ebnf.py 27.7 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, \
32
33
    replace_by_single_child, TOKEN_PTYPE, remove_expendables, remove_tokens, flatten, \
    forbid, assert_content, WHITESPACE_PTYPE, key_parser_name, key_tag_name
Eckhart Arnold's avatar
Eckhart Arnold committed
34
from .versionnumber import __version__
35
36


37
38
39
40
41
__all__ = ['get_ebnf_scanner',
           'get_ebnf_grammar',
           'get_ebnf_transformer',
           'get_ebnf_compiler',
           'EBNFGrammar',
42
           'EBNFTransformer',
Eckhart Arnold's avatar
Eckhart Arnold committed
43
           'EBNFCompilerError',
44
45
           'EBNFCompiler',
           'grammar_changed']
46
47


Eckhart Arnold's avatar
Eckhart Arnold committed
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
########################################################################
#
# EBNF scanning
#
########################################################################


def get_ebnf_scanner():
    return nil_scanner


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


66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
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
85
                | [flowmarker] regexchain
86
87
88
89
90
91
92
93
94
                | [flowmarker] oneormore
                | repetition
                | option

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

    group      =  "(" expression §")"
95
    regexchain =  ">" expression §"<"                # compiles "expression" into a singular regular expression
96
97
    oneormore  =  "{" expression "}+"
    repetition =  "{" expression §"}"
98
99
    option     =  "[" expression §"]"

100
101
102
103
104
105
    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|$)/~
                                                     # '~' is a whitespace-marker, if present leading or trailing
                                                     # whitespace of a regular expression will be ignored tacitly.
106
    list_      =  /\w+/~ { "," /\w+/~ }              # comma separated list of symbols, e.g. BEGIN_LIST, END_LIST,
107
108
109
110
                                                     # BEGIN_QUOTE, END_QUOTE ; see CommonMark/markdown.py for an exmaple
    EOF =  !/./
    """
    expression = Forward()
111
    source_hash__ = "a410e1727fb7575e98ff8451dbf8f3bd"
112
    parser_initialization__ = "upon instatiation"
113
114
    COMMENT__ = r'#.*(?:\n|$)'
    WSP__ = mixin_comment(whitespace=r'\s*', comment=r'#.*(?:\n|$)')
115
    wspL__ = ''
116
    wspR__ = WSP__
117
    EOF = NegativeLookahead(RE('.', wR=''))
118
    list_ = Sequence(RE('\\w+'), ZeroOrMore(Sequence(Token(","), RE('\\w+'))))
119
120
121
    regexp = RE('~?/(?:[^/]|(?<=\\\\)/)*/~?')
    literal = Alternative(RE('"(?:[^"]|\\\\")*?"'), RE("'(?:[^']|\\\\')*?'"))
    symbol = RE('(?!\\d)\\w+')
122
    option = Sequence(Token("["), expression, Required(Token("]")))
123
124
    repetition = Sequence(Token("{"), expression, Required(Token("}")))
    oneormore = Sequence(Token("{"), expression, Token("}+"))
125
    regexchain = Sequence(Token("<"), expression, Required(Token(">")))
126
127
128
129
130
    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),
131
132
                         Sequence(Optional(flowmarker), group), Sequence(Optional(flowmarker), regexchain),
                         Sequence(Optional(flowmarker), oneormore), repetition, option)
133
134
135
136
    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)
137
    syntax = Sequence(Optional(RE('', wR='', wL=WSP__)), ZeroOrMore(Alternative(definition, directive)), Required(EOF))
138
139
140
    root__ = syntax


Eckhart Arnold's avatar
Eckhart Arnold committed
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
187
188
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
#
########################################################################


189
190
#TODO: Add Capture and Retrieve Validation: A variable mustn't be captured twice before retrival?!?

