syntaxtree.py

# syntaxtree.py - syntax tree classes 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.


"""
Module ``syntaxtree`` defines the ``Node``-class for syntax trees as well
as an abstract base class for parser-objects. The latter is defined
here, because node-objects refer to parser-objects. All concrete
parser classes are defined in the ``parse`` module.
"""


import collections.abc
from collections import OrderedDict
import copy

from DHParser.error import Error, linebreaks, line_col
from DHParser.stringview import StringView
from DHParser.toolkit import re, typing
from typing import Callable, cast, Iterator, List, AbstractSet, Set, Union, Tuple, Optional


__all__ = ('ParserBase',
           'WHITESPACE_PTYPE',
           'TOKEN_PTYPE',
           'MockParser',
           'ZombieParser',
           'ZOMBIE_PARSER',
           'ZOMBIE_NODE',
           'Node',
           'parse_sxpr',
           'flatten_sxpr')


#######################################################################
#
# parser base and mock parsers
#
#######################################################################


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 __repr__(self):
         return self.name + self.ptype

    def __str__(self):
        return self.name + (' = ' if self.name else '') + repr(self)

    def __call__(self, text: StringView) -> Tuple[Optional['Node'], StringView]:
        return None, text

    @property
    def name(self):
        """Returns the name of the parser or the empty string '' for unnamed
        parsers."""
        return self._name

    @property
    def ptype(self) -> str:
        """Returns the type of the parser. By default this is the parser's
        class name preceded by a colon, e.g. ':ZeroOrMore'."""
        return self._ptype

    @property
    def repr(self) -> str:
        """Returns the parser's name if it has a name and repr()"""
        return self.name if self.name else repr(self)

    def reset(self):
        """Resets any parser variables. (Should be overridden.)"""
        pass

    def grammar(self) -> Optional[object]:
        """Returns the Grammar object to which the parser belongs. If not
        yet connected to any Grammar object, None is returned."""
        return None

    def apply(self, func: Callable) -> bool:
        """Applies the function `func` to the parser. Returns False, if
        - for whatever reason - the functions has not been applied, True
        otherwise."""
        return False


WHITESPACE_PTYPE = ':Whitespace'
TOKEN_PTYPE = ':Token'


class MockParser(ParserBase):
    """
    MockParser objects can be used to reconstruct syntax trees from a
    serialized form like S-expressions or XML. Mock objects can mimic
    different parser types by assigning them a ptype on initialization.

    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.
    """
    def __init__(self, name='', ptype=''):  # , pbases=frozenset()):
        assert not ptype or ptype[0] == ':'
        super().__init__(name)
        self._ptype = ptype or ':' + self.__class__.__name__


class ZombieParser(MockParser):
    """
    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):
        super(ZombieParser, self).__init__("__ZOMBIE__")
        assert not self.__class__.alive, "There can be only one!"
        assert self.__class__ == ZombieParser, "No derivatives, please!"
        self.__class__.alive = True

    def __copy__(self):
        return self

    def __deepcopy__(self, memo):
        return self

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


ZOMBIE_PARSER = ZombieParser()


#######################################################################
#
# syntaxtree nodes
#
#######################################################################


ChildrenType = Tuple['Node', ...]
NoChildren = cast(ChildrenType, ())  # type: ChildrenType
StrictResultType = Union[ChildrenType, StringView, str]
ResultType = Union[ChildrenType, 'Node', StringView, str, None]


def flatten_sxpr(sxpr: str) -> str:
    """Returns S-expression ``sxpr`` as a one-liner without unnecessary
    whitespace.

    Example:
    >>> flatten_sxpr('(a\\n    (b\\n        c\\n    )\\n)\\n')
    '(a (b c))'
    """
    return re.sub(r'\s(?=\))', '', re.sub(r'\s+', ' ', sxpr)).strip()


class Node(collections.abc.Sized):
    """
    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!

        content (str):  Yields the contents of the tree as string. The
            difference to ``str(node)`` is that ``node.content`` does
            not add the error messages to the returned string.

        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, ...)``.

        errors (list):  A list of parser- or compiler-errors:
            tuple(position, string) attached to this node

        error_flag (int):  0 if no error occurred in either the node
            itself or any of its descendants. Otherwise contains the
            highest warning or error level or all errors that occurred.

        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 and 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 during the parsing
            stage and never during or after the AST-transformation.

        attributes (dict): An optional dictionary of XML-attributes. This
            dictionary is created lazily upon first usage. The attributes
            will only be shown in the XML-Representation, not in the
            S-Expression-output.


