transform.py 42.3 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',
Eckhart Arnold's avatar
Eckhart Arnold committed
48
           'always',
49
           'is_named',
50
           'update_attr',
51
           'replace_by_single_child',
Eckhart Arnold's avatar
Eckhart Arnold committed
52
           'reduce_single_child',
53
           'replace_or_reduce',
Eckhart Arnold's avatar
Eckhart Arnold committed
54
           'change_tag_name',
55
           'collapse',
56
           'collapse_if',
57
           'replace_content',
58
           'replace_content_by',
59
           'normalize_whitespace',
60
           'move_adjacent',
61
           'left_associative',
62
           'lean_left',
63
           'apply_if',
eckhart's avatar
eckhart committed
64
           'apply_unless',
65
           'traverse_locally',
66
           'is_anonymous',
67
68
69
           'is_insignificant_whitespace',
           'contains_only_whitespace',
           'is_any_kind_of_whitespace',
70
71
           'is_empty',
           '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
           'remove_anonymous_tokens',
94
           '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
def key_tag_name(node: Node) -> str:
Eckhart Arnold's avatar
Eckhart Arnold committed
235
236
237
238
    """
    Returns the tag name of the node as key of selecting transformations
    from the transformation table.
    """
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
    """
Eckhart Arnold's avatar
Eckhart Arnold committed
246
    Traverses the syntax 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
#######################################################################
#
# 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.
#
#######################################################################


Eckhart Arnold's avatar
Eckhart Arnold committed
388
389
390
391
392
def always(context: List[Node]) -> bool:
    """Always returns True, no matter that the state of the context."""
    return True


393
def is_single_child(context: List[Node]) -> bool:
eckhart's avatar
eckhart committed
394
    """Returns ``True`` if the current node does not have any siblings."""
395
396
397
398
    return len(context[-2].children) == 1


def is_named(context: List[Node]) -> bool:
eckhart's avatar
eckhart committed
399
    """Returns ``True`` if the current node's parser is a named parser."""
400
    return not context[-1].is_anonymous()
401
402
403


def is_anonymous(context: List[Node]) -> bool:
eckhart's avatar
eckhart committed
404
    """Returns ``True`` if the current node's parser is an anonymous parser."""
405
    return context[-1].is_anonymous()
406
407


408
def is_insignificant_whitespace(context: List[Node]) -> bool:
di68kap's avatar
di68kap committed
409
    """Returns ``True`` for whitespace and comments defined with the
410
    ``@comment``-directive."""
411
    return context[-1].tag_name == WHITESPACE_PTYPE
412
413


414
RX_WHITESPACE = re.compile(r'\s+')
415
416
417
418
419
420
421


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."""
422
    content = context[-1].content
423
    return bool(not content or RX_WHITESPACE.match(content))
424
425
426
427
428
429
430
431
432
433


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)


434
def is_empty(context: List[Node]) -> bool:
eckhart's avatar
eckhart committed
435
    """Returns ``True`` if the current node's content is empty."""
436
437
438
    return not context[-1].result


Eckhart Arnold's avatar
Eckhart Arnold committed
439
440
441
442
443
# DEPRECATED, because name is too ambiguous
# def is_expendable(context: List[Node]) -> bool:
#     """Returns ``True`` if the current node either is a node containing
#     whitespace or an empty node."""
#     return is_empty(context) or is_insignificant_whitespace(context)
444
445


446
@transformation_factory(collections.abc.Set)
447
def is_token(context: List[Node], tokens: AbstractSet[str] = frozenset()) -> bool:
448
449
    """
    Checks whether the last node in the context has `ptype == TOKEN_PTYPE`
450
451
    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
452
    any token is a match.
453
454
    """
    node = context[-1]
455
    return node.tag_name == TOKEN_PTYPE and (not tokens or node.content in tokens)
456
457


458
@transformation_factory(collections.abc.Set)
459
460
461
462
463
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


464
465
466
467
468
469
@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
470
471
@transformation_factory(collections.abc.Set)
def matches_re(context: List[Node], patterns: AbstractSet[str]) -> bool:
472
473
    """
    Returns true, if the node's tag_name matches one of the regular
Eckhart Arnold's avatar
Eckhart Arnold committed
474
475
476
477
478
479
480
481
482
    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
483
@transformation_factory
484
def has_content(context: List[Node], regexp: str) -> bool:
485
486
487
488
489
490
491
492
    """
    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('$'):
        regexp += "$"
493
494
495
    return bool(re.match(regexp, context[-1].content))


