ebnf.py 27.4 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
19

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.
"""

from functools import partial
20
21
import keyword
import os
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
Eckhart Arnold's avatar
Eckhart Arnold committed
28
from .parsers import GrammarBase, mixin_comment, nil_scanner, Forward, RE, NegativeLookahead, \
29
30
    Alternative, Sequence, Optional, Required, OneOrMore, ZeroOrMore, Token, CompilerBase, \
    Capture, Retrieve
31
from .syntaxtree import Node, traverse, remove_enclosing_delimiters, reduce_single_child, \
di68kap's avatar
di68kap committed
32
    replace_by_single_child, TOKEN_KEYWORD, remove_expendables, remove_tokens, flatten, \
33
    forbid, assert_content, WHITESPACE_KEYWORD
Eckhart Arnold's avatar
Eckhart Arnold committed
34
from .versionnumber import __version__
35
36


Eckhart Arnold's avatar
Eckhart Arnold committed
37
__all__ = ['EBNFGrammar',
38
           'EBNFTransformer',
Eckhart Arnold's avatar
Eckhart Arnold committed
39
           'EBNFCompilerError',
40
41
           'EBNFCompiler',
           'grammar_changed']
42
43


Eckhart Arnold's avatar
Eckhart Arnold committed
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63


########################################################################
#
# EBNF scanning
#
########################################################################


def get_ebnf_scanner():
    return nil_scanner


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


64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
class EBNFGrammar(GrammarBase):
    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
83
                | [flowmarker] regexchain
84
85
86
87
88
89
90
91
92
                | [flowmarker] oneormore
                | repetition
                | option

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

    group      =  "(" expression §")"
93
    regexchain =  ">" expression §"<"                # compiles "expression" into a singular regular expression
94
95
    oneormore  =  "{" expression "}+"
    repetition =  "{" expression §"}"
96
97
    option     =  "[" expression §"]"

98
99
100
101
102
103
    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|$)/~
                                                     # '~' is a whitespace-marker, if present leading or trailing
                                                     # whitespace of a regular expression will be ignored tacitly.
104
    list_      =  /\w+/~ { "," /\w+/~ }              # comma separated list of symbols, e.g. BEGIN_LIST, END_LIST,
105
106
107
108
                                                     # BEGIN_QUOTE, END_QUOTE ; see CommonMark/markdown.py for an exmaple
    EOF =  !/./
    """
    expression = Forward()
109
    source_hash__ = "a410e1727fb7575e98ff8451dbf8f3bd"
110
    parser_initialization__ = "upon instatiation"
111
112
    COMMENT__ = r'#.*(?:\n|$)'
    WSP__ = mixin_comment(whitespace=r'\s*', comment=r'#.*(?:\n|$)')
113
    wspL__ = ''
114
    wspR__ = WSP__
115
    EOF = NegativeLookahead(RE('.', wR=''))
116
    list_ = Sequence(RE('\\w+'), ZeroOrMore(Sequence(Token(","), RE('\\w+'))))
117
118
119
    regexp = RE('~?/(?:[^/]|(?<=\\\\)/)*/~?')
    literal = Alternative(RE('"(?:[^"]|\\\\")*?"'), RE("'(?:[^']|\\\\')*?'"))
    symbol = RE('(?!\\d)\\w+')
120
    option = Sequence(Token("["), expression, Required(Token("]")))
121
122
    repetition = Sequence(Token("{"), expression, Required(Token("}")))
    oneormore = Sequence(Token("{"), expression, Token("}+"))
123
    regexchain = Sequence(Token("<"), expression, Required(Token(">")))
124
125
126
127
128
    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),
129
130
                         Sequence(Optional(flowmarker), group), Sequence(Optional(flowmarker), regexchain),
                         Sequence(Optional(flowmarker), oneormore), repetition, option)
131
132
133
134
    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)
135
    syntax = Sequence(Optional(RE('', wR='', wL=WSP__)), ZeroOrMore(Alternative(definition, directive)), Required(EOF))
136
137
138
    root__ = syntax


Eckhart Arnold's avatar
Eckhart Arnold committed
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
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
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()
        m = re.search('class \w*\(GrammarBase\)', pycode)
        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
#
########################################################################


187
188
#TODO: Add Capture and Retrieve Validation: A variable mustn't be captured twice before retrival?!?

