transform.py 44.1 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',
52
           'replace_by_children',
Eckhart Arnold's avatar
Eckhart Arnold committed
53
           'reduce_single_child',
54
           'replace_or_reduce',
Eckhart Arnold's avatar
Eckhart Arnold committed
55
           'change_tag_name',
56
           'collapse',
57
           'collapse_if',
58
           'replace_content',
59
           'replace_content_by',
60
           'normalize_whitespace',
61
           'merge_adjacent',
62
           'move_adjacent',
63
           'left_associative',
64
           'lean_left',
65
           'apply_if',
eckhart's avatar
eckhart committed
66
           'apply_unless',
67
           'traverse_locally',
68
           'is_anonymous',
69
70
71
           'is_insignificant_whitespace',
           'contains_only_whitespace',
           'is_any_kind_of_whitespace',
72
73
           'is_empty',
           'is_token',
74
           'is_one_of',
75
           'not_one_of',
Eckhart Arnold's avatar
Eckhart Arnold committed
76
           'matches_re',
77
           'has_content',
di68kap's avatar
di68kap committed
78
           'has_parent',
79
80
81
82
83
84
85
86
           'lstrip',
           'rstrip',
           'strip',
           'keep_children',
           'keep_children_if',
           'keep_tokens',
           'keep_nodes',
           'keep_content',
87
           'remove_children_if',
eckhart's avatar
eckhart committed
88
           'remove_nodes',
89
90
91
92
93
           'remove_content',
           'remove_first',
           'remove_last',
           'remove_whitespace',
           'remove_empty',
di68kap's avatar
di68kap committed
94
           'remove_anonymous_empty',
95
           'remove_anonymous_tokens',
96
           'remove_brackets',
97
           'remove_infix_operator',
98
99
100
101
           'remove_tokens',
           'flatten',
           'forbid',
           'require',
102
           'assert_content',
103
           'error_on',
di68kap's avatar
di68kap committed
104
105
           'assert_has_children',
           'peek')
106
107


108
TransformationProc = Callable[[List[Node]], None]
Eckhart Arnold's avatar
Eckhart Arnold committed
109
TransformationDict = Dict[str, Sequence[Callable]]
110
TransformationFunc = Union[Callable[[Node], Any], partial]
Eckhart Arnold's avatar
Eckhart Arnold committed
111
ProcessingTableType = Dict[str, Union[Sequence[Callable], TransformationDict]]
112
113
ConditionFunc = Callable  # Callable[[List[Node]], bool]
KeyFunc = Callable[[Node], str]
eckhart's avatar
eckhart committed
114
CriteriaType = Union[int, str, Callable]
115
116


117
def transformation_factory(t1=None, t2=None, t3=None, t4=None, t5=None):
118
119
    """
    Creates factory functions from transformation-functions that
120
    dispatch on the first parameter after the context parameter.
121
122

    Decorating a transformation-function that has more than merely the
Eckhart Arnold's avatar
Eckhart Arnold committed
123
    ``context``-parameter with ``transformation_factory`` creates a
124
    function with the same name, which returns a partial-function that
125
    takes just the context-parameter.
126
127
128
129
130
131
132
133
134

    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.

135
136
    Usage::

eckhart's avatar
eckhart committed
137
        @transformation_factory(AbstractSet[str])
138
        def remove_tokens(context, tokens):
139
            ...
140
141
142

    or, alternatively::

143
        @transformation_factory
144
        def remove_tokens(context, tokens: AbstractSet[str]):
145
146
            ...

147
148
    Example::

149
        trans_table = { 'expression': remove_tokens('+', '-') }
150
151
152

    instead of::

153
        trans_table = { 'expression': partial(remove_tokens, tokens={'+', '-'}) }
154
155

    Parameters:
156
        t1:  type of the second argument of the transformation function,
157
158
            only necessary if the transformation functions' parameter list
            does not have type annotations.
159
160
    """

