ebnf.py 27.6 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
21
from functools import partial

22
23
24
25
26
try:
    import regex as re
except ImportError:
    import re

27
from .toolkit import load_if_file, escape_re, md5, sane_parser_name
28
29
from .parsers import Grammar, mixin_comment, nil_scanner, Forward, RE, NegativeLookahead, \
    Alternative, Sequence, Optional, Required, OneOrMore, ZeroOrMore, Token, CompilerBase
30
from .syntaxtree import Node, traverse, remove_enclosing_delimiters, reduce_single_child, \
31
    replace_by_single_child, TOKEN_PTYPE, remove_expendables, remove_tokens, flatten, \
32
    forbid, assert_content, WHITESPACE_PTYPE, key_tag_name
Eckhart Arnold's avatar
Eckhart Arnold committed
33
from .versionnumber import __version__
34
35


36
37
38
39
40
__all__ = ['get_ebnf_scanner',
           'get_ebnf_grammar',
           'get_ebnf_transformer',
           'get_ebnf_compiler',
           'EBNFGrammar',
41
           'EBNFTransformer',
Eckhart Arnold's avatar
Eckhart Arnold committed
42
           'EBNFCompilerError',
43
44
           'EBNFCompiler',
           'grammar_changed']
45
46


Eckhart Arnold's avatar
Eckhart Arnold committed
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
########################################################################
#
# EBNF scanning
#
########################################################################


def get_ebnf_scanner():
    return nil_scanner


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


65
class EBNFGrammar(Grammar):
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
    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 }+
    factor     =  [flowmarker] [retrieveop] symbol !"="   # negative lookahead to be sure it's not a definition
                | [flowmarker] literal
                | [flowmarker] regexp
                | [flowmarker] group
84
                | [flowmarker] regexchain
85
86
87
88
89
90
91
92
93
                | [flowmarker] oneormore
                | repetition
                | option

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

    group      =  "(" expression §")"
94
    regexchain =  ">" expression §"<"                # compiles "expression" into a singular regular expression
95
96
    oneormore  =  "{" expression "}+"
    repetition =  "{" expression §"}"
97
98
    option     =  "[" expression §"]"

99
100
101
102
103
104
    symbol     =  /(?!\d)\w+/~                       # e.g. expression, factor, parameter_list
    literal    =  /"(?:[^"]|\\")*?"/~                # e.g. "(", '+', 'while'
                | /'(?:[^']|\\')*?'/~                # whitespace following literals will be ignored tacitly.
    regexp     =  /~?\/(?:[^\/]|(?<=\\)\/)*\/~?/~    # e.g. /\w+/, ~/#.*(?:\n|$)/~
                                                     # '~' is a whitespace-marker, if present leading or trailing
                                                     # whitespace of a regular expression will be ignored tacitly.
105
    list_      =  /\w+/~ { "," /\w+/~ }              # comma separated list of symbols, e.g. BEGIN_LIST, END_LIST,
106
107
108
109
                                                     # BEGIN_QUOTE, END_QUOTE ; see CommonMark/markdown.py for an exmaple
    EOF =  !/./
    """
    expression = Forward()
110
    source_hash__ = "a410e1727fb7575e98ff8451dbf8f3bd"
111
    parser_initialization__ = "upon instantiation"
112
113
    COMMENT__ = r'#.*(?:\n|$)'
    WSP__ = mixin_comment(whitespace=r'\s*', comment=r'#.*(?:\n|$)')
114
    wspL__ = ''
115
    wspR__ = WSP__
116
    EOF = NegativeLookahead(RE('.', wR=''))
117
    list_ = Sequence(RE('\\w+'), ZeroOrMore(Sequence(Token(","), RE('\\w+'))))
118
119
120
    regexp = RE('~?/(?:[^/]|(?<=\\\\)/)*/~?')
    literal = Alternative(RE('"(?:[^"]|\\\\")*?"'), RE("'(?:[^']|\\\\')*?'"))
    symbol = RE('(?!\\d)\\w+')
121
    option = Sequence(Token("["), expression, Required(Token("]")))
122
123
    repetition = Sequence(Token("{"), expression, Required(Token("}")))
    oneormore = Sequence(Token("{"), expression, Token("}+"))
124
    regexchain = Sequence(Token("<"), expression, Required(Token(">")))
125
126
127
128
129
    group = Sequence(Token("("), expression, Required(Token(")")))
    retrieveop = Alternative(Token("::"), Token(":"))
    flowmarker = Alternative(Token("!"), Token("&"), Token("§"), Token("-!"), Token("-&"))
    factor = Alternative(Sequence(Optional(flowmarker), Optional(retrieveop), symbol, NegativeLookahead(Token("="))),
                         Sequence(Optional(flowmarker), literal), Sequence(Optional(flowmarker), regexp),
130
131
                         Sequence(Optional(flowmarker), group), Sequence(Optional(flowmarker), regexchain),
                         Sequence(Optional(flowmarker), oneormore), repetition, option)
132
133
134
135
    term = OneOrMore(factor)
    expression.set(Sequence(term, ZeroOrMore(Sequence(Token("|"), term))))
    directive = Sequence(Token("@"), Required(symbol), Required(Token("=")), Alternative(regexp, literal, list_))
    definition = Sequence(symbol, Required(Token("=")), expression)
136
    syntax = Sequence(Optional(RE('', wR='', wL=WSP__)), ZeroOrMore(Alternative(definition, directive)), Required(EOF))
137
138
139
    root__ = syntax


Eckhart Arnold's avatar
Eckhart Arnold committed
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
def grammar_changed(grammar_class, grammar_source):
    """Returns ``True`` if ``grammar_class`` does not reflect the latest
    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()
160
        m = re.search('class \w*\(Grammar\)', pycode)
Eckhart Arnold's avatar
Eckhart Arnold committed
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
        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__


def get_ebnf_grammar():
    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
#
########################################################################


188
189
#TODO: Add Capture and Retrieve Validation: A variable mustn't be captured twice before retrival?!?

190
EBNF_transformation_table = {
191
192
193
194
195
    # AST Transformations for EBNF-grammar
    "syntax":
        remove_expendables,
    "directive, definition":
        partial(remove_tokens, tokens={'@', '='}),
Eckhart Arnold's avatar
Eckhart Arnold committed
196
    "expression":
197
        [replace_by_single_child, flatten,
Eckhart Arnold's avatar
Eckhart Arnold committed
198
         partial(remove_tokens, tokens={'|'})],
199
200
201
202
203
    "term":
        [replace_by_single_child, flatten],  # supports both idioms:  "{ factor }+" and "factor { factor }"
    "factor, flowmarker, retrieveop":
        replace_by_single_child,
    "group":
204
205
206
        [remove_enclosing_delimiters, replace_by_single_child],
    "oneormore, repetition, option, regexchain":
        [reduce_single_child, remove_enclosing_delimiters],
207
    "symbol, literal, regexp":
208
        [remove_expendables, reduce_single_child],
209
    (TOKEN_PTYPE, WHITESPACE_PTYPE):
210
        [remove_expendables, reduce_single_child],
211
    "list_":
212
        [flatten, partial(remove_tokens, tokens={','})],
213
    "*":
214
215
216
        [remove_expendables, replace_by_single_child]
}

217

218
EBNF_validation_table = {
219
220
    # Semantic validation on the AST
    "repetition, option, oneormore":
221
        [partial(forbid, child_tags=['repetition', 'option', 'oneormore']),
222
223
         partial(assert_content, regex=r'(?!§)')],
}
224

225

226
def EBNFTransformer(syntax_tree):
227
    for processing_table, key_func in [(EBNF_transformation_table, key_tag_name),
228
                                       (EBNF_validation_table, key_tag_name)]:
229
        traverse(syntax_tree, processing_table, key_func)
di68kap's avatar
di68kap committed
230
231


Eckhart Arnold's avatar
Eckhart Arnold committed
232
def get_ebnf_transformer():
233
    return EBNFTransformer
Eckhart Arnold's avatar
Eckhart Arnold committed
234
235
236
237
238
239
240
241


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

242
SCANNER_FACTORY = '''
243
def get_scanner():
244
245
246
247
248
    return {NAME}Scanner
'''


GRAMMAR_FACTORY = '''
249
def get_grammar():
250
251
252
253
254
255
256
257
258
259
260
    global thread_local_{NAME}_grammar_singleton
    try:
        grammar = thread_local_{NAME}_grammar_singleton
        return grammar
    except NameError:
        thread_local_{NAME}_grammar_singleton = {NAME}Grammar()
        return thread_local_{NAME}_grammar_singleton
'''


TRANSFORMER_FACTORY = '''
261
def get_transformer():
262
263
264
265
266
    return {NAME}Transform
'''


