transform.py 19.2 KB
Newer Older
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
"""transformation.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.
"""

import inspect
from functools import partial, reduce, singledispatch

from DHParser.syntaxtree import WHITESPACE_PTYPE, TOKEN_PTYPE, MockParser, Node

try:
    import regex as re
except ImportError:
    import re
try:
    from typing import AbstractSet, Any, ByteString, Callable, cast, Container, Dict, \
        Iterator, List, NamedTuple, Sequence, Union, Text, Tuple
except ImportError:
    from .typing34 import AbstractSet, Any, ByteString, Callable, cast, Container, Dict, \
        Iterator, List, NamedTuple, Sequence, Union, Text, Tuple

from DHParser.toolkit import expand_table, smart_list

__all__ = ('transformation_factory',
           'key_parser_name',
           'key_tag_name',
           'traverse',
42
           'is_named',
43
44
           'replace_by_single_child',
           'reduce_single_child',
45
           'replace_or_reduce',
46
47
           'replace_parser',
           'collapse',
48
           'merge_children',
49
50
           'replace_content',
           'apply_if',
51
           'is_anonymous',
52
53
54
55
           'is_whitespace',
           'is_empty',
           'is_expendable',
           'is_token',
56
           'is_one_of',
57
           'has_content',
58
           'remove_children_if',
59
60
61
62
63
64
65
66
           'remove_parser',
           'remove_content',
           'remove_first',
           'remove_last',
           'remove_whitespace',
           'remove_empty',
           'remove_expendables',
           'remove_brackets',
67
68
           'remove_infix_operator',
           'remove_single_child',
69
           'remove_tokens',
70
           'keep_children',
71
72
73
           'flatten',
           'forbid',
           'require',
74
75
76
77
           'assert_content',
           'assert_condition',
           'assert_has_children',
           'TRUE_CONDITION')
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98


def transformation_factory(t=None):
    """Creates factory functions from transformation-functions that
    dispatch on the first parameter after the node parameter.

    Decorating a transformation-function that has more than merely the
    ``node``-parameter with ``transformation_factory`` creates a
    function with the same name, which returns a partial-function that
    takes just the node-parameter.

    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.

    Usage:
        @transformation_factory(AbtractSet[str])
99
        def remove_tokens(context, tokens):
100
101
102
            ...
      or, alternatively:
        @transformation_factory
103
        def remove_tokens(context, tokens: AbstractSet[str]):
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
            ...

    Example:
        trans_table = { 'expression': remove_tokens('+', '-') }
      instead of:
        trans_table = { 'expression': partial(remove_tokens, tokens={'+', '-'}) }
    """

    def decorator(f):
        sig = inspect.signature(f)
        params = list(sig.parameters.values())[1:]
        if len(params) == 0:
            return f  # '@transformer' not needed w/o free parameters
        assert t or params[0].annotation != params[0].empty, \
            "No type information on second parameter found! Please, use type " \
            "annotation or provide the type information via transfomer-decorator."
        p1type = t or params[0].annotation
        f = singledispatch(f)
        if len(params) == 1 and issubclass(p1type, Container) and not issubclass(p1type, Text) \
                and not issubclass(p1type, ByteString):
            def gen_special(*args):
                c = set(args) if issubclass(p1type, AbstractSet) else \
                    list(args) if issubclass(p1type, Sequence) else args
                d = {params[0].name: c}
                return partial(f, **d)

            f.register(p1type.__args__[0], gen_special)

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

        f.register(p1type, gen_partial)
        return f

    if isinstance(t, type(lambda: 1)):
        # Provide for the case that transformation_factory has been
        # written as plain decorator and not as a function call that
        # returns the decorator proper.
144
        func = t
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
        t = None
        return decorator(func)
    else:
        return decorator


def key_parser_name(node) -> str:
    return node.parser.name


def key_tag_name(node) -> str:
    return node.tag_name


def traverse(root_node, processing_table, key_func=key_tag_name) -> None:
    """Traverses the snytax tree starting with the given ``node`` depth
    first and applies the sequences of callback-functions registered
    in the ``calltable``-dictionary.

