21.10.2021, 9:00 - 11:00: Due to updates GitLab may be unavailable for some minutes between 09:00 and 11:00.

ebnf.py 59 KB
Newer Older
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
# ebnf.py - EBNF -> Python-Parser compilation for DHParser
#
# Copyright 2016  by Eckhart Arnold (arnold@badw.de)
#                 Bavarian Academy of Sciences an Humanities (badw.de)
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
# implied.  See the License for the specific language governing
# permissions and limitations under the License.
17
18


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

26

27
from collections import OrderedDict
28
from functools import partial
eckhart's avatar
eckhart committed
29
30
import keyword
import os
31

eckhart's avatar
eckhart committed
32
from DHParser.compile import CompilerError, Compiler, compile_source, visitor_name
33
from DHParser.error import Error
34
from DHParser.parse import Grammar, mixin_comment, Forward, RegExp, Whitespace, \
35
36
    NegativeLookahead, Alternative, Series, Option, OneOrMore, ZeroOrMore, Token, \
    GrammarError
37
from DHParser.preprocess import nil_preprocessor, PreprocessorFunc
eckhart's avatar
eckhart committed
38
from DHParser.syntaxtree import Node, WHITESPACE_PTYPE, TOKEN_PTYPE
eckhart's avatar
eckhart committed
39
from DHParser.toolkit import load_if_file, escape_re, md5, sane_parser_name, re, expand_table, \
40
    GLOBALS, CONFIG_PRESET, get_config_value, unrepr, compile_python_object, typing
eckhart's avatar
eckhart committed
41
from DHParser.transform import TransformationFunc, traverse, remove_brackets, \
Eckhart Arnold's avatar
Eckhart Arnold committed
42
    reduce_single_child, replace_by_single_child, remove_expendables, \
eckhart's avatar
eckhart committed
43
    remove_tokens, flatten, forbid, assert_content
44
from DHParser.versionnumber import __version__
eckhart's avatar
eckhart committed
45
from typing import Callable, Dict, List, Set, Tuple, Sequence, Union, Optional, Any, cast
eckhart's avatar
eckhart committed
46

47

48
__all__ = ('get_ebnf_preprocessor',
49
50
51
52
           'get_ebnf_grammar',
           'get_ebnf_transformer',
           'get_ebnf_compiler',
           'EBNFGrammar',
53
           'EBNFTransform',
Eckhart Arnold's avatar
Eckhart Arnold committed
54
           'EBNFCompilerError',
55
           'EBNFCompiler',
56
           'grammar_changed',
eckhart's avatar
eckhart committed
57
           'compile_ebnf',
58
           'PreprocessorFactoryFunc',
59
60
           'ParserFactoryFunc',
           'TransformerFactoryFunc',
61
           'CompilerFactoryFunc')
62
63


64
65
66
67
68
69
70
########################################################################
#
# Presets
#
########################################################################

CONFIG_PRESET['add_grammar_source_to_parser_docstring'] = False
Eckhart Arnold's avatar
Eckhart Arnold committed
71
72
73
74
# CONFIG_PRESET['static_analysis'] = "early" # do a static analysis right
#                                            # after ebnf compilation
# already set in parse.py - config vars should probably moved to a
#                           a dedicated global module
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114

########################################################################
#
# source code support
#
########################################################################


dhparserdir = os.path.dirname(os.path.dirname(os.path.realpath(__file__)))


DHPARSER_IMPORTS = '''
import collections
from functools import partial
import os
import sys

sys.path.append(r'{dhparserdir}')

try:
    import regex as re
except ImportError:
    import re
from DHParser import logging, is_filename, load_if_file, \\
    Grammar, Compiler, nil_preprocessor, PreprocessorToken, Whitespace, DropWhitespace, \\
    Lookbehind, Lookahead, Alternative, Pop, Token, DropToken, Synonym, AllOf, SomeOf, \\
    Unordered, Option, NegativeLookbehind, OneOrMore, RegExp, Retrieve, Series, Capture, \\
    ZeroOrMore, Forward, NegativeLookahead, Required, mixin_comment, compile_source, \\
    grammar_changed, last_value, counterpart, accumulate, PreprocessorFunc, is_empty, \\
    Node, TransformationFunc, TransformationDict, transformation_factory, traverse, \\
    remove_children_if, move_adjacent, normalize_whitespace, is_anonymous, matches_re, \\
    reduce_single_child, replace_by_single_child, replace_or_reduce, remove_whitespace, \\
    remove_expendables, remove_empty, remove_tokens, flatten, is_insignificant_whitespace, \\
    is_expendable, collapse, collapse_if, replace_content, WHITESPACE_PTYPE, TOKEN_PTYPE, \\
    remove_nodes, remove_content, remove_brackets, replace_parser, remove_anonymous_tokens, \\
    keep_children, is_one_of, not_one_of, has_content, apply_if, remove_first, remove_last, \\
    remove_anonymous_empty, keep_nodes, traverse_locally, strip, lstrip, rstrip, \\
    replace_content, replace_content_by, forbid, assert_content, remove_infix_operator, \\
    flatten_anonymous_nodes, error_on, recompile_grammar, GLOBALS
'''.format(dhparserdir=dhparserdir)
115
116


Eckhart Arnold's avatar
Eckhart Arnold committed
117
118
119
120
121
122
123
########################################################################
#
# EBNF scanning
#
########################################################################


124
125
def get_ebnf_preprocessor() -> PreprocessorFunc:
    return nil_preprocessor
Eckhart Arnold's avatar
Eckhart Arnold committed
126
127
128
129
130
131
132
133


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

134

di68kap's avatar
di68kap committed
135
class EBNFGrammar(Grammar):
eckhart's avatar
eckhart committed
136
    r"""
eckhart's avatar
eckhart committed
137
138
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
    Parser for an EBNF source file, with this grammar:

