transform.py 41.4 KB
Newer Older
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
# transform.py - transformation functions for converting the
#                concrete into the abstract syntax tree
#
# 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.
18
19


20
21
22
"""
Module ``transform`` contains the functions for transforming the
concrete syntax tree (CST) into an abstract syntax tree (AST).
23

24
25
26
As these functions are very generic, they can in principle be
used for any kind of tree transformations, not necessarily only
for CST -> AST transformations.
27
28
"""

29

30
import collections.abc
31
import inspect
eckhart's avatar
eckhart committed
32
from functools import partial, singledispatch
33

34
from DHParser.error import Error, ErrorCode
35
from DHParser.syntaxtree import Node, WHITESPACE_PTYPE, TOKEN_PTYPE, PLACEHOLDER, RootNode, parse_sxpr, flatten_sxpr
36
from DHParser.toolkit import issubtype, isgenerictype, expand_table, smart_list, re, typing
37
from typing import AbstractSet, Any, ByteString, Callable, cast, Container, Dict, \
38
    Tuple, List, Sequence, Union, Text, Generic
39

40
41
__all__ = ('TransformationDict',
           'TransformationProc',
42
           'TransformationFunc',
43
44
45
           'ConditionFunc',
           'KeyFunc',
           'transformation_factory',
46
47
           'key_tag_name',
           'traverse',
48
           'is_named',
49
           'update_attr',
50
           'flatten_anonymous_nodes',
51
           'replace_by_single_child',
Eckhart Arnold's avatar
Eckhart Arnold committed
52
           'reduce_single_child',
53
           'replace_or_reduce',
54
55
           'replace_parser',
           'collapse',
56
           'collapse_if',
57
           # 'merge_children',
58
           'replace_content',
59
           'replace_content_by',
60
           'normalize_whitespace',
61
           'move_adjacent',
62
           'apply_if',
eckhart's avatar
eckhart committed
63
           'apply_unless',
64
           'traverse_locally',
65
           'is_anonymous',
66
67
68
           'is_insignificant_whitespace',
           'contains_only_whitespace',
           'is_any_kind_of_whitespace',
69
70
71
           'is_empty',
           'is_expendable',
           'is_token',
72
           'is_one_of',
73
           'not_one_of',
Eckhart Arnold's avatar
Eckhart Arnold committed
74
           'matches_re',
75
           'has_content',
di68kap's avatar
di68kap committed
76
           'has_parent',
77
78
79
80
81
82
83
84
           'lstrip',
           'rstrip',
           'strip',
           'keep_children',
           'keep_children_if',
           'keep_tokens',
           'keep_nodes',
           'keep_content',
85
           'remove_children_if',
eckhart's avatar
eckhart committed
86
           'remove_nodes',
87
88
89
90
91
           'remove_content',
           'remove_first',
           'remove_last',
           'remove_whitespace',
           'remove_empty',
di68kap's avatar
di68kap committed
92
           'remove_anonymous_empty',
93
94
           'remove_anonymous_expendables',
           'remove_anonymous_tokens',
95
96
           'remove_expendables',
           'remove_brackets',
97
98
           'remove_infix_operator',
           'remove_single_child',
99
100
101
102
           'remove_tokens',
           'flatten',
           'forbid',
           'require',
103
           'assert_content',
104
           'error_on',
di68kap's avatar
di68kap committed
105
106
           'assert_has_children',
           'peek')
107
108


109
TransformationProc = Callable[[List[Node]], None]
Eckhart Arnold's avatar
Eckhart Arnold committed
110
TransformationDict = Dict[str, Sequence[Callable]]
111
TransformationFunc = Union[Callable[[Node], Any], partial]
Eckhart Arnold's avatar
Eckhart Arnold committed
112
ProcessingTableType = Dict[str, Union[Sequence[Callable], TransformationDict]]
113
114
ConditionFunc = Callable  # Callable[[List[Node]], bool]
KeyFunc = Callable[[Node], str]
eckhart's avatar
eckhart committed
115
CriteriaType = Union[int, str, Callable]
116
117


118
def transformation_factory(t1=None, t2=None, t3=None, t4=None, t5=None):
119
120
    """
    Creates factory functions from transformation-functions that
121
    dispatch on the first parameter after the context parameter.
122
123

    Decorating a transformation-function that has more than merely the
Eckhart Arnold's avatar
Eckhart Arnold committed
124
    ``context``-parameter with ``transformation_factory`` creates a
125
    function with the same name, which returns a partial-function that
126
    takes just the context-parameter.
127
128
129
130
131
132
133
134
135

    Additionally, there is some some syntactic sugar for
    transformation-functions that receive a collection as their second
    parameter and do not have any further parameters. In this case a
    list of parameters passed to the factory function will be converted
    into a collection.

    Main benefit is readability of processing tables.