191
EBNF_transformation_table = {
192
193
194
195
196
    # AST Transformations for EBNF-grammar
    "syntax":
        remove_expendables,
    "directive, definition":
        partial(remove_tokens, tokens={'@', '='}),
Eckhart Arnold's avatar
Eckhart Arnold committed
197
    "expression":
198
        [replace_by_single_child, flatten,
Eckhart Arnold's avatar
Eckhart Arnold committed
199
         partial(remove_tokens, tokens={'|'})],
200
201
202
203
204
    "term":
        [replace_by_single_child, flatten],  # supports both idioms:  "{ factor }+" and "factor { factor }"
    "factor, flowmarker, retrieveop":
        replace_by_single_child,
    "group":
205
206
207
        [remove_enclosing_delimiters, replace_by_single_child],
    "oneormore, repetition, option, regexchain":
        [reduce_single_child, remove_enclosing_delimiters],
208
    "symbol, literal, regexp":
209
        [remove_expendables, reduce_single_child],
210
    (TOKEN_PTYPE, WHITESPACE_PTYPE):
211
        [remove_expendables, reduce_single_child],
212
    "list_":
213
        [flatten, partial(remove_tokens, tokens={','})],
214
    "*":
215
216
217
        [remove_expendables, replace_by_single_child]
}

218

219
EBNF_validation_table = {
220
221
    # Semantic validation on the AST
    "repetition, option, oneormore":
222
        [partial(forbid, child_tags=['repetition', 'option', 'oneormore']),
223
224
         partial(assert_content, regex=r'(?!§)')],
}
225

226

227
def EBNFTransformer(syntax_tree):
228
    for processing_table, key_func in [(EBNF_transformation_table, key_tag_name),
229
                                       (EBNF_validation_table, key_tag_name)]:
230
        traverse(syntax_tree, processing_table, key_func)
di68kap's avatar
di68kap committed
231
232


Eckhart Arnold's avatar
Eckhart Arnold committed
233
def get_ebnf_transformer():
234
    return EBNFTransformer
Eckhart Arnold's avatar
Eckhart Arnold committed
235
236
237
238
239
240
241
242


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

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
278
279
280
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
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__"
293
294
    WHITESPACE_KEYWORD = "WSP__"
    RESERVED_SYMBOLS = {WHITESPACE_KEYWORD, COMMENT_KEYWORD}
295
296
    AST_ERROR = "Badly structured syntax tree. " \
                "Potentially due to erroneuos AST transformation."
297
298
299
300
    PREFIX_TABLE = {'§': 'Required',
                    '&': 'Lookahead', '!': 'NegativeLookahead',
                    '-&': 'Lookbehind', '-!': 'NegativeLookbehind',
                    '::': 'Pop', ':': 'Retrieve'}
301
302
303
    WHITESPACE = {'horizontal': r'[\t ]*',  # default: horizontal
                  'linefeed': r'[ \t]*\n?(?!\s*\n)[ \t]*',
                  'vertical': r'\s*'}
304

305
    def __init__(self, grammar_name="", grammar_source=""):
Eckhart Arnold's avatar
Eckhart Arnold committed
306
        super(EBNFCompiler, self).__init__(grammar_name, grammar_source)
307
308
309
        self._reset()

    def _reset(self):
Eckhart Arnold's avatar
Eckhart Arnold committed
310
        self._result = None
311
312
        self.rules = set()
        self.variables = set()
313
        self.symbol_nodes = []
314
315
316
        self.definition_names = []
        self.recursive = set()
        self.root = ""
317
        self.directives = {'whitespace': self.WHITESPACE['horizontal'],
318
                           'comment': '',
319
                           'literalws': ['right'],
320
                           'tokens': set(),     # alt. 'scanner_tokens'
321
                           'counterpart': set()}     # alt. 'retrieve_counterpart'
322

Eckhart Arnold's avatar
Eckhart Arnold committed
323
324
325
326
    @property
    def result(self):
        return self._result