    @ 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 | symbol) { "," (regexp | literal | symbol) }

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

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

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

    symbol     = /(?!\d)\w+/~                       # e.g. expression, factor, parameter_list
    literal    = /"(?:[^"]|\\")*?"/~                # e.g. "(", '+', 'while'
               | /'(?:[^']|\\')*?'/~                # whitespace following literals will be ignored tacitly.
    plaintext  = /`(?:[^"]|\\")*?`/~                # like literal but does not eat whitespace
    regexp     = /\/(?:\\\/|[^\/])*?\//~            # e.g. /\w+/, ~/#.*(?:\n|$)/~
    whitespace = /~/~                               # insignificant whitespace

    EOF = !/./
178
    """
di68kap's avatar
di68kap committed
179
    expression = Forward()
eckhart's avatar
eckhart committed
180
    source_hash__ = "82a7c668f86b83f86515078e6c9093ed"
181
    static_analysis_pending__ = False
eckhart's avatar
eckhart committed
182
    parser_initialization__ = ["upon instantiation"]
di68kap's avatar
di68kap committed
183
184
    COMMENT__ = r'#.*(?:\n|$)'
    WHITESPACE__ = r'\s*'
185
    WSP_RE__ = mixin_comment(whitespace=WHITESPACE__, comment=COMMENT__)
186
    wsp__ = Whitespace(WSP_RE__)
di68kap's avatar
di68kap committed
187
    EOF = NegativeLookahead(RegExp('.'))
188
189
190
191
192
193
194
195
196
197
198
199
200
201
    whitespace = Series(RegExp('~'), wsp__)
    regexp = Series(RegExp('/(?:\\\\/|[^/])*?/'), wsp__)
    plaintext = Series(RegExp('`(?:[^"]|\\\\")*?`'), wsp__)
    literal = Alternative(Series(RegExp('"(?:[^"]|\\\\")*?"'), wsp__), Series(RegExp("'(?:[^']|\\\\')*?'"), wsp__))
    symbol = Series(RegExp('(?!\\d)\\w+'), wsp__)
    option = Series(Series(Token("["), wsp__), expression, Series(Token("]"), wsp__), mandatory=1)
    repetition = Series(Series(Token("{"), wsp__), expression, Series(Token("}"), wsp__), mandatory=1)
    oneormore = Series(Series(Token("{"), wsp__), expression, Series(Token("}+"), wsp__))
    unordered = Series(Series(Token("<"), wsp__), expression, Series(Token(">"), wsp__), mandatory=1)
    group = Series(Series(Token("("), wsp__), expression, Series(Token(")"), wsp__), mandatory=1)
    retrieveop = Alternative(Series(Token("::"), wsp__), Series(Token(":"), wsp__))
    flowmarker = Alternative(Series(Token("!"), wsp__), Series(Token("&"), wsp__),
                             Series(Token("-!"), wsp__), Series(Token("-&"), wsp__))
    factor = Alternative(Series(Option(flowmarker), Option(retrieveop), symbol,
eckhart's avatar
eckhart committed
202
203
204
205
                                NegativeLookahead(Series(Token("="), wsp__))), Series(Option(flowmarker), literal),
                         Series(Option(flowmarker), plaintext), Series(Option(flowmarker), regexp),
                         Series(Option(flowmarker), whitespace), Series(Option(flowmarker), oneormore),
                         Series(Option(flowmarker), group), Series(Option(flowmarker), unordered), repetition, option)
206
207
208
    term = OneOrMore(Series(Option(Series(Token("§"), wsp__)), factor))
    expression.set(Series(term, ZeroOrMore(Series(Series(Token("|"), wsp__), term))))
    directive = Series(Series(Token("@"), wsp__), symbol, Series(Token("="), wsp__),
eckhart's avatar
eckhart committed
209
210
                       Alternative(regexp, literal, symbol),
                       ZeroOrMore(Series(Series(Token(","), wsp__), Alternative(regexp, literal, symbol))), mandatory=1)
211
    definition = Series(symbol, Series(Token("="), wsp__), expression, mandatory=1)
eckhart's avatar
eckhart committed
212
    syntax = Series(Option(Series(wsp__, RegExp(''))), ZeroOrMore(Alternative(definition, directive)), EOF, mandatory=2)
di68kap's avatar
di68kap committed
213
214
215
    root__ = syntax


216
def grammar_changed(grammar_class, grammar_source: str) -> bool:
217
218
    """
    Returns ``True`` if ``grammar_class`` does not reflect the latest
Eckhart Arnold's avatar
Eckhart Arnold committed
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
    changes of ``grammar_source``

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

    Returns (bool):
        True, if the source text of the grammar is different from the
        source from which the grammar class was generated
    """
    grammar = load_if_file(grammar_source)
    chksum = md5(grammar, __version__)
    if isinstance(grammar_class, str):
        # grammar_class = load_compiler_suite(grammar_class)[1]
        with open(grammar_class, 'r', encoding='utf8') as f:
            pycode = f.read()
eckhart's avatar
eckhart committed
237
        m = re.search(r'class \w*\(Grammar\)', pycode)
Eckhart Arnold's avatar
Eckhart Arnold committed
238
239
240
241
242
243
244
245
246
247
        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__


248
def get_ebnf_grammar() -> EBNFGrammar:
Eckhart Arnold's avatar
Eckhart Arnold committed
249
    try:
250
        grammar = GLOBALS.ebnf_grammar_singleton
Eckhart Arnold's avatar
Eckhart Arnold committed
251
        return grammar
252
    except AttributeError:
253
254
        GLOBALS.ebnf_grammar_singleton = EBNFGrammar()
        return GLOBALS.ebnf_grammar_singleton
Eckhart Arnold's avatar
Eckhart Arnold committed
255
256
257
258
259
260
261
262
263