136
137
    Usage::

eckhart's avatar
eckhart committed
138
        @transformation_factory(AbstractSet[str])
139
        def remove_tokens(context, tokens):
140
            ...
141
142
143

    or, alternatively::

144
        @transformation_factory
145
        def remove_tokens(context, tokens: AbstractSet[str]):
146
147
            ...

148
149
    Example::

150
        trans_table = { 'expression': remove_tokens('+', '-') }
151
152
153

    instead of::

154
        trans_table = { 'expression': partial(remove_tokens, tokens={'+', '-'}) }
155
156

    Parameters:
157
        t1:  type of the second argument of the transformation function,
158
159
            only necessary if the transformation functions' parameter list
            does not have type annotations.
160
161
    """

162
163
164
165
    def type_guard(t):
        """Raises an error if type `t` is a generic type or could be mistaken
        for the type of the canonical first parameter "List[Node] of
        transformation functions. Returns `t`."""
166
167
168
169
        # if isinstance(t, GenericMeta):
        #     raise TypeError("Generic Type %s not permitted\n in transformation_factory "
        #                     "decorator. Use the equivalent non-generic type instead!"
        #                     % str(t))
Eckhart Arnold's avatar
Eckhart Arnold committed
170
171
        if isinstance(t, str):          # ensure compatibility with python versions
            t = eval(t.replace('unicode', 'str'))  # with alternative type handling.
172
        if isgenerictype(t):
173
174
175
            raise TypeError("Generic Type %s not permitted\n in transformation_factory "
                            "decorator. Use the equivalent non-generic type instead!"
                            % str(t))
176
        if issubtype(List[Node], t):
177
178
            raise TypeError("Sequence type %s not permitted\nin transformation_factory "
                            "decorator, because it could be mistaken for a base class "
179
180
181
                            "of List[Node]\nwhich is the type of the canonical first "
                            "argument of transformation functions. Try 'tuple' instead!"
                            % str(t))
182
183
        return t

184
    def decorator(f):
185
        nonlocal t1
186
187
188
189
        sig = inspect.signature(f)
        params = list(sig.parameters.values())[1:]
        if len(params) == 0:
            return f  # '@transformer' not needed w/o free parameters
190
        assert t1 or params[0].annotation != params[0].empty, \
191
            "No type information on second parameter found! Please, use type " \
eckhart's avatar
eckhart committed
192
            "annotation or provide the type information via transformer-decorator."
193
        f = singledispatch(f)
194
195
196
        p1type = params[0].annotation
        if t1 is None:
            t1 = type_guard(p1type)
197
        elif issubtype(p1type, type_guard(t1)):
198
            try:
199
                if len(params) == 1 and issubtype(p1type, Container) \
200
                        and not (issubtype(p1type, Text) or issubtype(p1type, ByteString)):
201
                    def gen_special(*args):
202
203
                        c = set(args) if issubtype(p1type, AbstractSet) else \
                            tuple(args) if issubtype(p1type, Sequence) else args
204
205
206
207
208
209
210
211
                        d = {params[0].name: c}
                        return partial(f, **d)
                    f.register(type_guard(p1type.__args__[0]), gen_special)
            except AttributeError:
                pass  # Union Type does not allow subclassing, but is not needed here
        else:
            raise TypeError("Annotated type %s is not a subclass of decorated type %s !"
                            % (str(p1type), str(t1)))
212
213
214
215
216
217

        def gen_partial(*args, **kwargs):
            d = {p.name: arg for p, arg in zip(params, args)}
            d.update(kwargs)
            return partial(f, **d)

218
        for t in (t1, t2, t3, t4, t5):
219
            if t:
220
                f.register(type_guard(t), gen_partial)
221
222
            else:
                break
223
224
        return f

225
    if isinstance(t1, type(lambda: 1)):
226
227
228
        # Provide for the case that transformation_factory has been
        # written as plain decorator and not as a function call that
        # returns the decorator proper.
229
230
        func = t1
        t1 = None
231
232
233
234
235
        return decorator(func)
    else:
        return decorator


236
237
# def key_parser_name(node: Node) -> str:
#     return node.parser.name
238
239


240
def key_tag_name(node: Node) -> str:
241
242
243
    return node.tag_name


244
def traverse(root_node: Node,
Eckhart Arnold's avatar
Eckhart Arnold committed
245
             processing_table: ProcessingTableType,
eckhart's avatar
eckhart committed
246
             key_func: KeyFunc = key_tag_name) -> None:
247
248
    """
    Traverses the snytax tree starting with the given ``node`` depth
249
    first and applies the sequences of callback-functions registered
250
    in the ``processing_table``-dictionary.
251
252
253
254
255
256
257
258
259

    The most important use case is the transformation of a concrete
    syntax tree into an abstract tree (AST). But it is also imaginable
    to employ tree-traversal for the semantic analysis of the AST.

