10.12., 9:00 - 11:00: Due to updates GitLab may be unavailable for some minutes between 09:00 and 11:00.

syntaxtree.py 32 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, reduce, 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, repr_call, 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
           'replace_parser',
52
           'collapse',
53
           'join',
54
           'replace_content',
55
           'apply_if',
Eckhart Arnold's avatar
Eckhart Arnold committed
56 57 58 59
           'is_whitespace',
           'is_empty',
           'is_expendable',
           'is_token',
60 61
           'has_name',
           'has_content',
62
           'remove_parser',
63
           'remove_content',
Eckhart Arnold's avatar
Eckhart Arnold committed
64 65
           'remove_first',
           'remove_last',
Eckhart Arnold's avatar
Eckhart Arnold committed
66
           'remove_whitespace',
Eckhart Arnold's avatar
Eckhart Arnold committed
67
           'remove_empty',
Eckhart Arnold's avatar
Eckhart Arnold committed
68
           'remove_expendables',
69
           'remove_brackets',
Eckhart Arnold's avatar
Eckhart Arnold committed
70 71
           'remove_tokens',
           'flatten',
72 73 74
           'forbid',
           'require',
           'assert_content']
Eckhart Arnold's avatar
Eckhart Arnold committed
75 76


Eckhart Arnold's avatar
Eckhart Arnold committed
77 78 79 80 81 82 83 84 85 86
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

87 88 89
    def __repr__(self):
        return self.name + self.ptype

Eckhart Arnold's avatar
Eckhart Arnold committed
90
    def __str__(self):
91
        return self.name + (' = ' if self.name else '') + repr(self)
Eckhart Arnold's avatar
Eckhart Arnold committed
92 93 94 95 96

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

97 98 99
    @property
    def repr(self) -> str:
        return self.name if self.name else repr(self)
Eckhart Arnold's avatar
Eckhart Arnold committed
100 101

class MockParser(ParserBase):
102 103
    """
    MockParser objects can be used to reconstruct syntax trees from a
Eckhart Arnold's avatar
Eckhart Arnold committed
104 105
    serialized form like S-expressions or XML. Mock objects can mimic
    different parser types by assigning them a ptype on initialization.
106 107 108 109 110
    
    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
111
    def __init__(self, name='', ptype=''):  # , pbases=frozenset()):
112
        assert not ptype or ptype[0] == ':'
Eckhart Arnold's avatar
Eckhart Arnold committed
113
        super(MockParser, self).__init__(name)
114
        self.name = name
Eckhart Arnold's avatar
Eckhart Arnold committed
115
        self._ptype = ptype or ':' + self.__class__.__name__
116 117 118


class ZombieParser(MockParser):
119 120 121 122 123 124 125 126 127 128 129 130
    """
    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):
131
        super(ZombieParser, self).__init__("__ZOMBIE__")
132 133 134 135
        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
136 137 138 139 140 141
    def __copy__(self):
        return self

    def __deepcopy__(self, memo):
        return self

142 143 144 145
    def __call__(self, text):
        """Better call Saul ;-)"""
        return None, text

146 147 148 149

ZOMBIE_PARSER = ZombieParser()


150 151 152 153 154
# # Python 3.6:
# class Error(NamedTuple):
#     pos: int
#     msg: str
Error = NamedTuple('Error', [('pos', int), ('msg', str)])
155

Eckhart Arnold's avatar
Eckhart Arnold committed
156 157
StrictResultType = Union[Tuple['Node', ...], str]
ResultType = Union[Tuple['Node', ...], 'Node', str, None]
158 159


160 161 162 163 164 165 166 167 168 169 170 171
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!
172 173 174 175 176 177
        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, ...)``.
178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199
        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.
    """
200

