syntaxtree.py 24.5 KB
Newer Older
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
"""syntaxtree.py - syntax tree classes and transformation functions for 
converting the concrete into the abstract syntax tree for DHParser

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.
"""

20
import copy
21
import itertools
22
23
import os
from functools import partial
24
25
26
27
try:
    import regex as re
except ImportError:
    import re
28
from typing import Any, Callable, cast, Iterator, NamedTuple, Union, Tuple, List
29

30
from DHParser.toolkit import log_dir, expand_table, line_col, smart_list
31

Eckhart Arnold's avatar
Eckhart Arnold committed
32

33
34
__all__ = ['WHITESPACE_PTYPE',
           'TOKEN_PTYPE',
Eckhart Arnold's avatar
Eckhart Arnold committed
35
36
37
           'ZOMBIE_PARSER',
           'Error',
           'Node',
38
           'TransformerFunc',
39
40
           'key_parser_name',
           'key_tag_name',
41
           'traverse',
di68kap's avatar
di68kap committed
42
           'no_operation',
Eckhart Arnold's avatar
Eckhart Arnold committed
43
44
           'replace_by_single_child',
           'reduce_single_child',
45
           'replace_parser',
Eckhart Arnold's avatar
Eckhart Arnold committed
46
47
48
49
50
51
52
53
54
           'is_whitespace',
           'is_empty',
           'is_expendable',
           'is_token',
           'remove_children_if',
           'remove_whitespace',
           'remove_expendables',
           'remove_tokens',
           'flatten',
55
56
57
58
           'remove_enclosing_delimiters',
           'forbid',
           'require',
           'assert_content']
Eckhart Arnold's avatar
Eckhart Arnold committed
59
60


61
class MockParser:
62
63
64
65
66
67
68
69
70
71
    """
    MockParser objects can be used to reconstruct syntax trees from a
    serialized form like S-expressions or XML. Mock objects are needed,
    because Node objects require a parser object for instantiation.
    Mock objects have just enough properties to serve that purpose. 
    
    Mock objects should not be used for anything other than 
    syntax tree (re-)construction. In all other cases where a parser
    object substitute is needed, chose the singleton ZOMBIE_PARSER.
    """
72
    def __init__(self, name='', ptype='', pbases=frozenset()):
73
        assert not ptype or ptype[0] == ':'
74
        self.name = name
75
        self.ptype = ptype or ':' + self.__class__.__name__
76
        # self.pbases = pbases or {cls.__name__ for cls in inspect.getmro(self.__class__)}
77
78

    def __str__(self):
79
        return self.name or self.ptype
80
81
82


class ZombieParser(MockParser):
83
84
85
86
87
88
89
90
91
92
93
94
    """
    Serves as a substitute for a Parser instance.

    ``ZombieParser`` is the class of the singelton object
    ``ZOMBIE_PARSER``. The  ``ZOMBIE_PARSER`` has a name and can be
    called, but it never matches. It serves as a substitute where only
    these (or one of these properties) is needed, but no real Parser-
    object is instantiated.
    """
    alive = False

    def __init__(self):
95
        super(ZombieParser, self).__init__("__ZOMBIE__")
96
97
98
99
        assert not self.__class__.alive, "There can be only one!"
        assert self.__class__ == ZombieParser, "No derivatives, please!"
        self.__class__.alive = True

Eckhart Arnold's avatar
Eckhart Arnold committed
100
101
102
103
104
105
    def __copy__(self):
        return self

    def __deepcopy__(self, memo):
        return self

106
107
108
109
    def __call__(self, text):
        """Better call Saul ;-)"""
        return None, text

110
111
112
113

ZOMBIE_PARSER = ZombieParser()


114
115
116
117
118
# # Python 3.6:
# class Error(NamedTuple):
#     pos: int
#     msg: str
Error = NamedTuple('Error', [('pos', int), ('msg', str)])
119

120
121
ResultType = Union[Tuple['Node', ...], str]
SloppyResultType = Union[Tuple['Node', ...], 'Node', str, None]
122
123


124
125
126
127
128
129
130
131
132
133
134
135
class Node:
    """
    Represents a node in the concrete or abstract syntax tree.

