transform.py 39.2 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
50
           'update_attr',
           'eliminate_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
61
           'normalize_whitespace',
           'move_whitespace',
62
           'apply_if',
eckhart's avatar
eckhart committed
63
           'apply_unless',
64
           'traverse_locally',
65
           'is_anonymous',
66
67
68
69
           'is_whitespace',
           'is_empty',
           'is_expendable',
           'is_token',
70
           'is_one_of',
71
           'not_one_of',
Eckhart Arnold's avatar
Eckhart Arnold committed
72
           'matches_re',
73
           'has_content',
di68kap's avatar
di68kap committed
74
           'has_parent',
75
76
77
78
79
80
81
82
           'lstrip',
           'rstrip',
           'strip',
           'keep_children',
           'keep_children_if',
           'keep_tokens',
           'keep_nodes',
           'keep_content',
83
           'remove_children_if',
eckhart's avatar
eckhart committed
84
           'remove_nodes',
85
86
87
88
89
           'remove_content',
           'remove_first',
           'remove_last',
           'remove_whitespace',
           'remove_empty',
di68kap's avatar
di68kap committed
90
           'remove_anonymous_empty',
91
92
           'remove_anonymous_expendables',
           'remove_anonymous_tokens',
93
94
           'remove_expendables',
           'remove_brackets',
95
96
           'remove_infix_operator',
           'remove_single_child',
97
98
99
100
           'remove_tokens',
           'flatten',
           'forbid',
           'require',
101
           'assert_content',
102
           'error_on',
di68kap's avatar
di68kap committed
103
104
           'assert_has_children',
           'peek')
105
106


107
TransformationProc = Callable[[List[Node]], None]
Eckhart Arnold's avatar
Eckhart Arnold committed
108
TransformationDict = Dict[str, Sequence[Callable]]
109
TransformationFunc = Union[Callable[[Node], Any], partial]
Eckhart Arnold's avatar
Eckhart Arnold committed
110
ProcessingTableType = Dict[str, Union[Sequence[Callable], TransformationDict]]
111
112
ConditionFunc = Callable  # Callable[[List[Node]], bool]
KeyFunc = Callable[[Node], str]
eckhart's avatar
eckhart committed
113
CriteriaType = Union[int, str, Callable]
114
115


116
def transformation_factory(t1=None, t2=None, t3=None, t4=None, t5=None):
117
118
    """
    Creates factory functions from transformation-functions that
119
    dispatch on the first parameter after the context parameter.
120
121

    Decorating a transformation-function that has more than merely the
Eckhart Arnold's avatar
Eckhart Arnold committed
122
    ``context``-parameter with ``transformation_factory`` creates a
123
    function with the same name, which returns a partial-function that
124
    takes just the context-parameter.
125
126
127
128
129
130
131
132
133

    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.

134
135
    Usage::

eckhart's avatar
eckhart committed
136
        @transformation_factory(AbstractSet[str])
137
        def remove_tokens(context, tokens):
138
            ...
139
140
141

    or, alternatively::

142
        @transformation_factory
143
        def remove_tokens(context, tokens: AbstractSet[str]):
144
145
            ...

146
147
    Example::

148
        trans_table = { 'expression': remove_tokens('+', '-') }
149
150
151

    instead of::

152
        trans_table = { 'expression': partial(remove_tokens, tokens={'+', '-'}) }
153
154

    Parameters:
155
        t1:  type of the second argument of the transformation function,
156
157
            only necessary if the transformation functions' parameter list
            does not have type annotations.
158
159
    """

160
161
162
163
    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`."""
164
165
166
167
        # 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
168
169
        if isinstance(t, str):          # ensure compatibility with python versions
            t = eval(t.replace('unicode', 'str'))  # with alternative type handling.
170
        if isgenerictype(t):
171
172
173
            raise TypeError("Generic Type %s not permitted\n in transformation_factory "
                            "decorator. Use the equivalent non-generic type instead!"
                            % str(t))
174
        if issubtype(List[Node], t):
175
176
            raise TypeError("Sequence type %s not permitted\nin transformation_factory "
                            "decorator, because it could be mistaken for a base class "
177
178
179
                            "of List[Node]\nwhich is the type of the canonical first "
                            "argument of transformation functions. Try 'tuple' instead!"
                            % str(t))
180
181
        return t

