ebnf.py 61.6 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
Eckhart Arnold's avatar
Eckhart Arnold committed
31
from typing import Callable, Dict, List, Set, Tuple, Sequence, Union, Optional, Any, cast
32

eckhart's avatar
eckhart committed
33
from DHParser.compile import CompilerError, Compiler, compile_source, visitor_name
eckhart's avatar
eckhart committed
34
from DHParser.configuration import THREAD_LOCALS, get_config_value
35
from DHParser.error import Error
Eckhart Arnold's avatar
Eckhart Arnold committed
36
37
38
from DHParser.parse import Grammar, mixin_comment, mixin_noempty, Forward, RegExp, \
    DropWhitespace, NegativeLookahead, Alternative, Series, Option, OneOrMore, ZeroOrMore, \
    Token, GrammarError
39
from DHParser.preprocess import nil_preprocessor, PreprocessorFunc
eckhart's avatar
eckhart committed
40
from DHParser.syntaxtree import Node, WHITESPACE_PTYPE, TOKEN_PTYPE
eckhart's avatar
eckhart committed
41
from DHParser.toolkit import load_if_file, escape_re, md5, sane_parser_name, re, expand_table, \
42
    unrepr, compile_python_object, DHPARSER_PARENTDIR, NEVER_MATCH_PATTERN
eckhart's avatar
eckhart committed
43
from DHParser.transform import TransformationFunc, traverse, remove_brackets, \
Eckhart Arnold's avatar
Eckhart Arnold committed
44
    reduce_single_child, replace_by_single_child, remove_whitespace, remove_empty, \
eckhart's avatar
eckhart committed
45
    remove_tokens, flatten, forbid, assert_content
46
from DHParser.versionnumber import __version__
Eckhart Arnold's avatar
Eckhart Arnold committed
47

eckhart's avatar
eckhart committed
48

49
__all__ = ('get_ebnf_preprocessor',
50
51
52
53
           'get_ebnf_grammar',
           'get_ebnf_transformer',
           'get_ebnf_compiler',
           'EBNFGrammar',
54
           'EBNFTransform',
Eckhart Arnold's avatar
Eckhart Arnold committed
55
           'EBNFCompilerError',
56
           'EBNFCompiler',
57
           'grammar_changed',
eckhart's avatar
eckhart committed
58
           'compile_ebnf',
59
           'PreprocessorFactoryFunc',
60
61
           'ParserFactoryFunc',
           'TransformerFactoryFunc',
62
           'CompilerFactoryFunc')
63
64


65
66
67
68
69
70
71
72
73
74
75
76
77
########################################################################
#
# source code support
#
########################################################################


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

78
79
if r'{dhparser_parentdir}' not in sys.path:
    sys.path.append(r'{dhparser_parentdir}')
80
81
82
83
84

try:
    import regex as re
except ImportError:
    import re
85
from DHParser import start_logging, suspend_logging, resume_logging, is_filename, load_if_file, \\
86
87
88
89
    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, \\
90
    grammar_changed, last_value, counterpart, PreprocessorFunc, is_empty, remove_if, \\
91
92
93
    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, \\
di68kap's avatar
di68kap committed
94
    replace_by_children, remove_empty, remove_tokens, flatten, is_insignificant_whitespace, \\
95
    merge_adjacent, collapse, collapse_children_if, replace_content, WHITESPACE_PTYPE, TOKEN_PTYPE, \\
Eckhart Arnold's avatar
Eckhart Arnold committed
96
    remove_nodes, remove_content, remove_brackets, change_tag_name, remove_anonymous_tokens, \\
97
98
99
    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, \\
100
101
    error_on, recompile_grammar, left_associative, lean_left, set_config_value, \\
    get_config_value, XML_SERIALIZATION, SXPRESSION_SERIALIZATION, COMPACT_SERIALIZATION, \\
102
103
    JSON_SERIALIZATION, access_thread_locals, access_presets, finalize_presets, ErrorCode, \\
    RX_NEVER_MATCH
eckhart's avatar
eckhart committed
104
'''.format(dhparser_parentdir=DHPARSER_PARENTDIR)
105
106


Eckhart Arnold's avatar
Eckhart Arnold committed
107
108
109
110
111
112
113
########################################################################
#
# EBNF scanning
#
########################################################################