327
328
    def gen_scanner_skeleton(self):
        name = self.grammar_name + "Scanner"
329
330
        return "def %s(text):\n    return text\n" % name \
               + SCANNER_FACTORY.format(NAME=self.grammar_name)
331

332
    def gen_transformer_skeleton(self):
333
        if not self.definition_names:
Eckhart Arnold's avatar
Eckhart Arnold committed
334
335
            raise EBNFCompilerError('Compiler must be run before calling '
                                    '"gen_transformer_Skeleton()"!')
336
337
        tt_name = self.grammar_name + '_AST_transformation_table'
        tf_name = self.grammar_name + 'Transform'
di68kap's avatar
di68kap committed
338
        transtable = [tt_name + ' = {',
339
340
341
                      '    # AST Transformations for the ' +
                      self.grammar_name + '-grammar']
        for name in self.definition_names:
di68kap's avatar
di68kap committed
342
            transtable.append('    "' + name + '": no_operation,')
343
344
        transtable += ['    "": no_operation', '}', '',  tf_name +
                       ' = partial(traverse, processing_table=%s)' % tt_name, '']
345
        transtable += [TRANSFORMER_FACTORY.format(NAME=self.grammar_name)]
346
347
348
349
350
351
352
353
354
355
356
        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
357
                    self.grammar_name + '", grammar_source=""):',
358
                    '        super(' + self.grammar_name +
Eckhart Arnold's avatar
Eckhart Arnold committed
359
                    'Compiler, self).__init__(grammar_name, grammar_source)',
360
361
                    "        assert re.match('\w+\Z', grammar_name)", '']
        for name in self.definition_names:
362
            method_name = CompilerBase.derive_method_name(name)
363
            if name == self.root:
364
                compiler += ['    def ' + method_name + '(self, node):',
365
366
                             '        return node', '']
            else:
367
                compiler += ['    def ' + method_name + '(self, node):',
368
                             '        pass', '']
369
        compiler += [COMPILER_FACTORY.format(NAME=self.grammar_name)]
370
        return '\n'.join(compiler)
371

372
    def assemble_parser(self, definitions, root_node):
373
        # fix capture of variables that have been defined before usage [sic!]
374

375
376
377
        if self.variables:
            for i in range(len(definitions)):
                if definitions[i][0] in self.variables:
378
                    definitions[i] = (definitions[i][0], 'Capture(%s)' % definitions[1])
379
380

        self.definition_names = [defn[0] for defn in definitions]
381
        definitions.append(('wspR__', self.WHITESPACE_KEYWORD
Eckhart Arnold's avatar
Eckhart Arnold committed
382
                            if 'right' in self.directives['literalws'] else "''"))
383
        definitions.append(('wspL__', self.WHITESPACE_KEYWORD
Eckhart Arnold's avatar
Eckhart Arnold committed
384
                            if 'left' in self.directives['literalws'] else "''"))
385
        definitions.append((self.WHITESPACE_KEYWORD,
386
387
388
389
390
391
392
                            ("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
393

394
        article = 'an ' if self.grammar_name[0:1] in "AaEeIiOoUu" else 'a '  # what about 'hour', 'universe' etc.?
395
396
397
398
        declarations = ['class ' + self.grammar_name +
                        'Grammar(GrammarBase):',
                        'r"""Parser for ' + article + self.grammar_name +
                        ' source file' +
399
                        (', with this grammar:' if self.grammar_source else '.')]
400
        definitions.append(('parser_initialization__', '"upon instatiation"'))
401
        if self.grammar_source:
402
            definitions.append(('source_hash__',
403
                                '"%s"' % md5(self.grammar_source, __version__)))
404
            declarations.append('')
405
            declarations += [line for line in self.grammar_source.split('\n')]
406
407
408
409
            while declarations[-1].strip() == '':
                declarations = declarations[:-1]
        declarations.append('"""')