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


264
EBNF_AST_transformation_table = {
265
    # AST Transformations for EBNF-grammar
266
    "<":
267
        remove_expendables,
268
    "syntax":
269
        [],  # otherwise '"*": replace_by_single_child' would be applied
270
    "directive, definition":
eckhart's avatar
eckhart committed
271
        [flatten, remove_tokens('@', '=', ',')],
Eckhart Arnold's avatar
Eckhart Arnold committed
272
    "expression":
273
        [replace_by_single_child, flatten, remove_tokens('|')],  # remove_infix_operator],
274
    "term":
275
276
        [replace_by_single_child, flatten],  # supports both idioms:
                                             # "{ factor }+" and "factor { factor }"
277
    "factor, flowmarker, retrieveop":
278
        replace_by_single_child,
279
    "group":
280
        [remove_brackets, replace_by_single_child],
281
282
    "unordered":
        remove_brackets,
283
    "oneormore, repetition, option":
Eckhart Arnold's avatar
Eckhart Arnold committed
284
        [reduce_single_child, remove_brackets,
285
         forbid('repetition', 'option', 'oneormore'), assert_content(r'(?!§)(?:.|\n)*')],
286
    "symbol, literal, regexp":
Eckhart Arnold's avatar
Eckhart Arnold committed
287
        reduce_single_child,
288
    (TOKEN_PTYPE, WHITESPACE_PTYPE):
Eckhart Arnold's avatar
Eckhart Arnold committed
289
        reduce_single_child,
eckhart's avatar
eckhart committed
290
291
    # "list_":
    #     [flatten, remove_infix_operator],
292
    "*":
293
        replace_by_single_child
294
295
}

296

Eckhart Arnold's avatar
Eckhart Arnold committed
297
def EBNFTransform() -> TransformationFunc:
298
    return partial(traverse, processing_table=EBNF_AST_transformation_table.copy())
di68kap's avatar
di68kap committed
299

eckhart's avatar
eckhart committed
300

301
def get_ebnf_transformer() -> TransformationFunc:
302
    try:
303
        transformer = GLOBALS.EBNF_transformer_singleton
304
    except AttributeError:
305
306
        GLOBALS.EBNF_transformer_singleton = EBNFTransform()
        transformer = GLOBALS.EBNF_transformer_singleton
307
    return transformer
Eckhart Arnold's avatar
Eckhart Arnold committed
308
309
310
311
312
313
314
315


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

316

317
PreprocessorFactoryFunc = Callable[[], PreprocessorFunc]
318
ParserFactoryFunc = Callable[[], Grammar]
319
TransformerFactoryFunc = Callable[[], TransformationFunc]
320
321
CompilerFactoryFunc = Callable[[], Compiler]

322
323
324
PREPROCESSOR_FACTORY = '''
def get_preprocessor() -> PreprocessorFunc:
    return {NAME}Preprocessor
325
326
327
328
'''


GRAMMAR_FACTORY = '''
329
def get_grammar() -> {NAME}Grammar:
eckhart's avatar
eckhart committed
330
    global GLOBALS
331
    try:
332
        grammar = GLOBALS.{NAME}_{ID:08d}_grammar_singleton
333
    except AttributeError:
334
        GLOBALS.{NAME}_{ID:08d}_grammar_singleton = {NAME}Grammar()
eckhart's avatar
eckhart committed
335
        if hasattr(get_grammar, 'python_src__'):
336
337
            GLOBALS.{NAME}_{ID:08d}_grammar_singleton.python_src__ = get_grammar.python_src__
        grammar = GLOBALS.{NAME}_{ID:08d}_grammar_singleton
338
    return grammar
339
340
341
342
'''


TRANSFORMER_FACTORY = '''
343
344
345
def {NAME}Transform() -> TransformationDict:
    return partial(traverse, processing_table={NAME}_AST_transformation_table.copy())

346
def get_transformer() -> TransformationFunc:
347
    try:
348
        transformer = GLOBALS.{NAME}_{ID:08d}_transformer_singleton
349
    except AttributeError:
350
351
        GLOBALS.{NAME}_{ID:08d}_transformer_singleton = {NAME}Transform()
        transformer = GLOBALS.{NAME}_{ID:08d}_transformer_singleton
352
    return transformer
353
354
355
356
'''


COMPILER_FACTORY = '''
eckhart's avatar
eckhart committed
357
def get_compiler() -> {NAME}Compiler:
358
    try:
359
        compiler = GLOBALS.{NAME}_{ID:08d}_compiler_singleton
360
    except AttributeError:
361
362
        GLOBALS.{NAME}_{ID:08d}_compiler_singleton = {NAME}Compiler()
        compiler = GLOBALS.{NAME}_{ID:08d}_compiler_singleton
363
    return compiler
364
365
'''

Eckhart Arnold's avatar
Eckhart Arnold committed
366

367
368
369
370
WHITESPACE_TYPES = {'horizontal': r'[\t ]*',  # default: horizontal
                    'linefeed': r'[ \t]*\n?(?!\s*\n)[ \t]*',
                    'vertical': r'\s*'}

371
372
373
DROP_TOKEN  = 'token'
DROP_WSPC   = 'whitespace'
DROP_VALUES = {DROP_TOKEN, DROP_WSPC}
374

eckhart's avatar
eckhart committed
375
376
# Representation of Python code or, rather, something that will be output as Python code
ReprType = Union[str, unrepr]
eckhart's avatar
eckhart committed
377
378


379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
class EBNFDirectives:
    """
    A Record that keeps information about compiler directives
    during the compilation process.