189
EBNF_transformation_table = {
190
191
192
193
194
    # AST Transformations for EBNF-grammar
    "syntax":
        remove_expendables,
    "directive, definition":
        partial(remove_tokens, tokens={'@', '='}),
Eckhart Arnold's avatar
Eckhart Arnold committed
195
    "expression":
196
        [replace_by_single_child, flatten,
Eckhart Arnold's avatar
Eckhart Arnold committed
197
         partial(remove_tokens, tokens={'|'})],
198
199
200
201
202
    "term":
        [replace_by_single_child, flatten],  # supports both idioms:  "{ factor }+" and "factor { factor }"
    "factor, flowmarker, retrieveop":
        replace_by_single_child,
    "group":
203
204
205
        [remove_enclosing_delimiters, replace_by_single_child],
    "oneormore, repetition, option, regexchain":
        [reduce_single_child, remove_enclosing_delimiters],
206
    "symbol, literal, regexp":
207
208
209
        [remove_expendables, reduce_single_child],
    (TOKEN_KEYWORD, WHITESPACE_KEYWORD):
        [remove_expendables, reduce_single_child],
210
    "list_":
211
        [flatten, partial(remove_tokens, tokens={','})],
212
213
214
215
    "":
        [remove_expendables, replace_by_single_child]
}

216

217
EBNF_validation_table = {
218
219
220
221
222
    # Semantic validation on the AST
    "repetition, option, oneormore":
        [partial(forbid, child_tags=['repetition', 'option', 'oneormore']),
         partial(assert_content, regex=r'(?!§)')],
}
223

224

225
def EBNFTransformer(syntax_tree):
226
227
    for processing_table in [EBNF_transformation_table, EBNF_validation_table]:
        traverse(syntax_tree, processing_table)
di68kap's avatar
di68kap committed
228
229


Eckhart Arnold's avatar
Eckhart Arnold committed
230
def get_ebnf_transformer():
231
    return EBNFTransformer
Eckhart Arnold's avatar
Eckhart Arnold committed
232
233
234
235
236
237
238
239


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

240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
SCANNER_FACTORY = '''
def get_{NAME}_scanner():
    return {NAME}Scanner
'''


GRAMMAR_FACTORY = '''
def get_{NAME}_grammar():
    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 = '''
def get_{NAME}_transformer():
    return {NAME}Transform
'''


COMPILER_FACTORY = '''
def get_{NAME}_compiler(grammar_name="{NAME}", 
                        grammar_source=""):
    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
278

279
280
281
282
283
284
285
286
287
288
289
290
291
292
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__"
    RESERVED_SYMBOLS = {TOKEN_KEYWORD, WHITESPACE_KEYWORD, COMMENT_KEYWORD}
    AST_ERROR = "Badly structured syntax tree. " \
                "Potentially due to erroneuos AST transformation."
293
294
295
296
    PREFIX_TABLE = {'§': 'Required',
                    '&': 'Lookahead', '!': 'NegativeLookahead',
                    '-&': 'Lookbehind', '-!': 'NegativeLookbehind',
                    '::': 'Pop', ':': 'Retrieve'}
297
298
299
    WHITESPACE = {'horizontal': r'[\t ]*',  # default: horizontal
                  'linefeed': r'[ \t]*\n?(?!\s*\n)[ \t]*',
                  'vertical': r'\s*'}
300