496
@transformation_factory(collections.abc.Set)
497
def has_parent(context: List[Node], tag_name_set: AbstractSet[str]) -> bool:
498
499
500
501
    """
    Checks whether a node with one of the given tag names appears somewhere
     in the context before the last node in the context.
     """
502
    for i in range(2, len(context) + 1):
503
504
505
506
507
508
509
510
511
512
513
514
        if context[-i].tag_name in tag_name_set:
            return True
    return False


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


515
def update_attr(node: Node, child: Node):
516
517
518
519
    """
    Adds all attributes from `child` to `node`.This is needed, in order
    to keep the attributes if the child node is going to be eliminated.
    """
520
521
522
523
524
525
526
527
528
    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


529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
def swap_attributes(node: Node, other: Node):
    """
    Exchanges the attributes between node and other. This might be
    needed when rearanging trees.
    """
    NA = node.attr_active()
    OA = other.attr_active()
    if NA or OA:
        save = node._xml_attr if NA else None
        if OA:
            node._xml_attr = other._xml_attr
        elif NA:
            node._xml_attr = None
        if NA:
            other._xml_attr = node._xml_attr
        elif OA:
            other._xml_attr = None


548
def _replace_by(node: Node, child: Node):
549
550
551
    """
    Replaces node's contents by child's content including the tag name.
    """
552
553
554
555
    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)
556
557
        # parser names must not be overwritten, else: child.parser.name = node.parser.name
    node.result = child.result
558
    update_attr(node, child)
559
560


561
def _reduce_child(node: Node, child: Node):
562
563
564
    """
    Sets node's results to the child's result, keeping node's tag_name.
    """
565
    node.result = child.result
566
    update_attr(node, child)
567
568


569
570
571
572
573
574
575
576
577
578
579
580
#######################################################################
#
# 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)
#
#######################################################################

Eckhart Arnold's avatar
Eckhart Arnold committed
581
582
# DEPRECATED
# def flatten_anonymous_nodes(context: List[Node]):
583
#     """
Eckhart Arnold's avatar
Eckhart Arnold committed
584
585
586
587
588
#     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.
589
#     """
Eckhart Arnold's avatar
Eckhart Arnold committed
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
#     node = context[-1]
#     if node.children:
#         new_result = []  # type: List[Node]
#         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
#                 node.result = child.result
#                 update_attr(node, child)
#                 return
#             elif child.is_anonymous():
#                 node.result = child.result
#                 update_attr(node, child)
#                 return
#         node.result = tuple(new_result)
614
615


616
617
def replace_by_single_child(context: List[Node]):
    """
618
619
620
    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.
621
622
623
    """
    node = context[-1]
    if len(node.children) == 1:
624
        _replace_by(node, node.children[0])
625
626


Eckhart Arnold's avatar
Eckhart Arnold committed
627
def reduce_single_child(context: List[Node]):
628
    """
629
    Reduces a single branch node by transferring the result of its
630
    immediate descendant to this node, but keeping this node's parser entry.
631
632
    Reduction only takes place if the last node in the context has
    exactly one child.
633
634
635
    """
    node = context[-1]
    if len(node.children) == 1:
636
        _reduce_child(node, node.children[0])
637
638


639
@transformation_factory(collections.abc.Callable)
eckhart's avatar
eckhart committed
640
def replace_or_reduce(context: List[Node], condition: Callable = is_named):
641
642
    """
    Replaces node by a single child, if condition is met on child,
643
644
    otherwise (i.e. if the child is anonymous) reduces the child.
    """
645
    node = context[-1]
646
    if len(node.children) == 1:
di68kap's avatar
di68kap committed
647
        child = node.children[0]
Eckhart Arnold's avatar
Eckhart Arnold committed
648
        if condition(context):
649
            _replace_by(node, child)
650
        else:
651
            _reduce_child(node, child)
652
653


Eckhart Arnold's avatar
Eckhart Arnold committed
654
655
@transformation_factory(str)
def change_tag_name(context: List[Node], tag_name: str, restriction: Callable = always):
656
    """
Eckhart Arnold's avatar
Eckhart Arnold committed
657
    Changes the tag name of a node.
658
659