182
    def decorator(f):
183
        nonlocal t1
184
185
186
187
        sig = inspect.signature(f)
        params = list(sig.parameters.values())[1:]
        if len(params) == 0:
            return f  # '@transformer' not needed w/o free parameters
188
        assert t1 or params[0].annotation != params[0].empty, \
189
            "No type information on second parameter found! Please, use type " \
eckhart's avatar
eckhart committed
190
            "annotation or provide the type information via transformer-decorator."
191
        f = singledispatch(f)
192
193
194
        p1type = params[0].annotation
        if t1 is None:
            t1 = type_guard(p1type)
195
        elif issubtype(p1type, type_guard(t1)):
196
            try:
197
                if len(params) == 1 and issubtype(p1type, Container) \
198
                        and not (issubtype(p1type, Text) or issubtype(p1type, ByteString)):
199
                    def gen_special(*args):
200
201
                        c = set(args) if issubtype(p1type, AbstractSet) else \
                            tuple(args) if issubtype(p1type, Sequence) else args
202
203
204
205
206
207
208
209
                        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)))
210
211
212
213
214
215

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

216
        for t in (t1, t2, t3, t4, t5):
217
            if t:
218
                f.register(type_guard(t), gen_partial)
219
220
            else:
                break
221
222
        return f

223
    if isinstance(t1, type(lambda: 1)):
224
225
226
        # Provide for the case that transformation_factory has been
        # written as plain decorator and not as a function call that
        # returns the decorator proper.
227
228
        func = t1
        t1 = None
229
230
231
232
233
        return decorator(func)
    else:
        return decorator


234
235
# def key_parser_name(node: Node) -> str:
#     return node.parser.name
236
237


238
def key_tag_name(node: Node) -> str:
239
240
241
    return node.tag_name


242
def traverse(root_node: Node,
Eckhart Arnold's avatar
Eckhart Arnold committed
243
             processing_table: ProcessingTableType,
eckhart's avatar
eckhart committed
244
             key_func: KeyFunc = key_tag_name) -> None:
245
246
    """
    Traverses the snytax tree starting with the given ``node`` depth
247
    first and applies the sequences of callback-functions registered
248
    in the ``processing_table``-dictionary.
249
250
251
252
253
254
255
256
257

    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:
258

259
    - '<': always called (before any other processing function)
260
261
    - '*': called for those nodes for which no (other) processing
      function appears in the table
262
    - '>': always called (after any other processing function)
263
264
265
266
267

    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
268
269
            is interpreted as a ``compact_table``. See
            :func:`expand_table` or :func:`EBNFCompiler.EBNFTransTable`
270
        key_func (function): A mapping key_func(node) -> keystr. The default
271
            key_func yields node.tag_name.
272

273
274
    Example::

275
        table = { "term": [replace_by_single_child, flatten],
276
                  "factor, flowmarker, retrieveop": replace_by_single_child }
277
        traverse(node, table)
278

279
    """
280

281
282
283
    # Is this optimazation really needed?
    if '__cache__' in processing_table:
        # assume that processing table has already been expanded
eckhart's avatar
eckhart committed
284
        table = processing_table               # type: ProcessingTableType
eckhart's avatar
eckhart committed
285
        cache = cast(TransformationDict, processing_table['__cache__'])  # type: TransformationDict
286
    else:
287
288
        # normalize processing_table entries by turning single values
        # into lists with a single value
289
290
        table = {name: cast(Sequence[Callable], smart_list(call))
                 for name, call in list(processing_table.items())}
291
        table = expand_table(table)
292
        # substitute key for insiginificant whitespace
293
        assert '+' not in table, 'Symbol "+" in processing table is obsolete, use "<" instead'
294
295
        if '~' in table:
            if ':Whitespace' in table:
eckhart's avatar
eckhart committed
296
297
298
299
                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!')
300
301
302
303
            whitespace_transformation = table['~']
            del table['~']
            table[':Whitespace'] = whitespace_transformation
        # cache expanded table
eckhart's avatar
eckhart committed
304
305
        cache = cast(TransformationDict,
                     table.setdefault('__cache__', cast(TransformationDict, dict())))
306
307
        # change processing table in place, so its already expanded and cache filled next time
        processing_table.clear()
308
309
        processing_table.update(table)