161
162
163
164
    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`."""
165
166
167
168
        # 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
169
170
        if isinstance(t, str):          # ensure compatibility with python versions
            t = eval(t.replace('unicode', 'str'))  # with alternative type handling.
171
        if isgenerictype(t):
172
173
174
            raise TypeError("Generic Type %s not permitted\n in transformation_factory "
                            "decorator. Use the equivalent non-generic type instead!"
                            % str(t))
175
        if issubtype(List[Node], t):
176
177
            raise TypeError("Sequence type %s not permitted\nin transformation_factory "
                            "decorator, because it could be mistaken for a base class "
178
179
180
                            "of List[Node]\nwhich is the type of the canonical first "
                            "argument of transformation functions. Try 'tuple' instead!"
                            % str(t))
181
182
        return t

183
    def decorator(f):
184
        nonlocal t1
185
186
187
188
        sig = inspect.signature(f)
        params = list(sig.parameters.values())[1:]
        if len(params) == 0:
            return f  # '@transformer' not needed w/o free parameters
189
        assert t1 or params[0].annotation != params[0].empty, \
190
            "No type information on second parameter found! Please, use type " \
eckhart's avatar
eckhart committed
191
            "annotation or provide the type information via transformer-decorator."
192
        f = singledispatch(f)
193
194
195
        p1type = params[0].annotation
        if t1 is None:
            t1 = type_guard(p1type)
196
        elif issubtype(p1type, type_guard(t1)):
197
            try:
198
                if len(params) == 1 and issubtype(p1type, Container) \
199
                        and not (issubtype(p1type, Text) or issubtype(p1type, ByteString)):
200
                    def gen_special(*args):
201
202
                        c = set(args) if issubtype(p1type, AbstractSet) else \
                            tuple(args) if issubtype(p1type, Sequence) else args
203
204
205
206
207
208
209
210
                        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)))
211
212
213
214
215
216

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

217
        for t in (t1, t2, t3, t4, t5):
218
            if t:
219
                f.register(type_guard(t), gen_partial)
220
221
            else:
                break
222
223
        return f

224
    if isinstance(t1, type(lambda: 1)):
225
226
227
        # Provide for the case that transformation_factory has been
        # written as plain decorator and not as a function call that
        # returns the decorator proper.
228
229
        func = t1
        t1 = None
230
231
232
233
234
        return decorator(func)
    else:
        return decorator


235
def key_tag_name(node: Node) -> str:
Eckhart Arnold's avatar
Eckhart Arnold committed
236
237
238
239
    """
    Returns the tag name of the node as key of selecting transformations
    from the transformation table.
    """
240
241
242
    return node.tag_name


243
def traverse(root_node: Node,
Eckhart Arnold's avatar
Eckhart Arnold committed
244
             processing_table: ProcessingTableType,
eckhart's avatar
eckhart committed
245
             key_func: KeyFunc = key_tag_name) -> None:
246
    """
Eckhart Arnold's avatar
Eckhart Arnold committed
247
    Traverses the syntax tree starting with the given ``node`` depth
248
    first and applies the sequences of callback-functions registered
249
    in the ``processing_table``-dictionary.
250
251
252
253
254
255
256
257
258

    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:
259

260
    - '<': always called (before any other processing function)
261
262
    - '*': called for those nodes for which no (other) processing
      function appears in the table
263
    - '>': always called (after any other processing function)
264
265
266
267
268

    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
269
270
            is interpreted as a ``compact_table``. See
            :func:`expand_table` or :func:`EBNFCompiler.EBNFTransTable`
271
        key_func (function): A mapping key_func(node) -> keystr. The default
272
            key_func yields node.tag_name.
273

274
275
    Example::

276
        table = { "term": [replace_by_single_child, flatten],
277
                  "factor, flowmarker, retrieveop": replace_by_single_child }
278
        traverse(node, table)
279

