syntaxtree.py 27.9 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.
"""

Eckhart Arnold's avatar
Eckhart Arnold committed
20
import abc
21
import copy
22
import inspect
23
import itertools
24
import os
25
from functools import partial, singledispatch
26 27 28 29
try:
    import regex as re
except ImportError:
    import re
30 31
from typing import AbstractSet, Any, ByteString, Callable, cast, Container, Iterator, List, \
    NamedTuple, Sequence, Union, Text, Tuple
32

33
from DHParser.toolkit import log_dir, expand_table, line_col, smart_list
34

Eckhart Arnold's avatar
Eckhart Arnold committed
35

36 37
__all__ = ['WHITESPACE_PTYPE',
           'TOKEN_PTYPE',
Eckhart Arnold's avatar
Eckhart Arnold committed
38
           'ZOMBIE_PARSER',
Eckhart Arnold's avatar
Eckhart Arnold committed
39
           'ParserBase',
Eckhart Arnold's avatar
Eckhart Arnold committed
40 41
           'Error',
           'Node',
42
           'TransformationFunc',
Eckhart Arnold's avatar
Eckhart Arnold committed
43
           'transformation_factory',
44 45
           'key_parser_name',
           'key_tag_name',
46
           'traverse',
47
           'no_transformation',
Eckhart Arnold's avatar
Eckhart Arnold committed
48 49
           'replace_by_single_child',
           'reduce_single_child',
50
           'replace_parser',
Eckhart Arnold's avatar
Eckhart Arnold committed
51 52 53 54 55 56 57 58 59
           'is_whitespace',
           'is_empty',
           'is_expendable',
           'is_token',
           'remove_children_if',
           'remove_whitespace',
           'remove_expendables',
           'remove_tokens',
           'flatten',
60 61 62 63
           'remove_enclosing_delimiters',
           'forbid',
           'require',
           'assert_content']
Eckhart Arnold's avatar
Eckhart Arnold committed
64 65


Eckhart Arnold's avatar
Eckhart Arnold committed
66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84
class ParserBase:
    """
    ParserBase is the base class for all real and mock parser classes.
    It is defined here, because Node objects require a parser object
    for instantiation.
    """
    def __init__(self, name=''):  # , pbases=frozenset()):
        self.name = name  # type: str
        self._ptype = ':' + self.__class__.__name__  # type: str

    def __str__(self):
        return self.name or self.ptype

    @property
    def ptype(self) -> str:
        return self._ptype


class MockParser(ParserBase):
85 86
    """
    MockParser objects can be used to reconstruct syntax trees from a
Eckhart Arnold's avatar
Eckhart Arnold committed
87 88
    serialized form like S-expressions or XML. Mock objects can mimic
    different parser types by assigning them a ptype on initialization.
89 90 91 92 93
    
    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.
    """
Eckhart Arnold's avatar
Eckhart Arnold committed
94
    def __init__(self, name='', ptype=''):  # , pbases=frozenset()):
95
        assert not ptype or ptype[0] == ':'
Eckhart Arnold's avatar
Eckhart Arnold committed
96
        super(MockParser, self).__init__(name)
97
        self.name = name
Eckhart Arnold's avatar
Eckhart Arnold committed
98
        self._ptype = ptype or ':' + self.__class__.__name__
99 100

    def __str__(self):
101
        return self.name or self.ptype
102 103 104


class ZombieParser(MockParser):
105 106 107 108 109 110 111 112 113 114 115 116
    """
    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):
117
        super(ZombieParser, self).__init__("__ZOMBIE__")
118 119 120 121
        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
122 123 124 125 126 127
    def __copy__(self):
        return self

    def __deepcopy__(self, memo):
        return self

128 129 130 131
    def __call__(self, text):
        """Better call Saul ;-)"""
        return None, text

132 133 134 135

ZOMBIE_PARSER = ZombieParser()


136 137 138 139 140
# # Python 3.6:
# class Error(NamedTuple):
#     pos: int
#     msg: str
Error = NamedTuple('Error', [('pos', int), ('msg', str)])
141

Eckhart Arnold's avatar
Eckhart Arnold committed
142 143
StrictResultType = Union[Tuple['Node', ...], str]
ResultType = Union[Tuple['Node', ...], 'Node', str, None]
144 145