    Attributes:
        whitespace:  the regular expression string for (insignificant)
                whitespace

        comment:  the regular expression string for comments

        literalws:  automatic whitespace eating next to literals. Can
                be either 'left', 'right', 'none', 'both'

        tokens:  set of the names of preprocessor tokens
        filter:  mapping of symbols to python filter functions that
                will be called on any retrieve / pop - operations on
                these symbols

        error:  mapping of symbols to tuples of match conditions and
                customized error messages. A match condition can be
                either a string or a regular expression. The first
                error message where the search condition matches will
                be displayed. An empty string '' as search condition
                always matches, so in case of multiple error messages,
                this condition should be placed at the end.

406
407
408
409
410
411
412
413
414
415
416
417
        skip:  mapping of symbols to a list of search expressions. A
                search expressions can be either a string ot a regular
                expression. The closest match is the point of reentry
                for the series-parser when a mandatory item failed to
                match the following text.

        resume:  mapping of symbols to a list of search expressions. A
                search expressions can be either a string ot a regular
                expression. The closest match is the point of reentry
                for after a parsing error has error occurred. Other
                than the skip field, this configures resuming after
                the failing parser has returned.
418
    """
419
    __slots__ = ['whitespace', 'comment', 'literalws', 'tokens', 'filter', 'error', 'skip',
420
                 'resume', 'drop']
eckhart's avatar
eckhart committed
421

422
423
424
425
    def __init__(self):
        self.whitespace = WHITESPACE_TYPES['vertical']  # type: str
        self.comment = ''     # type: str
        self.literalws = {'right'}  # type: Collection[str]
eckhart's avatar
eckhart committed
426
        self.tokens = set()   # type: Collection[str]
427
        self.filter = dict()  # type: Dict[str, str]
eckhart's avatar
eckhart committed
428
        self.error = dict()   # type: Dict[str, List[Tuple[ReprType, ReprType]]]
429
        self.skip = dict()    # type: Dict[str, List[Union[unrepr, str]]]
430
        self.resume = dict()  # type: Dict[str, List[Union[unrepr, str]]]
431
        self.drop = set()     # type: Set[str]
432
433
434
435
436
437
438
439
440
441
442
443

    def __getitem__(self, key):
        return getattr(self, key)

    def __setitem__(self, key, value):
        assert hasattr(self, key)
        setattr(self, key, value)

    def keys(self):
        return self.__dict__.keys()


eckhart's avatar
eckhart committed
444
class EBNFCompilerError(CompilerError):
445
    """Error raised by `EBNFCompiler` class. (Not compilation errors
446
    in the strict sense, see `CompilationError` in module ``dsl.py``)"""
447
448
449
    pass


450
class EBNFCompiler(Compiler):
451
452
    """
    Generates a Parser from an abstract syntax tree of a grammar specified
453
    in EBNF-Notation.
454
455
456
457
458
459
460

    Instances of this class must be called with the root-node of the
    abstract syntax tree from an EBNF-specification of a formal language.
    The returned value is the Python-source-code of a Grammar class for
    this language that can be used to parse texts in this language.
    See classes `parser.Compiler` and `parser.Grammar` for more information.

461
    Additionally, class EBNFCompiler provides helper methods to generate
462
463
464
465
466
    code-skeletons for a preprocessor, AST-transformation and full
    compilation of the formal language. These method's names start with
    the prefix `gen_`.

    Attributes:
467
        current_symbols:  During compilation, a list containing the root
468
469
470
471
                node of the currently compiled definition as first element
                and then the nodes of the symbols that are referred to in
                the currently compiled definition.

472
        rules:  Dictionary that maps rule names to a list of Nodes that
473
474
475
476
477
478
                contain symbol-references in the definition of the rule.
                The first item in the list is the node of the rule-
                definition itself. Example:

                           `alternative = a | b`

479
                Now `[node.content for node in self.rules['alternative']]`
480
481
                yields `['alternative = a | b', 'a', 'b']`

482
        symbols:  A mapping of symbol names to their first usage (not
483
484
                their definition!) in the EBNF source.

485
        variables:  A set of symbols names that are used with the
486
487
488
489
                Pop or Retrieve operator. Because the values of these
                symbols need to be captured they are called variables.
                See `test_parser.TestPopRetrieve` for an example.

490
        recursive:  A set of symbols that are used recursively and
491
492
                therefore require a `Forward`-operator.

493
        definitions:  A dictionary of definitions. Other than `rules`
494
495
                this maps the symbols to their compiled definienda.

496
        deferred_tasks:  A list of callables that is filled during
497
498
499
500
501
                compilatation, but that will be executed only after
                compilation has finished. Typically, it contains
                sementatic checks that require information that
                is only available upon completion of compilation.

502
        root_symbol: The name of the root symbol.
503

504
505
506
507
508
509
510
511
512
513
514
515
        directives:  A record of all directives and their default values.

        defined_directives:  A set of all directives that have already been
                defined. With the exception of those directives contained
                in EBNFCompiler.REPEATABLE_DIRECTIVES, directives must only
                be defined once.

        consumed_custom_errors:  A set of symbols for which a custom error
                has been defined and(!) consumed during compilation. This
                allows to add a compiler error in those cases where (i) an
                error message has been defined but will never used or (ii)
                an error message is accidently used twice. For examples, see
516
517
518
519
                `test_ebnf.TestErrorCustomization`.

        consumed_skip_rules: The same as `consumed_custom_errors` only for
                in-series-resume-rules (aka 'skip-rules') for Series-parsers.
520
521
522

        re_flags:  A set of regular expression flags to be added to all
                regular expressions found in the current parsing process
523

eckhart's avatar
eckhart committed
524
525
526
527
        grammar_name:  The name of the grammar to be compiled