280
    """
281

282
283
284
    # Is this optimazation really needed?
    if '__cache__' in processing_table:
        # assume that processing table has already been expanded
eckhart's avatar
eckhart committed
285
        table = processing_table               # type: ProcessingTableType
eckhart's avatar
eckhart committed
286
        cache = cast(TransformationDict, processing_table['__cache__'])  # type: TransformationDict
287
    else:
288
289
        # normalize processing_table entries by turning single values
        # into lists with a single value
290
291
        table = {name: cast(Sequence[Callable], smart_list(call))
                 for name, call in list(processing_table.items())}
292
        table = expand_table(table)
293
        # substitute key for insiginificant whitespace
294
        assert '+' not in table, 'Symbol "+" in processing table is obsolete, use "<" instead'
295
296
        if '~' in table:
            if ':Whitespace' in table:
eckhart's avatar
eckhart committed
297
298
299
300
                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!')
301
302
303
304
            whitespace_transformation = table['~']
            del table['~']
            table[':Whitespace'] = whitespace_transformation
        # cache expanded table
eckhart's avatar
eckhart committed
305
306
        cache = cast(TransformationDict,
                     table.setdefault('__cache__', cast(TransformationDict, dict())))
307
308
        # change processing table in place, so its already expanded and cache filled next time
        processing_table.clear()
309
310
        processing_table.update(table)

311
    def traverse_recursive(context):
eckhart's avatar
eckhart committed
312
        nonlocal cache
313
        node = context[-1]
314
        if node.children:
315
            context.append(PLACEHOLDER)
di68kap's avatar
di68kap committed
316
            for child in node.children:
eckhart's avatar
eckhart committed
317
                context[-1] = child
318
                traverse_recursive(context)  # depth first
eckhart's avatar
eckhart committed
319
            context.pop()
320
321

        key = key_func(node)
322
323
324
        try:
            sequence = cache[key]
        except KeyError:
325
            sequence = table.get('<', []) \
326
                + table.get(key, table.get('*', [])) \
327
                + table.get('>', [])
328
329
330
            cache[key] = sequence

        for call in sequence:
331
            call(context)
332

333
    traverse_recursive([root_node])
334
335
    # assert processing_table['__cache__']

336

337
#######################################################################
338
#
339
340
# meta transformations, i.e. transformations that call other
# transformations
341
#
342
#######################################################################
343
344


eckhart's avatar
eckhart committed
345
@transformation_factory(dict)
346
def traverse_locally(context: List[Node],
eckhart's avatar
eckhart committed
347
348
                     processing_table: Dict,              # actually: ProcessingTableType
                     key_func: Callable = key_tag_name):  # actually: KeyFunc
349
350
    """
    Transforms the syntax tree starting from the last node in the context
351
352
353
354
355
356
357
    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)


358
@transformation_factory(collections.abc.Callable)
359
def apply_if(context: List[Node], transformation: Callable, condition: Callable):
360
361
362
    """
    Applies a transformation only if a certain condition is met.
    """
363
364
365
366
    if condition(context):
        transformation(context)


367
@transformation_factory(collections.abc.Callable)
eckhart's avatar
eckhart committed
368
def apply_unless(context: List[Node], transformation: Callable, condition: Callable):
369
370
371
    """
    Applies a transformation if a certain condition is *not* met.
    """
eckhart's avatar
eckhart committed
372
373
374
375
    if not condition(context):
        transformation(context)


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.
#
#######################################################################


Eckhart Arnold's avatar
Eckhart Arnold committed
389
390
391
392
393
def always(context: List[Node]) -> bool:
    """Always returns True, no matter that the state of the context."""
    return True


394
def is_single_child(context: List[Node]) -> bool:
eckhart's avatar
eckhart committed
395
    """Returns ``True`` if the current node does not have any siblings."""
396
397
398
399
    return len(context[-2].children) == 1


def is_named(context: List[Node]) -> bool:
eckhart's avatar
eckhart committed
400
    """Returns ``True`` if the current node's parser is a named parser."""
401
    return not context[-1].is_anonymous()
402
403
404


def is_anonymous(context: List[Node]) -> bool:
eckhart's avatar
eckhart committed
405
    """Returns ``True`` if the current node's parser is an anonymous parser."""
406
    return context[-1].is_anonymous()
407
408


409
def is_insignificant_whitespace(context: List[Node]) -> bool:
di68kap's avatar
di68kap committed
410
    """Returns ``True`` for whitespace and comments defined with the
411
    ``@comment``-directive."""
412
    return context[-1].tag_name == WHITESPACE_PTYPE
413
414