        _parent (Node): SLOT RESERVED FOR FUTURE USE!
    """

    __slots__ = ['_result', 'children', '_errors', '_len', '_pos', 'parser', 'error_flag',
                 '_xml_attr', '_content', '_parent']

    def __init__(self, parser, result: ResultType, leafhint: bool = False) -> None:
        """
        Initializes the ``Node``-object with the ``Parser``-Instance
        that generated the node and the parser's result.
        """
        self.error_flag = 0             # type: int
        self._errors = []               # type: List[Error]
        self._pos = -1                  # type: int
        # Assignment to self.result initializes the attributes _result, children and _len
        # The following if-clause is merely an optimization, i.e. a fast-path for leaf-Nodes
        if leafhint:
            self._result = result       # type: StrictResultType
            self._content = None        # type: Optional[str]
            self.children = NoChildren  # type: ChildrenType
            self._len = -1              # type: int  # lazy evaluation
        else:
            self.result = result
        self.parser = parser or ZOMBIE_PARSER


    def __str__(self):
        s = "".join(str(child) for child in self.children) if self.children else self.content
        if self._errors:
            return ' <<< Error on "%s" | %s >>> ' % \
                   (s, '; '.join(e.message for e in self._errors))
        return s


    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)


    def __len__(self):
        if self._len < 0:
            self._len = sum(len(child) for child in self.children) \
                if self.children else len(self._result)
        return self._len


    def __bool__(self):
        # A node that is not None is always True, even if it's empty
        return True


    def __eq__(self, other):
        """
        Equality of nodes: Two nodes are considered as equal, if their tag
        name is the same and if their results are equal.
        """
        return self.tag_name == other.tag_name and self.result == other.result


    def __hash__(self):
        return hash(self.tag_name)


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


    def __getitem__(self, index_or_tagname: Union[int, str]) -> Union['Node', Iterator['Node']]:
        """
        Returns the child node with the given index if ``index_or_tagname`` is
        an integer or the first child node with the given tag name. Examples::

            >>> tree = parse_sxpr('(a (b "X") (X (c "d")) (e (X "F")))')
            >>> flatten_sxpr(tree[0].as_sxpr())
            '(b "X")'
            >>> flatten_sxpr(tree["X"].as_sxpr())
            '(X (c "d"))'

        Args:
            index_or_tagname(str): Either an index of a child node or a
                tag name.
        Returns:
            Node: All nodes which have a given tag name.
        """
        if self.children:
            if isinstance(index_or_tagname, int):
                return self.children[index_or_tagname]
            else:
                for child in self.children:
                    if child.tag_name == index_or_tagname:
                        return child
                raise KeyError(index_or_tagname)
        raise ValueError('Leave nodes have no children that can be indexed!')


    def __contains__(self, tag_name: str) -> bool:
        """
        Returns true if a child with the given tag name exists.
        Args:
            tag_name (str): tag_name which will be searched among to immediate
                descendants of this node.
        Returns:
            bool:  True, if at least one descendant node with the given tag
                name exists, False otherwise
        """
        # assert isinstance(tag_name, str)
        if self.children:
            for child in self.children:
                if child.tag_name == tag_name:
                    return True
            return False
        raise ValueError('Leave node cannot contain other nodes')
        # generator = self.select_by_tag(tag_name, False)
        # try:
        #     generator.__next__()
        #     return True
        # except StopIteration:
        #     return False


    @property   # this needs to be a (dynamic) property, in case sef.parser gets updated
    def tag_name(self) -> str:
        """
        Returns the tage name of Node, i.e. the name for XML or
        S-expression representation. By default the tag name is the
        name of the node's parser or, if the node's parser is unnamed, the
        node's parser's `ptype`.
        """
        return self.parser.name or self.parser.ptype


    @property
    def result(self) -> StrictResultType:
        """
        Returns the result from the parser that created the node.
        Error messages are not included in the result. Use `self.content()`
        if the result plus any error messages is needed.
        """
        return self._result


    @result.setter
    def result(self, result: ResultType):
        # # made obsolete by static type checking with mypy
        # assert ((isinstance(result, tuple) and all(isinstance(child, Node) for child in result))
        #         or isinstance(result, Node)
        #         or isinstance(result, str)), str(result)
        # Possible optimization: Do not allow single nodes as argument:
        # assert not isinstance(result, Node)
        self._len = -1        # lazy evaluation
        self._content = None
        if isinstance(result, Node):
            self.children = (result,)
            self._result = self.children
            self.error_flag = result.error_flag
        else:
            if isinstance(result, tuple):
                self.children = result
                self._result = result or ''
                if result:
                    if self.error_flag == 0:
                        self.error_flag = max(child.error_flag for child in self.children)
            else:
                self.children = NoChildren
                self._result = result