    In order to assign sequences of callback-functions to nodes, a
    dictionary ("processing table") is used. The keys usually represent
    tag names, but any other key function is possible. There exist
    three special keys:
260

261
    - '<': always called (before any other processing function)
262
263
    - '*': called for those nodes for which no (other) processing
      function appears in the table
264
    - '>': always called (after any other processing function)
265
266
267
268
269

    Args:
        root_node (Node): The root-node of the syntax tree to be traversed
        processing_table (dict): node key -> sequence of functions that
            will be applied to matching nodes in order. This dictionary
270
271
            is interpreted as a ``compact_table``. See
            :func:`expand_table` or :func:`EBNFCompiler.EBNFTransTable`
272
        key_func (function): A mapping key_func(node) -> keystr. The default
273
            key_func yields node.tag_name.
274

275
276
    Example::

277
        table = { "term": [replace_by_single_child, flatten],
278
                  "factor, flowmarker, retrieveop": replace_by_single_child }
279
        traverse(node, table)
280

281
    """
282

283
284
285
    # Is this optimazation really needed?
    if '__cache__' in processing_table:
        # assume that processing table has already been expanded
eckhart's avatar
eckhart committed
286
        table = processing_table               # type: ProcessingTableType
eckhart's avatar
eckhart committed
287
        cache = cast(TransformationDict, processing_table['__cache__'])  # type: TransformationDict
288
    else:
289
290
        # normalize processing_table entries by turning single values
        # into lists with a single value
291
292
        table = {name: cast(Sequence[Callable], smart_list(call))
                 for name, call in list(processing_table.items())}
293
        table = expand_table(table)
294
        # substitute key for insiginificant whitespace
295
        assert '+' not in table, 'Symbol "+" in processing table is obsolete, use "<" instead'
296
297
        if '~' in table:
            if ':Whitespace' in table:
eckhart's avatar
eckhart committed
298
299
300
301
                raise AssertionError(
                    '"~" is a synonym for ":Whitespace" in the processing table. '
                    'To avoid confusion, choose either of the two, but do not use '
                    'both at the same time!')
302
303
304
305
            whitespace_transformation = table['~']
            del table['~']
            table[':Whitespace'] = whitespace_transformation
        # cache expanded table
eckhart's avatar
eckhart committed
306
307
        cache = cast(TransformationDict,
                     table.setdefault('__cache__', cast(TransformationDict, dict())))
308
309
        # change processing table in place, so its already expanded and cache filled next time
        processing_table.clear()
310
311
        processing_table.update(table)

312
    def traverse_recursive(context):
eckhart's avatar
eckhart committed
313
        nonlocal cache
314
        node = context[-1]
315
        if node.children:
316
            context.append(PLACEHOLDER)
di68kap's avatar
di68kap committed
317
            for child in node.children:
eckhart's avatar
eckhart committed
318
                context[-1] = child
319
                traverse_recursive(context)  # depth first
eckhart's avatar
eckhart committed
320
            context.pop()
321
322

        key = key_func(node)
323
324
325
        try:
            sequence = cache[key]
        except KeyError:
326
            sequence = table.get('<', []) \
327
                + table.get(key, table.get('*', [])) \
328
                + table.get('>', [])
329
330
331
            cache[key] = sequence

        for call in sequence:
332
            call(context)
333

334
    traverse_recursive([root_node])
335
336
    # assert processing_table['__cache__']

337

338
#######################################################################
339
#
340
341
# meta transformations, i.e. transformations that call other
# transformations
342
#
343
#######################################################################
344
345


eckhart's avatar
eckhart committed
346
@transformation_factory(dict)
347
def traverse_locally(context: List[Node],
eckhart's avatar
eckhart committed
348
349
                     processing_table: Dict,              # actually: ProcessingTableType
                     key_func: Callable = key_tag_name):  # actually: KeyFunc
350
351
    """
    Transforms the syntax tree starting from the last node in the context
352
353
354
355
356
357
358
    according to the given processing table. The purpose of this function is
    to apply certain transformations locally, i.e. only for those nodes that
    have the last node in the context as their parent node.
    """
    traverse(context[-1], processing_table, key_func)


359
@transformation_factory(collections.abc.Callable)
360
def apply_if(context: List[Node], transformation: Callable, condition: Callable):
361
362
363
    """
    Applies a transformation only if a certain condition is met.
    """
364
365
366
367
    if condition(context):
        transformation(context)


368
@transformation_factory(collections.abc.Callable)
eckhart's avatar
eckhart committed
369
def apply_unless(context: List[Node], transformation: Callable, condition: Callable):
370
371
372
    """
    Applies a transformation if a certain condition is *not* met.
    """
eckhart's avatar
eckhart committed
373
374
375
376
    if not condition(context):
        transformation(context)