114
def get_ebnf_preprocessor() -> PreprocessorFunc:
eckhart's avatar
eckhart committed
115
116
117
118
119
    """
    Returns the preprocessor function for the EBNF compiler.
    As of now, no preprocessing is needed for EBNF-sources. Therefore,
    just a dummy function is returned.
    """
120
    return nil_preprocessor
Eckhart Arnold's avatar
Eckhart Arnold committed
121
122
123
124
125
126
127
128


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

129

di68kap's avatar
di68kap committed
130
class EBNFGrammar(Grammar):
eckhart's avatar
eckhart committed
131
    r"""
eckhart's avatar
eckhart committed
132
133
134
135
136
    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
137
138
139
    @ drop       = whitespace                       # do not include whitespace in concrete syntax tree

    #: top-level
eckhart's avatar
eckhart committed
140
141
142
143
144

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

145
146
    #: components

eckhart's avatar
eckhart committed
147
148
149
150
151
152
153
154
155
156
157
158
159
    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

160
161
    #: flow-operators

eckhart's avatar
eckhart committed
162
163
164
165
    flowmarker = "!"  | "&"                         # '!' negative lookahead, '&' positive lookahead
               | "-!" | "-&"                        # '-' negative lookbehind, '-&' positive lookbehind
    retrieveop = "::" | ":"                         # '::' pop, ':' retrieve

166
167
    #: groups

eckhart's avatar
eckhart committed
168
169
170
171
172
173
    group      = "(" §expression ")"
    unordered  = "<" §expression ">"                # elements of expression in arbitrary order
    oneormore  = "{" expression "}+"
    repetition = "{" §expression "}"
    option     = "[" §expression "]"

174
175
    #: leaf-elements

eckhart's avatar
eckhart committed
176
    symbol     = /(?!\d)\w+/~                       # e.g. expression, factor, parameter_list
177
178
179
180
    literal    = /"(?:(?<!\\)\\"|[^"])*?"/~         # e.g. "(", '+', 'while'
               | /'(?:(?<!\\)\\'|[^'])*?'/~         # whitespace following literals will be ignored tacitly.
    plaintext  = /`(?:(?<!\\)\\`|[^`])*?`/~         # like literal but does not eat whitespace
    regexp     = /\/(?:(?<!\\)\\(?:\/)|[^\/])*?\//~     # e.g. /\w+/, ~/#.*(?:\n|$)/~
eckhart's avatar
eckhart committed
181
182
183
    whitespace = /~/~                               # insignificant whitespace

    EOF = !/./
184
    """
di68kap's avatar
di68kap committed
185
    expression = Forward()
eckhart's avatar
eckhart committed
186
    source_hash__ = "82a7c668f86b83f86515078e6c9093ed"
187
    static_analysis_pending__ = []
eckhart's avatar
eckhart committed
188
    parser_initialization__ = ["upon instantiation"]
di68kap's avatar
di68kap committed
189
190
    COMMENT__ = r'#.*(?:\n|$)'
    WHITESPACE__ = r'\s*'
191
    WSP_RE__ = mixin_comment(whitespace=WHITESPACE__, comment=COMMENT__)
192
    wsp__ = DropWhitespace(WSP_RE__)
di68kap's avatar
di68kap committed
193
    EOF = NegativeLookahead(RegExp('.'))
194
    whitespace = Series(RegExp('~'), wsp__)
195
    regexp = Series(RegExp('/(?:(?<!\\\\)\\\\(?:/)|[^/])*?/'), wsp__)
eckhart's avatar
eckhart committed
196
    plaintext = Series(RegExp('`(?:(?<!\\\\)\\\\`|[^`])*?`'), wsp__)
197
198
    literal = Alternative(Series(RegExp('"(?:(?<!\\\\)\\\\"|[^"])*?"'), wsp__),
                          Series(RegExp("'(?:(?<!\\\\)\\\\'|[^'])*?'"), wsp__))
199
200
201
202
203
204
205
206
207
208
    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
209
210
211
212
                                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)
213
214
215
    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
216
217
                       Alternative(regexp, literal, symbol),
                       ZeroOrMore(Series(Series(Token(","), wsp__), Alternative(regexp, literal, symbol))), mandatory=1)
218
    definition = Series(symbol, Series(Token("="), wsp__), expression, mandatory=1)
eckhart's avatar
eckhart committed
219
    syntax = Series(Option(Series(wsp__, RegExp(''))), ZeroOrMore(Alternative(definition, directive)), EOF, mandatory=2)
di68kap's avatar
di68kap committed
220
221
222
    root__ = syntax