    @property
    def content(self) -> str:
        """
        Returns content as string, omitting error messages.
        """
        if self._content is None:
            if self.children:
                self._content = "".join(child.content for child in self.children)
            else:
                # self._content = self._result
                self._content = str(self._result)
                self._result = self._content  # self._result might be more efficient as a string!?
        return self._content


    @property
    def structure(self) -> str:
        """
        Return structure (and content) as S-expression on a single line
        without any line breaks.
        """
        return flatten_sxpr(self.as_sxpr(showerrors=False))


    @property
    def pos(self) -> int:
        """Returns the position of the Node's content in the source text."""
        if self._pos < 0:
            raise AssertionError("Position value not initialized!")
        return self._pos


    def init_pos(self, pos: int, overwrite: bool = False) -> 'Node':
        """
        (Re-)initialize position value. Usually, the parser guard
        (`parsers.add_parser_guard()`) takes care of assigning the
        position in the document to newly created nodes. However,
        where Nodes are created outside the reach of the parser
        guard, their document-position must be assigned manually.
        This function recursively reassigns the position values
        of the child nodes, too.
        """
        if overwrite or self._pos < 0:
            self._pos = pos
            for err in self._errors:
                err.pos = pos
        else:
            assert self._pos == pos, str("%i != %i" % (self._pos, pos))
        # recursively adjust pos-values of all children
        offset = self.pos
        for child in self.children:
            child.init_pos(offset, overwrite)
            offset = child.pos + len(child)
        return self


    @property
    def errors(self) -> List[Error]:
        """
        Returns the errors that occurred at this Node,
        not including any errors from child nodes.
        """
        return self._errors.copy()


    def add_error(self,
                  message: str,
                  code: int = Error.ERROR) -> 'Node':
        """
        Adds an error to this Node.
        Parameters:
            message(str): A string with the error message.abs
            code(int):    An error code to identify the kind of error
        """
        self._errors.append(Error(message, code))
        self.error_flag = max(self.error_flag, self._errors[-1].code)
        return self


    def collect_errors(self, clear_errors=False) -> List[Error]:
        """
        Recursively adds line- and column-numbers to all error objects.
        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.
        """
        errors = self.errors
        for err in errors:
            err.pos = self.pos
        if self.children:
            for child in self.children:
                errors.extend(child.collect_errors(clear_errors))
        if clear_errors:
            self._errors = []
            self.error_flag = 0
        else:
            if self._errors:
                self.error_flag = max(err.code for err in self.errors)
            if self.children:
                max_child_error = max(child.error_flag for child in self.children)
                self.error_flag = max(self.error_flag, max_child_error)
        return errors


    @property
    def attributes(self):
        """Returns a dictionary of XML-Attributes attached to the Node."""
        if not hasattr(self, '_xml_attr'):
            self._xml_attr = OrderedDict()
        return self._xml_attr


    def _tree_repr(self, tab, open_fn, close_fn, data_fn=lambda i: i, density=0) -> str:
        """
        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.

        Args:
            tab (str):  The indentation string, e.g. '\t' or '    '
            open_fn:   (Node->str) A function that returns an opening
                string (e.g. an XML-tag_name) for a given node
            close_fn:  (Node->str) A function that returns a closeF
                string (e.g. an XML-tag_name) for a given node.
            data_fn:   (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 = open_fn(self)
        tail = close_fn(self)

        if not self.result:
            return head.rstrip() + tail.lstrip()

        tail = tail.lstrip(None if density & 2 else '')

        if self.children:
            content = []
            for child in self.children:
                subtree = child._tree_repr(tab, open_fn, close_fn, data_fn, density).split('\n')
                content.append('\n'.join((tab + s) for s in subtree))
            return head + '\n'.join(content) + tail

        res = cast(str, self.result)  # safe, because if there are no children, result is a string
        if density & 1 and res.find('\n') < 0:  # and head[0] == "<":
            # except for XML, add a gap between opening statement and content
            gap = ' ' if head.rstrip()[-1] != '>' else ''
            return head.rstrip() + gap + data_fn(self.result) + tail.lstrip()
        else:
            return head + '\n'.join([tab + data_fn(s) for s in res.split('\n')]) + tail


    def as_sxpr(self, src: str = None, compact: bool = False, showerrors: bool = True) -> str:
        """
        Returns content as S-expression, i.e. in lisp-like form.