    Parameters:
Eckhart Arnold's avatar
Eckhart Arnold committed
660
661
        restriction: A function of the context that returns False in cases
                where the tag name shall not be exchanged
662
        context: the context where the parser shall be replaced
Eckhart Arnold's avatar
Eckhart Arnold committed
663
        tag_name: The new tag name.
664
    """
Eckhart Arnold's avatar
Eckhart Arnold committed
665
666
667
    if restriction(context):
        node = context[-1]
        node.tag_name = tag_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


Eckhart Arnold's avatar
Eckhart Arnold committed
878
879
def left_associative(context: List[Node]):
    """
880
    Rearranges a flat node with infix operators into a left associative tree.
Eckhart Arnold's avatar
Eckhart Arnold committed
881
    """
882
    node = context[-1]
883
884
885
886
887
888
889
890
891
892
    if len(node.children) >= 3:
        assert (len(node.children) + 1) % 2 == 0
        rest = list(node.result)
        left, rest = rest[0], rest[1:]
        while rest:
            infix, right, rest = rest[0], rest[1], rest[2:]
            assert not infix.children
            assert infix.tag_name[0:1] != ":"
            left = Node(infix.tag_name, (left, right))
        node.result = left
893

Eckhart Arnold's avatar
Eckhart Arnold committed
894

895
@transformation_factory(collections.abc.Set)
896
def lean_left(context: List[Node], operators: AbstractSet[str]):
897
    """
898
899
900
901
902
903
904
905
906
    Turns a right leaning tree into a left leaning tree:
    (op1 a (op2 b c))  ->  (op2 (op1 a b) c)
    If a left-associative operator is parsed with a right-recursive
    parser, `lean_left' can be used to rearrange the tree structure
    so that it properly reflects the order of association.

    ATTENTION: This transformation function moves forward recursively,
    so grouping nodes must not be eliminated during traversal! This
    must be done in a second pass.
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
    """
    node = context[-1]
    assert node.children and len(node.children) == 2
    assert node.tag_name in operators
    right = node.children[1]
    if right.tag_name in operators:
        assert right.children and len(right.children) == 2
        a, b, c = node.children[0], right.children[0], right.children[1]
        op1 = node.tag_name
        op2 = right.tag_name
        right.result = (a, b)
        right.tag_name = op1
        node.result = (right, c)
        node.tag_name = op2
        swap_attributes(node, right)
922
923
        # continue recursively on the left branch
        lean_left([right], operators)
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949


# @transformation_factory(collections.abc.Set)
# def left_associative_tree(context: List[Node], operators: AbstracSet[str]):
#     """
#     Rearranges a right associative tree into a left associative tree.
#     ``operators`` is a list of tag names of nodes that shall be rearranged.
#     Other nodes will be lelft untouched.
#     """
#     node = context[-1]
#     assert node.tag_name in operators
#     right = node.children[1]
#     while right.tag_name in operators:
#         node.result = (node.children[0], right.children[0])
#         right.result = (node, right.children[1])
#         node = right
#         right = node.children[1]
#     parent = context[-2]
#     result = list(parent.result)
#     for i in range(len(result)):
#         if result[i] == contexnt[-1]:
#             result[i] = node
#             parent.result = tuple(result)
#             break
#     else:
#         assert False, "???"
Eckhart Arnold's avatar
Eckhart Arnold committed
950
951


952
953
954
955
956
#######################################################################
#
# destructive transformations:
#
# - leaves may be dropped (e.g. if deemed irrelevant)
957
# - errors of dropped leaves may be be lost
958
959
960
# - no promise that order will be preserved
#
#######################################################################
961
962


963
@transformation_factory(collections.abc.Callable)
964
def lstrip(context: List[Node], condition: Callable = contains_only_whitespace):
965
966
967
968
969
970
971
972
973
974
975
976
    """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:]


977
@transformation_factory(collections.abc.Callable)
978
def rstrip(context: List[Node], condition: Callable = contains_only_whitespace):
979
980
981
982
983
984
985
    """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
986
        while i > 0 and condition(context + [node.children[i - 1]]):
987
988
989
990
991
            i -= 1
        if i < L:
            node.result = node.children[:i]


992
@transformation_factory(collections.abc.Callable)
993
def strip(context: List[Node], condition: Callable = contains_only_whitespace):
994
995
996
997
998
    """Removes leading and trailing child-nodes that fulfill a given condition."""
    lstrip(context, condition)
    rstrip(context, condition)


999
@transformation_factory  # (slice)
1000
def keep_children(context: List[Node], section: slice = slice(None)):
1001
    """Keeps only child-nodes which fall into a slice of the result field."""
1002
    node = context[-1]
1003
    if node.children: