syntaxtree.py 30.5 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 inspect
22
import os
23
from functools import partial, singledispatch
24 25 26 27
try:
    import regex as re
except ImportError:
    import re
28 29 30 31 32 33
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
34

35
from DHParser.toolkit import log_dir, expand_table, line_col, smart_list
36

Eckhart Arnold's avatar
Eckhart Arnold committed
37

38 39
__all__ = ['WHITESPACE_PTYPE',
           'TOKEN_PTYPE',
Eckhart Arnold's avatar
Eckhart Arnold committed
40
           'ZOMBIE_PARSER',
Eckhart Arnold's avatar
Eckhart Arnold committed
41
           'ParserBase',
Eckhart Arnold's avatar
Eckhart Arnold committed
42 43
           'Error',
           'Node',
44
           'TransformationFunc',
Eckhart Arnold's avatar
Eckhart Arnold committed
45
           'transformation_factory',
46 47
           'key_parser_name',
           'key_tag_name',
48
           'traverse',
Eckhart Arnold's avatar
Eckhart Arnold committed
49 50
           'replace_by_single_child',
           'reduce_single_child',
51
           'reduce_children',
52
           'replace_parser',
Eckhart Arnold's avatar
Eckhart Arnold committed
53 54 55 56
           'is_whitespace',
           'is_empty',
           'is_expendable',
           'is_token',
57 58
           'has_name',
           'has_content',
Eckhart Arnold's avatar
Eckhart Arnold committed
59
           'remove_children_if',
60 61 62 63
           'remove_children',
           'remove_content',
           'remove_first',
           'remove_last',
Eckhart Arnold's avatar
Eckhart Arnold committed
64
           'remove_whitespace',
Eckhart Arnold's avatar
Eckhart Arnold committed
65
           'remove_empty',
Eckhart Arnold's avatar
Eckhart Arnold committed
66 67 68
           'remove_expendables',
           'remove_tokens',
           'flatten',
69 70 71
           'forbid',
           'require',
           'assert_content']
Eckhart Arnold's avatar
Eckhart Arnold committed
72 73


Eckhart Arnold's avatar
Eckhart Arnold committed
74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92
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):
93 94
    """
    MockParser objects can be used to reconstruct syntax trees from a
Eckhart Arnold's avatar
Eckhart Arnold committed
95 96
    serialized form like S-expressions or XML. Mock objects can mimic
    different parser types by assigning them a ptype on initialization.
97 98 99 100 101
    
    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
102
    def __init__(self, name='', ptype=''):  # , pbases=frozenset()):
103
        assert not ptype or ptype[0] == ':'
Eckhart Arnold's avatar
Eckhart Arnold committed
104
        super(MockParser, self).__init__(name)
105
        self.name = name
Eckhart Arnold's avatar
Eckhart Arnold committed
106
        self._ptype = ptype or ':' + self.__class__.__name__
107 108

    def __str__(self):
109
        return self.name or self.ptype
110 111 112


class ZombieParser(MockParser):
113 114 115 116 117 118 119 120 121 122 123 124
    """
    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):
125
        super(ZombieParser, self).__init__("__ZOMBIE__")
126 127 128 129
        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
130 131 132 133 134 135
    def __copy__(self):
        return self

    def __deepcopy__(self, memo):
        return self

136 137 138 139
    def __call__(self, text):
        """Better call Saul ;-)"""
        return None, text

140 141 142 143

ZOMBIE_PARSER = ZombieParser()


144 145 146 147 148
# # Python 3.6:
# class Error(NamedTuple):
#     pos: int
#     msg: str
Error = NamedTuple('Error', [('pos', int), ('msg', str)])
149

Eckhart Arnold's avatar
Eckhart Arnold committed
150 151
StrictResultType = Union[Tuple['Node', ...], str]
ResultType = Union[Tuple['Node', ...], 'Node', str, None]
152 153


154 155 156 157 158 159 160 161 162 163 164 165
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!
166 167 168 169 170 171
        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, ...)``.
172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193
        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.
    """
194

Eckhart Arnold's avatar
Eckhart Arnold committed
195
    def __init__(self, parser, result: ResultType) -> None:
196 197 198
        """Initializes the ``Node``-object with the ``Parser``-Instance
        that generated the node and the parser's result.
        """
Eckhart Arnold's avatar
Eckhart Arnold committed
199
        self._result = ''  # type: StrictResultType
200
        self._errors = []  # type: List[str]
201
        self._children = ()  # type: Tuple['Node', ...]
202
        self.result = result
203 204 205 206
        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
207
        self.parser = parser or ZOMBIE_PARSER
208 209
        self.error_flag = any(r.error_flag for r in self.children) \
            if self.children else False  # type: bool
210 211 212

    def __str__(self):
        if self.children:
213
            return "".join(str(child) for child in self.children)
214 215 216 217 218 219 220 221
        return str(self.result)

    def __repr__(self):
        mpargs = {'name': self.parser.name, 'ptype': self.parser.ptype}
        parg = "MockParser({name}, {ptype})".format(**mpargs)
        rarg = str(self) if not self.children else \
               "(" + ", ".join(repr(child) for child in self.children) + ")"
        return "Node(%s, %s)" % (parg, rarg)
222