377
378
379
380
381
382
383
384
385
386
387
388
389
390
#######################################################################
#
# conditionals that determine whether the context (or the last node in
# the context for that matter) fulfill a specific condition.
# ---------------------------------------------------------------------
#
# The context of a node is understood as a list of all parent nodes
# leading up to and including the node itself. If represented as list,
# the last element of the list is the node itself.
#
#######################################################################


def is_single_child(context: List[Node]) -> bool:
eckhart's avatar
eckhart committed
391
    """Returns ``True`` if the current node does not have any siblings."""
392
393
394
395
    return len(context[-2].children) == 1


def is_named(context: List[Node]) -> bool:
eckhart's avatar
eckhart committed
396
    """Returns ``True`` if the current node's parser is a named parser."""
397
    return not context[-1].is_anonymous()
398
399
400


def is_anonymous(context: List[Node]) -> bool:
eckhart's avatar
eckhart committed
401
    """Returns ``True`` if the current node's parser is an anonymous parser."""
402
    return context[-1].is_anonymous()
403
404


405
def is_insignificant_whitespace(context: List[Node]) -> bool:
di68kap's avatar
di68kap committed
406
    """Returns ``True`` for whitespace and comments defined with the
407
    ``@comment``-directive."""
408
    return context[-1].tag_name == WHITESPACE_PTYPE
409
410


411
RX_WHITESPACE = re.compile(r'\s+')
412
413
414
415
416
417
418


def contains_only_whitespace(context: List[Node]) -> bool:
    """Returns ``True`` for nodes that contain only whitespace regardless
    of the tag_name, i.e. nodes the content of which matches the regular
    expression /\s*/, including empty nodes. Note, that this is not true
    for anonymous whitespace nodes that contain comments."""
419
    content = context[-1].content
420
    return bool(not content or RX_WHITESPACE.match(content))
421
422
423
424
425
426
427
428
429
430


def is_any_kind_of_whitespace(context: List[Node]) -> bool:
    """Returns ``True`` for nodes that either contain only whitespace or
    are insignificant whitespace nodes, i.e. nodes with the ``tag_name``
    ``PTYPE_WHITESPACE``, including those that contain comment-text."""
    node = context[-1]
    return node.tag_name == WHITESPACE_PTYPE or RX_WHITESPACE.match(node.content)


431
def is_empty(context: List[Node]) -> bool:
eckhart's avatar
eckhart committed
432
    """Returns ``True`` if the current node's content is empty."""
433
434
435
436
    return not context[-1].result


def is_expendable(context: List[Node]) -> bool:
eckhart's avatar
eckhart committed
437
438
    """Returns ``True`` if the current node either is a node containing
    whitespace or an empty node."""
439
    return is_empty(context) or is_insignificant_whitespace(context)
440
441


442
@transformation_factory(collections.abc.Set)
443
def is_token(context: List[Node], tokens: AbstractSet[str] = frozenset()) -> bool:
444
445
    """
    Checks whether the last node in the context has `ptype == TOKEN_PTYPE`
446
447
    and it's content matches one of the given tokens. Leading and trailing
    whitespace-tokens will be ignored. In case an empty set of tokens is passed,
eckhart's avatar
eckhart committed
448
    any token is a match.
449
450
    """
    node = context[-1]
451
    return node.tag_name == TOKEN_PTYPE and (not tokens or node.content in tokens)
452
453


454
@transformation_factory(collections.abc.Set)
455
456
457
458
459
def is_one_of(context: List[Node], tag_name_set: AbstractSet[str]) -> bool:
    """Returns true, if the node's tag_name is one of the given tag names."""
    return context[-1].tag_name in tag_name_set


460
461
462
463
464
465
@transformation_factory(collections.abc.Set)
def not_one_of(context: List[Node], tag_name_set: AbstractSet[str]) -> bool:
    """Returns true, if the node's tag_name is not one of the given tag names."""
    return context[-1].tag_name not in tag_name_set


Eckhart Arnold's avatar
Eckhart Arnold committed
466
467
@transformation_factory(collections.abc.Set)
def matches_re(context: List[Node], patterns: AbstractSet[str]) -> bool:
468
469
    """
    Returns true, if the node's tag_name matches one of the regular
Eckhart Arnold's avatar
Eckhart Arnold committed
470
471
472
473
474
475
476
477
478
    expressions in `patterns`. For example, ':.*' matches all anonymous nodes.
    """
    tn = context[-1].tag_name
    for pattern in patterns:
        if re.match(pattern, tn):
            return True
    return False


eckhart's avatar
eckhart committed
479
@transformation_factory
480
def has_content(context: List[Node], regexp: str) -> bool:
481
482
483
484
485
486
487
488
    """
    Checks a node's content against a regular expression.