301
    def __init__(self, grammar_name="", grammar_source=""):
Eckhart Arnold's avatar
Eckhart Arnold committed
302
        super(EBNFCompiler, self).__init__(grammar_name, grammar_source)
303
304
305
        self._reset()

    def _reset(self):
Eckhart Arnold's avatar
Eckhart Arnold committed
306
        self._result = None
307
308
        self.rules = set()
        self.variables = set()
309
        self.symbol_nodes = []
310
311
312
        self.definition_names = []
        self.recursive = set()
        self.root = ""
313
        self.directives = {'whitespace': self.WHITESPACE['horizontal'],
314
                           'comment': '',
315
                           'literalws': ['right'],
316
                           'tokens': set(),     # alt. 'scanner_tokens'
317
                           'counterpart': set()}     # alt. 'retrieve_counterpart'
318

Eckhart Arnold's avatar
Eckhart Arnold committed
319
320
321
322
    @property
    def result(self):
        return self._result

323
324
    def gen_scanner_skeleton(self):
        name = self.grammar_name + "Scanner"
325
326
        return "def %s(text):\n    return text\n" % name \
               + SCANNER_FACTORY.format(NAME=self.grammar_name)
327

328
    def gen_transformer_skeleton(self):
329
        if not self.definition_names:
Eckhart Arnold's avatar
Eckhart Arnold committed
330
331
            raise EBNFCompilerError('Compiler must be run before calling '
                                    '"gen_transformer_Skeleton()"!')
332
333
        tt_name = self.grammar_name + '_AST_transformation_table'
        tf_name = self.grammar_name + 'Transform'
di68kap's avatar
di68kap committed
334
        transtable = [tt_name + ' = {',
335
336
337
                      '    # AST Transformations for the ' +
                      self.grammar_name + '-grammar']
        for name in self.definition_names:
di68kap's avatar
di68kap committed
338
            transtable.append('    "' + name + '": no_operation,')
339
340
        transtable += ['    "": no_operation', '}', '',  tf_name +
                       ' = partial(traverse, processing_table=%s)' % tt_name, '']
341
        transtable += [TRANSFORMER_FACTORY.format(NAME=self.grammar_name)]
342
343
344
345
346
347
348
349
350
351
352
        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
353
                    self.grammar_name + '", grammar_source=""):',
354
                    '        super(' + self.grammar_name +
Eckhart Arnold's avatar
Eckhart Arnold committed
355
                    'Compiler, self).__init__(grammar_name, grammar_source)',
356
357
                    "        assert re.match('\w+\Z', grammar_name)", '']
        for name in self.definition_names:
358
            method_name = CompilerBase.derive_method_name(name)
359
            if name == self.root:
360
                compiler += ['    def ' + method_name + '(self, node):',
361
362
                             '        return node', '']
            else:
363
                compiler += ['    def ' + method_name + '(self, node):',
364
                             '        pass', '']
365
        compiler += [COMPILER_FACTORY.format(NAME=self.grammar_name)]
366
        return '\n'.join(compiler)
367

Eckhart Arnold's avatar
Eckhart Arnold committed
368
    def assemble_parser(self, definitions):
369
        # fix capture of variables that have been defined before usage [sic!]
370

371
372
373
        if self.variables:
            for i in range(len(definitions)):
                if definitions[i][0] in self.variables:
374
                    definitions[i] = (definitions[i][0], 'Capture(%s)' % definitions[1])
375
376

        self.definition_names = [defn[0] for defn in definitions]
Eckhart Arnold's avatar
Eckhart Arnold committed
377
378
379
380
        definitions.append(('wspR__', WHITESPACE_KEYWORD
                            if 'right' in self.directives['literalws'] else "''"))
        definitions.append(('wspL__', WHITESPACE_KEYWORD
                            if 'left' in self.directives['literalws'] else "''"))
381
382
383
384
385
386
387
388
        definitions.append((WHITESPACE_KEYWORD,
                            ("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
389

390
        article = 'an ' if self.grammar_name[0:1] in "AaEeIiOoUu" else 'a '  # what about 'hour', 'universe' etc.?
391
392
393
394
        declarations = ['class ' + self.grammar_name +
                        'Grammar(GrammarBase):',
                        'r"""Parser for ' + article + self.grammar_name +
                        ' source file' +
395
                        (', with this grammar:' if self.grammar_source else '.')]
396
        definitions.append(('parser_initialization__', '"upon instatiation"'))
397
        if self.grammar_source:
398
            definitions.append(('source_hash__',
399
                                '"%s"' % md5(self.grammar_source, __version__)))
400
            declarations.append('')
401
            declarations += [line for line in self.grammar_source.split('\n')]
402
403
404
405
            while declarations[-1].strip() == '':
                declarations = declarations[:-1]
        declarations.append('"""')

406
        # add default functions for counterpart filters of pop or retrieve operators
407

408
409
410
411
412
413
        for symbol in self.directives['counterpart']:
            # declarations.append('def %s_counterpart(value): \n' % symbol +
            #                     '        return value.replace("(", ")").replace("[", "]")'
            #                     '.replace("{", "}").replace(">", "<")\n')
            declarations.append(symbol + '_counterpart = lambda value: value.replace("(", ")")'
                                '.replace("[", "]").replace("{", "}").replace(">", "<")')
414

415
        # turn definitions into declarations in reverse order
416

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

436
    def on_syntax(self, node):
437
438
439
440
441
442
443
444
445
446
447
        self._reset()
        definitions = []

        # drop the wrapping sequence node
        if isinstance(node.parser, Sequence) and \
                isinstance(node.result[0].parser, ZeroOrMore):
            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

Eckhart Arnold's avatar
Eckhart Arnold committed
453
        return self.assemble_parser(definitions)
454

455
    def on_definition(self, node):
456
        rule = node.result[0].result
457
458
459
        if rule in self.rules:
            node.add_error('A rule with name "%s" has already been defined.' % rule)
        elif rule in EBNFCompiler.RESERVED_SYMBOLS:
460
461
462
463
            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)
464
        elif rule in self.directives['tokens']:
465
466
467
468
469
470
471
            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)
472
            defn = self._compile(node.result[1])
473
            if rule in self.variables:
474
                defn = 'Capture(%s)' % defn
475
476
477
478
479
                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
480
        return rule, defn
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495

    @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

496
    def on_directive(self, node):
497
        key = node.result[0].result.lower()
498
        assert key not in self.directives['tokens']
499
        if key in {'comment', 'whitespace'}:
500
501
            if node.result[1].parser.name == "list_":
                if len(node.result[1].result) != 1:
Eckhart Arnold's avatar
Eckhart Arnold committed
502
                    node.add_error('Directive "%s" must have one, but not %i values.' %
503
                                   (key, len(node.result[1])))
504
                value = self._compile(node.result[1]).pop()
505
506
                if key == 'whitespace' and value in EBNFCompiler.WHITESPACE:
                    value = EBNFCompiler.WHITESPACE[value]  # replace whitespace-name by regex
507
                else:
508
                    node.add_error('Value "%s" not allowed for directive "%s".' % (value, key))
509
            else:
510
511
512
513
514
515
516
517
                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])
518
519
520
                if key == 'whitespace' and not re.match(value, ''):
                    node.add_error("Implicit whitespace should always match the empty string, "
                                   "/%s/ does not." % value)
521
            self.directives[key] = value
522

523
        elif key == 'literalws':
524
            value = {item.lower() for item in self._compile(node.result[1])}
525
            if (len(value - {'left', 'right', 'both', 'none'}) > 0
Eckhart Arnold's avatar
Eckhart Arnold committed
526
                    or ('none' in value and len(value) > 1)):
527
528
529
530
531
532
533
                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)

534
        elif key in {'tokens', 'scanner_tokens'}:
535
            self.directives['tokens'] |= self._compile(node.result[1])
536

537
        elif key in {'counterpart', 'retrieve_counterpart'}:
538
            self.directives['counterpart'] |= self._compile(node.result[1])
539