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

528
529
530
        grammar_id: a unique id for every compiled grammar. (Required for
                disambiguation of of thread local variables storing
                compiled texts.)
531
532
    """
    COMMENT_KEYWORD = "COMMENT__"
533
    WHITESPACE_KEYWORD = "WSP_RE__"
Eckhart Arnold's avatar
Eckhart Arnold committed
534
    RAW_WS_KEYWORD = "WHITESPACE__"
535
    WHITESPACE_PARSER_KEYWORD = "wsp__"
536
    DROP_WHITESPACE_PARSER_KEYWORD = "dwsp__"
537
    RESUME_RULES_KEYWORD = "resume_rules__"
538
    SKIP_RULES_SUFFIX = '_skip__'
eckhart's avatar
eckhart committed
539
540
541
    ERR_MSG_SUFFIX = '_err_msg__'
    RESERVED_SYMBOLS = {WHITESPACE_KEYWORD, RAW_WS_KEYWORD, COMMENT_KEYWORD,
                        RESUME_RULES_KEYWORD, ERR_MSG_SUFFIX}
542
    AST_ERROR = "Badly structured syntax tree. " \
Eckhart Arnold's avatar
Eckhart Arnold committed
543
                "Potentially due to erroneous AST transformation."
544
545
546
547
    PREFIX_TABLE = {'§': 'Required',
                    '&': 'Lookahead', '!': 'NegativeLookahead',
                    '-&': 'Lookbehind', '-!': 'NegativeLookbehind',
                    '::': 'Pop', ':': 'Retrieve'}
548
    REPEATABLE_DIRECTIVES = {'tokens'}
549

550

eckhart's avatar
eckhart committed
551
    def __init__(self, grammar_name="DSL", grammar_source=""):
552
        self.grammar_id = 0
eckhart's avatar
eckhart committed
553
554
        super(EBNFCompiler, self).__init__()  # calls the _reset()-method
        self.set_grammar_name(grammar_name, grammar_source)
555

556

557
    def _reset(self):
558
        super(EBNFCompiler, self)._reset()
559
        self._result = ''           # type: str
560
        self.re_flags = set()       # type: Set[str]
561
562
563
        self.rules = OrderedDict()  # type: OrderedDict[str, List[Node]]
        self.current_symbols = []   # type: List[Node]
        self.symbols = {}           # type: Dict[str, Node]
564
565
        self.variables = set()      # type: Set[str]
        self.recursive = set()      # type: Set[str]
566
        self.definitions = {}       # type: Dict[str, str]
567
        self.deferred_tasks = []    # type: List[Callable]
568
        self.root_symbol = ""       # type: str
569
570
571
        self.directives = EBNFDirectives()   # type: EBNFDirectives
        self.defined_directives = set()      # type: Set[str]
        self.consumed_custom_errors = set()  # type: Set[str]
572
        self.consumed_skip_rules = set()     # type: Set[str]
573
574
        self.grammar_id += 1

575

Eckhart Arnold's avatar
Eckhart Arnold committed
576
    @property
577
    def result(self) -> str:
Eckhart Arnold's avatar
Eckhart Arnold committed
578
579
        return self._result

eckhart's avatar
eckhart committed
580
581
582
583
584
585
586
587
588
589
590
591

    def set_grammar_name(self, grammar_name: str = "", grammar_source: str = ""):
        """
        Changes the grammar name and source.

        The grammar name and the source text are metadata that do not affect the
        compilation process. It is used to name and annotate the output.
        Returns `self`.
        """
        assert grammar_name == "" or re.match(r'\w+\Z', grammar_name)
        if not grammar_name and re.fullmatch(r'[\w/:\\]+', grammar_source):
            grammar_name = os.path.splitext(os.path.basename(grammar_source))[0]
Eckhart Arnold's avatar
Eckhart Arnold committed
592
        self.grammar_name = grammar_name or "NameUnknown"
eckhart's avatar
eckhart committed
593
594
595
596
        self.grammar_source = load_if_file(grammar_source)
        return self


597
    # methods for generating skeleton code for preprocessor, transformer, and compiler
598

599
    def gen_preprocessor_skeleton(self) -> str:
600
601
602
603
        """
        Returns Python-skeleton-code for a preprocessor-function for
        the previously compiled formal language.
        """
604
        name = self.grammar_name + "Preprocessor"
605
        return "def %s(text):\n    return text, lambda i: i\n" % name \
606
               + PREPROCESSOR_FACTORY.format(NAME=self.grammar_name)
607

608

609
    def gen_transformer_skeleton(self) -> str:
610
611
612
613
        """
        Returns Python-skeleton-code for the AST-transformation for the
        previously compiled formal language.
        """
614
        if not self.rules:
Eckhart Arnold's avatar
Eckhart Arnold committed
615
616
            raise EBNFCompilerError('Compiler must be run before calling '
                                    '"gen_transformer_Skeleton()"!')
617
        tt_name = self.grammar_name + '_AST_transformation_table'
di68kap's avatar
di68kap committed
618
        transtable = [tt_name + ' = {',
Eckhart Arnold's avatar
Eckhart Arnold committed
619
                      '    # AST Transformations for the ' + self.grammar_name + '-grammar']
620
        transtable.append('    "<": flatten_anonymous_nodes,')
621
        for name in self.rules:
eckhart's avatar
eckhart committed
622
            transformations = '[]'
623
624
625
626
627
            # rule = self.definitions[name]
            # if rule.startswith('Alternative'):
            #     transformations = '[replace_or_reduce]'
            # elif rule.startswith('Synonym'):
            #     transformations = '[reduce_single_child]'
eckhart's avatar
eckhart committed
628
            transtable.append('    "' + name + '": %s,' % transformations)
629
        # transtable.append('    ":Token": reduce_single_child,')
eckhart's avatar
eckhart committed
630
        transtable += ['    "*": replace_by_single_child', '}', '']
631
        transtable += [TRANSFORMER_FACTORY.format(NAME=self.grammar_name, ID=self.grammar_id)]
632
633
        return '\n'.join(transtable)