410
        # add default functions for counterpart filters of pop or retrieve operators
411

412
413
414
415
416
417
        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(">", "<")')
418

419
        # turn definitions into declarations in reverse order
420

421
422
423
424
425
426
427
428
429
        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]
430
        known_symbols = self.rules | self.RESERVED_SYMBOLS
431
        for nd in self.symbol_nodes:
432
            if nd.result not in known_symbols:
433
                nd.add_error("Missing production for symbol '%s'" % nd.result)
434
                root_node.error_flag = True
435
        if self.root and 'root__' not in self.rules:
436
437
            declarations.append('root__ = ' + self.root)
        declarations.append('')
Eckhart Arnold's avatar
Eckhart Arnold committed
438
439
440
        self._result = '\n    '.join(declarations) \
                       + GRAMMAR_FACTORY.format(NAME=self.grammar_name)
        return self._result
441

442
    def on_syntax(self, node):
443
444
445
446
        self._reset()
        definitions = []

        # drop the wrapping sequence node
447
        if len(node.children) == 1 and not node.result[0].parser.name:
448
449
450
451
452
            node = node.result[0]

        # compile definitions and directives and collect definitions
        for nd in node.result:
            if nd.parser.name == "definition":
453
                definitions.append(self._compile(nd))
454
455
            else:
                assert nd.parser.name == "directive", nd.as_sexpr()
456
                self._compile(nd)
457
                node.error_flag |= nd.error_flag
458

459
        return self.assemble_parser(definitions, node)
460

461
    def on_definition(self, node):
462
        rule = node.result[0].result
463
464
465
        if rule in self.rules:
            node.add_error('A rule with name "%s" has already been defined.' % rule)
        elif rule in EBNFCompiler.RESERVED_SYMBOLS:
466
467
468
469
            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)
470
        elif rule in self.directives['tokens']:
471
472
473
474
475
476
477
            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)
478
            defn = self._compile(node.result[1])
479
            if rule in self.variables:
480
                defn = 'Capture(%s)' % defn
481
482
483
484
485
                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
486
        return rule, defn
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501

    @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

502
    def on_directive(self, node):
503
        key = node.result[0].result.lower()
504
        assert key not in self.directives['tokens']
505
        if key in {'comment', 'whitespace'}:
506
507
            if node.result[1].parser.name == "list_":
                if len(node.result[1].result) != 1:
Eckhart Arnold's avatar
Eckhart Arnold committed
508
                    node.add_error('Directive "%s" must have one, but not %i values.' %
509
                                   (key, len(node.result[1])))
510
                value = self._compile(node.result[1]).pop()
511
512
                if key == 'whitespace' and value in EBNFCompiler.WHITESPACE:
                    value = EBNFCompiler.WHITESPACE[value]  # replace whitespace-name by regex
513
                else:
514
                    node.add_error('Value "%s" not allowed for directive "%s".' % (value, key))
515
            else:
516
517
518
519
520
521
522
523
                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])
524
525
526
                if key == 'whitespace' and not re.match(value, ''):
                    node.add_error("Implicit whitespace should always match the empty string, "
                                   "/%s/ does not." % value)
527
            self.directives[key] = value
528

529
        elif key == 'literalws':
530
            value = {item.lower() for item in self._compile(node.result[1])}
531
            if (len(value - {'left', 'right', 'both', 'none'}) > 0
Eckhart Arnold's avatar
Eckhart Arnold committed
532
                    or ('none' in value and len(value) > 1)):
533
534
535
536
537
538
539
                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)

540
        elif key in {'tokens', 'scanner_tokens'}:
541
            self.directives['tokens'] |= self._compile(node.result[1])
542

543
        elif key in {'counterpart', 'retrieve_counterpart'}:
544
            self.directives['counterpart'] |= self._compile(node.result[1])
545