223
def grammar_changed(grammar_class, grammar_source: str) -> bool:
224
225
    """
    Returns ``True`` if ``grammar_class`` does not reflect the latest
Eckhart Arnold's avatar
Eckhart Arnold committed
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
    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
244
        m = re.search(r'class \w*\(Grammar\)', pycode)
Eckhart Arnold's avatar
Eckhart Arnold committed
245
246
247
248
249
250
251
252
253
254
        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__


255
def get_ebnf_grammar() -> EBNFGrammar:
Eckhart Arnold's avatar
Eckhart Arnold committed
256
    try:
257
        grammar = THREAD_LOCALS.ebnf_grammar_singleton
Eckhart Arnold's avatar
Eckhart Arnold committed
258
        return grammar
259
    except AttributeError:
260
261
        THREAD_LOCALS.ebnf_grammar_singleton = EBNFGrammar()
        return THREAD_LOCALS.ebnf_grammar_singleton
Eckhart Arnold's avatar
Eckhart Arnold committed
262
263
264
265
266
267
268
269
270


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


271
EBNF_AST_transformation_table = {
272
    # AST Transformations for EBNF-grammar
273
    "<":
274
        [remove_empty],  # remove_whitespace
275
    "syntax":
276
        [],  # otherwise '"*": replace_by_single_child' would be applied
277
    "directive, definition":
eckhart's avatar
eckhart committed
278
        [flatten, remove_tokens('@', '=', ',')],
Eckhart Arnold's avatar
Eckhart Arnold committed
279
    "expression":
280
        [replace_by_single_child, flatten, remove_tokens('|')],  # remove_infix_operator],
281
    "term":
282
283
        [replace_by_single_child, flatten],  # supports both idioms:
                                             # "{ factor }+" and "factor { factor }"
284
    "factor, flowmarker, retrieveop":
285
        replace_by_single_child,
286
    "group":
287
        [remove_brackets, replace_by_single_child],
288
289
    "unordered":
        remove_brackets,
290
    "oneormore, repetition, option":
Eckhart Arnold's avatar
Eckhart Arnold committed
291
        [reduce_single_child, remove_brackets,
292
         forbid('repetition', 'option', 'oneormore'), assert_content(r'(?!§)(?:.|\n)*')],
293
    "symbol, literal, regexp":
Eckhart Arnold's avatar
Eckhart Arnold committed
294
        reduce_single_child,
295
    (TOKEN_PTYPE, WHITESPACE_PTYPE):
Eckhart Arnold's avatar
Eckhart Arnold committed
296
        reduce_single_child,
297
    "*":
298
        replace_by_single_child
299
300
}

301

Eckhart Arnold's avatar
Eckhart Arnold committed
302
def EBNFTransform() -> TransformationFunc:
303
    return partial(traverse, processing_table=EBNF_AST_transformation_table.copy())
di68kap's avatar
di68kap committed
304

eckhart's avatar
eckhart committed
305

306
def get_ebnf_transformer() -> TransformationFunc:
307
    try:
308
        transformer = THREAD_LOCALS.EBNF_transformer_singleton
309
    except AttributeError:
310
311
        THREAD_LOCALS.EBNF_transformer_singleton = EBNFTransform()
        transformer = THREAD_LOCALS.EBNF_transformer_singleton
312
    return transformer
Eckhart Arnold's avatar
Eckhart Arnold committed
313
314
315
316
317
318
319
320


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

321

322
PreprocessorFactoryFunc = Callable[[], PreprocessorFunc]
323
ParserFactoryFunc = Callable[[], Grammar]
324
TransformerFactoryFunc = Callable[[], TransformationFunc]
325
326
CompilerFactoryFunc = Callable[[], Compiler]

327
328
329
PREPROCESSOR_FACTORY = '''
def get_preprocessor() -> PreprocessorFunc:
    return {NAME}Preprocessor
330
331
332
333
'''