415
RX_WHITESPACE = re.compile(r'\s*$')
416
417
418
419
420
421
422


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."""
423
    return bool(RX_WHITESPACE.match(context[-1].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
def swap_attributes(node: Node, other: Node):
    """
    Exchanges the attributes between node and other. This might be
    needed when rearanging trees.
    """
534
535
    NA = node.has_attr()
    OA = other.has_attr()
536
537
538
539
540
541
542
543
544
545
546
547
    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(child, node)
567
    if child.has_attr():
568
        node._xml_attr = child._xml_attr
569
570


571
572
573
574
575
576
577
578
579
580
581
582
#######################################################################
#
# 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
583
584
# DEPRECATED
# def flatten_anonymous_nodes(context: List[Node]):
585
#     """
Eckhart Arnold's avatar
Eckhart Arnold committed
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.
591
#     """
Eckhart Arnold's avatar
Eckhart Arnold committed
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
#     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)
616
617


618
619
def replace_by_single_child(context: List[Node]):
    """
620
621
    Removes single branch node, replacing it by its immediate descendant.
    Replacement only takes place, if the last node in the context has
622
623
624
    exactly one child. Attributes will be merged. In case one and the same
    attribute is defined for the child as well as the parent, the child's
    attribute value take precedence.
625
626
627
    """
    node = context[-1]
    if len(node.children) == 1:
628
        _replace_by(node, node.children[0])
629
630


631
632
633
def replace_by_children(context: List[Node]):
    """
    Eliminates the last node in the context by replacing it with its children.
di68kap's avatar
di68kap committed
634
635
    The attributes of this node will be dropped. In case the last node is
    the root-note (i.e. len(context) == 1), it will not be eliminated.
636
    """
di68kap's avatar
di68kap committed
637
638
639
640
    try:
        parent = context[-2]
    except IndexError:
        return
641
642
643
644
645
646
647
    node = context[-1]
    assert node.children
    result = parent.result
    i = result.index(node)
    parent.result = result[:i] + node.children + result[i + 1:]


Eckhart Arnold's avatar
Eckhart Arnold committed
648
def reduce_single_child(context: List[Node]):
649
    """
650
    Reduces a single branch node by transferring the result of its
651
    immediate descendant to this node, but keeping this node's parser entry.
652
    Reduction only takes place if the last node in the context has
653
654
655
    exactly one child. Attributes will be merged. In case one and the same
    attribute is defined for the child as well as the parent, the parent's
    attribute value take precedence.
656
657
658
    """
    node = context[-1]
    if len(node.children) == 1:
659
        _reduce_child(node, node.children[0])
660
661


662
@transformation_factory(collections.abc.Callable)
eckhart's avatar
eckhart committed
663
def replace_or_reduce(context: List[Node], condition: Callable = is_named):
664
665
    """
    Replaces node by a single child, if condition is met on child,
666
667
    otherwise (i.e. if the child is anonymous) reduces the child.
    """
668
    node = context[-1]
669
    if len(node.children) == 1:
di68kap's avatar
di68kap committed
670
        child = node.children[0]
Eckhart Arnold's avatar
Eckhart Arnold committed
671
        if condition(context):
672
            _replace_by(node, child)
673
        else:
674
            _reduce_child(node, child)
675
676


Eckhart Arnold's avatar
Eckhart Arnold committed
677
678
@transformation_factory(str)
def change_tag_name(context: List[Node], tag_name: str, restriction: Callable = always):
679
    """
Eckhart Arnold's avatar
Eckhart Arnold committed
680
    Changes the tag name of a node.
681
682

    Parameters:
Eckhart Arnold's avatar
Eckhart Arnold committed
683
684
        restriction: A function of the context that returns False in cases
                where the tag name shall not be exchanged
685
        context: the context where the parser shall be replaced
Eckhart Arnold's avatar
Eckhart Arnold committed
686
        tag_name: The new tag name.
687
    """
Eckhart Arnold's avatar
Eckhart Arnold committed
688
689
690
    if restriction(context):
        node = context[-1]
        node.tag_name = tag_name
691
692


693
@transformation_factory(collections.abc.Callable)
eckhart's avatar
eckhart committed
694
def flatten(context: List[Node], condition: Callable = is_anonymous, recursive: bool = True):
695
    """
eckhart's avatar
eckhart committed
696
    Flattens all children, that fulfill the given ``condition``
697
698
699
    (default: all unnamed children). Flattening means that wherever a
    node has child nodes, the child nodes are inserted in place of the
    node.
700
701