    In contrast to ``re.match`` the regular expression must match the complete
    string and not just the beginning of the string to succeed!
    """
    if not regexp.endswith('$'):
        regexp += "$"
489
490
491
    return bool(re.match(regexp, context[-1].content))


492
@transformation_factory(collections.abc.Set)
493
def has_parent(context: List[Node], tag_name_set: AbstractSet[str]) -> bool:
494
495
496
497
    """
    Checks whether a node with one of the given tag names appears somewhere
     in the context before the last node in the context.
     """
498
    for i in range(2, len(context) + 1):
499
500
501
502
503
504
505
506
507
508
509
510
        if context[-i].tag_name in tag_name_set:
            return True
    return False


#######################################################################
#
# utility functions (private)
#
#######################################################################


511
512
513
514
515
516
517
518
519
520
def update_attr(node: Node, child: Node):
    if hasattr(child, '_xml_attr'):
        for k, v in child.attr:
            if k in node.attr and v != node.attr[k]:
                raise ValueError('Conflicting attribute values %s and %s for key %s '
                                 'when reducing %s to %s ! Tree transformation stopped.'
                                 % (v, node.attr[k], k, str(child), str(node)))
            node.attr[k] = v


521
def _replace_by(node: Node, child: Node):
522
523
524
525
    if node.is_anonymous() or not child.is_anonymous():
        node.tag_name = child.tag_name
        # name, ptype = (node.tag_name.split(':') + [''])[:2]
        # child.parser = MockParser(name, ptype)
526
527
        # parser names must not be overwritten, else: child.parser.name = node.parser.name
    node.result = child.result
528
    update_attr(node, child)
529
530


531
def _reduce_child(node: Node, child: Node):
532
    node.result = child.result
533
    update_attr(node, child)
534
535


536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
#######################################################################
#
# rearranging transformations
#
# - tree may be rearranged (e.g.flattened)
# - nodes that are not leaves may be dropped
# - order is preserved
# - leave content is preserved (though not necessarily the leaves
#   themselves)
#
#######################################################################


# @transformation_factory(int, str, Callable)
# def replace_by_child(context: List[Node], criteria: CriteriaType=is_single_child):
#     """
#     Replaces a node by the first of its immediate descendants
#     that meets the `criteria`. The criteria can either be the
#     index of the child (counting from zero), or the tag name or
#     a boolean-valued function on the context of the child.
#     If no child matching the criteria is found, the node will
#     not be replaced.
#     With the default value for `criteria` the same semantics is
#     the same that of `replace_by_single_child`.
#     """
#     child = _pick_child(context, criteria)
#     if child:
#         _replace_by(context[-1], child)
#
#
# @transformation_factory(int, str, Callable)
# def content_from_child(context: List[Node], criteria: CriteriaType = is_single_child):
#     """
#     Reduces a node, by transferring the result of the first of its
#     immediate descendants that meets the `criteria` to this node,
#     but keeping this node's parser entry. The criteria can either
#     be the index of the child (counting from zero), or the tag
#     name or a boolean-valued function on the context of the child.
#     If no child matching the criteria is found, the node will
#     not be replaced.
#     With the default value for `criteria` this has the same semantics
#     as `content_from_single_child`.
#     """
#     child = _pick_child(context, criteria)
#     if child:
#         _reduce_child(context[-1], child)
582
583


584
def flatten_anonymous_nodes(context: List[Node]):
Eckhart Arnold's avatar
Eckhart Arnold committed
585
    """
586
587
588
589
590
    Flattens non-recursively all anonymous non-leaf children by adding
    their result to the result of the parent node. Empty anonymous children
    will be dropped altogether. If the parent node (i.e. `context[-1]) is
    anonymous itself and has only one child node, it will be replaced by
    its single child node.
Eckhart Arnold's avatar
Eckhart Arnold committed
591
    """
592
593
    node = context[-1]
    if node.children:
594
        new_result = []  # type: List[Node]
595
596
597
598
599
600
601
602
603
604
605
606
607
        for child in node.children:
            if child.is_anonymous():
                if child.children:
                    new_result.extend(child.children)
                    update_attr(node, child)
                elif child.result:
                    new_result.append(child)
            else:
                new_result.append(child)
        if len(new_result) == 1:
            child = new_result[0]
            if node.is_anonymous():
                node.tag_name = child.tag_name
Eckhart Arnold's avatar
Eckhart Arnold committed
608
                node.result = child.result
609
                update_attr(node, child)
Eckhart Arnold's avatar
Eckhart Arnold committed
610
                return
611
            elif child.is_anonymous():
Eckhart Arnold's avatar
Eckhart Arnold committed
612
                node.result = child.result
613
                update_attr(node, child)
Eckhart Arnold's avatar
Eckhart Arnold committed
614
                return
615
616
617
        node.result = tuple(new_result)


618
619
def replace_by_single_child(context: List[Node]):
    """
620
621
622
    Removes single branch node, replacing it by its immediate descendant.
    Replacement only takes place, if the last node in the context has
    exactly one child.
623
624
625
    """
    node = context[-1]
    if len(node.children) == 1:
626
        _replace_by(node, node.children[0])
627
628