GRAMMAR_FACTORY = '''
334
def get_grammar() -> {NAME}Grammar:
335
    """Returns a thread/process-exclusive {NAME}Grammar-singleton."""
336
    THREAD_LOCALS = access_thread_locals()    
337
    try:
338
        grammar = THREAD_LOCALS.{NAME}_{ID:08d}_grammar_singleton
339
    except AttributeError:
340
        THREAD_LOCALS.{NAME}_{ID:08d}_grammar_singleton = {NAME}Grammar()
eckhart's avatar
eckhart committed
341
        if hasattr(get_grammar, 'python_src__'):
342
343
            THREAD_LOCALS.{NAME}_{ID:08d}_grammar_singleton.python_src__ = get_grammar.python_src__
        grammar = THREAD_LOCALS.{NAME}_{ID:08d}_grammar_singleton
344
    return grammar
345
346
347
348
'''


TRANSFORMER_FACTORY = '''
349
350
351
def Create{NAME}Transformer() -> TransformationFunc:
    """Creates a transformation function that does not share state with other
    threads or processes."""
352
353
    return partial(traverse, processing_table={NAME}_AST_transformation_table.copy())

354
def get_transformer() -> TransformationFunc:
355
    """Returns a thread/process-exclusive transformation function."""
356
    THREAD_LOCALS = access_thread_locals()
357
    try:
358
        transformer = THREAD_LOCALS.{NAME}_{ID:08d}_transformer_singleton
359
    except AttributeError:
360
361
        THREAD_LOCALS.{NAME}_{ID:08d}_transformer_singleton = Create{NAME}Transformer()
        transformer = THREAD_LOCALS.{NAME}_{ID:08d}_transformer_singleton
362
    return transformer
363
364
365
366
'''


COMPILER_FACTORY = '''
eckhart's avatar
eckhart committed
367
def get_compiler() -> {NAME}Compiler:
368
    """Returns a thread/process-exclusive {NAME}Compiler-singleton."""
369
    THREAD_LOCALS = access_thread_locals()
370
    try:
371
        compiler = THREAD_LOCALS.{NAME}_{ID:08d}_compiler_singleton
372
    except AttributeError:
373
374
        THREAD_LOCALS.{NAME}_{ID:08d}_compiler_singleton = {NAME}Compiler()
        compiler = THREAD_LOCALS.{NAME}_{ID:08d}_compiler_singleton
375
    return compiler
376
377
'''

Eckhart Arnold's avatar
Eckhart Arnold committed
378

379
380
381
382
WHITESPACE_TYPES = {'horizontal': r'[\t ]*',  # default: horizontal
                    'linefeed': r'[ \t]*\n?(?!\s*\n)[ \t]*',
                    'vertical': r'\s*'}

383
384
385
DROP_TOKEN  = 'token'
DROP_WSPC   = 'whitespace'
DROP_VALUES = {DROP_TOKEN, DROP_WSPC}
386

eckhart's avatar
eckhart committed
387
388
# Representation of Python code or, rather, something that will be output as Python code
ReprType = Union[str, unrepr]
eckhart's avatar
eckhart committed
389
390


391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
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.

418
419
420
421
422
423
424
425
426
427
428
429
        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.
430
431
432
433
434
435
436
437
438

        drop:   A set that may contain the elements `DROP_TOKEN` and
                `DROP_WSP'

        super_ws(property): Cache for the "super whitespace" which
                is a regular expression that merges whitespace and
                comments. This property should only be accessed after
                the `whitespace` and `comment` field have been filled
                with the values parsed from the EBNF source.
439
    """
440
    __slots__ = ['whitespace', 'comment', 'literalws', 'tokens', 'filter', 'error', 'skip',
441
                 'resume', 'drop', '_super_ws']
eckhart's avatar
eckhart committed
442

443
444
445
446
    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
447
        self.tokens = set()   # type: Collection[str]
448
        self.filter = dict()  # type: Dict[str, str]
eckhart's avatar
eckhart committed
449
        self.error = dict()   # type: Dict[str, List[Tuple[ReprType, ReprType]]]
450
        self.skip = dict()    # type: Dict[str, List[Union[unrepr, str]]]
451
        self.resume = dict()  # type: Dict[str, List[Union[unrepr, str]]]
452
        self.drop = set()     # type: Set[str]
453
        self._super_ws = None # type: Optional[str]
454
455
456
457
458
459
460
461

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

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

462
463
464
465
466
467
    @property
    def super_ws(self):
        if self._super_ws is None:
            self._super_ws = mixin_comment(self.whitespace, self.comment)
        return self._super_ws

468
    def keys(self):
469
        return self.__slots__
470
471


eckhart's avatar
eckhart committed
472
class EBNFCompilerError(CompilerError):
473
    """Error raised by `EBNFCompiler` class. (Not compilation errors
474
    in the strict sense, see `CompilationError` in module ``dsl.py``)"""
475
476
477
    pass


478
class EBNFCompiler(Compiler):
479
480
    """
    Generates a Parser from an abstract syntax tree of a grammar specified
481
    in EBNF-Notation.
482
483
484
485
486
487
488

    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.