    Attributes:
        tag_name (str):  The name of the node, which is either its
            parser's name or, if that is empty, the parser's class name
        result (str or tuple):  The result of the parser which
            generated this node, which can be either a string or a
            tuple of child nodes.
        children (tuple):  The tuple of child nodes or an empty tuple
            if there are no child nodes. READ ONLY!
136
137
138
139
140
141
        parser (Parser):  The parser which generated this node. 
            WARNING: In case you use mock syntax trees for testing or
            parser replacement during the AST-transformation: DO NOT
            rely on this being a real parser object in any phase after 
            parsing (i.e. AST-transformation and compiling), for 
            example by calling ``isinstance(node.parer, ...)``.
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
        errors (list):  A list of parser- or compiler-errors:
            tuple(position, string) attached to this node
        len (int):  The full length of the node's string result if the
            node is a leaf node or, otherwise, the concatenated string
            result's of its descendants. The figure always represents
            the length before AST-transformation ans will never change
            through AST-transformation. READ ONLY!
        pos (int):  the position of the node within the parsed text.

            The value of ``pos`` is -1 meaning invalid by default. 
            Setting this value will set the positions of all child
            nodes relative to this value.  

            To set the pos values of all nodes in a syntax tree, the
            pos value of the root node should be set to 0 right 
            after parsing.

            Other than that, this value should be considered READ ONLY. 
            At any rate, it should only be reassigned only during
            parsing stage and never during or after the
            AST-transformation.
    """
164
165

    def __init__(self, parser, result: SloppyResultType) -> None:
166
167
168
        """Initializes the ``Node``-object with the ``Parser``-Instance
        that generated the node and the parser's result.
        """
169
170
        self._result = ''  # type: ResultType
        self._errors = []  # type: List[str]
171
        self._children = ()  # type: Tuple['Node', ...]
172
173
174
175
        self._len = len(self.result) if not self.children else \
            sum(child._len for child in self.children)  # type: int
        # self.pos: int  = 0  # continuous updating of pos values
        self._pos = -1  # type: int
176
177
        self.result = result
        self.parser = parser or ZOMBIE_PARSER
178
179
        self.error_flag = any(r.error_flag for r in self.children) \
            if self.children else False  # type: bool
180
181
182

    def __str__(self):
        if self.children:
183
            return "".join(str(child) for child in self.children)
184
185
        return str(self.result)

186
    def __eq__(self, other):
187
188
        # return str(self.parser) == str(other.parser) and self.result == other.result
        return self.tag_name == other.tag_name and self.result == other.result
189

190
    def __hash__(self):
191
        return hash(self.tag_name)
192
193
194
195
196
197
198

    def __deepcopy__(self, memodict={}):
        result = copy.deepcopy(self.result)
        other = Node(self.parser, result)
        other._pos = self._pos
        return other

199
    @property
200
    def tag_name(self) -> str:
201
202
        return self.parser.name or self.parser.ptype
        # ONLY FOR DEBUGGING: return self.parser.name + ':' + self.parser.ptype
203
204

    @property
205
    def result(self) -> ResultType:
206
207
208
        return self._result

    @result.setter
209
    def result(self, result: SloppyResultType):
210
211
212
213
        # # made obsolete by static type checking with mypy is done
        # assert ((isinstance(result, tuple) and all(isinstance(child, Node) for child in result))
        #         or isinstance(result, Node)
        #         or isinstance(result, str)), str(result)
214
        self._result = (result,) if isinstance(result, Node) else result or ''
215
216
        self._children = cast(Tuple['Node', ...], self._result) \
            if isinstance(self._result, tuple) else cast(Tuple['Node', ...], ())
217
218

    @property
219
    def children(self) -> Tuple['Node', ...]:
220
221
222
        return self._children

    @property
223
    def len(self) -> int:
224
        # DEBUGGING:  print(self.tag_name, str(self.pos), str(self._len), str(self)[:10].replace('\n','.'))
225
226
227
        return self._len

    @property
228
    def pos(self) -> int:
229
230
231
232
        assert self._pos >= 0, "position value not initialized!"
        return self._pos

    @pos.setter
233
234
    def pos(self, pos: int):
        # assert isinstance(pos, int)
235
236
237
238
239
240
241
        self._pos = pos
        offset = 0
        for child in self.children:
            child.pos = pos + offset
            offset += child.len

    @property
242
    def errors(self) -> List[Error]:
243
244
        return [Error(self.pos, err) for err in self._errors]

245
    def _tree_repr(self, tab, openF, closeF, dataF=lambda s: s) -> str:
246
247
248
249
250
251
252
253
        """
        Generates a tree representation of this node and its children
        in string from.

        The kind ot tree-representation that is determined by several
        function parameters. This could be an XML-representation or a
        lisp-like S-expression.