223
    def __eq__(self, other):
224 225
        # return str(self.parser) == str(other.parser) and self.result == other.result
        return self.tag_name == other.tag_name and self.result == other.result
226

227
    def __hash__(self):
228
        return hash(self.tag_name)
229 230 231 232 233 234 235

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

236
    @property   # this needs to be a (dynamic) property, in case sef.parser gets updated
237
    def tag_name(self) -> str:
238 239
        return self.parser.name or self.parser.ptype
        # ONLY FOR DEBUGGING: return self.parser.name + ':' + self.parser.ptype
240 241

    @property
Eckhart Arnold's avatar
Eckhart Arnold committed
242
    def result(self) -> StrictResultType:
243 244 245
        return self._result

    @result.setter
Eckhart Arnold's avatar
Eckhart Arnold committed
246
    def result(self, result: ResultType):
247 248 249 250
        # # 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)
251
        self._result = (result,) if isinstance(result, Node) else result or ''
252 253
        self._children = cast(Tuple['Node', ...], self._result) \
            if isinstance(self._result, tuple) else cast(Tuple['Node', ...], ())
254 255

    @property
256
    def children(self) -> Tuple['Node', ...]:
257 258 259
        return self._children

    @property
260
    def len(self) -> int:
261
        # DEBUGGING:  print(self.tag_name, str(self.pos), str(self._len), str(self)[:10].replace('\n','.'))
262 263 264
        return self._len

    @property
265
    def pos(self) -> int:
266 267 268 269
        assert self._pos >= 0, "position value not initialized!"
        return self._pos

    @pos.setter
270 271
    def pos(self, pos: int):
        # assert isinstance(pos, int)
272 273 274 275 276 277 278
        self._pos = pos
        offset = 0
        for child in self.children:
            child.pos = pos + offset
            offset += child.len

    @property
279
    def errors(self) -> List[Error]:
280 281
        return [Error(self.pos, err) for err in self._errors]

282 283
    def show(self) -> str:
        """Returns content as string, inserting error messages where
284
        errors occurred.
285
        """
286
        s = "".join(child.show() for child in self.children) if self.children \
287 288 289
            else str(self.result)
        return (' <<< Error on "%s" | %s >>> ' % (s, '; '.join(self._errors))) if self._errors else s

290
    def _tree_repr(self, tab, openF, closeF, dataF=lambda s: s) -> str:
291 292 293 294 295 296 297 298
        """
        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.

299
        Args:
300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322
            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 = []
323
            for child in self.children:
324 325 326 327
                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

328 329
        res = cast(str, self.result)  # safe, because if there are no children, result is a string
        if head[0] == "<" and res.find('\n') < 0:
330 331 332
            # for XML: place tags for leaf-nodes on one line if possible
            return head[:-1] + self.result + tail[1:]
        else:
333
            return head + '\n'.join([tab + dataF(s) for s in res.split('\n')]) + tail
334

335
    def as_sexpr(self, src: str=None) -> str:
336 337 338
        """
        Returns content as S-expression, i.e. in lisp-like form.

339
        Args:
340 341 342 343 344
            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.
        """

345
        def opening(node) -> str:
346 347 348
            s = '(' + node.tag_name
            # s += " '(pos %i)" % node.pos
            if src:
Eckhart Arnold's avatar
Eckhart Arnold committed
349
                s += " '(pos %i " % node.pos + " %i %i)" % line_col(src, node.pos)
350 351 352 353 354 355 356 357 358 359
            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'\"')

360
        return self._tree_repr('    ', opening, lambda node: ')', pretty)
361

362
    def as_xml(self, src: str=None) -> str:
363 364 365
        """
        Returns content as XML-tree.