634

635
    def gen_compiler_skeleton(self) -> str:
636
637
638
639
        """
        Returns Python-skeleton-code for a Compiler-class for the
        previously compiled formal language.
        """
640
        if not self.rules:
641
642
            raise EBNFCompilerError('Compiler has not been run before calling '
                                    '"gen_Compiler_Skeleton()"!')
643
        compiler = ['class ' + self.grammar_name + 'Compiler(Compiler):',
eckhart's avatar
eckhart committed
644
645
                    '    """Compiler for the abstract-syntax-tree of a '
                    + self.grammar_name + ' source file.',
646
                    '    """', '',
eckhart's avatar
eckhart committed
647
648
649
                    '    def __init__(self):',
                    '        super(' + self.grammar_name + 'Compiler, self).__init__()',
                    '',
650
651
652
                    '    def _reset(self):',
                    '        super()._reset()',
                    '        # initialize your variables here, not in the constructor!']
653
        for name in self.rules:
eckhart's avatar
eckhart committed
654
            method_name = visitor_name(name)
655
            if name == self.root_symbol:
656
                compiler += ['    def ' + method_name + '(self, node):',
657
                             '        return self.fallback_compiler(node)', '']
658
            else:
di68kap's avatar
di68kap committed
659
                compiler += ['    # def ' + method_name + '(self, node):',
eckhart's avatar
eckhart committed
660
                             '    #     return node', '']
661
        compiler += [COMPILER_FACTORY.format(NAME=self.grammar_name, ID=self.grammar_id)]
662
        return '\n'.join(compiler)
663

664
    def verify_transformation_table(self, transtable):
665
666
667
668
669
670
        """
        Checks for symbols that occur in the transformation-table but have
        never been defined in the grammar. Usually, this kind of
        inconsistency results from an error like a typo in the transformation
        table.
        """
671
        assert self._dirty_flag
672
        table_entries = set(expand_table(transtable).keys()) - {'*', '<', '>', '~'}
673
674
675
676
677
678
        symbols = self.rules.keys()
        messages = []
        for entry in table_entries:
            if entry not in symbols and not entry.startswith(":"):
                messages.append(Error(('Symbol "%s" is not defined in grammar %s but appears in '
                                       'the transformation table!') % (entry, self.grammar_name),
679
                                      0, Error.UNDEFINED_SYMBOL_IN_TRANSTABLE_WARNING))
680
681
        return messages

682
683
684
685
686
687
688
    def verify_compiler(self, compiler):
        """
        Checks for on_XXXX()-methods that occur in the compiler, although XXXX
        has never been defined in the grammar. Usually, this kind of
        inconsistency results from an error like a typo in the compiler-code.
        """
        pass  # TODO: add verification code here
689

690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712

    def _check_rx(self, node: Node, rx: str) -> str:
        """
        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.
        """
        flags = self.re_flags | {'x'} if rx.find('\n') >= 0 else self.re_flags
        if flags:
            rx = "(?%s)%s" % ("".join(flags), rx)
        try:
            re.compile(rx)
        except Exception as re_error:
            self.tree.new_error(node, "malformed regular expression %s: %s" %
                                (repr(rx), str(re_error)))
        return rx


    def _extract_regex(self, node: Node) -> str:
        """Extracts regular expression string from regexp-Node."""
        value = node.content.strip("~")
        if value[0] + value[-1] in {'""', "''"}:
            value = escape_re(value[1:-1])
713
        elif value[0] + value[-1] == '//' and value != '//':
714
715
716
717
            value = self._check_rx(node, value[1:-1])
        return value


eckhart's avatar
eckhart committed
718
    def _gen_search_rule(self, nd: Node) -> ReprType:
719
720
721
        """Generates a search rule, which can be either a string for simple
        string search or a regular expression from the nodes content. Returns
        an empty string in case the node is neither regexp nor literal.
722
        """
723
        if nd.tag_name == 'regexp':
724
            return unrepr("re.compile(r'%s')" % self._extract_regex(nd))
725
        elif nd.tag_name == 'literal':
726
727
728
729
            s = nd.content.strip()
            return s.strip('"') if s[0] == '"' else s.strip("'")
        return ''

730
731
732
733
734
735
736
    def _gen_search_list(self, nodes: Sequence[Node]) -> List[Union[unrepr, str]]:
        search_list = []  # type: List[Union[unrepr, str]]
        for child in nodes:
            rule = self._gen_search_rule(child)
            search_list.append(rule if rule else unrepr(child.content.strip()))
        return search_list

737

738
739
740
741
742
    def assemble_parser(self, definitions: List[Tuple[str, str]], root_node: Node) -> str:
        """
        Creates the Python code for the parser after compilation of
        the EBNF-Grammar
        """
743
744
745
746
747
748
749
750
751
752

        # execute deferred tasks, for example semantic checks that cannot
        # be done before the symbol table is complete

        for task in self.deferred_tasks:
            task()

        # provide for capturing of symbols that are variables, i.e. the
        # value of will be retrieved at some point during the parsing process

753
754
755
        if self.variables:
            for i in range(len(definitions)):
                if definitions[i][0] in self.variables:
756
                    definitions[i] = (definitions[i][0], 'Capture(%s)' % definitions[i][1])
757

758
759
        # add special fields for Grammar class

760
761
        definitions.append((self.WHITESPACE_PARSER_KEYWORD,
                            'Whitespace(%s)' % self.WHITESPACE_KEYWORD))
762
763
764
        if DROP_WSPC in self.directives.drop:
            definitions.append((self.DROP_WHITESPACE_PARSER_KEYWORD,
                                'DropWhitespace(%s)' % self.WHITESPACE_KEYWORD))
765
        definitions.append((self.WHITESPACE_KEYWORD,
eckhart's avatar
eckhart committed
766
767
                            ("mixin_comment(whitespace=" + self.RAW_WS_KEYWORD
                             + ", comment=" + self.COMMENT_KEYWORD + ")")))
768
769
        definitions.append((self.RAW_WS_KEYWORD, "r'{}'".format(self.directives.whitespace)))
        definitions.append((self.COMMENT_KEYWORD, "r'{}'".format(self.directives.comment)))
770
771
772