146 147 148 149 150 151 152 153 154 155 156 157
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!
158 159 160 161 162 163
        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, ...)``.
164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185
        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.
    """
186

Eckhart Arnold's avatar
Eckhart Arnold committed
187
    def __init__(self, parser, result: ResultType) -> None:
188 189 190
        """Initializes the ``Node``-object with the ``Parser``-Instance
        that generated the node and the parser's result.
        """
Eckhart Arnold's avatar
Eckhart Arnold committed
191
        self._result = ''  # type: StrictResultType
192
        self._errors = []  # type: List[str]
193
        self._children = ()  # type: Tuple['Node', ...]
194 195 196 197
        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
198 199
        self.result = result
        self.parser = parser or ZOMBIE_PARSER
200 201
        self.error_flag = any(r.error_flag for r in self.children) \
            if self.children else False  # type: bool
202 203 204

    def __str__(self):
        if self.children:
205
            return "".join(str(child) for child in self.children)
Eckhart Arnold's avatar
Eckhart Arnold committed
206
        return str(self.result) if self.parser.name != "__ZOMBIE__" else ''
207

208
    def __eq__(self, other):
209 210
        # return str(self.parser) == str(other.parser) and self.result == other.result
        return self.tag_name == other.tag_name and self.result == other.result
211

212
    def __hash__(self):
213
        return hash(self.tag_name)
214 215 216 217 218 219 220

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

221
    @property
222
    def tag_name(self) -> str:
223 224
        return self.parser.name or self.parser.ptype
        # ONLY FOR DEBUGGING: return self.parser.name + ':' + self.parser.ptype
225 226

    @property
Eckhart Arnold's avatar
Eckhart Arnold committed
227
    def result(self) -> StrictResultType:
228 229 230
        return self._result

    @result.setter
Eckhart Arnold's avatar
Eckhart Arnold committed
231
    def result(self, result: ResultType):
232 233 234 235
        # # 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)
236
        self._result = (result,) if isinstance(result, Node) else result or ''
237 238
        self._children = cast(Tuple['Node', ...], self._result) \
            if isinstance(self._result, tuple) else cast(Tuple['Node', ...], ())
239 240

    @property
241
    def children(self) -> Tuple['Node', ...]:
242 243 244
        return self._children

    @property
245
    def len(self) -> int:
246
        # DEBUGGING:  print(self.tag_name, str(self.pos), str(self._len), str(self)[:10].replace('\n','.'))
247 248 249
        return self._len

    @property
250
    def pos(self) -> int:
251 252 253 254
        assert self._pos >= 0, "position value not initialized!"
        return self._pos

    @pos.setter
255 256
    def pos(self, pos: int):
        # assert isinstance(pos, int)
257 258 259 260 261 262 263
        self._pos = pos
        offset = 0
        for child in self.children:
            child.pos = pos + offset
            offset += child.len

    @property
264
    def errors(self) -> List[Error]:
265 266
        return [Error(self.pos, err) for err in self._errors]

267
    def _tree_repr(self, tab, openF, closeF, dataF=lambda s: s) -> str:
268 269 270 271 272 273 274 275
        """
        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.

276
        Args:
277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299
            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 = []
300
            for child in self.children:
301 302 303 304
                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

305 306
        res = cast(str, self.result)  # safe, because if there are no children, result is a string
        if head[0] == "<" and res.find('\n') < 0:
307 308 309
            # for XML: place tags for leaf-nodes on one line if possible
            return head[:-1] + self.result + tail[1:]
        else:
310
            return head + '\n'.join([tab + dataF(s) for s in res.split('\n')]) + tail
311

312
    def as_sexpr(self, src=None) -> str:
313 314 315
        """
        Returns content as S-expression, i.e. in lisp-like form.

316
        Args:
317 318 319
            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.
320 321
            prettyprint(bool):  True (default), if pretty printing 
                of leaf nodes shall be applied for better readability.
322 323
        """

324
        def opening(node) -> str:
325 326 327 328 329 330 331 332 333 334 335 336 337 338
            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'\"')

339
        return self._tree_repr('    ', opening, lambda node: ')', pretty)  # pretty if prettyprint else lambda s: s)
340

341
    def as_xml(self, src=None) -> str:
342 343 344
        """
        Returns content as XML-tree.

345
        Args:
346 347 348 349 350
            src:  The source text or `None`. In case the source text is
                given the position will also be reported as line and
                column.
        """

351
        def opening(node) -> str:
352 353 354 355 356 357 358 359 360 361 362 363 364 365 366
            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)

367
    def add_error(self, error_str) -> 'Node':
368 369 370 371
        self._errors.append(error_str)
        self.error_flag = True
        return self

372 373 374 375 376 377 378 379 380
    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

381
    def collect_errors(self, clear_errors=False) -> List[Error]:
382 383 384 385 386
        """
        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.
        """
387 388 389 390 391
        errors = self.errors
        if clear_errors:
            self._errors = []
            self.error_flag = False
        if self.children:
392
            for child in self.children:
393
                errors.extend(child.collect_errors(clear_errors))
394
        return errors
395

Eckhart Arnold's avatar
Eckhart Arnold committed
396 397 398 399
    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())
400

401
    def find(self, match_function) -> Iterator['Node']:
402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421
        """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

422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438
    # 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
    #     """


439
    def navigate(self, path):
440 441 442 443 444
        """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)
445

446
        Args:
447 448
            path (str):  The path of the object, e.g. 'a/b/c'. The
                components of ``path`` can be regular expressions
449 450 451 452 453

        Returns:
            The object at the path, either a string or a Node or
            ``None``, if the path did not match.
        """
454 455 456 457
        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))
458
            else:
459 460
                return self.result,
        return nav(path.split('/'))
461 462 463 464 465 466 467 468 469


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


470 471 472 473
TransformationFunc = Union[Callable[[Node], Any], partial]


def transformation_factory(t=None):
Eckhart Arnold's avatar
Eckhart Arnold committed
474 475 476 477 478 479 480 481 482 483 484 485 486 487 488
    """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.
489 490 491 492 493

    Usage:
        @transformation_factory(AbtractSet[str])
        def remove_tokens(node, tokens):
            ...
Eckhart Arnold's avatar
Eckhart Arnold committed
494
      or, alternatively:
495 496 497
        @transformation_factory
        def remove_tokens(node, tokens: AbstractSet[str]):
            ...
Eckhart Arnold's avatar
Eckhart Arnold committed
498 499 500 501 502

    Example:
        trans_table = { 'expression': remove_tokens('+', '-') }
      instead of:
        trans_table = { 'expression': partial(remove_tokens, tokens={'+', '-'}) }
503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533
    """

    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)):
Eckhart Arnold's avatar
Eckhart Arnold committed
534 535 536 537
        # Provide for the case that transformation_factory has been
        # written as plain decorator and not as a function call that
        # returns the decorator proper.
        func = t;  t = None
538 539 540
        return decorator(func)
    else:
        return decorator
541 542


543 544
WHITESPACE_PTYPE = ':Whitespace'
TOKEN_PTYPE = ':Token'
545 546


547
def key_parser_name(node) -> str:
548 549 550
    return node.parser.name


551
def key_tag_name(node) -> str:
552 553 554
    return node.tag_name


555
def traverse(root_node, processing_table, key_func=key_tag_name) -> None:
556
    """Traverses the snytax tree starting with the given ``node`` depth
557
    first and applies the sequences of callback-functions registered
558 559
    in the ``calltable``-dictionary.
    
560 561
    The most important use case is the transformation of a concrete
    syntax tree into an abstract tree (AST). But it is also imaginable
562
    to employ tree-traversal for the semantic analysis of the AST.
563 564 565 566 567 568 569 570 571 572

    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)

573
    Args:
di68kap's avatar
di68kap committed
574
        root_node (Node): The root-node of the syntax tree to be traversed 
575
        processing_table (dict): node key -> sequence of functions that
di68kap's avatar
di68kap committed
576 577
            will be applied to matching nodes in order. This dictionary
            is interpreted as a ``compact_table``. See 
578
            ``toolkit.expand_table`` or ``EBNFCompiler.EBNFTransTable``
579 580
        key_func (function): A mapping key_func(node) -> keystr. The default
            key_func yields node.parser.name.
581 582 583 584 585
            