        Args:
            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.
            compact:  If True a compact representation is returned where
                brackets are omitted and only the indentation indicates the
                tree structure.
        """

        left_bracket, right_bracket, density = ('', '', 1) if compact else ('(', '\n)', 0)

        def opening(node) -> str:
            """Returns the opening string for the representation of `node`."""
            txt = [left_bracket,  node.tag_name]
            # s += " '(pos %i)" % node.pos
            if hasattr(node, '_xml_attr'):
                txt.extend(' `(%s "%s")' % (k, v) for k, v in node.attributes.items())
            if src:
                txt.append(" `(pos %i %i %i)" % (node.pos, *line_col(src, node.pos)))
            # if node.error_flag:   # just for debugging error collecting
            #     txt += " HAS ERRORS"
            if showerrors and node.errors:
                txt.append(" `(err `%s)" % ' '.join(str(err) for err in node.errors))
            return "".join(txt) + '\n'

        def closing(node) -> str:
            """Returns the closing string for the representation of `node`."""
            return right_bracket

        def pretty(strg):
            """Encloses `strg` with the right kind of quotation marks."""
            return '"%s"' % strg if strg.find('"') < 0 \
                else "'%s'" % strg if strg.find("'") < 0 \
                else '"%s"' % strg.replace('"', r'\"')

        return self._tree_repr('    ', opening, closing, pretty, density=density)


    def as_xml(self, src: str = None, showerrors: bool = True) -> str:
        """
        Returns content as XML-tree.

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

        def opening(node) -> str:
            """Returns the opening string for the representation of `node`."""            
            txt = ['<', node.tag_name]
            has_reserved_attrs = hasattr(node, '_xml_attr') \
                and any (r in node.attributes for r in {'err', 'line', 'col'})
            if hasattr(node, '_xml_attr'):
                txt.extend(' %s="%s"' % (k, v) for k, v in node.attributes.items())
            if src and not has_reserved_attrs:
                txt.append(' line="%i" col="%i"' % line_col(line_breaks, node.pos))
            if showerrors and node.errors and not has_reserved_attrs:
                txt.append(' err="%s"' % ''.join(str(err).replace('"', r'\"')
                                                 for err in node.errors))
            return "".join(txt + [">\n"])

        def closing(node):
            """Returns the closing string for the representation of `node`."""            
            return '\n</' + node.tag_name + '>'

        line_breaks = linebreaks(src) if src else []
        return self._tree_repr('    ', opening, closing, density=1)


    def select(self, match_function: Callable, include_root: bool=True) -> Iterator['Node']:
        """
        Finds nodes in the tree that fulfill a given criterion.

        `select` is a generator that yields all nodes for which the
        given `match_function` evaluates to True. The tree is
        traversed pre-order.

        See function `Node.select_by_tag` for some examples.

        Args:
            match_function (function): A function  that takes as Node
                object as argument and returns True or False
            include_root (bool): If False, only descendant nodes will be
                checked for a match.
        Yields:
            Node: All nodes of the tree for which
            ``match_function(node)`` returns True
        """
        if include_root and match_function(self):
            yield self
        else:
            for child in self.children:
                for node in child.select(match_function, True):
                    yield node


    def select_by_tag(self, tag_names: Union[str, AbstractSet[str]],
                      include_root: bool=True) -> Iterator['Node']:
        """
        Returns an iterator that runs through all descendants that have one
        of the given tag names.

        Examples::

            >>> tree = parse_sxpr('(a (b "X") (X (c "d")) (e (X "F")))')
            >>> list(flatten_sxpr(item.as_sxpr()) for item in tree.select_by_tag("X", False))
            ['(X (c "d"))', '(X "F")']
            >>> list(flatten_sxpr(item.as_sxpr()) for item in tree.select_by_tag({"X", "b"}, False))
            ['(b "X")', '(X (c "d"))', '(X "F")']
            >>> any(tree.select_by_tag('a', False))
            False
            >>> list(flatten_sxpr(item.as_sxpr()) for item in tree.select_by_tag('a', True))
            ['(a (b "X") (X (c "d")) (e (X "F")))']
            >>> flatten_sxpr(next(tree.select_by_tag("X", False)).as_sxpr())
            '(X (c "d"))'

        Args:
            tag_name(set): A tag name or set of tag names that is being
                searched for
            include_root (bool): If False, only descendant nodes will be
                checked for a match.
        Yields:
            Node: All nodes which have a given tag name.
        """
        if isinstance(tag_names, str):
            tag_names = frozenset({tag_names})
        return self.select(lambda node: node.tag_name in tag_names, include_root)


    def pick(self, tag_names: Union[str, Set[str]]) -> Optional['Node']:
        """
        Picks the first descendant with one of the given tag_names.