310
    def traverse_recursive(context):
eckhart's avatar
eckhart committed
311
        nonlocal cache
312
        node = context[-1]
313
        if node.children:
314
            context.append(PLACEHOLDER)
di68kap's avatar
di68kap committed
315
            for child in node.children:
eckhart's avatar
eckhart committed
316
                context[-1] = child
317
                traverse_recursive(context)  # depth first
eckhart's avatar
eckhart committed
318
            context.pop()
319
320

        key = key_func(node)
321
322
323
        try:
            sequence = cache[key]
        except KeyError:
324
            sequence = table.get('<', []) \
325
                + table.get(key, table.get('*', [])) \
326
                + table.get('>', [])
327
328
329
            cache[key] = sequence

        for call in sequence:
330
            call(context)
331

332
    traverse_recursive([root_node])
333
334
    # assert processing_table['__cache__']

335

336
#######################################################################
337
#
338
339
# meta transformations, i.e. transformations that call other
# transformations
340
#
341
#######################################################################
342
343


eckhart's avatar
eckhart committed
344
@transformation_factory(dict)
345
def traverse_locally(context: List[Node],
eckhart's avatar
eckhart committed
346
347
                     processing_table: Dict,              # actually: ProcessingTableType
                     key_func: Callable = key_tag_name):  # actually: KeyFunc
348
349
    """
    Transforms the syntax tree starting from the last node in the context
350
351
352
353
354
355
356
    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)


357
@transformation_factory(collections.abc.Callable)
358
def apply_if(context: List[Node], transformation: Callable, condition: Callable):
359
360
361
    """
    Applies a transformation only if a certain condition is met.
    """
362
363
364
365
    if condition(context):
        transformation(context)


366
@transformation_factory(collections.abc.Callable)
eckhart's avatar
eckhart committed
367
def apply_unless(context: List[Node], transformation: Callable, condition: Callable):
368
369
370
    """
    Applies a transformation if a certain condition is *not* met.
    """
eckhart's avatar
eckhart committed
371
372
373
374
    if not condition(context):
        transformation(context)


375
376
377
378
379
380
381
382
383
384
385
386
387
388
#######################################################################
#
# 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
389
    """Returns ``True`` if the current node does not have any siblings."""
390
391
392
393
    return len(context[-2].children) == 1


def is_named(context: List[Node]) -> bool:
eckhart's avatar
eckhart committed
394
    """Returns ``True`` if the current node's parser is a named parser."""
395
    return not context[-1].is_anonymous()
396
397
398


def is_anonymous(context: List[Node]) -> bool:
eckhart's avatar
eckhart committed
399
    """Returns ``True`` if the current node's parser is an anonymous parser."""
400
    return context[-1].is_anonymous()
401
402
403


def is_whitespace(context: List[Node]) -> bool:
di68kap's avatar
di68kap committed
404
    """Returns ``True`` for whitespace and comments defined with the
405
    ``@comment``-directive."""
406
    return context[-1].tag_name == WHITESPACE_PTYPE
407
408
409


def is_empty(context: List[Node]) -> bool:
eckhart's avatar
eckhart committed
410
    """Returns ``True`` if the current node's content is empty."""
411
412
413
414
    return not context[-1].result


def is_expendable(context: List[Node]) -> bool:
eckhart's avatar
eckhart committed
415
416
    """Returns ``True`` if the current node either is a node containing
    whitespace or an empty node."""
417
418
419
    return is_empty(context) or is_whitespace(context)


420
@transformation_factory(collections.abc.Set)
421
def is_token(context: List[Node], tokens: AbstractSet[str] = frozenset()) -> bool:
422
423
    """
    Checks whether the last node in the context has `ptype == TOKEN_PTYPE`
424
425
    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
426
    any token is a match.
427
428
    """
    node = context[-1]
429
    return node.tag_name == TOKEN_PTYPE and (not tokens or node.content in tokens)
430
431


432
@transformation_factory(collections.abc.Set)
433
434
435
436
437
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


438
439
440
441
442
443
@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
444
445
@transformation_factory(collections.abc.Set)
def matches_re(context: List[Node], patterns: AbstractSet[str]) -> bool:
446
447
    """
    Returns true, if the node's tag_name matches one of the regular
Eckhart Arnold's avatar
Eckhart Arnold committed
448
449
450
451
452
453
454
455
456
    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
457
@transformation_factory
458
def has_content(context: List[Node], regexp: str) -> bool:
459
460
461
462
463
464
465
466
    """
    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('$'):
        regexp += "$"
467
468
469
    return bool(re.match(regexp, context[-1].content))


470
@transformation_factory(collections.abc.Set)
471
def has_parent(context: List[Node], tag_name_set: AbstractSet[str]) -> bool:
472
473
474
475
    """
    Checks whether a node with one of the given tag names appears somewhere
     in the context before the last node in the context.
     """
476
    for i in range(2, len(context) + 1):
477
478
479
480
481
482
483
484
485
486
487
488
        if context[-i].tag_name in tag_name_set:
            return True
    return False