489
    Additionally, class EBNFCompiler provides helper methods to generate
490
491
492
493
494
    code-skeletons for a preprocessor, AST-transformation and full
    compilation of the formal language. These method's names start with
    the prefix `gen_`.

    Attributes:
495
        current_symbols:  During compilation, a list containing the root
496
497
498
499
                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.

500
        rules:  Dictionary that maps rule names to a list of Nodes that
501
502
503
504
505
506
                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`

507
                Now `[node.content for node in self.rules['alternative']]`
508
509
                yields `['alternative = a | b', 'a', 'b']`

510
        symbols:  A mapping of symbol names to their first usage (not
511
512
                their definition!) in the EBNF source.

513
        variables:  A set of symbols names that are used with the
514
515
516
517
                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.

518
        recursive:  A set of symbols that are used recursively and
519
520
                therefore require a `Forward`-operator.

521
        definitions:  A dictionary of definitions. Other than `rules`
522
523
                this maps the symbols to their compiled definienda.

524
        deferred_tasks:  A list of callables that is filled during
525
526
527
528
529
                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.

530
        root_symbol: The name of the root symbol.
531

532
533
534
535
536
537
538
539
540
541
542
543
        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
544
545
546
547
                `test_ebnf.TestErrorCustomization`.

        consumed_skip_rules: The same as `consumed_custom_errors` only for
                in-series-resume-rules (aka 'skip-rules') for Series-parsers.
548
549
550

        re_flags:  A set of regular expression flags to be added to all
                regular expressions found in the current parsing process
551

eckhart's avatar
eckhart committed
552
553
554
555
        grammar_name:  The name of the grammar to be compiled

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

556
557
558
        grammar_id: a unique id for every compiled grammar. (Required for
                disambiguation of of thread local variables storing
                compiled texts.)
559
560
    """
    COMMENT_KEYWORD = "COMMENT__"
561
    COMMENT_RX_KEYWORD = "comment_rx__"
562
    WHITESPACE_KEYWORD = "WSP_RE__"
Eckhart Arnold's avatar
Eckhart Arnold committed
563
    RAW_WS_KEYWORD = "WHITESPACE__"
564
    WHITESPACE_PARSER_KEYWORD = "wsp__"
565
    DROP_WHITESPACE_PARSER_KEYWORD = "dwsp__"
566
    RESUME_RULES_KEYWORD = "resume_rules__"
567
    SKIP_RULES_SUFFIX = '_skip__'
eckhart's avatar
eckhart committed
568
    ERR_MSG_SUFFIX = '_err_msg__'
569
570
571
    RESERVED_SYMBOLS = {COMMENT_KEYWORD, COMMENT_RX_KEYWORD, WHITESPACE_KEYWORD, RAW_WS_KEYWORD,
                        WHITESPACE_PARSER_KEYWORD, DROP_WHITESPACE_PARSER_KEYWORD,
                        RESUME_RULES_KEYWORD}
572
    AST_ERROR = "Badly structured syntax tree. " \
Eckhart Arnold's avatar
Eckhart Arnold committed
573
                "Potentially due to erroneous AST transformation."
574
575
576
577
    PREFIX_TABLE = {'§': 'Required',
                    '&': 'Lookahead', '!': 'NegativeLookahead',
                    '-&': 'Lookbehind', '-!': 'NegativeLookbehind',
                    '::': 'Pop', ':': 'Retrieve'}
578
    REPEATABLE_DIRECTIVES = {'tokens'}
579

580

eckhart's avatar
eckhart committed
581
    def __init__(self, grammar_name="DSL", grammar_source=""):
582
        self.grammar_id = 0
583
        super(EBNFCompiler, self).__init__()  # calls the reset()-method
eckhart's avatar
eckhart committed
584
        self.set_grammar_name(grammar_name, grammar_source)
585