    Example:
        table = { "term": [replace_by_single_child, flatten], 
            "factor, flowmarker, retrieveop": replace_by_single_child }
        traverse(node, table)
586
    """
587 588 589 590
    # commented, because this approach is too error prone!
    # def funclist(call):
    #     return [as_partial(func) for func in smart_list(call)]

di68kap's avatar
di68kap committed
591
    # normalize processing_table entries by turning single values into lists
592
    # with a single value
593
    table = {name: smart_list(call) for name, call in list(processing_table.items())}
594 595
    table = expand_table(table)

di68kap's avatar
di68kap committed
596 597 598
    def traverse_recursive(node):
        if node.children:
            for child in node.result:
599
                traverse_recursive(child)
600
                node.error_flag |= child.error_flag  # propagate error flag
601 602
        sequence = table.get('+', []) + \
                   table.get(key_func(node), table.get('*', [])) + \
603
                   table.get('~', [])
604
        # '+' always called (before any other processing function)
605
        # '*' called for those nodes for which no (other) processing function appears in the table
606
        # '~' always called (after any other processing function)
607
        for call in sequence:
di68kap's avatar
di68kap committed
608
            call(node)
609

di68kap's avatar
di68kap committed
610
    traverse_recursive(root_node)
611 612


613
def no_transformation(node):
614 615 616 617 618 619
    pass


# ------------------------------------------------
#
# rearranging transformations:
di68kap's avatar
di68kap committed
620 621
#     - tree may be rearranged (e.g.flattened)
#     - nodes that are not leaves may be dropped
622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649
#     - 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


650 651
@transformation_factory
def replace_parser(node, name: str):
652
    """Replaces the parser of a Node with a mock parser with the given
653 654 655 656 657
    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
658
    """
659
    name, ptype = (name.split(':') + [''])[:2]
660
    node.parser = MockParser(name, ptype)
di68kap's avatar
di68kap committed
661 662


663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694
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)


695 696 697 698 699 700 701 702 703 704 705 706 707 708
# ------------------------------------------------
#
# 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."""
709
    return node.parser.ptype == WHITESPACE_PTYPE
710 711 712 713 714 715 716


def is_empty(node):
    return not node.result


def is_expendable(node):
717
    return is_empty(node) or is_whitespace(node)
718 719


720
def is_token(node, tokens: AbstractSet[str] = frozenset()) -> bool:
721
    return node.parser.ptype == TOKEN_PTYPE and (not tokens or node.result in tokens)
722 723


724
@transformation_factory(Callable)  # @singledispatch
725
def remove_children_if(node, condition):
726 727 728 729 730 731
    """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))


732 733
remove_whitespace = partial(remove_children_if, condition=is_whitespace)
remove_expendables = partial(remove_children_if, condition=is_expendable)
734 735


736 737 738 739 740
# remove_scanner_tokens = partial(remove_children_if, condition=is_scanner_token)


@transformation_factory
def remove_tokens(node, tokens: AbstractSet[str] = frozenset()):
741 742 743 744
    """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.
    """
745
    remove_children_if(node, partial(is_token, tokens=tokens))
746 747


748
def remove_enclosing_delimiters(node):
749 750 751 752 753 754 755 756
    """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]


Eckhart Arnold's avatar
Eckhart Arnold committed
757
def map_content(node, func: Callable[[Node], ResultType]):
758 759 760 761 762 763
    """Replaces the content of the node. ``func`` takes the node
    as an argument an returns the mapped result.
    """
    node.result = func(node.result)


764 765
########################################################################
#
766
# AST semantic validation functions
di68kap's avatar
di68kap committed
767
# EXPERIMENTAL!
768 769 770 771
#
########################################################################


772 773
@transformation_factory
def require(node, child_tags: AbstractSet[str]):
774
    for child in node.children:
775
        if child.tag_name not in child_tags:
776
            node.add_error('Element "%s" is not allowed inside "%s".' %
777
                           (child.parser.name, node.parser.name))
778 779


780 781
@transformation_factory
def forbid(node, child_tags: AbstractSet[str]):
782
    for child in node.children:
783
        if child.tag_name in child_tags:
784
            node.add_error('Element "%s" cannot be nested inside "%s".' %
785
                           (child.parser.name, node.parser.name))
786 787


788 789
@transformation_factory
def assert_content(node, regex: str):
790 791 792
    content = str(node)
    if not re.match(regex, content):
        node.add_error('Element "%s" violates %s on %s' %
793
                       (node.parser.name, str(regex), content))
794