COMPILER_FACTORY = '''
267
def get_compiler(grammar_name="{NAME}", grammar_source=""):
268
269
270
271
272
273
274
275
276
277
278
    global thread_local_{NAME}_compiler_singleton
    try:
        compiler = thread_local_{NAME}_compiler_singleton
        compiler.set_grammar_name(grammar_name, grammar_source)
        return compiler
    except NameError:
        thread_local_{NAME}_compiler_singleton = \\
            {NAME}Compiler(grammar_name, grammar_source)
        return thread_local_{NAME}_compiler_singleton 
'''

Eckhart Arnold's avatar
Eckhart Arnold committed
279

280
281
282
283
284
285
286
287
288
289
290
class EBNFCompilerError(Exception):
    """Error raised by `EBNFCompiler` class. (Not compilation errors
    in the strict sense, see `CompilationError` below)"""
    pass


class EBNFCompiler(CompilerBase):
    """Generates a Parser from an abstract syntax tree of a grammar specified
    in EBNF-Notation.
    """
    COMMENT_KEYWORD = "COMMENT__"
291
292
    WHITESPACE_KEYWORD = "WSP__"
    RESERVED_SYMBOLS = {WHITESPACE_KEYWORD, COMMENT_KEYWORD}
293
294
    AST_ERROR = "Badly structured syntax tree. " \
                "Potentially due to erroneuos AST transformation."
295
296
297
298
    PREFIX_TABLE = {'§': 'Required',
                    '&': 'Lookahead', '!': 'NegativeLookahead',
                    '-&': 'Lookbehind', '-!': 'NegativeLookbehind',
                    '::': 'Pop', ':': 'Retrieve'}
299
300
301
    WHITESPACE = {'horizontal': r'[\t ]*',  # default: horizontal
                  'linefeed': r'[ \t]*\n?(?!\s*\n)[ \t]*',
                  'vertical': r'\s*'}
302

303
    def __init__(self, grammar_name="", grammar_source=""):
Eckhart Arnold's avatar
Eckhart Arnold committed
304
        super(EBNFCompiler, self).__init__(grammar_name, grammar_source)
305
306
307
        self._reset()

    def _reset(self):
Eckhart Arnold's avatar
Eckhart Arnold committed
308
        self._result = None
309
310
        self.rules = set()
        self.variables = set()
311
        self.symbol_nodes = []
312
313
314
        self.definition_names = []
        self.recursive = set()
        self.root = ""
315
        self.directives = {'whitespace': self.WHITESPACE['horizontal'],
316
                           'comment': '',
317
                           'literalws': ['right'],
318
                           'tokens': set(),     # alt. 'scanner_tokens'
319
                           'filter': dict()}     # alt. 'retrieve_filter'
320

Eckhart Arnold's avatar
Eckhart Arnold committed
321
322
323
324
    @property
    def result(self):
        return self._result

325
326
    def gen_scanner_skeleton(self):
        name = self.grammar_name + "Scanner"
327
328
        return "def %s(text):\n    return text\n" % name \
               + SCANNER_FACTORY.format(NAME=self.grammar_name)
329

330
    def gen_transformer_skeleton(self):
331
        if not self.definition_names:
Eckhart Arnold's avatar
Eckhart Arnold committed
332
333
            raise EBNFCompilerError('Compiler must be run before calling '
                                    '"gen_transformer_Skeleton()"!')
334
335
        tt_name = self.grammar_name + '_AST_transformation_table'
        tf_name = self.grammar_name + 'Transform'
di68kap's avatar
di68kap committed
336
        transtable = [tt_name + ' = {',
337
338
339
                      '    # AST Transformations for the ' +
                      self.grammar_name + '-grammar']
        for name in self.definition_names:
di68kap's avatar
di68kap committed
340
            transtable.append('    "' + name + '": no_operation,')
341
        transtable += ['    "*": no_operation', '}', '', tf_name +
342
                       ' = partial(traverse, processing_table=%s)' % tt_name, '']
343
        transtable += [TRANSFORMER_FACTORY.format(NAME=self.grammar_name)]
344
345
346
347
348
349
350
351
352
353
354
        return '\n'.join(transtable)

