
613
            node.error_flag = max(node.error_flag, nd.error_flag)
614
        self.definitions.update(definitions)
615

616
        return self.assemble_parser(definitions, node)
617

618

619
    def on_definition(self, node: Node) -> Tuple[str, str]:
620
        rule = node.children[0].content
621
        if rule in self.rules:
Eckhart Arnold's avatar
Eckhart Arnold committed
622
623
624
625
626
            first = self.rules[rule][0]
            if not first._errors:
                first.add_error('First definition of rule "%s" '
                                'followed by illegal redefinitions.' % rule)
            node.add_error('A rule with name "%s" has already been defined earlier.' % rule)
627
        elif rule in EBNFCompiler.RESERVED_SYMBOLS:
628
629
630
631
            node.add_error('Symbol "%s" is a reserved symbol.' % rule)
        elif not sane_parser_name(rule):
            node.add_error('Illegal symbol "%s". Symbols must not start or '
                           ' end with a doube underscore "__".' % rule)
632
        elif rule in self.directives['tokens']:
633
            node.add_error('Symbol "%s" has already been defined as '
634
                           'a preprocessor token.' % rule)
635
636
        elif keyword.iskeyword(rule):
            node.add_error('Python keyword "%s" may not be used as a symbol. '
637
                           % rule + '(This may change in the future.)')
638
        try:
639
640
            self.current_symbols = [node]
            self.rules[rule] = self.current_symbols
641
            defn = self.compile(node.children[1])
642
            if rule in self.variables:
643
                defn = 'Capture(%s)' % defn
644
                self.variables.remove(rule)
645
646
647
            elif defn.find("(") < 0:
                # assume it's a synonym, like 'page = REGEX_PAGE_NR'
                defn = 'Synonym(%s)' % defn
648
        except TypeError as error:
649
650
651
652
            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())
653
654
            node.add_error(errmsg)
            rule, defn = rule + ':error', '"' + errmsg + '"'
Eckhart Arnold's avatar
Eckhart Arnold committed
655
        return rule, defn
656

657

658
    def _check_rx(self, node: Node, rx: str) -> str:
659
660
        """
        Checks whether the string `rx` represents a valid regular
661
662
663
        expression. Makes sure that multiline regular expressions are
        prepended by the multiline-flag. Returns the regular expression string.
        """
664
        flags = self.re_flags | {'x'} if rx.find('\n') >= 0 else self.re_flags
665
        if flags:  rx = "(?%s)%s" % ("".join(flags), rx)
666
667
668
669
670
671
672
        try:
            re.compile(rx)
        except Exception as re_error:
            node.add_error("malformed regular expression %s: %s" %
                           (repr(rx), str(re_error)))
        return rx

673

674
    def on_directive(self, node: Node) -> str:
675
        key = node.children[0].content.lower()
676
        assert key not in self.directives['tokens']
677

678
679
680
681
682
683
684
685
        if key not in self.REPEATABLE_DIRECTIVES:
            if key in self.defined_directives:
                node.add_error('Directive "%s" has already been defined earlier. ' % key + \
                               'Later definition will be ignored!',
                               code=Error.REDEFINED_DIRECTIVE_WARNING)
                return ""
            self.defined_directives.add(key)

686
        if key in {'comment', 'whitespace'}:
687
688
            if node.children[1].parser.name == "list_":
                if len(node.children[1].result) != 1:
Eckhart Arnold's avatar
Eckhart Arnold committed
689
                    node.add_error('Directive "%s" must have one, but not %i values.' %
690
                                   (key, len(node.children[1].result)))
691
                value = self.compile(node.children[1]).pop()
692
693
                if key == 'whitespace' and value in EBNFCompiler.WHITESPACE:
                    value = EBNFCompiler.WHITESPACE[value]  # replace whitespace-name by regex
694
                else:
695
                    node.add_error('Value "%s" not allowed for directive "%s".' % (value, key))
696
            else:
697
698
699
                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):
700
701
702
703
704
705
                    node.add_error("Whitespace marker '~' not allowed in definition of "
                                   "%s regular expression." % key)
                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])
706
707
708
                if key == 'whitespace' and not re.match(value, ''):
                    node.add_error("Implicit whitespace should always match the empty string, "
                                   "/%s/ does not." % value)
709
            self.directives[key] = value
710

711
        elif key == 'ignorecase':
712
            if node.children[1].content.lower() not in {"off", "false", "no"}:
713
714
                self.re_flags.add('i')

Eckhart Arnold's avatar
Eckhart Arnold committed
715
        # elif key == 'testing':
716
        #     value = node.children[1].content
Eckhart Arnold's avatar
Eckhart Arnold committed
717
        #     self.directives['testing'] = value.lower() not in {"off", "false", "no"}
718

719
        elif key == 'literalws':
720
            value = {item.lower() for item in self.compile(node.children[1])}
eckhart's avatar
eckhart committed
721
            if ((value - {'left', 'right', 'both', 'none'})
Eckhart Arnold's avatar
Eckhart Arnold committed
722
                    or ('none' in value and len(value) > 1)):
723
724
725
                node.add_error('Directive "literalws" allows the values '
                               '`left`, `right`, `both` or `none`, '
                               'but not `%s`' % ", ".join(value))
eckhart's avatar
eckhart committed
726
            wsp = {'left', 'right'} if 'both' in value \
727
                else {} if 'none' in value else value
eckhart's avatar
eckhart committed
728
            self.directives[key] = list(wsp)
729