    If the parameter ``recursive`` is ``True`` the same will recursively be
702
703
704
    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.
705
706
707
708

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

di68kap's avatar
di68kap committed
709
        (1 (+ 2) (+ 3))    ->   (1 + 2 + 3)
710
711
        (1 (+ (2 + (3))))  ->   (1 + 2 + 3)
    """
712

713
    node = context[-1]
714
    if node.children:
Eckhart Arnold's avatar
Eckhart Arnold committed
715
        new_result = []     # type: List[Node]
716
        context.append(PLACEHOLDER)
717
        for child in node.children:
eckhart's avatar
eckhart committed
718
            context[-1] = child
719
            if child.children and condition(context):
720
                if recursive:
721
                    flatten(context, condition, recursive)
722
                new_result.extend(child.children)
723
                update_attr(node, child)
724
725
            else:
                new_result.append(child)
eckhart's avatar
eckhart committed
726
        context.pop()
727
728
729
        node.result = tuple(new_result)


730
def collapse(context: List[Node]):
731
732
733
734
    """
    Collapses all sub-nodes of a node by replacing them with the
    string representation of the node. USE WITH CARE!
    """
735
    node = context[-1]
736
    node.result = node.content
737
738


739
@transformation_factory(collections.abc.Callable)
740
def collapse_if(context: List[Node], condition: Callable, target_tag: str):
741
742
    """
    (Recursively) merges the content of all adjacent child nodes that
743
744
745
746
747
    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)
748
    >>> collapse_if([tree], not_one_of({'superscript', 'subscript'}), 'text')
749
    >>> print(flatten_sxpr(tree.as_sxpr()))
750
    (place (text "p.26") (superscript "b") (text ",18"))
751
752
753

    See `test_transform.TestComplexTransformations` for examples.
    """
754

755
756
    assert isinstance(target_tag, str)  # TODO: Delete this when safe

757
    node = context[-1]
eckhart's avatar
eckhart committed
758
759
    package = []  # type: List[Node]
    result = []  # type: List[Node]
760
761
762
763

    def close_package():
        nonlocal package
        if package:
764
            s = "".join(nd.content for nd in package)
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
            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)


788
@transformation_factory(collections.abc.Callable)
789
def replace_content(context: List[Node], func: Callable):  # Callable[[Node], ResultType]
790
791
    """
    Replaces the content of the node. ``func`` takes the node's result
792
793
    as an argument an returns the mapped result.
    """
794
    node = context[-1]
795
796
797
    node.result = func(node.result)


798
@transformation_factory  # (str)
799
def replace_content_by(context: List[Node], content: str):  # Callable[[Node], ResultType]
800
801
    """
    Replaces the content of the node with the given text content.
802
803
804
805
806
    """
    node = context[-1]
    node.result = content


807
def normalize_whitespace(context):
di68kap's avatar
di68kap committed
808
809
    """
    Normalizes Whitespace inside a leaf node, i.e. any sequence of
810
811
    whitespaces, tabs and linefeeds will be replaced by a single
    whitespace. Empty (i.e. zero-length) Whitespace remains empty,
di68kap's avatar
di68kap committed
812
813
    however.
    """
814
815
    node = context[-1]
    assert not node.children
816
    if is_insignificant_whitespace(context):
817
818
819
        if node.result:
            node.result = ' '
    else:
eckhart's avatar
eckhart committed
820
        node.result = re.sub(r'\s+', ' ', node.result)
821
822


823
def merge_adjacent(context, condition: Callable):
di68kap's avatar
di68kap committed
824
    """
825
826
827
    Merges adjacent nodes that fulfill the given `condition`. It is
    is assumed that `condition` is never true for leaf-nodes and non-leaf-nodes
    alike. Otherwise a type-error might ensue.
di68kap's avatar
di68kap committed
828
829
830
831
832
833
834
    """
    node = context[-1]
    children = node.children
    new_result = []
    i = 0
    L = len(children)
    while i < L:
835
836
        if condition([children[i]]):
            initial = () if children[i].children else ''
di68kap's avatar
di68kap committed
837
            k = i
838
            while i < L and condition([children[k]]):
di68kap's avatar
di68kap committed
839
840
                i += 1
            if i > k:
841
                children[k].result = sum((children[n].result for n in range(k, i + 1)), initial)
di68kap's avatar
di68kap committed
842
843
844
845
846
            new_result.append(children[k])
        i += 1
    node.result = tuple(new_result)