    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:
        '+': always called (before any other processing function)
        '*': called for those nodes for which no (other) processing
             function appears in the table
        '~': always called (after any other processing function)

    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
            is interpreted as a ``compact_table``. See
            ``toolkit.expand_table`` or ``EBNFCompiler.EBNFTransTable``
        key_func (function): A mapping key_func(node) -> keystr. The default
            key_func yields node.parser.name.

    Example:
        table = { "term": [replace_by_single_child, flatten],
            "factor, flowmarker, retrieveop": replace_by_single_child }
        traverse(node, table)
    """
    # commented, because this approach is too error prone!
    # def funclist(call):
    #     return [as_partial(func) for func in smart_list(call)]

    # normalize processing_table entries by turning single values into lists
    # with a single value
    table = {name: smart_list(call) for name, call in list(processing_table.items())}
    table = expand_table(table)
    cache = {}  # type: Dict[str, List[Callable]]

201
202
    def traverse_recursive(context):
        node = context[-1]
203
204
        if node.children:
            for child in node.result:
205
206
                context.append(child)
                traverse_recursive(context)  # depth first
207
                node.error_flag |= child.error_flag  # propagate error flag
208
                context.pop()
209
210
211
212
213
214
215
216
217
218
219
220
221
222

        key = key_func(node)
        sequence = cache.get(key, None)
        if sequence is None:
            sequence = table.get('+', []) + \
                       table.get(key, table.get('*', [])) + \
                       table.get('~', [])
            # '+' always called (before any other processing function)
            # '*' called for those nodes for which no (other) processing function
            #     appears in the table
            # '~' always called (after any other processing function)
            cache[key] = sequence

        for call in sequence:
223
            call(context)
224

225
    traverse_recursive([root_node])
226
227
228
229
230
231
232
233
234
235
236
237
238


# ------------------------------------------------
#
# 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)
#
# ------------------------------------------------


239
240
241
242
def TRUE_CONDITION(node):
    return True


243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
def replace_child(node):
    assert len(node.children) == 1
    if not node.children[0].parser.name:
        node.children[0].parser.name = node.parser.name
    node.parser = node.children[0].parser
    node._errors.extend(node.children[0]._errors)
    node.result = node.result[0].result


def reduce_child(node):
    assert len(node.children) == 1
    node._errors.extend(node.children[0]._errors)
    node.result = node.result[0].result


258
@transformation_factory(Callable)
259
def replace_by_single_child(context, condition=TRUE_CONDITION):
260
261
    """Remove single branch node, replacing it by its immediate descendant
    if and only if the condision on the descendant is true.
262
263
264
    (In case the descendant's name is empty (i.e. anonymous) the
    name of this node's parser is kept.)
    """
265
    node = context[-1]
266
267
    if len(node.children) == 1 and condition(node.children[0]):
        replace_child(node)
268
269


270
@transformation_factory(Callable)
271
def reduce_single_child(context, condition=TRUE_CONDITION):
272
273
    """Reduce a single branch node, by transferring the result of its
    immediate descendant to this node, but keeping this node's parser entry.
274
275
    If the condition evaluates to false on the descendant, it will not
    be reduced.
276
    """
277
    node = context[-1]
278
279
280
281
282
283
284
285
286
    if len(node.children) == 1 and condition(node.children[0]):
        reduce_child(node)


def is_named(node):
    return node.parser.name


@transformation_factory(Callable)
287
def replace_or_reduce(context, condition=is_named):
288
289
290
    """Replaces node by a single child, if condition is met on child,
    otherwise (i.e. if the child is anonymous) reduces the child.
    """
291
    node = context[-1]
292
293
294
295
    if len(node.children) == 1 and condition(node.children[0]):
        replace_child(node)
    else:
        reduce_child(node)
296
297
298


@transformation_factory
299
def replace_parser(context, name: str):
300
301
302
303
304
305
306
    """Replaces the parser of a Node with a mock parser with the given
    name.