366
        Args:
367 368 369 370 371
            src:  The source text or `None`. In case the source text is
                given the position will also be reported as line and
                column.
        """

372
        def opening(node) -> str:
373 374 375 376 377 378 379 380 381 382 383 384 385 386 387
            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)

388
    def add_error(self, error_str) -> 'Node':
389 390 391 392
        self._errors.append(error_str)
        self.error_flag = True
        return self

393 394
    def propagate_error_flags(self) -> None:
        """Recursively propagates error flags set on child nodes to its
395 396 397 398 399
        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()
400
            self.error_flag = self.error_flag or child.error_flag
401

402
    def collect_errors(self, clear_errors=False) -> List[Error]:
403 404 405 406 407
        """
        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.
        """
408 409 410 411 412
        errors = self.errors
        if clear_errors:
            self._errors = []
            self.error_flag = False
        if self.children:
413
            for child in self.children:
414
                errors.extend(child.collect_errors(clear_errors))
415
        return errors
416

Eckhart Arnold's avatar
Eckhart Arnold committed
417 418 419 420
    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())
421

422
    def find(self, match_function) -> Iterator['Node']:
423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442
        """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

443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459
    # 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
    #     """


460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481
    # def navigate(self, path):
    #     """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)
    #
    #     Args:
    #         path (str):  The path of the object, e.g. 'a/b/c'. The
    #             components of ``path`` can be regular expressions
    #
    #     Returns:
    #         The object at the path, either a string or a Node or
    #         ``None``, if the path did not match.
    #     """
    #     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))
    #         else:
    #             return self.result,
    #     return nav(path.split('/'))
482 483 484 485 486 487 488 489 490


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


491 492 493 494
TransformationFunc = Union[Callable[[Node], Any], partial]


def transformation_factory(t=None):
Eckhart Arnold's avatar
Eckhart Arnold committed
495 496 497 498 499 500 501 502 503 504 505 506 507 508 509
    """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.
510 511 512 513 514

    Usage:
        @transformation_factory(AbtractSet[str])
        def remove_tokens(node, tokens):
            ...
Eckhart Arnold's avatar
Eckhart Arnold committed
515
      or, alternatively:
516 517 518
        @transformation_factory
        def remove_tokens(node, tokens: AbstractSet[str]):
            ...
Eckhart Arnold's avatar
Eckhart Arnold committed
519 520 521 522 523

    Example:
        trans_table = { 'expression': remove_tokens('+', '-') }
      instead of:
        trans_table = { 'expression': partial(remove_tokens, tokens={'+', '-'}) }
524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554
    """

    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
555 556 557 558
        # 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
559 560 561
        return decorator(func)
    else:
        return decorator
562 563


564 565
WHITESPACE_PTYPE = ':Whitespace'
TOKEN_PTYPE = ':Token'
566 567


568
def key_parser_name(node) -> str:
569 570 571
    return node.parser.name


572
def key_tag_name(node) -> str:
573 574 575
    return node.tag_name


576
def traverse(root_node, processing_table, key_func=key_tag_name) -> None:
577
    """Traverses the snytax tree starting with the given ``node`` depth
578
    first and applies the sequences of callback-functions registered
579 580
    in the ``calltable``-dictionary.
    
581 582
    The most important use case is the transformation of a concrete
    syntax tree into an abstract tree (AST). But it is also imaginable
583
    to employ tree-traversal for the semantic analysis of the AST.
584 585 586 587 588 589 590 591 592 593

    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)

594
    Args:
di68kap's avatar
di68kap committed
595
        root_node (Node): The root-node of the syntax tree to be traversed 
596
        processing_table (dict): node key -> sequence of functions that
di68kap's avatar
di68kap committed
597 598
            will be applied to matching nodes in order. This dictionary
            is interpreted as a ``compact_table``. See 
599
            ``toolkit.expand_table`` or ``EBNFCompiler.EBNFTransTable``
600 601
        key_func (function): A mapping key_func(node) -> keystr. The default
            key_func yields node.parser.name.
602 603 604 605 606
            
    Example:
        table = { "term": [replace_by_single_child, flatten], 
            "factor, flowmarker, retrieveop": replace_by_single_child }
        traverse(node, table)
607
    """
608 609 610 611
    # 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
612
    # normalize processing_table entries by turning single values into lists
613
    # with a single value
614
    table = {name: smart_list(call) for name, call in list(processing_table.items())}
615
    table = expand_table(table)
616
    cache = {}  # type: Dict[str, List[Callable]]
617

di68kap's avatar
di68kap committed
618 619 620
    def traverse_recursive(node):
        if node.children:
            for child in node.result:
621
                traverse_recursive(child)            # depth first
622
                node.error_flag |= child.error_flag  # propagate error flag
623 624 625 626 627 628 629 630 631 632 633 634 635

        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

636
        for call in sequence:
di68kap's avatar
di68kap committed
637
            call(node)
638

di68kap's avatar
di68kap committed
639
    traverse_recursive(root_node)
640 641 642 643 644


# ------------------------------------------------
#
# rearranging transformations:
di68kap's avatar
di68kap committed
645 646
#     - tree may be rearranged (e.g.flattened)
#     - nodes that are not leaves may be dropped
647 648 649 650 651 652
#     - order is preserved
#     - all leaves are kept
#
# ------------------------------------------------