586

587
588
    def reset(self):
        super(EBNFCompiler, self).reset()
589
        self._result = ''           # type: str
590
        self.re_flags = set()       # type: Set[str]
591
592
593
        self.rules = OrderedDict()  # type: OrderedDict[str, List[Node]]
        self.current_symbols = []   # type: List[Node]
        self.symbols = {}           # type: Dict[str, Node]
594
595
        self.variables = set()      # type: Set[str]
        self.recursive = set()      # type: Set[str]
596
        self.definitions = {}       # type: Dict[str, str]
597
        self.deferred_tasks = []    # type: List[Callable]
598
        self.root_symbol = ""       # type: str
599
600
601
        self.directives = EBNFDirectives()   # type: EBNFDirectives
        self.defined_directives = set()      # type: Set[str]
        self.consumed_custom_errors = set()  # type: Set[str]
602
        self.consumed_skip_rules = set()     # type: Set[str]
603
604
        self.grammar_id += 1

605

Eckhart Arnold's avatar
Eckhart Arnold committed
606
    @property
607
    def result(self) -> str:
Eckhart Arnold's avatar
Eckhart Arnold committed
608
609
        return self._result

eckhart's avatar
eckhart committed
610
611
612
613
614
615
616
617
618
619
620
621

    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
622
        self.grammar_name = grammar_name or "NameUnknown"
eckhart's avatar
eckhart committed
623
624
625
626
        self.grammar_source = load_if_file(grammar_source)
        return self


627
    # methods for generating skeleton code for preprocessor, transformer, and compiler
628

629
    def gen_preprocessor_skeleton(self) -> str:
630
631
632
633
        """
        Returns Python-skeleton-code for a preprocessor-function for
        the previously compiled formal language.
        """
634
        name = self.grammar_name + "Preprocessor"
635
        return "def %s(text):\n    return text, lambda i: i\n" % name \
636
               + PREPROCESSOR_FACTORY.format(NAME=self.grammar_name)
637

638

639
    def gen_transformer_skeleton(self) -> str:
640
641
642
643
        """
        Returns Python-skeleton-code for the AST-transformation for the
        previously compiled formal language.
        """
644
        if not self.rules:
Eckhart Arnold's avatar
Eckhart Arnold committed
645
646
            raise EBNFCompilerError('Compiler must be run before calling '
                                    '"gen_transformer_Skeleton()"!')
647
        tt_name = self.grammar_name + '_AST_transformation_table'
di68kap's avatar
di68kap committed
648
        transtable = [tt_name + ' = {',
Eckhart Arnold's avatar
Eckhart Arnold committed
649
                      '    # AST Transformations for the ' + self.grammar_name + '-grammar']
Eckhart Arnold's avatar
Eckhart Arnold committed
650
        transtable.append('    "<": flatten,')
651
        for name in self.rules:
eckhart's avatar
eckhart committed
652
            transformations = '[]'
653
654
655
656
657
            # 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
658
            transtable.append('    "' + name + '": %s,' % transformations)
659
        # transtable.append('    ":Token": reduce_single_child,')
eckhart's avatar
eckhart committed
660
        transtable += ['    "*": replace_by_single_child', '}', '']
661
        transtable += [TRANSFORMER_FACTORY.format(NAME=self.grammar_name, ID=self.grammar_id)]
662
663
        return '\n'.join(transtable)

664

665
    def gen_compiler_skeleton(self) -> str:
666
667
668
669
        """
        Returns Python-skeleton-code for a Compiler-class for the
        previously compiled formal language.
        """
670
        if not self.rules:
671
672
            raise EBNFCompilerError('Compiler has not been run before calling '
                                    '"gen_Compiler_Skeleton()"!')
673
        compiler = ['class ' + self.grammar_name + 'Compiler(Compiler):',
eckhart's avatar
eckhart committed
674
675
                    '    """Compiler for the abstract-syntax-tree of a '
                    + self.grammar_name + ' source file.',
676
                    '    """', '',
eckhart's avatar
eckhart committed
677
678
679
                    '    def __init__(self):',
                    '        super(' + self.grammar_name + 'Compiler, self).__init__()',
                    '',
680
681
                    '    def reset(self):',
                    '        super().reset()',
eckhart's avatar
eckhart committed
682
683
                    '        # initialize your variables here, not in the constructor!',
                    '']
684
        for name in self.rules:
eckhart's avatar
eckhart committed
685
            method_name = visitor_name(name)
686
            if name == self.root_symbol:
687
                compiler += ['    def ' + method_name + '(self, node):',
688
                             '        return self.fallback_compiler(node)', '']
689
            else:
di68kap's avatar
di68kap committed
690
                compiler += ['    # def ' + method_name + '(self, node):',
eckhart's avatar
eckhart committed
691
                             '    #     return node', '']
692
        compiler += [COMPILER_FACTORY.format(NAME=self.grammar_name, ID=self.grammar_id)]
693
        return '\n'.join(compiler)
694