730
        elif key in {'tokens', 'preprocessor_tokens'}:
731
            tokens = self.compile(node.children[1])
732
            redeclared = self.directives['tokens'] & tokens
733
734
735
736
737
            if redeclared:
                node.add_error('Tokens %s have already been declared earlier. '
                               % str(redeclared) + 'Later declaration will be ignored',
                               code=Error.REDECLARED_TOKEN_WARNING)
            self.directives['tokens'] |= tokens - redeclared
738

739
        elif key.endswith('_filter'):
740
            filter_set = self.compile(node.children[1])
741
742
743
744
            if not isinstance(filter_set, set) or len(filter_set) != 1:
                node.add_error('Directive "%s" accepts exactly on symbol, not %s'
                               % (key, str(filter_set)))
            self.directives['filter'][key[:-7]] = filter_set.pop()
745

746
747
        else:
            node.add_error('Unknown directive %s ! (Known ones are %s .)' %
748
                           (key, ', '.join(list(self.directives.keys()))))
749
750
        return ""

751

752
    def non_terminal(self, node: Node, parser_class: str, custom_args: List[str]=[]) -> str:
753
754
        """
        Compiles any non-terminal, where `parser_class` indicates the Parser class
755
756
        name for the particular non-terminal.
        """
757
        arguments = [self.compile(r) for r in node.children] + custom_args
758
        node.error_flag = max(node.error_flag, max(t.error_flag for t in node.children))
759
760
        return parser_class + '(' + ', '.join(arguments) + ')'

761

762
    def on_expression(self, node) -> str:
di68kap's avatar
di68kap committed
763
        # TODO: Add check for errors like "a" | "ab" (which will always yield a, even for ab)
764
765
        return self.non_terminal(node, 'Alternative')

766

767
    def on_term(self, node) -> str:
di68kap's avatar
di68kap committed
768
769
770
771
772
773
774
775
        # 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 = []
        filtered_children = []
        i = 0
        for nd in node.children:
776
            if nd.parser.ptype == TOKEN_PTYPE and nd.content == "§":
di68kap's avatar
di68kap committed
777
778
779
780
781
782
783
784
785
786
787
788
                mandatory_marker.append(i)
                if i == 0:
                    nd.add_error('First item of a series should not be mandatory.',
                                 Error.WARNING)
                elif len(mandatory_marker) > 1:
                    nd.add_error('One mandatory marker (§) sufficient to declare the '
                                 'rest of the series as mandatory.', Error.WARNING)
            else:
                filtered_children.append(nd)
                i += 1
        saved_result = node.result
        node.result = tuple(filtered_children)
eckhart's avatar
eckhart committed
789
        custom_args = ['mandatory=%i' % mandatory_marker[0]] if mandatory_marker else []
di68kap's avatar
di68kap committed
790
791
792
        compiled = self.non_terminal(node, 'Series', custom_args)
        node.result = saved_result
        return compiled
793

794

795
    def on_factor(self, node: Node) -> str:
796
        assert node.children
797
        assert len(node.children) >= 2, node.as_sxpr()
798
        prefix = node.children[0].content
799
        custom_args = []  # type: List[str]
800
801

        if prefix in {'::', ':'}:
802
803
            assert len(node.children) == 2
            arg = node.children[-1]
804
            if arg.parser.name != 'symbol':
Eckhart Arnold's avatar
Eckhart Arnold committed
805
                node.add_error(('Retrieve Operator "%s" requires a symbol, '
806
807
                                'and not a %s.') % (prefix, str(arg.parser)))
                return str(arg.result)
808
            if str(arg) in self.directives['filter']:
809
                custom_args = ['rfilter=%s' % self.directives['filter'][str(arg)]]
810
            self.variables.add(str(arg))  # cast(str, arg.result)
811

812
        elif len(node.children) > 2:
813
814
            # shift = (Node(node.parser, node.result[1].result),)
            # node.result[1].result = shift + node.result[2:]
815
816
817
818
            node.children[1].result = (Node(node.children[1].parser, node.children[1].result),) \
                                    + node.children[2:]
            node.children[1].parser = node.parser
            node.result = (node.children[0], node.children[1])
819

820
        node.result = node.children[1:]
821
822
        try:
            parser_class = self.PREFIX_TABLE[prefix]
823
824
825
826
            result = self.non_terminal(node, parser_class, custom_args)
            if prefix[:1] == '-':
                def check(node):
                    nd = node
827
828
829
                    if len(nd.children) >= 1:
                        nd = nd.children[0]
                    while nd.parser.name == "symbol":
830
                        symlist = self.rules.get(nd.content, [])
831
832
833
834
835
836
                        if len(symlist) == 2:
                            nd = symlist[1]
                        else:
                            if len(symlist) == 1:
                                nd = symlist[0].children[1]
                            break
837
838
                    if (nd.parser.name != "regexp" or nd.content[:1] != '/'
                            or nd.content[-1:] != '/'):
839
                        node.add_error("Lookbehind-parser can only be used with plain RegExp-"
840
                                       "parsers, not with: " + nd.parser.name + nd.parser.ptype)
841
842
843
844

                if not result.startswith('RegExp('):
                    self.deferred_tasks.append(lambda: check(node))
            return result
845
846
        except KeyError:
            node.add_error('Unknown prefix "%s".' % prefix)
847
        return ""
848

849

850
    def on_option(self, node) -> str:
851
        return self.non_terminal(node, 'Option')
852

853

854
    def on_repetition(self, node) -> str:
855
856
        return self.non_terminal(node, 'ZeroOrMore')

857

858
    def on_oneormore(self, node) -> str:
859
860
        return self.non_terminal(node, 'OneOrMore')

861