    def gen_compiler_skeleton(self):
        if not self.definition_names:
            raise EBNFCompilerError('Compiler has not been run before calling '
                                    '"gen_Compiler_Skeleton()"!')
        compiler = ['class ' + self.grammar_name + 'Compiler(CompilerBase):',
                    '    """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
355
                    self.grammar_name + '", grammar_source=""):',
356
                    '        super(' + self.grammar_name +
Eckhart Arnold's avatar
Eckhart Arnold committed
357
                    'Compiler, self).__init__(grammar_name, grammar_source)',
358
359
                    "        assert re.match('\w+\Z', grammar_name)", '']
        for name in self.definition_names:
360
            method_name = CompilerBase.derive_method_name(name)
361
            if name == self.root:
362
                compiler += ['    def ' + method_name + '(self, node):',
363
364
                             '        return node', '']
            else:
365
                compiler += ['    def ' + method_name + '(self, node):',
366
                             '        pass', '']
367
        compiler += [COMPILER_FACTORY.format(NAME=self.grammar_name)]
368
        return '\n'.join(compiler)
369

370
    def assemble_parser(self, definitions, root_node):
371
        # fix capture of variables that have been defined before usage [sic!]
372

373
374
375
        if self.variables:
            for i in range(len(definitions)):
                if definitions[i][0] in self.variables:
376
                    definitions[i] = (definitions[i][0], 'Capture(%s)' % definitions[1])
377
378

        self.definition_names = [defn[0] for defn in definitions]
379
        definitions.append(('wspR__', self.WHITESPACE_KEYWORD
Eckhart Arnold's avatar
Eckhart Arnold committed
380
                            if 'right' in self.directives['literalws'] else "''"))
381
        definitions.append(('wspL__', self.WHITESPACE_KEYWORD
Eckhart Arnold's avatar
Eckhart Arnold committed
382
                            if 'left' in self.directives['literalws'] else "''"))
383
        definitions.append((self.WHITESPACE_KEYWORD,
384
385
386
387
388
389
390
                            ("mixin_comment(whitespace="
                             "r'{whitespace}', comment=r'{comment}')").
                            format(**self.directives)))
        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
391

392
        article = 'an ' if self.grammar_name[0:1] in "AaEeIiOoUu" else 'a '  # what about 'hour', 'universe' etc.?
393
        declarations = ['class ' + self.grammar_name +
394
                        'Grammar(Grammar):',
395
396
                        'r"""Parser for ' + article + self.grammar_name +
                        ' source file' +
397
                        (', with this grammar:' if self.grammar_source else '.')]
398
        definitions.append(('parser_initialization__', '"upon instatiation"'))
399
        if self.grammar_source:
400
            definitions.append(('source_hash__',
401
                                '"%s"' % md5(self.grammar_source, __version__)))
402
            declarations.append('')
403
            declarations += [line for line in self.grammar_source.split('\n')]
404
405
406
407
            while declarations[-1].strip() == '':
                declarations = declarations[:-1]
        declarations.append('"""')

408
        # add default functions for retrieve_filter filters of pop or retrieve operators
409

410
411
412
        # for symbol, fun in self.directives['filter']:
        #     declarations.append(symbol + '_filter = lambda value: value.replace("(", ")")'
        #                         '.replace("[", "]").replace("{", "}").replace(">", "<")')
413

414
        # turn definitions into declarations in reverse order
415

416
417
418
419
420
421
422
423
424
        self.root = definitions[0][0] if definitions else ""
        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]
425
        known_symbols = self.rules | self.RESERVED_SYMBOLS
426
        for nd in self.symbol_nodes:
427
            if nd.result not in known_symbols:
428
                nd.add_error("Missing production for symbol '%s'" % nd.result)
429
                root_node.error_flag = True
430
        if self.root and 'root__' not in self.rules:
431
432
            declarations.append('root__ = ' + self.root)
        declarations.append('')
Eckhart Arnold's avatar
Eckhart Arnold committed
433
434
435
        self._result = '\n    '.join(declarations) \
                       + GRAMMAR_FACTORY.format(NAME=self.grammar_name)
        return self._result
436

437
    def on_syntax(self, node):
438
439
440
441
        self._reset()
        definitions = []

        # drop the wrapping sequence node
442
        if len(node.children) == 1 and not node.result[0].parser.name:
443
444
445
446
447
            node = node.result[0]

        # compile definitions and directives and collect definitions
        for nd in node.result:
            if nd.parser.name == "definition":
448
                definitions.append(self._compile(nd))
449
450
            else:
                assert nd.parser.name == "directive", nd.as_sexpr()
451
                self._compile(nd)
452
                node.error_flag |= nd.error_flag
453