254
        Args:
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
            tab (str):  The indentation string, e.g. '\t' or '    '
            openF:  (Node->str) A function that returns an opening
                string (e.g. an XML-tag_name) for a given node
            closeF:  (Node->str) A function that returns a closeF
                string (e.g. an XML-tag_name) for a given node.
            dataF:  (str->str) A function that filters the data string
                before printing, e.g. to add quotation marks

        Returns (str):
            A string that contains a (serialized) tree representation
            of the node and its children.
        """
        head = openF(self)
        tail = closeF(self)

        if not self.result:
            return head + tail

        head = head + '\n'  # place the head, tail and content
        tail = '\n' + tail  # of the node on different lines

        if self.children:
            content = []
278
            for child in self.children:
279
280
281
282
                subtree = child._tree_repr(tab, openF, closeF, dataF).split('\n')
                content.append('\n'.join((tab + s) for s in subtree))
            return head + '\n'.join(content) + tail

283
284
        res = cast(str, self.result)  # safe, because if there are no children, result is a string
        if head[0] == "<" and res.find('\n') < 0:
285
286
287
            # for XML: place tags for leaf-nodes on one line if possible
            return head[:-1] + self.result + tail[1:]
        else:
288
            return head + '\n'.join([tab + dataF(s) for s in res.split('\n')]) + tail
289

290
    def as_sexpr(self, src=None) -> str:
291
292
293
        """
        Returns content as S-expression, i.e. in lisp-like form.

294
        Args:
295
296
297
            src:  The source text or `None`. In case the source text is
                given the position of the element in the text will be
                reported as line and column.
298
299
            prettyprint(bool):  True (default), if pretty printing 
                of leaf nodes shall be applied for better readability.
300
301
        """

302
        def opening(node) -> str:
303
304
305
306
307
308
309
310
311
312
313
314
315
316
            s = '(' + node.tag_name
            # s += " '(pos %i)" % node.pos
            if src:
                s += " '(pos %i  %i %i)" % (node.pos, *line_col(src, node.pos))
            if node.errors:
                s += " '(err '(%s))" % ' '.join(str(err).replace('"', r'\"')
                                                for err in node.errors)
            return s

        def pretty(s):
            return '"%s"' % s if s.find('"') < 0 \
                else "'%s'" % s if s.find("'") < 0 \
                else '"%s"' % s.replace('"', r'\"')

317
        return self._tree_repr('    ', opening, lambda node: ')', pretty)  # pretty if prettyprint else lambda s: s)
318

319
    def as_xml(self, src=None) -> str:
320
321
322
        """
        Returns content as XML-tree.

323
        Args:
324
325
326
327
328
            src:  The source text or `None`. In case the source text is
                given the position will also be reported as line and
                column.
        """

329
        def opening(node) -> str:
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
            s = '<' + node.tag_name
            # s += ' pos="%i"' % node.pos
            if src:
                s += ' line="%i" col="%i"' % line_col(src, node.pos)
            if node.errors:
                s += ' err="%s"' % ''.join(str(err).replace('"', r'\"') for err in node.errors)
            s += ">"
            return s

        def closing(node):
            s = '</' + node.tag_name + '>'
            return s

        return self._tree_repr('    ', opening, closing)

345
    def add_error(self, error_str) -> 'Node':
346
347
348
349
        self._errors.append(error_str)
        self.error_flag = True
        return self

350
351
352
353
354
355
356
357
358
    def propagate_error_flags(self):
        """ Recursively propagates error flags set on child nodes to its
        parents. This can be used if errors are added to descendant 
        nodes after syntaxtree construction, i.e. in the compile phase.
        """
        for child in self.children:
            child.propagate_error_flags()
            self.error_flag |= child.error_flag

359
    def collect_errors(self, clear_errors=False) -> List[Error]:
360
361
362
363
364
        """
        Returns all errors of this node or any child node in the form
        of a set of tuples (position, error_message), where position
        is always relative to this node.
        """
365
366
367
368
369
        errors = self.errors
        if clear_errors:
            self._errors = []
            self.error_flag = False
        if self.children:
370
            for child in self.children:
371
                errors.extend(child.collect_errors(clear_errors))
372
        return errors
373

Eckhart Arnold's avatar
Eckhart Arnold committed
374
375
376
377
    def log(self, log_file_name):
        st_file_name = log_file_name
        with open(os.path.join(log_dir(), st_file_name), "w", encoding="utf-8") as f:
            f.write(self.as_sexpr())
378

379
    def find(self, match_function) -> Iterator['Node']:
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
        """Finds nodes in the tree that match a specific criterion.
        
        ``find`` is a generator that yields all nodes for which the
        given ``match_function`` evaluates to True. The tree is 
        traversed pre-order.
        