540
541
542
        else:
            node.add_error('Unknown directive %s ! (Known ones are %s .)' %
                           (key,
543
                            ', '.join(list(self.directives.keys()))))
544
545
        return ""

546
    def non_terminal(self, node, parser_class, custom_args=[]):
547
548
549
        """Compiles any non-terminal, where `parser_class` indicates the Parser class
        name for the particular non-terminal.
        """
550
        arguments = [self._compile(r) for r in node.result] + custom_args
551
552
        return parser_class + '(' + ', '.join(arguments) + ')'

553
    def on_expression(self, node):
554
555
        return self.non_terminal(node, 'Alternative')

556
    def on_term(self, node):
557
558
        return self.non_terminal(node, 'Sequence')

559
    def on_factor(self, node):
560
561
562
563
        assert isinstance(node.parser, Sequence), node.as_sexpr()  # these assert statements can be removed
        assert node.children
        assert len(node.result) >= 2, node.as_sexpr()
        prefix = node.result[0].result
564
        custom_args = []
565
566
567

        if prefix in {'::', ':'}:
            assert len(node.result) == 2
568
            arg = node.result[-1]
569
            if arg.parser.name != 'symbol':
Eckhart Arnold's avatar
Eckhart Arnold committed
570
                node.add_error(('Retrieve Operator "%s" requires a symbol, '
571
572
                                'and not a %s.') % (prefix, str(arg.parser)))
                return str(arg.result)
573
574
            if str(arg) in self.directives['counterpart']:
                custom_args = ['counterpart=%s_counterpart' % str(arg)]
575
576
            self.variables.add(arg.result)

577
        elif len(node.result) > 2:
578
579
            # shift = (Node(node.parser, node.result[1].result),)
            # node.result[1].result = shift + node.result[2:]
580
            node.result[1].result = (Node(node.result[1].parser, node.result[1].result),) \
581
582
583
584
585
                                    + node.result[2:]
            node.result[1].parser = node.parser
            node.result = (node.result[0], node.result[1])

        node.result = node.result[1:]
586
587
588
589
590
        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)
591

592
    def on_option(self, node):
593
594
        return self.non_terminal(node, 'Optional')

595
    def on_repetition(self, node):
596
597
        return self.non_terminal(node, 'ZeroOrMore')

598
    def on_oneormore(self, node):
599
600
        return self.non_terminal(node, 'OneOrMore')

601
    def on_regexchain(self, node):
602
603
        raise EBNFCompilerError("Not yet implemented!")

604
    def on_group(self, node):
605
606
607
        raise EBNFCompilerError("Group nodes should have been eliminated by "
                                "AST transformation!")

608
    def on_symbol(self, node):
609
        if node.result in self.directives['tokens']:
610
611
            return 'ScannerToken("' + node.result + '")'
        else:
612
            self.symbol_nodes.append(node)
613
614
615
616
            if node.result in self.rules:
                self.recursive.add(node.result)
            return node.result

617
    def on_literal(self, node):
618
        return 'Token(' + node.result.replace('\\', r'\\') + ')'  # return 'Token(' + ', '.join([node.result]) + ')' ?
619

620
    def on_regexp(self, node):
621
622
623
624
625
626
627
628
629
        rx = node.result
        name = []
        if rx[:2] == '~/':
            if not 'left' in self.directives['literalws']:
                name = ['wL=' + WHITESPACE_KEYWORD] + name
            rx = rx[1:]
        elif 'left' in self.directives['literalws']:
            name = ["wL=''"] + name
        if rx[-2:] == '/~':
Eckhart Arnold's avatar
Eckhart Arnold committed
630
            if 'right' not in self.directives['literalws']:
631
632
633
634
635
636
637
638
639
640
641
642
643
                name = ['wR=' + WHITESPACE_KEYWORD] + name
            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) + ')'

644
    def on_list_(self, node):
645
646
        assert node.children
        return set(item.result.strip() for item in node.result)
647
648


Eckhart Arnold's avatar
Eckhart Arnold committed
649
650
651
652
653
654
655
656
657
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