454
        return self.assemble_parser(definitions, node)
455

456
    def on_definition(self, node):
457
        rule = node.result[0].result
458
459
460
        if rule in self.rules:
            node.add_error('A rule with name "%s" has already been defined.' % rule)
        elif rule in EBNFCompiler.RESERVED_SYMBOLS:
461
462
463
464
            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)
465
        elif rule in self.directives['tokens']:
466
467
468
469
470
471
472
            node.add_error('Symbol "%s" has already been defined as '
                           'a scanner token.' % rule)
        elif keyword.iskeyword(rule):
            node.add_error('Python keyword "%s" may not be used as a symbol. '
                           % rule + '(This may change in the furute.)')
        try:
            self.rules.add(rule)
473
            defn = self._compile(node.result[1])
474
            if rule in self.variables:
475
                defn = 'Capture(%s)' % defn
476
477
478
479
480
                self.variables.remove(rule)
        except TypeError as error:
            errmsg = EBNFCompiler.AST_ERROR + " (" + str(error) + ")\n" + node.as_sexpr()
            node.add_error(errmsg)
            rule, defn = rule + ':error', '"' + errmsg + '"'
Eckhart Arnold's avatar
Eckhart Arnold committed
481
        return rule, defn
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496

    @staticmethod
    def _check_rx(node, rx):
        """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.
        """
        rx = rx if rx.find('\n') < 0 or rx[0:4] == '(?x)' else '(?x)' + rx
        try:
            re.compile(rx)
        except Exception as re_error:
            node.add_error("malformed regular expression %s: %s" %
                           (repr(rx), str(re_error)))
        return rx

497
    def on_directive(self, node):
498
        key = node.result[0].result.lower()
499
        assert key not in self.directives['tokens']
500
        if key in {'comment', 'whitespace'}:
501
502
            if node.result[1].parser.name == "list_":
                if len(node.result[1].result) != 1:
Eckhart Arnold's avatar
Eckhart Arnold committed
503
                    node.add_error('Directive "%s" must have one, but not %i values.' %
504
                                   (key, len(node.result[1])))
505
                value = self._compile(node.result[1]).pop()
506
507
                if key == 'whitespace' and value in EBNFCompiler.WHITESPACE:
                    value = EBNFCompiler.WHITESPACE[value]  # replace whitespace-name by regex
508
                else:
509
                    node.add_error('Value "%s" not allowed for directive "%s".' % (value, key))
510
            else:
511
512
513
514
515
516
517
518
                value = node.result[1].result.strip("~")
                if value != node.result[1].result:
                    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])
519
520
521
                if key == 'whitespace' and not re.match(value, ''):
                    node.add_error("Implicit whitespace should always match the empty string, "
                                   "/%s/ does not." % value)
522
            self.directives[key] = value
523

524
        elif key == 'literalws':
525
            value = {item.lower() for item in self._compile(node.result[1])}
526
            if (len(value - {'left', 'right', 'both', 'none'}) > 0
Eckhart Arnold's avatar
Eckhart Arnold committed
527
                    or ('none' in value and len(value) > 1)):
528
529
530
531
532
533
534
                node.add_error('Directive "literalws" allows the values '
                               '`left`, `right`, `both` or `none`, '
                               'but not `%s`' % ", ".join(value))
            ws = {'left', 'right'} if 'both' in value \
                else {} if 'none' in value else value
            self.directives[key] = list(ws)

535
        elif key in {'tokens', 'scanner_tokens'}:
536
            self.directives['tokens'] |= self._compile(node.result[1])
537

538
539
540
541
542
543
        elif key.endswith('_filter'):
            filter_set = self._compile(node.result[1])
            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()
544

545
546
547
        else:
            node.add_error('Unknown directive %s ! (Known ones are %s .)' %
                           (key,
548
                            ', '.join(list(self.directives.keys()))))
549
550
        return ""

551
    def non_terminal(self, node, parser_class, custom_args=[]):
552
553
554
        """Compiles any non-terminal, where `parser_class` indicates the Parser class
        name for the particular non-terminal.
        """
555
        arguments = [self._compile(r) for r in node.result] + custom_args
556
557
        return parser_class + '(' + ', '.join(arguments) + ')'

558
    def on_expression(self, node):
559
560
        return self.non_terminal(node, 'Alternative')

561
    def on_term(self, node):
562
563
        return self.non_terminal(node, 'Sequence')