695
    def verify_transformation_table(self, transtable):
696
697
698
699
700
701
        """
        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.
        """
702
        assert self._dirty_flag
703
        table_entries = set(expand_table(transtable).keys()) - {'*', '<', '>', '~'}
704
705
706
707
708
709
        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),
710
                                      0, Error.UNDEFINED_SYMBOL_IN_TRANSTABLE_WARNING))
711
712
        return messages

713
714
715
716
717
718
719
    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
720

721
722
723
724

    def _check_rx(self, node: Node, rx: str) -> str:
        """
        Checks whether the string `rx` represents a valid regular
725
726
        expression. Makes sure that multi-line regular expressions are
        prepended by the multi-line-flag. Returns the regular expression string.
727
        """
728
        # TODO: Support atomic grouping: https://stackoverflow.com/questions/13577372/do-python-regular-expressions-have-an-equivalent-to-rubys-atomic-grouping
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
        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])
745
        elif value[0] + value[-1] == '//' and value != '//':
746
747
748
749
            value = self._check_rx(node, value[1:-1])
        return value


eckhart's avatar
eckhart committed
750
    def _gen_search_rule(self, nd: Node) -> ReprType:
751
752
753
        """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.
754
        """
755
        if nd.tag_name == 'regexp':
Eckhart Arnold's avatar
Eckhart Arnold committed
756
757
758
759
            super_ws = self.directives.super_ws
            noempty_ws = mixin_noempty(super_ws)
            search_regex = self._extract_regex(nd)\
                .replace(r'\~!', noempty_ws).replace(r'\~', super_ws)
760
            return unrepr("re.compile(r'%s')" % search_regex)
761
        elif nd.tag_name == 'literal':
762
            s = nd.content[1:-1]  # remove quotation marks
763
            return unrepr("re.compile(r'(?=%s)')" % escape_re(s))
764
765
        return ''

766

767
768
769
770
771
772
773
    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

774

775
776
777
778
779
    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
        """
780
781
782
783
784
785
786
787

        # 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
di68kap's avatar
di68kap committed
788
        # value of which will be retrieved at some point during the parsing process
789

790
791
792
        if self.variables:
            for i in range(len(definitions)):
                if definitions[i][0] in self.variables:
793
                    definitions[i] = (definitions[i][0], 'Capture(%s)' % definitions[i][1])
794

795
796
        # add special fields for Grammar class

797
798
799
        if DROP_WSPC in self.directives.drop:
            definitions.append((self.DROP_WHITESPACE_PARSER_KEYWORD,
                                'DropWhitespace(%s)' % self.WHITESPACE_KEYWORD))
800
801
        definitions.append((self.WHITESPACE_PARSER_KEYWORD,
                            'Whitespace(%s)' % self.WHITESPACE_KEYWORD))
802
        definitions.append((self.WHITESPACE_KEYWORD,
eckhart's avatar
eckhart committed
803
804
                            ("mixin_comment(whitespace=" + self.RAW_WS_KEYWORD
                             + ", comment=" + self.COMMENT_KEYWORD + ")")))
805
        definitions.append((self.RAW_WS_KEYWORD, "r'{}'".format(self.directives.whitespace)))
806
807
        comment_rx = "re.compile(COMMENT__)" if self.directives.comment else "RX_NEVER_MATCH"
        definitions.append((self.COMMENT_RX_KEYWORD, comment_rx))
808
        definitions.append((self.COMMENT_KEYWORD, "r'{}'".format(self.directives.comment)))
809
810
811