Eckhart Arnold's avatar
Eckhart Arnold committed
201
    def __init__(self, parser, result: ResultType) -> None:
202 203 204
        """Initializes the ``Node``-object with the ``Parser``-Instance
        that generated the node and the parser's result.
        """
Eckhart Arnold's avatar
Eckhart Arnold committed
205
        self._result = ''  # type: StrictResultType
206
        self._errors = []  # type: List[str]
207
        self._children = ()  # type: Tuple['Node', ...]
208
        self.result = result
209 210 211 212
        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
213
        self.parser = parser or ZOMBIE_PARSER
214 215
        self.error_flag = any(r.error_flag for r in self.children) \
            if self.children else False  # type: bool
216 217 218

    def __str__(self):
        if self.children:
219
            return "".join(str(child) for child in self.children)
220 221 222 223 224 225 226 227
        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)
228

229
    def __eq__(self, other):
230 231
        # return str(self.parser) == str(other.parser) and self.result == other.result
        return self.tag_name == other.tag_name and self.result == other.result
232

233
    def __hash__(self):
234
        return hash(self.tag_name)
235 236 237 238 239 240 241

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

242
    @property   # this needs to be a (dynamic) property, in case sef.parser gets updated
243
    def tag_name(self) -> str:
244
        return self.parser.name or self.parser.ptype
245 246

    @property
Eckhart Arnold's avatar
Eckhart Arnold committed
247
    def result(self) -> StrictResultType:
248 249 250
        return self._result

    @result.setter
Eckhart Arnold's avatar
Eckhart Arnold committed
251
    def result(self, result: ResultType):
252 253 254 255
        # # 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)
256
        self._result = (result,) if isinstance(result, Node) else result or ''
257 258
        self._children = cast(Tuple['Node', ...], self._result) \
            if isinstance(self._result, tuple) else cast(Tuple['Node', ...], ())
259 260

    @property
261
    def children(self) -> Tuple['Node', ...]:
262 263 264
        return self._children

    @property
265
    def len(self) -> int:
266
        # DEBUGGING:  print(self.tag_name, str(self.pos), str(self._len), str(self)[:10].replace('\n','.'))
267 268 269
        return self._len

    @property
270
    def pos(self) -> int:
271 272 273 274
        assert self._pos >= 0, "position value not initialized!"
        return self._pos

    @pos.setter
275 276
    def pos(self, pos: int):
        # assert isinstance(pos, int)
277 278 279 280 281 282 283
        self._pos = pos
        offset = 0
        for child in self.children:
            child.pos = pos + offset
            offset += child.len

    @property
284
    def errors(self) -> List[Error]:
285 286
        return [Error(self.pos, err) for err in self._errors]

287 288
    def show(self) -> str:
        """Returns content as string, inserting error messages where
289
        errors occurred.
290
        """
291
        s = "".join(child.show() for child in self.children) if self.children \
292 293 294
            else str(self.result)
        return (' <<< Error on "%s" | %s >>> ' % (s, '; '.join(self._errors))) if self._errors else s

295
    def _tree_repr(self, tab, openF, closeF, dataF=lambda s: s) -> str:
296 297 298 299 300 301 302 303
        """
        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.

304
        Args:
305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327
            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 = []
328
            for child in self.children:
329 330 331 332
                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

333 334
        res = cast(str, self.result)  # safe, because if there are no children, result is a string
        if head[0] == "<" and res.find('\n') < 0:
335 336 337
            # for XML: place tags for leaf-nodes on one line if possible
            return head[:-1] + self.result + tail[1:]
        else:
338
            return head + '\n'.join([tab + dataF(s) for s in res.split('\n')]) + tail
339

340
    def as_sxpr(self, src: str=None) -> str:
341 342 343
        """
        Returns content as S-expression, i.e. in lisp-like form.

344
        Args:
345 346 347 348 349
            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.
        """

350
        def opening(node) -> str:
351 352 353
            s = '(' + node.tag_name
            # s += " '(pos %i)" % node.pos
            if src:
Eckhart Arnold's avatar
Eckhart Arnold committed
354
                s += " '(pos %i " % node.pos + " %i %i)" % line_col(src, node.pos)
355 356 357 358 359 360 361 362 363 364
            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'\"')

365
        return self._tree_repr('    ', opening, lambda node: ')', pretty)
366

367
    def as_xml(self, src: str=None) -> str:
368 369 370
        """
        Returns content as XML-tree.