        Args:
            match_function (function): A function  that takes as Node
                object as argument and returns True or False
        Yields:
            Node: all nodes of the tree for which 
            ``match_function(node)`` returns True
        """
        if match_function(self):
            yield self
        else:
            for child in self.children:
                for nd in child.find(match_function):
                    yield nd

400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
    # def range(self, match_first, match_last):
    #     """Iterates over the range of nodes, starting from the first
    #     node for which ``match_first`` becomes True until the first node
    #     after this one for which ``match_last`` becomes true or until
    #     the end if it never does.
    #
    #     Args:
    #         match_first (function): A function  that takes as Node
    #             object as argument and returns True or False
    #         match_last (function): A function  that takes as Node
    #             object as argument and returns True or False
    #     Yields:
    #         Node: all nodes of the tree for which
    #         ``match_function(node)`` returns True
    #     """


417
    def navigate(self, path):
418
419
420
421
422
        """Yields the results of all descendant elements matched by
        ``path``, e.g.
        'd/s' yields 'l' from (d (s l)(e (r x1) (r x2))
        'e/r' yields 'x1', then 'x2'
        'e'   yields (r x1)(r x2)
423

424
        Args:
425
426
            path (str):  The path of the object, e.g. 'a/b/c'. The
                components of ``path`` can be regular expressions
427
428
429
430
431

        Returns:
            The object at the path, either a string or a Node or
            ``None``, if the path did not match.
        """
432
433
434
435
        def nav(node, pl):
            if pl:
                return itertools.chain(nav(child, pl[1:]) for child in node.children
                                       if re.match(pl[0], child.tag_name))
436
            else:
437
438
                return self.result,
        return nav(path.split('/'))
439
440
441
442
443
444
445
446
447


########################################################################
#
# syntax tree transformation functions
#
########################################################################


448
449
450
TransformerFunc = Union[Callable[[Node], Any], partial]


451
452
WHITESPACE_PTYPE = ':Whitespace'
TOKEN_PTYPE = ':Token'
453
454


455
def key_parser_name(node) -> str:
456
457
458
    return node.parser.name


459
def key_tag_name(node) -> str:
460
461
462
    return node.tag_name


463
def traverse(root_node, processing_table, key_func=key_tag_name):
464
    """Traverses the snytax tree starting with the given ``node`` depth
465
    first and applies the sequences of callback-functions registered
466
467
    in the ``calltable``-dictionary.
    
468
469
470
471
472
473
474
475
476
477
478
479
480
    The most important use case is the transformation of a concrete
    syntax tree into an abstract tree (AST). But it is also imaginable
    to emloy 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)

481
    Args:
di68kap's avatar
di68kap committed
482
        root_node (Node): The root-node of the syntax tree to be traversed 
483
        processing_table (dict): node key -> sequence of functions that
di68kap's avatar
di68kap committed
484
485
            will be applied to matching nodes in order. This dictionary
            is interpreted as a ``compact_table``. See 
486
            ``toolkit.expand_table`` or ``EBNFCompiler.EBNFTransTable``
487
488
        key_func (function): A mapping key_func(node) -> keystr. The default
            key_func yields node.parser.name.
489
490
491
492
493
            
    Example:
        table = { "term": [replace_by_single_child, flatten], 
            "factor, flowmarker, retrieveop": replace_by_single_child }
        traverse(node, table)
494
    """
di68kap's avatar
di68kap committed
495
    # normalize processing_table entries by turning single values into lists
496
    # with a single value
497
    table = {name: smart_list(call) for name, call in list(processing_table.items())}
498
499
    table = expand_table(table)

di68kap's avatar
di68kap committed
500
501
502
    def traverse_recursive(node):
        if node.children:
            for child in node.result:
503
                traverse_recursive(child)
504
                node.error_flag |= child.error_flag  # propagate error flag
505
506
        sequence = table.get('+', []) + \
                   table.get(key_func(node), table.get('*', [])) + \
507
                   table.get('~', [])
508
        # '+' always called (before any other processing function)
509
        # '*' called for those nodes for which no (other) processing function appears in the table
510
        # '~' always called (after any other processing function)
511
        for call in sequence:
di68kap's avatar
di68kap committed
512
            call(node)
513

di68kap's avatar
di68kap committed
514
    traverse_recursive(root_node)
515
516


517
518
519
520
521
522
523
524
# Note on processing functions: If processing functions receive more
# than one parameter, the ``node``-parameter should always be the
# last parameter to ease partial function application, e.g.:
# def replace_parser(name, node):
#    ...
# processing_func = partial(replace_parser, "special")


di68kap's avatar
di68kap committed
525
def no_operation(node):
526
527
528
529
530
531
    pass


# ------------------------------------------------
#
# rearranging transformations:
di68kap's avatar
di68kap committed
532
533
#     - tree may be rearranged (e.g.flattened)
#     - nodes that are not leaves may be dropped
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
#     - order is preserved
#     - all leaves are kept
#
# ------------------------------------------------


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
        node._errors.extend(node.result[0].errors)
        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:
        node._errors.extend(node.result[0].errors)
        node.result = node.result[0].result


562
def replace_parser(name, node):
563
    """Replaces the parser of a Node with a mock parser with the given
564
565
566
567
568
    name.