    Parameters:
        name(str): "NAME:PTYPE" of the surogate. The ptype is optional
        node(Node): The node where the parser shall be replaced
    """
307
    node = context[-1]
308
309
310
311
312
    name, ptype = (name.split(':') + [''])[:2]
    node.parser = MockParser(name, ptype)


@transformation_factory(Callable)
313
def flatten(context, condition=lambda node: not node.parser.name, recursive=True):
314
315
316
317
318
319
320
321
322
323
324
325
326
    """Flattens all children, that fulfil the given `condition`
    (default: all unnamed children). Flattening means that wherever a
    node has child nodes, the child nodes are inserted in place of the
    node.
     If the parameter `recursive` is `True` the same will recursively be
    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.
     Applying flatten recursively will result in these kinds of
    structural transformation:
        (1 (+ 2) (+ 3)     ->   (1 + 2 + 3)
        (1 (+ (2 + (3))))  ->   (1 + 2 + 3)
    """
327
    node = context[-1]
328
329
330
331
332
    if node.children:
        new_result = []
        for child in node.children:
            if child.children and condition(child):
                if recursive:
333
334
335
                    context.append(child)
                    flatten(context, condition, recursive)
                    context.pop()
336
337
338
339
340
341
                new_result.extend(child.children)
            else:
                new_result.append(child)
        node.result = tuple(new_result)


342
def collapse(context):
343
344
    """Collapses all sub-nodes of a node by replacing them with the
    string representation of the node.
345
    """
346
    node = context[-1]
347
348
349
350
    node.result = str(node)


@transformation_factory
351
def merge_children(context, tag_names: List[str]):
352
    """Joins all children next to each other and with particular tag-
353
354
    names into a single child node with mock parser with the name of
    the first tag name in the list.
355
    """
356
    node = context
357
    result = []
358
    name, ptype = ('', tag_names[0]) if tag_names[0][:1] == ':' else (tag_names[0], '')
359
    if node.children:
360
        i = 0
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
        L = len(node.children)
        while i < L:
            while i < L and not node.children[i].tag_name in tag_names:
                result.append(node.children[i])
                i += 1
            k = i + 1
            while (k < L and node.children[k].tag_name in tag_names
                   and bool(node.children[i].children) == bool(node.children[k].children)):
                k += 1
            if i < L:
                result.append(Node(MockParser(name, ptype),
                                   reduce(lambda a, b: a + b, (node.result for node in node.children[i:k]))))
            i = k
        node.result = tuple(result)


# ------------------------------------------------
#
# destructive transformations:
#     - tree may be rearranged (flattened),
#     - order is preserved
#     - but (irrelevant) leaves may be dropped
#     - errors of dropped leaves will be lost
#
# ------------------------------------------------


@transformation_factory
389
def replace_content(context, func: Callable):  # Callable[[Node], ResultType]
390
391
392
    """Replaces the content of the node. ``func`` takes the node
    as an argument an returns the mapped result.
    """
393
    node = context[-1]
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
    node.result = func(node.result)


def is_whitespace(node):
    """Removes whitespace and comments defined with the
    ``@comment``-directive."""
    return node.parser.ptype == WHITESPACE_PTYPE


def is_empty(node):
    return not node.result


def is_expendable(node):
    return is_empty(node) or is_whitespace(node)


def is_token(node, tokens: AbstractSet[str] = frozenset()) -> bool:
    return node.parser.ptype == TOKEN_PTYPE and (not tokens or node.result in tokens)


415
416
417
418
419
420
421
422
def is_anonymous(node):
    return not node.parser.name


def is_one_of(node, tag_name_set: AbstractSet[str]) -> bool:
    """Returns true, if the node's tag_name is on of the
    given tag names."""
    return node.tag_name in tag_name_set
423
424


425
426
427
def has_content(node, regexp: str) -> bool:
    """Checks a node's content against a regular expression."""
    return bool(re.match(regexp, str(node)))
428
429


430
@transformation_factory(Callable)
431
def apply_if(context, transformation: Callable, condition: Callable):
432
433
    """Applies a transformation only if a certain condition is met.
    """
434
    node = context[-1]
435
    if condition(node):
436
        transformation(context)
437
438


439
@transformation_factory(slice)
440
def keep_children(context, section: slice = slice(None)):
441
    """Keeps only child-nodes which fall into a slice of the result field."""
442
    node = context[-1]
443
    if node.children:
444
        node.result = node.children[section]
445
446
447