Eckhart Arnold's avatar
Eckhart Arnold committed
629
def reduce_single_child(context: List[Node]):
630
    """
631
    Reduces a single branch node by transferring the result of its
632
    immediate descendant to this node, but keeping this node's parser entry.
633
634
    Reduction only takes place if the last node in the context has
    exactly one child.
635
636
637
    """
    node = context[-1]
    if len(node.children) == 1:
638
        _reduce_child(node, node.children[0])
639
640


641
@transformation_factory(collections.abc.Callable)
eckhart's avatar
eckhart committed
642
def replace_or_reduce(context: List[Node], condition: Callable = is_named):
643
644
    """
    Replaces node by a single child, if condition is met on child,
645
646
    otherwise (i.e. if the child is anonymous) reduces the child.
    """
647
    node = context[-1]
648
    if len(node.children) == 1:
di68kap's avatar
di68kap committed
649
        child = node.children[0]
650
        if condition(context):   # TODO: bug here?
651
            _replace_by(node, child)
652
        else:
653
            _reduce_child(node, child)
654
655
656


@transformation_factory
657
658
659
def replace_parser(context: List[Node], name: str):
    """
    Replaces the parser of a Node with a mock parser with the given
660
661
662
    name.

    Parameters:
663
        context: the context where the parser shall be replaced
eckhart's avatar
eckhart committed
664
        name: "NAME:PTYPE" of the surrogate. The ptype is optional
665
    """
666
    node = context[-1]
667
    node.tag_name = name
668
669


670
@transformation_factory(collections.abc.Callable)
eckhart's avatar
eckhart committed
671
def flatten(context: List[Node], condition: Callable = is_anonymous, recursive: bool = True):
672
    """
eckhart's avatar
eckhart committed
673
    Flattens all children, that fulfill the given ``condition``
674
675
676
    (default: all unnamed children). Flattening means that wherever a
    node has child nodes, the child nodes are inserted in place of the
    node.
677
678

    If the parameter ``recursive`` is ``True`` the same will recursively be
679
680
681
    done with the child-nodes, first. In other words, all leaves of
    this node and its child nodes are collected in-order as direct
    children of this node.
682
683
684
685

    Applying flatten recursively will result in these kinds of
    structural transformation::

di68kap's avatar
di68kap committed
686
        (1 (+ 2) (+ 3))    ->   (1 + 2 + 3)
687
688
        (1 (+ (2 + (3))))  ->   (1 + 2 + 3)
    """
689

690
    node = context[-1]
691
    if node.children:
Eckhart Arnold's avatar
Eckhart Arnold committed
692
        new_result = []     # type: List[Node]
693
        context.append(PLACEHOLDER)
694
        for child in node.children:
eckhart's avatar
eckhart committed
695
            context[-1] = child
696
            if child.children and condition(context):
697
                if recursive:
698
                    flatten(context, condition, recursive)
699
                new_result.extend(child.children)
700
                update_attr(node, child)
701
702
            else:
                new_result.append(child)
eckhart's avatar
eckhart committed
703
        context.pop()
704
705
706
        node.result = tuple(new_result)


707
def collapse(context: List[Node]):
708
709
710
711
    """
    Collapses all sub-nodes of a node by replacing them with the
    string representation of the node. USE WITH CARE!
    """
712
    node = context[-1]
713
    node.result = node.content
714
715


716
@transformation_factory(collections.abc.Callable)
717
def collapse_if(context: List[Node], condition: Callable, target_tag: str):
718
719
    """
    (Recursively) merges the content of all adjacent child nodes that
720
721
722
723
724
    fulfil the given `condition` into a single leaf node with parser
    `target_tag`. Nodes that do not fulfil the condition will be preserved.

    >>> sxpr = '(place (abbreviation "p.") (page "26") (superscript "b") (mark ",") (page "18"))'
    >>> tree = parse_sxpr(sxpr)
725
    >>> collapse_if([tree], not_one_of({'superscript', 'subscript'}), 'text')
726
    >>> print(flatten_sxpr(tree.as_sxpr()))
727
    (place (text "p.26") (superscript "b") (text ",18"))
728
729
730

    See `test_transform.TestComplexTransformations` for examples.
    """
731

732
733
    assert isinstance(target_tag, str)  # TODO: Delete this when safe

734
    node = context[-1]
eckhart's avatar
eckhart committed
735
736
    package = []  # type: List[Node]
    result = []  # type: List[Node]
737
738
739
740

    def close_package():
        nonlocal package
        if package:
741
            s = "".join(nd.content for nd in package)
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
            result.append(Node(target_tag, s))
            package = []

    for child in node.children:
        if condition([child]):
            if child.children:
                collapse_if([child], condition, target_tag)
                for c in child.children:
                    if condition([c]):
                        package.append(c)
                    else:
                        close_package()
                        result.append(c)
                close_package()
            else:
                package.append(child)
        else:
            close_package()
            result.append(child)
    close_package()
    node.result = tuple(result)


765
@transformation_factory(collections.abc.Callable)
766
def replace_content(context: List[Node], func: Callable):  # Callable[[Node], ResultType]
767
768
    """
    Replaces the content of the node. ``func`` takes the node's result
769
770
    as an argument an returns the mapped result.
    """
771
    node = context[-1]
772
773
774
    node.result = func(node.result)


775
@transformation_factory  # (str)
776
def replace_content_by(context: List[Node], content: str):  # Callable[[Node], ResultType]
777
778
    """
    Replaces the content of the node with the given text content.
779
780
781
782
783
    """
    node = context[-1]
    node.result = content


784
def normalize_whitespace(context):
di68kap's avatar
di68kap committed
785
786
    """
    Normalizes Whitespace inside a leaf node, i.e. any sequence of
787
788
    whitespaces, tabs and linefeeds will be replaced by a single
    whitespace. Empty (i.e. zero-length) Whitespace remains empty,
di68kap's avatar
di68kap committed
789
790
    however.
    """
791
792
    node = context[-1]
    assert not node.children
793
    if is_insignificant_whitespace(context):
794
795
796
        if node.result:
            node.result = ' '
    else:
eckhart's avatar
eckhart committed
797
        node.result = re.sub(r'\s+', ' ', node.result)
798
799


di68kap's avatar
di68kap committed
800
801
802
803
804
805
806
807
808
809
def merge_whitespace(context):
    """
    Merges adjacent whitespace. UNTESTED!
    """
    node = context[-1]
    children = node.children
    new_result = []
    i = 0
    L = len(children)
    while i < L:
810
        if children[i].tag_name == WHITESPACE_PTYPE:
di68kap's avatar
di68kap committed
811
            k = i
812
            while i < L and children[k].tag_name == WHITESPACE_PTYPE:
di68kap's avatar
di68kap committed
813
814
815
816
817
818
819
820
                i += 1
            if i > k:
                children[k].result = sum(children[n].result for n in range(k, i + 1))
            new_result.append(children[k])
        i += 1
    node.result = tuple(new_result)


821
@transformation_factory(collections.abc.Callable)
822
def move_adjacent(context: List[Node], condition: Callable = is_insignificant_whitespace):
di68kap's avatar
di68kap committed
823
    """
824
    Moves adjacent nodes that fulfill the given condition to the parent node.
825
    """
826
827
828
829
830
831
832
833
834
835
836
837
838
    def join_results(a: Node, b: Node, c: Node) -> bool:
        """Joins the results of node `a` and `b` and write them to the result
        of `c` type-safely, if possible. Return True, if join was possible
        and done, False otherwise."""
        if a.children and b.children:
            c.result = cast(Tuple[Node, ...], a.result) + cast(Tuple[Node, ...], b.result)
            return True
        elif not a.children and not b.children:
            c.result = cast(str, a.result) + cast(str, b.result)
            return True
        return False


839
840
841
842
843
    node = context[-1]
    if len(context) <= 1 or not node.children:
        return
    parent = context[-2]
    children = node.children
844
    if condition([children[0]]):
845
        before = (children[0],)   # type: Tuple[Node, ...]
846
847
848
        children = children[1:]
    else:
        before = ()
849
    if children and condition([children[-1]]):
850
        after = (children[-1],)   # type: Tuple[Node, ...]
851
852
853
854
855
856
857
        children = children[:-1]
    else:
        after = tuple()

    if before or after:
        node.result = children
        for i, child in enumerate(parent.children):
858
            if id(child) == id(node):
859
860
861
                break

        # merge adjacent whitespace
eckhart's avatar
eckhart committed
862
863
        prevN = parent.children[i - 1] if i > 0 else None
        nextN = parent.children[i + 1] if i < len(parent.children) - 1 else None
864
        if before and prevN and condition([prevN]):
865
866
867
868
            # prevN.result = prevN.result + before[0].result
            # before = ()
            if join_results(prevN, before[0], prevN):
                before = ()
869
        if after and nextN and condition([nextN]):
870
871
872
873
            # nextN.result = after[0].result + nextN.result
            # after = ()
            if join_results(after[0], nextN, nextN):
                after = ()
874

875
        parent.result = parent.children[:i] + before + (node,) + after + parent.children[i+1:]
876
877


878
879
880
881
882
#######################################################################
#
# destructive transformations:
#
# - leaves may be dropped (e.g. if deemed irrelevant)
883
# - errors of dropped leaves may be be lost
884
885
886
# - no promise that order will be preserved
#
#######################################################################
887
888


889
@transformation_factory(collections.abc.Callable)
890
def lstrip(context: List[Node], condition: Callable = contains_only_whitespace):
891
892
893
894
895
896
897
898
899
900
901
902
    """Recursively removes all leading child-nodes that fulfill a given condition."""
    node = context[-1]
    i = 1
    while i > 0 and node.children:
        lstrip(context + [node.children[0]], condition)
        i, L = 0, len(node.children)
        while i < L and condition(context + [node.children[i]]):
            i += 1
        if i > 0:
            node.result = node.children[i:]


903
@transformation_factory(collections.abc.Callable)
904
def rstrip(context: List[Node], condition: Callable = contains_only_whitespace):
905
906
907
908
909
910
911
    """Recursively removes all leading nodes that fulfill a given condition."""
    node = context[-1]
    i, L = 0, len(node.children)
    while i < L and node.children:
        rstrip(context + [node.children[-1]], condition)
        L = len(node.children)
        i = L
eckhart's avatar
eckhart committed
912
        while i > 0 and condition(context + [node.children[i - 1]]):
913
914
915
916
917
            i -= 1
        if i < L:
            node.result = node.children[:i]


918
@transformation_factory(collections.abc.Callable)
919
def strip(context: List[Node], condition: Callable = contains_only_whitespace):
920
921
922
923
924
    """Removes leading and trailing child-nodes that fulfill a given condition."""
    lstrip(context, condition)
    rstrip(context, condition)


925
@transformation_factory  # (slice)
926
def keep_children(context: List[Node], section: slice = slice(None)):
927
    """Keeps only child-nodes which fall into a slice of the result field."""
928
    node = context[-1]
929
    if node.children:
930
        node.result = node.children[section]
931
932


933
@transformation_factory(collections.abc.Callable)
934
935
936
937
938
939
940
def keep_children_if(context: List[Node], condition: Callable):
    """Removes all children for which `condition()` returns `True`."""
    node = context[-1]
    if node.children:
        node.result = tuple(c for c in node.children if condition(context + [c]))


941
@transformation_factory(collections.abc.Set)
eckhart's avatar
eckhart committed
942
def keep_tokens(context: List[Node], tokens: AbstractSet[str] = frozenset()):
943
944
945
946
947
948
    """Removes any among a particular set of tokens from the immediate
    descendants of a node. If ``tokens`` is the empty set, all tokens
    are removed."""
    keep_children_if(context, partial(is_token, tokens=tokens))


949
@transformation_factory(collections.abc.Set)
950
951
952
953
954
955
956
957
958
959
960
def keep_nodes(context: List[Node], tag_names: AbstractSet[str]):
    """Removes children by tag name."""
    keep_children_if(context, partial(is_one_of, tag_name_set=tag_names))


@transformation_factory
def keep_content(context: List[Node], regexp: str):
    """Removes children depending on their string value."""
    keep_children_if(context, partial(has_content, regexp=regexp))


961
@transformation_factory(collections.abc.Callable)
962
def remove_children_if(context: List[Node], condition: Callable):
963
964
965
966
    """Removes all children for which `condition()` returns `True`."""
    node = context[-1]
    if node.children:
        node.result = tuple(c for c in node.children if not condition(context + [c]))
eckhart's avatar
eckhart committed
967
    pass
968

eckhart's avatar
eckhart committed
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
# @transformation_factory(Callable)
# def remove_children(context: List[Node],
#                     condition: Callable = TRUE_CONDITION,
#                     section: slice = slice(None)):
#     """Removes all nodes from a slice of the result field if the function
#     `condition(child_node)` evaluates to `True`."""
#     node = context[-1]
#     if node.children:
#         c = node.children
#         N = len(c)
#         rng = range(*section.indices(N))
#         node.result = tuple(c[i] for i in range(N)
#                             if i not in rng or not condition(context + [c[i]]))
#         # selection = []
#         # for i in range(N):
#         #     context.append(c[i])
#         #     if not i in rng or not condition(context):
#         #         selection.append(c[i])
#         #     context.pop()
#         # if len(selection) != c:
#         #     node.result = tuple(selection)
990
991


992
remove_whitespace = remove_children_if(is_insignificant_whitespace)
993
# partial(remove_children_if, condition=is_whitespace)
994
remove_empty = remove_children_if(is_empty)
di68kap's avatar
di68kap committed
995
remove_anonymous_empty = remove_children_if(lambda ctx: is_empty(ctx) and is_anonymous(ctx))
996
997
998
999
remove_expendables = remove_children_if(is_expendable)
# partial(remove_children_if, condition=is_expendable)
remove_anonymous_expendables = remove_children_if(lambda ctx: is_anonymous(ctx)
                                                  and is_expendable(ctx))
1000
remove_anonymous_tokens = remove_children_if(lambda ctx: is_token(ctx) and is_anonymous(ctx))
1001
1002
1003
# remove_first = apply_if(keep_children(slice(1, None)), lambda ctx: len(ctx[-1].children) > 1)
# remove_last = apply_if(keep_children(slice(None, -1)), lambda ctx: len(ctx[-1].children) > 1)
# remove_brackets = apply_if(keep_children(slice(1, -1)), lambda ctx: len(ctx[-1].children) >= 2)