564
    def on_factor(self, node):
565
566
567
        assert node.children
        assert len(node.result) >= 2, node.as_sexpr()
        prefix = node.result[0].result
568
        custom_args = []
569
570
571

        if prefix in {'::', ':'}:
            assert len(node.result) == 2
572
            arg = node.result[-1]
573
            if arg.parser.name != 'symbol':
Eckhart Arnold's avatar
Eckhart Arnold committed
574
                node.add_error(('Retrieve Operator "%s" requires a symbol, '
575
576
                                'and not a %s.') % (prefix, str(arg.parser)))
                return str(arg.result)
577
578
            if str(arg) in self.directives['filter']:
                custom_args = ['retrieve_filter=%s' % self.directives['filter'][str(arg)]]
579
580
            self.variables.add(arg.result)

581
        elif len(node.result) > 2:
582
583
            # shift = (Node(node.parser, node.result[1].result),)
            # node.result[1].result = shift + node.result[2:]
584
            node.result[1].result = (Node(node.result[1].parser, node.result[1].result),) \
585
586
587
588
589
                                    + node.result[2:]
            node.result[1].parser = node.parser
            node.result = (node.result[0], node.result[1])

        node.result = node.result[1:]
590
591
592
593
594
        try:
            parser_class = self.PREFIX_TABLE[prefix]
            return self.non_terminal(node, parser_class, custom_args)
        except KeyError:
            node.add_error('Unknown prefix "%s".' % prefix)
595

596
    def on_option(self, node):
597
598
        return self.non_terminal(node, 'Optional')

599
    def on_repetition(self, node):
600
601
        return self.non_terminal(node, 'ZeroOrMore')

602
    def on_oneormore(self, node):
603
604
        return self.non_terminal(node, 'OneOrMore')

605
    def on_regexchain(self, node):
606
607
        raise EBNFCompilerError("Not yet implemented!")

608
    def on_group(self, node):
609
610
611
        raise EBNFCompilerError("Group nodes should have been eliminated by "
                                "AST transformation!")

612
    def on_symbol(self, node):
613
        if node.result in self.directives['tokens']:
614
615
            return 'ScannerToken("' + node.result + '")'
        else:
616
            self.symbol_nodes.append(node)
617
618
619
620
            if node.result in self.rules:
                self.recursive.add(node.result)
            return node.result

621
    def on_literal(self, node):
622
        return 'Token(' + node.result.replace('\\', r'\\') + ')'  # return 'Token(' + ', '.join([node.result]) + ')' ?
623

624
    def on_regexp(self, node):
625
626
627
628
        rx = node.result
        name = []
        if rx[:2] == '~/':
            if not 'left' in self.directives['literalws']:
629
                name = ['wL=' + self.WHITESPACE_KEYWORD] + name
630
631
632
633
            rx = rx[1:]
        elif 'left' in self.directives['literalws']:
            name = ["wL=''"] + name
        if rx[-2:] == '/~':
Eckhart Arnold's avatar
Eckhart Arnold committed
634
            if 'right' not in self.directives['literalws']:
635
                name = ['wR=' + self.WHITESPACE_KEYWORD] + name
636
637
638
639
640
641
642
643
644
645
646
647
            rx = rx[:-1]
        elif 'right' in self.directives['literalws']:
            name = ["wR=''"] + name
        try:
            arg = repr(self._check_rx(node, rx[1:-1].replace(r'\/', '/')))
        except AttributeError as error:
            errmsg = EBNFCompiler.AST_ERROR + " (" + str(error) + ")\n" + \
                     node.as_sexpr()
            node.add_error(errmsg)
            return '"' + errmsg + '"'
        return 'RE(' + ', '.join([arg] + name) + ')'

648
    def on_list_(self, node):
649
650
        assert node.children
        return set(item.result.strip() for item in node.result)
651
652


Eckhart Arnold's avatar
Eckhart Arnold committed
653
654
655
656
657
658
659
660
661
def get_ebnf_compiler(grammar_name="", grammar_source=""):
    global thread_local_ebnf_compiler_singleton
    try:
        compiler = thread_local_ebnf_compiler_singleton
        compiler.set_grammar_name(grammar_name, grammar_source)
        return compiler
    except NameError:
        thread_local_ebnf_compiler_singleton = EBNFCompiler(grammar_name, grammar_source)
        return thread_local_ebnf_compiler_singleton