371
        Args:
372 373 374 375 376
            src:  The source text or `None`. In case the source text is
                given the position will also be reported as line and
                column.
        """

377
        def opening(node) -> str:
378 379 380 381 382 383 384 385 386 387 388 389 390 391 392
            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)

393
    def add_error(self, error_str) -> 'Node':
394 395 396 397
        self._errors.append(error_str)
        self.error_flag = True
        return self

398 399
    def propagate_error_flags(self) -> None:
        """Recursively propagates error flags set on child nodes to its
400 401 402 403 404
        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()
405
            self.error_flag = self.error_flag or child.error_flag
406

407
    def collect_errors(self, clear_errors=False) -> List[Error]:
408 409 410 411 412
        """
        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.
        """
413 414 415 416 417
        errors = self.errors
        if clear_errors:
            self._errors = []
            self.error_flag = False
        if self.children:
418
            for child in self.children:
419
                errors.extend(child.collect_errors(clear_errors))
420
        return errors
421

Eckhart Arnold's avatar
Eckhart Arnold committed
422 423 424
    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:
425
            f.write(self.as_sxpr())
426

427
    def find(self, match_function: Callable) -> Iterator['Node']:
428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447
        """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

448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464
    # 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
    #     """


465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486
    # 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('/'))
487 488 489 490 491 492 493 494 495


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


496 497 498 499
TransformationFunc = Union[Callable[[Node], Any], partial]


def transformation_factory(t=None):
Eckhart Arnold's avatar
Eckhart Arnold committed
500 501 502 503 504 505 506 507 508 509 510 511 512 513 514
    """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.
515 516 517 518 519

    Usage:
        @transformation_factory(AbtractSet[str])
        def remove_tokens(node, tokens):
            ...
Eckhart Arnold's avatar
Eckhart Arnold committed
520
      or, alternatively:
521 522 523
        @transformation_factory
        def remove_tokens(node, tokens: AbstractSet[str]):
            ...
Eckhart Arnold's avatar
Eckhart Arnold committed
524 525 526 527 528

    Example:
        trans_table = { 'expression': remove_tokens('+', '-') }
      instead of:
        trans_table = { 'expression': partial(remove_tokens, tokens={'+', '-'}) }
529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559
    """

    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
560 561 562 563
        # 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
564 565 566
        return decorator(func)
    else:
        return decorator
567 568


569 570
WHITESPACE_PTYPE = ':Whitespace'
TOKEN_PTYPE = ':Token'
571 572


573
def key_parser_name(node) -> str:
574 575 576
    return node.parser.name


577
def key_tag_name(node) -> str:
578 579 580
    return node.tag_name


581
def traverse(root_node, processing_table, key_func=key_tag_name) -> None:
582
    """Traverses the snytax tree starting with the given ``node`` depth
583
    first and applies the sequences of callback-functions registered
584 585
    in the ``calltable``-dictionary.
    
586 587
    The most important use case is the transformation of a concrete
    syntax tree into an abstract tree (AST). But it is also imaginable
588
    to employ tree-traversal for the semantic analysis of the AST.
589 590 591 592 593 594 595 596 597 598

    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)

599
    Args:
di68kap's avatar
di68kap committed
600
        root_node (Node): The root-node of the syntax tree to be traversed 
601
        processing_table (dict): node key -> sequence of functions that
di68kap's avatar
di68kap committed
602 603
            will be applied to matching nodes in order. This dictionary
            is interpreted as a ``compact_table``. See 
604
            ``toolkit.expand_table`` or ``EBNFCompiler.EBNFTransTable``
605 606
        key_func (function): A mapping key_func(node) -> keystr. The default
            key_func yields node.parser.name.
607 608 609 610 611
            