@transformation_factory(Callable)
448
def remove_children_if(context, condition: Callable, section: slice = slice(None)):
449
    """Removes all nodes from a slice of the result field if the function
450
    `condition(child_node)` evaluates to `True`."""
451
    node = context[-1]
452
    if node.children:
453
454
455
456
        c = node.children
        N = len(c)
        rng = range(*section.indices(N))
        node.result = tuple(c[i] for i in range(N) if not i in rng or not condition(c[i]))
457
458
459
460
461


remove_whitespace = remove_children_if(is_whitespace)  # partial(remove_children_if, condition=is_whitespace)
remove_empty = remove_children_if(is_empty)
remove_expendables = remove_children_if(is_expendable)  # partial(remove_children_if, condition=is_expendable)
462
463
464
465
466
remove_first = apply_if(keep_children(slice(1, None)), lambda nd: len(nd.children) > 1)
remove_last = apply_if(keep_children(slice(None, -1)), lambda nd: len(nd.children) > 1)
remove_brackets = apply_if(keep_children(slice(1, -1)), lambda nd: len(nd.children) >= 2)
remove_infix_operator = keep_children(slice(0, None, 2))
remove_single_child = apply_if(keep_children(slice(0)), lambda nd: len(nd.children) == 1)
467
468
469


@transformation_factory
470
def remove_tokens(context, tokens: AbstractSet[str] = frozenset()):
471
472
473
    """Reomoves any among a particular set of tokens from the immediate
    descendants of a node. If ``tokens`` is the empty set, all tokens
    are removed."""
474
    remove_children_if(context, partial(is_token, tokens=tokens))
475
476
477


@transformation_factory
478
def remove_parser(context, tag_names: AbstractSet[str]):
Eckhart Arnold's avatar
Eckhart Arnold committed
479
    """Removes children by tag name."""
480
    remove_children_if(context, partial(is_one_of, tag_name_set=tag_names))
481
482
483


@transformation_factory
484
def remove_content(context, regexp: str):
485
    """Removes children depending on their string value."""
486
    remove_children_if(context, partial(has_content, regexp=regexp))
487
488
489
490


########################################################################
#
491
# AST semantic validation functions (EXPERIMENTAL!!!)
492
493
494
#
########################################################################

495
@transformation_factory(Callable)
496
def assert_condition(context, condition: Callable, error_msg: str = '') -> bool:
497
    """Checks for `condition`; adds an error message if condition is not met."""
498
    node = context[-1]
499
500
501
502
503
504
505
506
507
508
509
510
511
512
    if not condition(node):
        if error_msg:
            node.add_error(error_msg % node.tag_name if error_msg.find("%s") > 0 else error_msg)
        else:
            cond_name = condition.__name__ if hasattr(condition, '__name__') \
                        else condition.__class__.__name__ if hasattr(condition, '__class__') \
                        else '<unknown>'
            node.add_error("transform.assert_condition: Failed to meet condition " + cond_name)


assert_has_children = assert_condition(lambda nd: nd.children, 'Element "%s" has no children')


@transformation_factory
513
514
def assert_content(context, regexp: str):
    node = context[-1]
515
516
517
518
519
520
521
    if not has_content(node, regexp):
        node.add_error('Element "%s" violates %s on %s' %
                       (node.parser.name, str(regexp), str(node)))

#
# @transformation_factory
# def assert_name(node, regexp: str):
522
#     if not is_one_of(node, regexp):
523
524
525
526
527
528
529
530
531
532
533
#         node.add_error('Element name "%s" does not match %s' % (node.tag_name), str(regexp))
#
#
# @transformation_factory(Callable)
# def assert_children(node, condition: Callable=lambda node: True,
#                     error_msg: str='', section: slice=slice(None)):
#     if node.children:
#         for child in node.children:
#             assert_condition(child, condition, error_msg)
#

534
535

@transformation_factory
536
537
def require(context, child_tags: AbstractSet[str]):
    node = context[-1]
538
539
540
541
542
543
544
    for child in node.children:
        if child.tag_name not in child_tags:
            node.add_error('Element "%s" is not allowed inside "%s".' %
                           (child.parser.name, node.parser.name))


@transformation_factory
545
546
def forbid(context, child_tags: AbstractSet[str]):
    node = context[-1]
547
548
549
550
    for child in node.children:
        if child.tag_name in child_tags:
            node.add_error('Element "%s" cannot be nested inside "%s".' %
                           (child.parser.name, node.parser.name))