    Parameters:
        name(str): "NAME:PTYPE" of the surogate. The ptype is optional
        node(Node): The node where the parser shall be replaced
di68kap's avatar
di68kap committed
569
    """
570
    name, ptype = (name.split(':') + [''])[:2]
571
    node.parser = MockParser(name, ptype)
di68kap's avatar
di68kap committed
572
573


574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
def flatten(node):
    """Recursively flattens all unnamed sub-nodes, in case there is more
    than one sub-node present. Flattening means that
    wherever a node has child nodes, the child nodes are inserted in place
    of the node. In other words, all leaves of this node and its child nodes
    are collected in-order as direct children of this node.
    This is meant to achieve these kinds of structural transformation:
        (1 (+ 2) (+ 3)     ->   (1 + 2 + 3)
        (1 (+ (2 + (3))))  ->   (1 + 2 + 3)

    Warning: Use with care. Du tue its recursive nature, flattening can
    have unexpected side-effects.
    """
    if node.children:
        new_result = []
        for child in node.children:
            if not child.parser.name and child.children:
                assert child.children, node.as_sexpr()
                flatten(child)
                new_result.extend(child.result)
            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)


606
607
608
609
610
611
612
613
614
615
616
617
618
619
# ------------------------------------------------
#
# destructive transformations:
#     - tree may be rearranged (flattened),
#     - order is preserved
#     - but (irrelevant) leaves may be dropped
#     - errors of dropped leaves will be lost
#
# ------------------------------------------------


def is_whitespace(node):
    """Removes whitespace and comments defined with the
    ``@comment``-directive."""
620
    return node.parser.ptype == WHITESPACE_PTYPE
621
622
623
624
625
626
627


def is_empty(node):
    return not node.result


def is_expendable(node):
628
    return is_empty(node) or is_whitespace(node)
629
630


631
def is_token(token_set, node):
632
    return node.parser.ptype == TOKEN_PTYPE and (not token_set or node.result in token_set)
633
634


635
def remove_children_if(condition, node):
636
637
638
639
640
641
    """Removes all nodes from the result field if the function 
    ``condition(child_node)`` evaluates to ``True``."""
    if node.children:
        node.result = tuple(c for c in node.children if not condition(c))


642
remove_whitespace = partial(remove_children_if, is_whitespace)
643
# remove_scanner_tokens = partial(remove_children_if, condition=is_scanner_token)
644
remove_expendables = partial(remove_children_if, is_expendable)
645
646


647
def remove_tokens(tokens, node):
648
649
650
651
    """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.
    """
652
    remove_children_if(partial(is_token, tokens), node)
653
654


655
def remove_enclosing_delimiters(node):
656
657
658
659
660
661
662
663
    """Removes any enclosing delimiters from a structure (e.g. quotation marks
    from a literal or braces from a group).
    """
    if len(node.children) >= 3:
        assert not node.children[0].children and not node.children[-1].children, node.as_sexpr()
        node.result = node.result[1:-1]


664
def map_content(func, node):
665
666
667
668
669
670
    """Replaces the content of the node. ``func`` takes the node
    as an argument an returns the mapped result.
    """
    node.result = func(node.result)


671
672
########################################################################
#
673
# AST semantic validation functions
di68kap's avatar
di68kap committed
674
# EXPERIMENTAL!
675
676
677
678
#
########################################################################


679
def require(child_tag, node):
680
    for child in node.children:
681
        if child.tag_name not in child_tag:
682
            node.add_error('Element "%s" is not allowed inside "%s".' %
683
                           (child.parser.name, node.parser.name))
684
685


686
def forbid(child_tags, node):
687
    for child in node.children:
688
        if child.tag_name in child_tags:
689
            node.add_error('Element "%s" cannot be nested inside "%s".' %
690
                           (child.parser.name, node.parser.name))
691
692


693
def assert_content(regex, node):
694
695
696
    content = str(node)
    if not re.match(regex, content):
        node.add_error('Element "%s" violates %s on %s' %
697
                       (node.parser.name, str(regex), content))