    Example:
        table = { "term": [replace_by_single_child, flatten], 
            "factor, flowmarker, retrieveop": replace_by_single_child }
        traverse(node, table)
612
    """
613 614 615 616
    # 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
617
    # normalize processing_table entries by turning single values into lists
618
    # with a single value
619
    table = {name: smart_list(call) for name, call in list(processing_table.items())}
620
    table = expand_table(table)
621
    cache = {}  # type: Dict[str, List[Callable]]
622

di68kap's avatar
di68kap committed
623 624 625
    def traverse_recursive(node):
        if node.children:
            for child in node.result:
626
                traverse_recursive(child)            # depth first
627
                node.error_flag |= child.error_flag  # propagate error flag
628 629 630 631 632 633 634 635 636 637 638 639 640

        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

641
        for call in sequence:
di68kap's avatar
di68kap committed
642
            call(node)
643

di68kap's avatar
di68kap committed
644
    traverse_recursive(root_node)
645 646 647 648 649


# ------------------------------------------------
#
# rearranging transformations:
di68kap's avatar
di68kap committed
650 651
#     - tree may be rearranged (e.g.flattened)
#     - nodes that are not leaves may be dropped
652
#     - order is preserved
653
#     - leave content is preserved (though not necessarily the leaves themselves)
654 655 656 657
#
# ------------------------------------------------


658 659 660 661 662 663 664 665 666 667 668 669 670
@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)


671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692
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


693
@transformation_factory(Callable)
694 695 696 697 698 699 700 701 702 703 704
def flatten(node, condition=lambda node: not node.parser.name, recursive=True):
    """Flattens all children, that fulfil the given `condition`
    (default: all unnamed children). Flattening means that wherever a
    node has child nodes, the child nodes are inserted in place of the
    node.
     If the parameter `recursive` is `True` the same will recursively be
    done with the child-nodes, first. In other words, all leaves of
    this node and its child nodes are collected in-order as direct
    children of this node.
     Applying flatten recursively will result in these kinds of
    structural transformation:
705 706 707 708 709 710
        (1 (+ 2) (+ 3)     ->   (1 + 2 + 3)
        (1 (+ (2 + (3))))  ->   (1 + 2 + 3)
    """
    if node.children:
        new_result = []
        for child in node.children:
711 712 713
            if child.children and condition(child):
                if recursive:
                    flatten(child, condition, recursive)
714
                new_result.extend(child.children)
715 716 717 718 719 720 721 722 723 724 725 726
            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)


727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750
@transformation_factory
def join(node, tag_names: List[str]):
    """Joins all children next to each other and with particular tag-
    names into a single child node with mock parser 'parser_name'.
    """
    result = []
    name, ptype = (tag_names[0].split(':') + [''])[:2]
    if node.children:
        i = 0;  L = len(node.children)
        while i < L:
            while i < L and not node.children[i].tag_name in tag_names:
                result.append(node.children[i])
                i += 1
            k = i + 1
            while (k < L and node.children[k].tag_name in tag_names
                   and bool(node.children[i].children) == bool(node.children[k].children)):
                k += 1
            if i < L:
                result.append(Node(MockParser(name, ptype),
                    reduce(lambda a, b: a + b, (node.result for node in node.children[i:k]))))
            i = k
        node.result = tuple(result)


751 752 753 754 755 756 757 758 759 760 761
# ------------------------------------------------
#
# destructive transformations:
#     - tree may be rearranged (flattened),
#     - order is preserved
#     - but (irrelevant) leaves may be dropped
#     - errors of dropped leaves will be lost
#
# ------------------------------------------------


762 763 764 765 766 767 768 769
@transformation_factory
def replace_content(node, func: Callable):      # Callable[[Node], ResultType]
    """Replaces the content of the node. ``func`` takes the node
    as an argument an returns the mapped result.
    """
    node.result = func(node.result)