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


489
490
491
492
493
494
495
496
497
498
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


499
def _replace_by(node: Node, child: Node):
500
501
502
503
    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)
504
505
        # parser names must not be overwritten, else: child.parser.name = node.parser.name
    node.result = child.result
506
    update_attr(node, child)
507
508


509
def _reduce_child(node: Node, child: Node):
510
    node.result = child.result
511
    update_attr(node, child)
512
513


514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
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
#######################################################################
#
# 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)
560
561


562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
@transformation_factory
def eliminate_anonymous_nodes(context: List[Node]):
    node = context[-1]
    if node.children:
        new_result = []
        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
                new_result = [child.result]
                update_attr(node, child)
            elif child.is_anonymous():
                new_result = [child.result]
                update_attr(node, child)
        node.result = tuple(new_result)


588
589
def replace_by_single_child(context: List[Node]):
    """
590
591
592
    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.
593
594
595
    """
    node = context[-1]
    if len(node.children) == 1:
596
        _replace_by(node, node.children[0])
597
598


Eckhart Arnold's avatar
Eckhart Arnold committed
599
def reduce_single_child(context: List[Node]):
600
    """
601
    Reduces a single branch node by transferring the result of its
602
    immediate descendant to this node, but keeping this node's parser entry.
603
604
    Reduction only takes place if the last node in the context has
    exactly one child.
605
606
607
    """
    node = context[-1]
    if len(node.children) == 1:
608
        _reduce_child(node, node.children[0])
609
610


611
@transformation_factory(collections.abc.Callable)
eckhart's avatar
eckhart committed
612
def replace_or_reduce(context: List[Node], condition: Callable = is_named):
613
614
    """
    Replaces node by a single child, if condition is met on child,
615
616
    otherwise (i.e. if the child is anonymous) reduces the child.
    """
617
    node = context[-1]
618
    if len(node.children) == 1:
di68kap's avatar
di68kap committed
619
        child = node.children[0]
620
        if condition(context):   # TODO: bug here?
621
            _replace_by(node, child)
622
        else:
623
            _reduce_child(node, child)
624
625
626


@transformation_factory
627
628
629
def replace_parser(context: List[Node], name: str):
    """
    Replaces the parser of a Node with a mock parser with the given
630
631
632
    name.

    Parameters:
633
        context: the context where the parser shall be replaced
eckhart's avatar
eckhart committed
634
        name: "NAME:PTYPE" of the surrogate. The ptype is optional
635
    """
636
    node = context[-1]
637
    node.tag_name = name
638
639


640
@transformation_factory(collections.abc.Callable)
eckhart's avatar
eckhart committed
641
def flatten(context: List[Node], condition: Callable = is_anonymous, recursive: bool = True):
642
    """
eckhart's avatar
eckhart committed
643
    Flattens all children, that fulfill the given ``condition``
644
645
646
    (default: all unnamed children). Flattening means that wherever a
    node has child nodes, the child nodes are inserted in place of the
    node.
647
648

    If the parameter ``recursive`` is ``True`` the same will recursively be
649
650
651
    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.
652
653
654
655

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

di68kap's avatar
di68kap committed
656
        (1 (+ 2) (+ 3))    ->   (1 + 2 + 3)
657
658
        (1 (+ (2 + (3))))  ->   (1 + 2 + 3)
    """
659

660
    node = context[-1]
661
    if node.children:
Eckhart Arnold's avatar
Eckhart Arnold committed
662
        new_result = []     # type: List[Node]
663
        context.append(PLACEHOLDER)
664
        for child in node.children:
eckhart's avatar
eckhart committed
665
            context[-1] = child
666
            if child.children and condition(context):
667
                if recursive:
668
                    flatten(context, condition, recursive)
669
                new_result.extend(child.children)
670
                update_attr(node, child)
671
672
            else:
                new_result.append(child)
eckhart's avatar
eckhart committed
673
        context.pop()
674
675
676
        node.result = tuple(new_result)


677
def collapse(context: List[Node]):
678
679
680
681
    """
    Collapses all sub-nodes of a node by replacing them with the
    string representation of the node. USE WITH CARE!
    """
682
    node = context[-1]
683
    node.result = node.content
684
685