546
547
548
        else:
            node.add_error('Unknown directive %s ! (Known ones are %s .)' %
                           (key,
549
                            ', '.join(list(self.directives.keys()))))
550
551
        return ""

552
    def non_terminal(self, node, parser_class, custom_args=[]):
553
554
555
        """Compiles any non-terminal, where `parser_class` indicates the Parser class
        name for the particular non-terminal.
        """
556
        arguments = [self._compile(r) for r in node.result] + custom_args
557
558
        return parser_class + '(' + ', '.join(arguments) + ')'

559
    def on_expression(self, node):
560
561
        return self.non_terminal(node, 'Alternative')

562
    def on_term(self, node):
563
564
        return self.non_terminal(node, 'Sequence')

565
    def on_factor(self, node):
566
567
568
        assert node.children
        assert len(node.result) >= 2, node.as_sexpr()
        prefix = node.result[0].result
569
        custom_args = []
570
571
572

        if prefix in {'::', ':'}:
            assert len(node.result) == 2
573
            arg = node.result[-1]
574
            if arg.parser.name != 'symbol':
Eckhart Arnold's avatar
Eckhart Arnold committed
575
                node.add_error(('Retrieve Operator "%s" requires a symbol, '
576
577
                                'and not a %s.') % (prefix, str(arg.parser)))
                return str(arg.result)
578
579
            if str(arg) in self.directives['counterpart']:
                custom_args = ['counterpart=%s_counterpart' % str(arg)]
580
581
            self.variables.add(arg.result)

582
        elif len(node.result) > 2:
583
584
            # shift = (Node(node.parser, node.result[1].result),)
            # node.result[1].result = shift + node.result[2:]
585
            node.result[1].result = (Node(node.result[1].parser, node.result[1].result),) \
586
587
588
589
590
                                    + node.result[2:]
            node.result[1].parser = node.parser
            node.result = (node.result[0], node.result[1])

        node.result = node.result[1:]
591
592
593
594
595
        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)
596

597
    def on_option(self, node):
598
599
        return self.non_terminal(node, 'Optional')

600
    def on_repetition(self, node):
601
602
        return self.non_terminal(node, 'ZeroOrMore')

603
    def on_oneormore(self, node):
604
605
        return self.non_terminal(node, 'OneOrMore')

606
    def on_regexchain(self, node):
607
608
        raise EBNFCompilerError("Not yet implemented!")

609
    def on_group(self, node):
610
611
612
        raise EBNFCompilerError("Group nodes should have been eliminated by "
                                "AST transformation!")

613
    def on_symbol(self, node):
614
        if node.result in self.directives['tokens']:
615
616
            return 'ScannerToken("' + node.result + '")'
        else:
617
            self.symbol_nodes.append(node)
618
619
620
621
            if node.result in self.rules:
                self.recursive.add(node.result)
            return node.result

622
    def on_literal(self, node):
623
        return 'Token(' + node.result.replace('\\', r'\\') + ')'  # return 'Token(' + ', '.join([node.result]) + ')' ?
624

625
    def on_regexp(self, node):
626
627
628
629
        rx = node.result
        name = []
        if rx[:2] == '~/':
            if not 'left' in self.directives['literalws']:
630
                name = ['wL=' + self.WHITESPACE_KEYWORD] + name
631
632
633
634
            rx = rx[1:]
        elif 'left' in self.directives['literalws']:
            name = ["wL=''"] + name
        if rx[-2:] == '/~':
Eckhart Arnold's avatar
Eckhart Arnold committed
635
            if 'right' not in self.directives['literalws']:
636
                name = ['wR=' + self.WHITESPACE_KEYWORD] + name
637
638
639
640
641
642
643
644
645
646
647
648
            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) + ')'

649
    def on_list_(self, node):
650
651
        assert node.children
        return set(item.result.strip() for item in node.result)
652
653


Eckhart Arnold's avatar
Eckhart Arnold committed
654
655
656
657
658
659
660
661
662
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