847
@transformation_factory(collections.abc.Callable)
848
def move_adjacent(context: List[Node], condition: Callable, merge: bool = True):
di68kap's avatar
di68kap committed
849
    """
850
    Moves adjacent nodes that fulfill the given condition to the parent node.
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
    If the `merge`-flag is set, a moved node will be merged with its
    predecessor (or successor, respectively) in the parent node in case it
    also fulfill the given `condition`.
    """
    def merge_results(a: Node, b: Node, c: Node) -> bool:
        """
        Merges the results of node `a` and `b` and writes them to the result
        of `c` type-safely, if b and c are either both leaf-nodes (in which case
        their result-strings are concatenated) or both non-leaf-nodes (in which
        case the tuples of children are concatenated).
        Returns `True` in case of a successful merge, `False` if only one node
        was a leaf node and the merge could thus not be done.

        Example:
            >>> head, tail = Node('head', '123'), Node('tail', '456')
            >>> merge_results(head, tail, head)  # merge head and tail (in that order) into head
            True
            >>> str(head)
            '123456'
        """
871
872
873
874
875
876
877
878
        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

879
880
881
882
883
    node = context[-1]
    if len(context) <= 1 or not node.children:
        return
    parent = context[-2]
    children = node.children
884
    if condition([children[0]]):
885
        before = (children[0],)   # type: Tuple[Node, ...]
886
887
888
        children = children[1:]
    else:
        before = ()
889
    if children and condition([children[-1]]):
890
        after = (children[-1],)   # type: Tuple[Node, ...]
891
892
893
894
895
896
897
        children = children[:-1]
    else:
        after = tuple()

    if before or after:
        node.result = children
        for i, child in enumerate(parent.children):
898
            if id(child) == id(node):
899
900
                break

901
902
903
904
905
906
907
908
909
910
911
912
913
914
        # merge adjacent nodes that fulfil the condition
        if merge:
            prevN = parent.children[i - 1] if i > 0 else None
            nextN = parent.children[i + 1] if i < len(parent.children) - 1 else None
            if before and prevN and condition([prevN]):
                # prevN.result = prevN.result + before[0].result
                # before = ()
                if merge_results(prevN, before[0], prevN):
                    before = ()
            if after and nextN and condition([nextN]):
                # nextN.result = after[0].result + nextN.result
                # after = ()
                if merge_results(after[0], nextN, nextN):
                    after = ()
915

916
        parent.result = parent.children[:i] + before + (node,) + after + parent.children[i+1:]
917
918


Eckhart Arnold's avatar
Eckhart Arnold committed
919
920
def left_associative(context: List[Node]):
    """
921
    Rearranges a flat node with infix operators into a left associative tree.
Eckhart Arnold's avatar
Eckhart Arnold committed
922
    """
923
    node = context[-1]
924
925
926
927
928
929
930
931
932
933
    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
934

Eckhart Arnold's avatar
Eckhart Arnold committed
935

936
@transformation_factory(collections.abc.Set)
937
def lean_left(context: List[Node], operators: AbstractSet[str]):
938
    """
939
940
941
942
943
944
945
946
947
    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.
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
    """
    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)
963
964
        # continue recursively on the left branch
        lean_left([right], operators)
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990


# @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
991
992


993
994
995
996
997
#######################################################################
#
# destructive transformations:
#
# - leaves may be dropped (e.g. if deemed irrelevant)
998
# - errors of dropped leaves may be be lost
999
1000
1001
# - no promise that order will be preserved
#
#######################################################################
1002
1003


1004
@transformation_factory(collections.abc.Callable)
1005
def lstrip(context: List[Node], condition: Callable = contains_only_whitespace):
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
    """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.