686
@transformation_factory(collections.abc.Callable)
687
def collapse_if(context: List[Node], condition: Callable, target_tag: str):
688
689
    """
    (Recursively) merges the content of all adjacent child nodes that
690
691
692
693
694
    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)
695
    >>> collapse_if([tree], not_one_of({'superscript', 'subscript'}), 'text')
696
    >>> print(flatten_sxpr(tree.as_sxpr()))
697
    (place (text "p.26") (superscript "b") (text ",18"))
698
699
700

    See `test_transform.TestComplexTransformations` for examples.
    """
701

702
703
    assert isinstance(target_tag, str)  # TODO: Delete this when safe

704
    node = context[-1]
eckhart's avatar
eckhart committed
705
706
    package = []  # type: List[Node]
    result = []  # type: List[Node]
707
708
709
710

    def close_package():
        nonlocal package
        if package:
711
            s = "".join(nd.content for nd in package)
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
            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)


735
@transformation_factory(collections.abc.Callable)
736
def replace_content(context: List[Node], func: Callable):  # Callable[[Node], ResultType]
737
738
    """
    Replaces the content of the node. ``func`` takes the node's result
739
740
    as an argument an returns the mapped result.
    """
741
    node = context[-1]
742
743
744
    node.result = func(node.result)


745
@transformation_factory  # (str)
746
def replace_content_by(context: List[Node], content: str):  # Callable[[Node], ResultType]
747
748
    """
    Replaces the content of the node with the given text content.
749
750
751
752
753
    """
    node = context[-1]
    node.result = content


754
def normalize_whitespace(context):
di68kap's avatar
di68kap committed
755
756
    """
    Normalizes Whitespace inside a leaf node, i.e. any sequence of
757
758
    whitespaces, tabs and linefeeds will be replaced by a single
    whitespace. Empty (i.e. zero-length) Whitespace remains empty,
di68kap's avatar
di68kap committed
759
760
    however.
    """
761
762
763
764
765
766
    node = context[-1]
    assert not node.children
    if is_whitespace(context):
        if node.result:
            node.result = ' '
    else:
eckhart's avatar
eckhart committed
767
        node.result = re.sub(r'\s+', ' ', node.result)
768
769


di68kap's avatar
di68kap committed
770
771
772
773
774
775
776
777
778
779
def merge_whitespace(context):
    """
    Merges adjacent whitespace. UNTESTED!
    """
    node = context[-1]
    children = node.children
    new_result = []
    i = 0
    L = len(children)
    while i < L:
780
        if children[i].tag_name == WHITESPACE_PTYPE:
di68kap's avatar
di68kap committed
781
            k = i
782
            while i < L and children[k].tag_name == WHITESPACE_PTYPE:
di68kap's avatar
di68kap committed
783
784
785
786
787
788
789
790
                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)


791
def move_whitespace(context):
di68kap's avatar
di68kap committed
792
793
    """
    Moves adjacent whitespace nodes to the parent node.
794
795
796
797
798
799
    """
    node = context[-1]
    if len(context) <= 1 or not node.children:
        return
    parent = context[-2]
    children = node.children
800
    if children[0].tag_name == WHITESPACE_PTYPE:
801
802
803
804
        before = (children[0],)
        children = children[1:]
    else:
        before = ()
805
    if children and children[-1].tag_name == WHITESPACE_PTYPE:
806
807
808
809
810
811
812
813
814
815
816
817
        after = (children[-1],)
        children = children[:-1]
    else:
        after = tuple()

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

        # merge adjacent whitespace
eckhart's avatar
eckhart committed
818
819
        prevN = parent.children[i - 1] if i > 0 else None
        nextN = parent.children[i + 1] if i < len(parent.children) - 1 else None
820
        if before and prevN and prevN.tag_name == WHITESPACE_PTYPE:
821
822
            prevN.result = prevN.result + before[0].result
            before = ()
823
        if after and nextN and nextN.tag_name == WHITESPACE_PTYPE:
824
825
826
            nextN.result = after[0].result + nextN.result
            after = ()

eckhart's avatar
eckhart committed
827
        parent.result = parent.children[:i] + before + (node,) + after + parent.children[i + 1:]
828
829


830
831
832
833
834
#######################################################################
#
# destructive transformations:
#
# - leaves may be dropped (e.g. if deemed irrelevant)
835
# - errors of dropped leaves may be be lost
836
837
838
# - no promise that order will be preserved
#
#######################################################################
839
840


841
@transformation_factory(collections.abc.Callable)
842
843
844
845
846
847
848
849
850
851
852
853
854
def lstrip(context: List[Node], condition: Callable = is_expendable):
    """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:]