        # prepare and add resume-rules

eckhart's avatar
eckhart committed
773
        resume_rules = dict()  # type: Dict[str, List[ReprType]]
774
        for symbol, raw_rules in self.directives.resume.items():
775
776
777
778
779
            refined_rules = []
            for rule in raw_rules:
                if isinstance(rule, unrepr) and rule.s.isidentifier():
                    try:
                        nd = self.rules[rule.s][0].children[1]
780
                        refined = self._gen_search_rule(nd)
781
782
783
784
785
786
787
788
789
790
791
                    except IndexError:
                        refined = ""
                    if refined:
                        refined_rules.append(refined)
                    else:
                        self.tree.new_error(nd, 'Symbol "%s" cannot be used in resume rule, since'
                                                ' it represents neither literal nor regexp!')
                else:
                    refined_rules.append(rule)
            resume_rules[symbol] = refined_rules
        definitions.append((self.RESUME_RULES_KEYWORD, repr(resume_rules)))
792

eckhart's avatar
eckhart committed
793
794
        # prepare and add customized error-messages

795
        for symbol, err_msgs in self.directives.error.items():
eckhart's avatar
eckhart committed
796
            custom_errors = []  # type: List[Tuple[ReprType, ReprType]]
eckhart's avatar
eckhart committed
797
798
799
800
801
802
803
804
805
806
            for search, message in err_msgs:
                if isinstance(search, unrepr) and search.s.isidentifier():
                    try:
                        nd = self.rules[search.s][0].children[1]
                        search = self._gen_search_rule(nd)
                    except IndexError:
                        search = ''
                custom_errors.append((search, message))
            definitions.append((symbol + self.ERR_MSG_SUFFIX, repr(custom_errors)))

807
808
809
810
811
812
        for symbol in self.directives.error.keys():
            if symbol not in self.consumed_custom_errors:
                def_node = self.rules[symbol][0]
                self.tree.new_error(
                    def_node, 'Customized error message for symbol "{}" will never be used, '
                    'because the mandatory marker "§" appears nowhere in its definiendum!'
813
814
815
816
817
                    .format(symbol), Error.UNUSED_ERROR_HANDLING_WARNING)

        # prepare and add skip-rules

        for symbol, skip in self.directives.skip.items():
Eckhart Arnold's avatar
Eckhart Arnold committed
818
            skip_rules = []  # type: List[ReprType]
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
            for search in skip:
                if isinstance(search, unrepr) and search.s.isidentifier():
                    try:
                        nd = self.rules[search.s][0].children[1]
                        search = self._gen_search_rule(nd)
                    except IndexError:
                        search = ''
                skip_rules.append(search)
            definitions.append((symbol + self.SKIP_RULES_SUFFIX, repr(skip_rules)))

        for symbol in self.directives.error.keys():
            if symbol not in self.consumed_skip_rules:
                def_node = self.rules[symbol][0]
                self.tree.new_error(
                    def_node, '"Skip-rules" for symbol "{}" will never be used, '
                    'because the mandatory marker "§" appears nowhere in its definiendum!'
                    .format(symbol), Error.UNUSED_ERROR_HANDLING_WARNING)
836

837
838
        # prepare parser class header and docstring and
        # add EBNF grammar to the doc string of the parser class
839

840
        article = 'an ' if self.grammar_name[0:1] in "AaEeIiOoUu" else 'a '  # what about 'hour', 'universe' etc.?
841
        show_source = get_config_value('add_grammar_source_to_parser_docstring')
eckhart's avatar
eckhart committed
842
843
844
845
        declarations = ['class ' + self.grammar_name
                        + 'Grammar(Grammar):',
                        'r"""Parser for ' + article + self.grammar_name
                        + ' source file'
846
                        + ('. Grammar:' if self.grammar_source and show_source else '.')]
eckhart's avatar
eckhart committed
847
        definitions.append(('parser_initialization__', '["upon instantiation"]'))
848
        definitions.append(('static_analysis_pending__', 'True'))
849
        if self.grammar_source:
850
            definitions.append(('source_hash__',
851
                                '"%s"' % md5(self.grammar_source, __version__)))
852
            declarations.append('')
853
854
            if show_source:
                declarations += [line for line in self.grammar_source.split('\n')]
855
856
857
858
859
            while declarations[-1].strip() == '':
                declarations = declarations[:-1]
        declarations.append('"""')

        # turn definitions into declarations in reverse order
860

861
        self.root_symbol = definitions[0][0] if definitions else ""
862
863
864
865
866
867
868
869
        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]
870
871
872
873
874
875

        # check for symbols used but never defined

        defined_symbols = set(self.rules.keys()) | self.RESERVED_SYMBOLS
        for symbol in self.symbols:
            if symbol not in defined_symbols:
eckhart's avatar
eckhart committed
876
                self.tree.new_error(self.symbols[symbol],
eckhart's avatar
eckhart committed
877
                                    "Missing definition for symbol '%s'" % symbol)
878
                # root_node.error_flag = True
879
880
881

        # check for unconnected rules

Eckhart Arnold's avatar
Eckhart Arnold committed
882
883
884
        defined_symbols.difference_update(self.RESERVED_SYMBOLS)

        def remove_connections(symbol):
885
            """Recursively removes all symbols which appear in the
eckhart's avatar
eckhart committed
886
            definiens of a particular symbol."""
Eckhart Arnold's avatar
Eckhart Arnold committed
887
888
889
890
891
892
893
            if symbol in defined_symbols:
                defined_symbols.remove(symbol)
                for related in self.rules[symbol][1:]:
                    remove_connections(str(related))

        remove_connections(self.root_symbol)
        for leftover in defined_symbols:
eckhart's avatar
eckhart committed
894
895
896
            self.tree.new_error(self.rules[leftover][0],
                                ('Rule "%s" is not connected to parser root "%s" !') %
                                (leftover, self.root_symbol), Error.WARNING)
897