653 654 655 656 657 658 659 660 661 662 663 664 665
@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)


666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687
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


688 689 690 691 692 693
@transformation_factory(Callable)
def reduce_children(node, condition=lambda node: not node.name):
    """Replaces those children of node that have children themselves
    ans fulfil the given condition (default unnamed nodes).
    In contrast to ``flatten`` (see below) this transformation does not
    operate recursively.
di68kap's avatar
di68kap committed
694
    """
695 696 697 698 699 700 701 702
    if node.children:
        new_result = []
        for child in node.children:
            if child.children and condition(child):
                new_result.extend(child.children)
            else:
                new_result.append(child)
        node.result = tuple(new_result)
di68kap's avatar
di68kap committed
703 704


705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722
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:
                flatten(child)
723
                new_result.extend(child.children)
724 725 726 727 728 729 730 731 732 733 734 735
            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)


736 737 738 739 740 741 742 743 744 745 746 747 748 749
# ------------------------------------------------
#
# 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."""
750
    return node.parser.ptype == WHITESPACE_PTYPE
751 752 753 754 755

def is_empty(node):
    return not node.result

def is_expendable(node):
756
    return is_empty(node) or is_whitespace(node)
757

758
def is_token(node, tokens: AbstractSet[str] = frozenset()) -> bool:
759
    return node.parser.ptype == TOKEN_PTYPE and (not tokens or node.result in tokens)
760 761


762
@transformation_factory
763 764 765 766
def has_name(node, tag_names: AbstractSet[str]) -> bool:
    return node.tag_name in tag_names


767
@transformation_factory
768 769 770 771
def has_content(node, contents: AbstractSet[str]) -> bool:
    return str(node) in contents


772
@transformation_factory(Callable)  # @singledispatch
773
def remove_children_if(node, condition):
774 775 776 777 778 779
    """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))


780
remove_whitespace = remove_children_if(is_whitespace)  # partial(remove_children_if, condition=is_whitespace)
Eckhart Arnold's avatar
Eckhart Arnold committed
781
remove_empty = remove_children_if(is_empty)
782
remove_expendables = remove_children_if(is_expendable)  # partial(remove_children_if, condition=is_expendable)
783 784


785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802
@transformation_factory(Callable)
def remove_first(node, condition=lambda node: True):
    """Removes the first child if the condition is met.
    Otherwise does nothing."""
    if node.children:
        if condition(node.children[0]):
            node.result = node.result[1:]


@transformation_factory(Callable)
def remove_last(node, condition=lambda node: True):
    """Removes the last child if the condition is met.
    Otherwise does nothing."""
    if node.children:
        if condition(node.children[-1]):
            node.result = node.result[:-1]


803 804
@transformation_factory
def remove_tokens(node, tokens: AbstractSet[str] = frozenset()):
805 806
    """Reomoves any among a particular set of tokens from the immediate
    descendants of a node. If ``tokens`` is the empty set, all tokens
807
    are removed."""
808
    remove_children_if(node, partial(is_token, tokens=tokens))
809 810


811 812 813 814 815 816 817 818 819 820
@transformation_factory
def remove_children(node, tag_names: AbstractSet[str]) -> bool:
    """Removes children by 'tag name'."""
    remove_children_if(node, partial(has_name, tag_names=tag_names))


@transformation_factory
def remove_content(node, contents: AbstractSet[str]):
    """Removes children depending on their string value."""
    remove_children_if(node, partial(has_content, contents=contents))
821 822


Eckhart Arnold's avatar
Eckhart Arnold committed
823 824
@transformation_factory
def map_content(node, func: Callable):      # Callable[[Node], ResultType]
825 826 827 828 829 830
    """Replaces the content of the node. ``func`` takes the node
    as an argument an returns the mapped result.
    """
    node.result = func(node.result)


831 832
########################################################################
#
833
# AST semantic validation functions
di68kap's avatar
di68kap committed
834
# EXPERIMENTAL!
835 836 837 838
#
########################################################################


839 840
@transformation_factory
def require(node, child_tags: AbstractSet[str]):
841
    for child in node.children:
842
        if child.tag_name not in child_tags:
843
            node.add_error('Element "%s" is not allowed inside "%s".' %
844
                           (child.parser.name, node.parser.name))
845 846


847 848
@transformation_factory
def forbid(node, child_tags: AbstractSet[str]):
849
    for child in node.children:
850
        if child.tag_name in child_tags:
851
            node.add_error('Element "%s" cannot be nested inside "%s".' %
852
                           (child.parser.name, node.parser.name))
853 854


855 856
@transformation_factory
def assert_content(node, regex: str):
857 858 859
    content = str(node)
    if not re.match(regex, content):
        node.add_error('Element "%s" violates %s on %s' %
860
                       (node.parser.name, str(regex), content))
861