        # prepare and add resume-rules

eckhart's avatar
eckhart committed
812
        resume_rules = dict()  # type: Dict[str, List[ReprType]]
813
        for symbol, raw_rules in self.directives.resume.items():
814
815
816
817
818
            refined_rules = []
            for rule in raw_rules:
                if isinstance(rule, unrepr) and rule.s.isidentifier():
                    try:
                        nd = self.rules[rule.s][0].children[1]
819
                        refined = self._gen_search_rule(nd)
820
821
822
823
824
825
826
827
828
829
830
                    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)))
831

eckhart's avatar
eckhart committed
832
833
        # prepare and add customized error-messages

834
        for symbol, err_msgs in self.directives.error.items():
eckhart's avatar
eckhart committed
835
            custom_errors = []  # type: List[Tuple[ReprType, ReprType]]
eckhart's avatar
eckhart committed
836
837
838
839
840
841
842
843
844
845
            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)))

846
847
848
849
850
851
        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!'
852
853
854
855
856
                    .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
857
            skip_rules = []  # type: List[ReprType]
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
            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)
875

876
877
        # prepare parser class header and docstring and
        # add EBNF grammar to the doc string of the parser class
878

879
        article = 'an ' if self.grammar_name[0:1] in "AaEeIiOoUu" else 'a '  # what about 'hour', 'universe' etc.?
880
        show_source = get_config_value('add_grammar_source_to_parser_docstring')
eckhart's avatar
eckhart committed
881
882
883
884
        declarations = ['class ' + self.grammar_name
                        + 'Grammar(Grammar):',
                        'r"""Parser for ' + article + self.grammar_name
                        + ' source file'
885
                        + ('. Grammar:' if self.grammar_source and show_source else '.')]
eckhart's avatar
eckhart committed
886
        definitions.append(('parser_initialization__', '["upon instantiation"]'))
887
        definitions.append(('static_analysis_pending__', '[True]'))
888
        if self.grammar_source:
889
            definitions.append(('source_hash__',
890
                                '"%s"' % md5(self.grammar_source, __version__)))
891
            declarations.append('')
892
893
            if show_source:
                declarations += [line for line in self.grammar_source.split('\n')]
894
895
896
897
898
            while declarations[-1].strip() == '':
                declarations = declarations[:-1]
        declarations.append('"""')

        # turn definitions into declarations in reverse order
899

900
        self.root_symbol = definitions[0][0] if definitions else ""
901
902
903
904
905
906
907
908
        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]
909
910
911
912
913
914

        # 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
915
                self.tree.new_error(self.symbols[symbol],
eckhart's avatar
eckhart committed
916
                                    "Missing definition for symbol '%s'" % symbol)
917
                # root_node.error_flag = True
918
919
920

        # check for unconnected rules

Eckhart Arnold's avatar
Eckhart Arnold committed
921
922
923
        defined_symbols.difference_update(self.RESERVED_SYMBOLS)

        def remove_connections(symbol):
924
            """Recursively removes all symbols which appear in the
eckhart's avatar
eckhart committed
925
            definiens of a particular symbol."""
Eckhart Arnold's avatar
Eckhart Arnold committed
926
927
928
929
930
931
932
            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
933
934
935
            self.tree.new_error(self.rules[leftover][0],
                                ('Rule "%s" is not connected to parser root "%s" !') %
                                (leftover, self.root_symbol), Error.WARNING)
936

937
        # set root_symbol parser and assemble python grammar definition
938

939
940
        if self.root_symbol and 'root__' not in self.rules:
            declarations.append('root__ = ' + self.root_symbol)
941
        declarations.append('')
Eckhart Arnold's avatar
Eckhart Arnold committed
942
        self._result = '\n    '.join(declarations) \
943
                       + GRAMMAR_FACTORY.format(NAME=self.grammar_name, ID=self.grammar_id)
Eckhart Arnold's avatar
Eckhart Arnold committed
944
        return self._result
945

946
947
948

    ## compilation methods

949
    def on_syntax(self, node: Node) -> str:
950
        definitions = []  # type: List[Tuple[str, str]]
951
952

        # drop the wrapping sequence node
953
        if len(node.children) == 1 and node.children[0].anonymous:
954
            node = node.children[0]
955
956

        # compile definitions and directives and collect definitions
957
        for nd in node.children:
958
            if nd.tag_name == "definition":
959
                definitions.append(self.compile(nd))