898
        # set root_symbol parser and assemble python grammar definition
899

900
901
        if self.root_symbol and 'root__' not in self.rules:
            declarations.append('root__ = ' + self.root_symbol)
902
        declarations.append('')
Eckhart Arnold's avatar
Eckhart Arnold committed
903
        self._result = '\n    '.join(declarations) \
904
                       + GRAMMAR_FACTORY.format(NAME=self.grammar_name, ID=self.grammar_id)
Eckhart Arnold's avatar
Eckhart Arnold committed
905
        return self._result
906

907
908
909

    ## compilation methods

910
    def on_syntax(self, node: Node) -> str:
911
        definitions = []  # type: List[Tuple[str, str]]
912
913

        # drop the wrapping sequence node
914
        if len(node.children) == 1 and node.children[0].is_anonymous():
915
            node = node.children[0]
916
917

        # compile definitions and directives and collect definitions
918
        for nd in node.children:
919
            if nd.tag_name == "definition":
920
                definitions.append(self.compile(nd))
921
            else:
922
                assert nd.tag_name == "directive", nd.as_sxpr()
923
                self.compile(nd)
924
            # node.error_flag = max(node.error_flag, nd.error_flag)
925
        self.definitions.update(definitions)
926

927
        grammar_python_src = self.assemble_parser(definitions, node)
Eckhart Arnold's avatar
Eckhart Arnold committed
928
        if get_config_value('static_analysis') == 'early':
929
            try:
930
931
                grammar_class = compile_python_object(DHPARSER_IMPORTS + grammar_python_src,
                                                      self.grammar_name)
932
                _ = grammar_class()
933
934
                grammar_python_src = grammar_python_src.replace(
                    'static_analysis_pending__ = True', 'static_analysis_pending__ = False', 1)
Eckhart Arnold's avatar
Eckhart Arnold committed
935
936
            except NameError:
                pass  # undefined name in the grammar are already cuaght and reported
937
938
939
940
941
942
            except GrammarError as error:
                for sym, prs, err in error.errors:
                    symdef_node = self.rules[sym][0]
                    err.pos = self.rules[sym][0].pos
                    self.tree.add_error(symdef_node, err)
        return grammar_python_src
943

944

945
    def on_definition(self, node: Node) -> Tuple[str, str]:
946
        rule = node.children[0].content
947
        if rule in self.rules:
Eckhart Arnold's avatar
Eckhart Arnold committed
948
            first = self.rules[rule][0]
eckhart's avatar
eckhart committed
949
            if not first.errors:
eckhart's avatar
eckhart committed
950
                self.tree.new_error(first, 'First definition of rule "%s" '
eckhart's avatar
eckhart committed
951
                                    'followed by illegal redefinitions.' % rule)
eckhart's avatar
eckhart committed
952
            self.tree.new_error(node, 'A rule "%s" has already been defined earlier.' % rule)
953
        elif rule in EBNFCompiler.RESERVED_SYMBOLS:
eckhart's avatar
eckhart committed
954
            self.tree.new_error(node, 'Symbol "%s" is a reserved symbol.' % rule)
955
        elif not sane_parser_name(rule):
eckhart's avatar
eckhart committed
956
            self.tree.new_error(node, 'Illegal symbol "%s". Symbols must not start or '
eckhart's avatar
eckhart committed
957
                                ' end with a doube underscore "__".' % rule)
958
        elif rule in self.directives.tokens:
eckhart's avatar
eckhart committed
959
            self.tree.new_error(node, 'Symbol "%s" has already been defined as '
eckhart's avatar
eckhart committed
960
                                'a preprocessor token.' % rule)
961
        elif keyword.iskeyword(rule):
eckhart's avatar
eckhart committed
962
            self.tree.new_error(node, 'Python keyword "%s" may not be used as a symbol. '
eckhart's avatar
eckhart committed
963
                                % rule + '(This may change in the future.)')
964
        try:
965
966
            self.current_symbols = [node]
            self.rules[rule] = self.current_symbols
967
            defn = self.compile(node.children[1])
968
            if rule in self.variables:
969
                defn = 'Capture(%s)' % defn
970
                self.variables.remove(rule)
971
972
973
            elif defn.find("(") < 0:
                # assume it's a synonym, like 'page = REGEX_PAGE_NR'
                defn = 'Synonym(%s)' % defn
974
        except TypeError as error:
975
976
977
978
            from traceback import extract_tb
            trace = str(extract_tb(error.__traceback__)[-1])
            errmsg = "%s (TypeError: %s; %s)\n%s" \
                     % (EBNFCompiler.AST_ERROR, str(error), trace, node.as_sxpr())
eckhart's avatar
eckhart committed
979
            self.tree.new_error(node, errmsg)
980
            rule, defn = rule + ':error', '"' + errmsg + '"'
Eckhart Arnold's avatar
Eckhart Arnold committed
981
        return rule, defn
982

983

984
    def on_directive(self, node: Node) -> str:
985
        key = node.children[0].content
986
        assert key not in self.directives.tokens
987

988
        if key not in self.REPEATABLE_DIRECTIVES and not key.endswith('_error'):
989
            if key in self.defined_directives:
eckhart's avatar
eckhart committed
990
                self.tree.new_error(node, 'Directive "%s" has already been defined earlier. '
eckhart's avatar
eckhart committed
991
                                    % key + 'Later definition will be ignored!',
992
                                    code=Error.REDEFINED_DIRECTIVE)
993
994