855
@transformation_factory(collections.abc.Callable)
856
857
858
859
860
861
862
863
def rstrip(context: List[Node], condition: Callable = is_expendable):
    """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
864
        while i > 0 and condition(context + [node.children[i - 1]]):
865
866
867
868
869
            i -= 1
        if i < L:
            node.result = node.children[:i]


870
@transformation_factory(collections.abc.Callable)
eckhart's avatar
eckhart committed
871
def strip(context: List[Node], condition: Callable = is_expendable):
872
873
874
875
876
    """Removes leading and trailing child-nodes that fulfill a given condition."""
    lstrip(context, condition)
    rstrip(context, condition)


877
@transformation_factory  # (slice)
878
def keep_children(context: List[Node], section: slice = slice(None)):
879
    """Keeps only child-nodes which fall into a slice of the result field."""
880
    node = context[-1]
881
    if node.children:
882
        node.result = node.children[section]
883
884


885
@transformation_factory(collections.abc.Callable)
886
887
888
889
890
891
892
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]))


893
@transformation_factory(collections.abc.Set)
eckhart's avatar
eckhart committed
894
def keep_tokens(context: List[Node], tokens: AbstractSet[str] = frozenset()):
895
896
897
898
899
900
    """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))


901
@transformation_factory(collections.abc.Set)
902
903
904
905
906
907
908
909
910
911
912
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))


913
@transformation_factory(collections.abc.Callable)
914
def remove_children_if(context: List[Node], condition: Callable):
915
916
917
918
    """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
919
    pass
920

eckhart's avatar
eckhart committed
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
# @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)
942
943


944
945
remove_whitespace = remove_children_if(is_whitespace)
# partial(remove_children_if, condition=is_whitespace)
946
remove_empty = remove_children_if(is_empty)
di68kap's avatar
di68kap committed
947
remove_anonymous_empty = remove_children_if(lambda ctx: is_empty(ctx) and is_anonymous(ctx))
948
949
950
951
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))
952
remove_anonymous_tokens = remove_children_if(lambda ctx: is_token(ctx) and is_anonymous(ctx))
953
954
955
# 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)
956
remove_infix_operator = keep_children(slice(0, None, 2))
957
remove_single_child = apply_if(keep_children(slice(0)), lambda ctx: len(ctx[-1].children) == 1)
958
959


960
961
962
963
964
def remove_first(context: List[Node]):
    """Removes the first non-whitespace child."""
    node = context[-1]
    if node.children:
        for i, child in enumerate(node.children):
965
            if child.tag_name != WHITESPACE_PTYPE:
966
967
968
                break
        else:
            return
eckhart's avatar
eckhart committed
969
        node.result = node.children[:i] + node.children[i + 1:]
970
971
972
973
974
975
976


def remove_last(context: List[Node]):
    """Removes the last non-whitespace child."""
    node = context[-1]
    if node.children:
        for i, child in enumerate(reversed(node.children)):
977
            if child.tag_name != WHITESPACE_PTYPE:
978
979
980
981
                break
        else:
            return
        i = len(node.children) - i - 1
eckhart's avatar
eckhart committed
982
        node.result = node.children[:i] + node.children[i + 1:]
983
984
985
986
987
988
989
990


def remove_brackets(context: List[Node]):
    """Removes the first and the last non-whitespace child."""
    remove_first(context)
    remove_last(context)


991
@transformation_factory(collections.abc.Set)
eckhart's avatar
eckhart committed
992
def remove_tokens(context: List[Node], tokens: AbstractSet[str] = frozenset()):
993
    """Removes any among a particular set of tokens from the immediate
994
995
    descendants of a node. If ``tokens`` is the empty set, all tokens
    are removed."""
996
    remove_children_if(context, partial(is_token, tokens=tokens))
997
998


999
@transformation_factory(collections.abc.Set)
eckhart's avatar
eckhart committed
1000
def remove_nodes(context: List[Node], tag_names: AbstractSet[str]):
Eckhart Arnold's avatar
Eckhart Arnold committed
1001
    """Removes children by tag name."""
1002
    remove_children_if(context, partial(is_one_of, tag_name_set=tag_names))
1003
1004
1005


@transformation_factory
1006
def remove_content(context: List[Node], regexp: str):
1007
    """Removes children depending on their string value."""
1008
    remove_children_if(context, partial(has_content, regexp=regexp))
1009
1010
1011
1012


########################################################################
#
1013
# AST semantic validation functions (EXPERIMENTAL!!!)
1014
1015
1016
#
########################################################################