transform.py 17 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
42
43
44
45
"""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',
           'replace_by_single_child',
           'reduce_single_child',
           'replace_parser',
           'collapse',
46
           'join_children',
47
48
49
50
51
52
53
54
           'replace_content',
           'apply_if',
           'is_whitespace',
           'is_empty',
           'is_expendable',
           'is_token',
           'has_name',
           'has_content',
55
           'remove_children_if',
56
57
58
59
60
61
62
63
64
           'remove_parser',
           'remove_content',
           'remove_first',
           'remove_last',
           'remove_whitespace',
           'remove_empty',
           'remove_expendables',
           'remove_brackets',
           'remove_tokens',
65
           'keep_children',
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
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
           'flatten',
           'forbid',
           'require',
           'assert_content')


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])
        def remove_tokens(node, tokens):
            ...
      or, alternatively:
        @transformation_factory
        def remove_tokens(node, tokens: AbstractSet[str]):
            ...

    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.
136
        func = t
137
138
139
140
141
142
143
144
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
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
        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]]

    def traverse_recursive(node):
        if node.children:
            for child in node.result:
                traverse_recursive(child)  # depth first
                node.error_flag |= child.error_flag  # propagate error flag

        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:
            call(node)

    traverse_recursive(root_node)


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


def replace_by_single_child(node):
    """Remove single branch node, replacing it by its immediate descendant.
    (In case the descendant's name is empty (i.e. anonymous) the
    name of this node's parser is kept.)
    """
    if node.children and len(node.result) == 1:
        if not node.result[0].parser.name:
            node.result[0].parser.name = node.parser.name
        node.parser = node.result[0].parser
237
        node._errors.extend(node.result[0]._errors)
238
239
240
241
242
243
244
245
        node.result = node.result[0].result


def reduce_single_child(node):
    """Reduce a single branch node, by transferring the result of its
    immediate descendant to this node, but keeping this node's parser entry.
    """
    if node.children and len(node.result) == 1:
246
        node._errors.extend(node.result[0]._errors)
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
        node.result = node.result[0].result


@transformation_factory
def replace_parser(node, name: str):
    """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
    """
    name, ptype = (name.split(':') + [''])[:2]
    node.parser = MockParser(name, ptype)


@transformation_factory(Callable)
def flatten(node, condition=lambda node: not node.parser.name, recursive=True):
    """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)
    """
    if node.children:
        new_result = []
        for child in node.children:
            if child.children and condition(child):
                if recursive:
                    flatten(child, condition, recursive)
                new_result.extend(child.children)
            else:
                new_result.append(child)
        node.result = tuple(new_result)


def collapse(node):
    """Collapses all sub-nodes by replacing the node's result with it's
    string representation.
    """
    node.result = str(node)


@transformation_factory
298
def join_children(node, tag_names: List[str]):
299
300
301
302
303
304
    """Joins all children next to each other and with particular tag-
    names into a single child node with mock parser 'parser_name'.
    """
    result = []
    name, ptype = (tag_names[0].split(':') + [''])[:2]
    if node.children:
305
        i = 0
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
        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
def replace_content(node, func: Callable):  # Callable[[Node], ResultType]
    """Replaces the content of the node. ``func`` takes the node
    as an argument an returns the mapped result.
    """
    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)


359
360
361
def has_name(node, regexp: str) -> bool:
    """Checks a node's tag name against a regular expression."""
    return bool(re.match(regexp, node.tag_name))
362
363


364
365
366
def has_content(node, regexp: str) -> bool:
    """Checks a node's content against a regular expression."""
    return bool(re.match(regexp, str(node)))
367
368


369
@transformation_factory(Callable)
370
371
372
373
374
375
376
def apply_if(node, transformation: Callable, condition: Callable):
    """Applies a transformation only if a certain condition is met.
    """
    if condition(node):
        transformation(node)


377
378
379
@transformation_factory(slice)
def keep_children(node, section: slice = slice(None)):
    """Keeps only child-nodes which fall into a slice of the result field."""
380
    if node.children:
381
        node.result = node.children[section]
382
383
384


@transformation_factory(Callable)
385
def remove_children_if(node, condition: Callable, section: slice = slice(None)):
386
    """Removes all nodes from a slice of the result field if the function
387
    `condition(child_node)` evaluates to `True`."""
388
    if node.children:
389
390
391
392
        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]))
393
394
395
396
397


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)
398
399
remove_first = keep_children(slice(1, None))
remove_last = keep_children(slice(None, -1))
400
401
402
403
404
405
406
407
408
409
410
411
remove_brackets = keep_children(slice(1, -1))


@transformation_factory
def remove_tokens(node, tokens: AbstractSet[str] = frozenset()):
    """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."""
    remove_children_if(node, partial(is_token, tokens=tokens))


@transformation_factory
412
def remove_parser(node, regexp: str):
413
    """Removes children by 'tag name'."""
414
    remove_children_if(node, partial(has_name, regexp=regexp))
415
416
417


@transformation_factory
418
def remove_content(node, regexp: str):
419
    """Removes children depending on their string value."""
420
    remove_children_if(node, partial(has_content, regexp=regexp))
421
422
423
424


########################################################################
#
425
# AST semantic validation functions (EXPERIMENTAL!!!)
426
427
428
#
########################################################################

429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
@transformation_factory(Callable)
def assert_condition(node, condition: Callable, error_msg: str='') -> bool:
    """Checks for `condition`; adds an error message if condition is not met."""
    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
def assert_content(node, regexp: str):
    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):
#     if not has_name(node, regexp):
#         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)
#

466
467
468
469
470
471
472
473
474
475
476
477
478
479
480

@transformation_factory
def require(node, child_tags: AbstractSet[str]):
    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
def forbid(node, child_tags: AbstractSet[str]):
    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))