770 771 772
def is_whitespace(node):
    """Removes whitespace and comments defined with the
    ``@comment``-directive."""
773
    return node.parser.ptype == WHITESPACE_PTYPE
774 775 776 777 778

def is_empty(node):
    return not node.result

def is_expendable(node):
779
    return is_empty(node) or is_whitespace(node)
780

781
def is_token(node, tokens: AbstractSet[str] = frozenset()) -> bool:
782
    return node.parser.ptype == TOKEN_PTYPE and (not tokens or node.result in tokens)
783 784


785
@transformation_factory
786
def has_name(node, tag_names: AbstractSet[str]) -> bool:
787 788
    """Checks if node has any of a given set of `tag names`.
    See property `Node.tagname`."""
789 790 791
    return node.tag_name in tag_names


792
@transformation_factory
793
def has_content(node, contents: AbstractSet[str]) -> bool:
794 795
    """Checks if the node's content (i.e. `str(node)`) matches any of
    a given set of strings."""
796 797 798
    return str(node) in contents


799 800 801 802 803 804 805 806
@transformation_factory
def apply_if(node, transformation: Callable, condition: Callable):
    """Applies a transformation only if a certain condition is met.
    """
    if condition(node):
        transformation(node)


807 808 809 810 811 812 813 814 815 816
@transformation_factory
def keep_children(node, section: slice=slice(None, None, None), condition=lambda node: True):
    """Keeps only the nodes which fall into a slice of the result field
    and for which the function `condition(child_node)` evaluates to
    `True`."""
    if node.children:
        node.result = tuple(c for c in node.children[section] if condition(c))


@transformation_factory(Callable)
817
def remove_children_if(node, condition):
818
    """Removes all nodes from a slice of the result field if the function
819 820 821 822 823
    ``condition(child_node)`` evaluates to ``True``."""
    if node.children:
        node.result = tuple(c for c in node.children if not condition(c))


824
remove_whitespace = remove_children_if(is_whitespace)  # partial(remove_children_if, condition=is_whitespace)
Eckhart Arnold's avatar
Eckhart Arnold committed
825
remove_empty = remove_children_if(is_empty)
826
remove_expendables = remove_children_if(is_expendable)  # partial(remove_children_if, condition=is_expendable)
827 828
remove_brackets = keep_children(slice(1,-1))

829

Eckhart Arnold's avatar
Eckhart Arnold committed
830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847
@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]


848 849
@transformation_factory
def remove_tokens(node, tokens: AbstractSet[str] = frozenset()):
850 851
    """Reomoves any among a particular set of tokens from the immediate
    descendants of a node. If ``tokens`` is the empty set, all tokens
852
    are removed."""
853
    remove_children_if(node, partial(is_token, tokens=tokens))
854 855


856
@transformation_factory
857
def remove_parser(node, tag_names: AbstractSet[str]):
858 859 860 861 862 863 864 865
    """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))
866 867


868 869
########################################################################
#
870
# AST semantic validation functions
di68kap's avatar
di68kap committed
871
# EXPERIMENTAL!
872 873 874 875
#
########################################################################


876 877
@transformation_factory
def require(node, child_tags: AbstractSet[str]):
878
    for child in node.children:
879
        if child.tag_name not in child_tags:
880
            node.add_error('Element "%s" is not allowed inside "%s".' %
881
                           (child.parser.name, node.parser.name))
882 883


884 885
@transformation_factory
def forbid(node, child_tags: AbstractSet[str]):
886
    for child in node.children:
887
        if child.tag_name in child_tags:
888
            node.add_error('Element "%s" cannot be nested inside "%s".' %
889
                           (child.parser.name, node.parser.name))
890 891


892 893
@transformation_factory
def assert_content(node, regex: str):
894 895 896
    content = str(node)
    if not re.match(regex, content):
        node.add_error('Element "%s" violates %s on %s' %
897
                       (node.parser.name, str(regex), content))
898