        This function is just syntactic sugar for
        ``next(node.select_by_tag(tag_names, False))``. However, rather than
        raising a StopIterationError if no descendant with the given tag-name
        exists, it returns None.
        """
        try:
            return next(self.select_by_tag(tag_names, False))
        except StopIteration:
            return None


    def tree_size(self) -> int:
        """
        Recursively counts the number of nodes in the tree including the root node.
        """
        return sum(child.tree_size() for child in self.children) + 1


ZOMBIE_NODE = Node(ZOMBIE_PARSER, '')


def parse_sxpr(sxpr: str) -> Node:
    """
    Generates a tree of nodes from an S-expression.

    This can - among other things - be used for deserialization of trees that
    have been serialized with `Node.as_sxpr()` or as a convenient way to
    generate test data.

    Example:
    >>> parse_sxpr("(a (b c))").as_sxpr()
    '(a\\n    (b\\n        "c"\\n    )\\n)'
    """
    sxpr = StringView(sxpr).strip()
    mock_parsers = dict()

    def next_block(s: StringView):
        """Generator that yields all characters until the next closing bracket
        that does not match an opening bracket matched earlier within the same
        package."""
        s = s.strip()
        try:
            while s[0] != ')':
                if s[0] != '(':
                    raise ValueError('"(" expected, not ' + s[:10])
                # assert s[0] == '(', s
                level = 1
                k = 1
                while level > 0:
                    if s[k] == '(':
                        level += 1
                    elif s[k] == ')':
                        level -= 1
                    k += 1
                yield s[:k]
                s = s[k:].strip()
        except IndexError:
            errmsg = ('Malformed S-expression. Unprocessed part: "%s"' % s) if s \
                else 'Malformed S-expression. Closing bracket(s) ")" missing.'
            raise AssertionError(errmsg)

    def inner_parser(sxpr: StringView) -> Node:
        if sxpr[0] != '(':
            raise ValueError('"(" expected, not ' + sxpr[:10])
        # assert sxpr[0] == '(', sxpr
        sxpr = sxpr[1:].strip()
        match = sxpr.match(re.compile(r'[\w:]+'))
        if match is None:
            raise AssertionError('Malformed S-expression Node-tagname or identifier expected, '
                                 'not "%s"' % sxpr[:40].replace('\n', ''))
        end = match.end() - sxpr.begin
        tagname = sxpr[:end]
        name, class_name = (tagname.split(':') + [''])[:2]
        sxpr = sxpr[end:].strip()
        pos = 0
        attributes = OrderedDict()
        if sxpr[0] == '(':
            result = tuple(inner_parser(block) for block in next_block(sxpr))
            for node in result:
                node._pos = pos
                pos += len(node)
        else:
            lines = []
            while sxpr and sxpr[0:1] != ')':
                # parse attributes
                while sxpr[:2] == "`(":
                    i = sxpr.find('"')
                    k = sxpr.find(')')
                    # read very special attribute pos
                    if sxpr[2:5] == "pos" and 0 < i < k:
                        pos = int(sxpr[5:k].strip().split(' ')[0])
                    # ignore very special attribute err
                    elif sxpr[2:5] == "err" and 0 <= sxpr.find('`', 5) < k:
                        m = sxpr.find('(', 5)
                        while m >= 0 and m < k:
                            m = sxpr.find('(', k)
                            k = max(k, sxpr.find(')', max(m, 0)))
                    # read attributes
                    else:
                        attr = sxpr[2:i].strip()
                        value = sxpr[i:k].strip()[1:-1]
                        attributes[attr] = value
                    sxpr = sxpr[k+1:].strip()
                # parse content
                for qtmark in ['"""', "'''", '"', "'"]:
                    match = sxpr.match(re.compile(qtmark + r'.*?' + qtmark, re.DOTALL))
                    if match:
                        end = match.end() - sxpr.begin
                        i = len(qtmark)
                        lines.append(str(sxpr[i:end - i]))
                        sxpr = sxpr[end:].strip()
                        break
                else:
                    match = sxpr.match(re.compile(r'(?:(?!\)).)*', re.DOTALL))
                    end = match.end() - sxpr.begin
                    lines.append(str(sxpr[:end]))
                    sxpr = sxpr[end:]
            result = "\n".join(lines)
        node = Node(mock_parsers.setdefault(tagname, MockParser(name, ':' + class_name)), result)
        if attributes:
            node.attributes.update(attributes)
        node._pos = pos
        return node

    return inner_parser(sxpr)


def parse_xml(xml: str) -> Node:
    """
    Generates a tree of nodes from a (Pseudo-)XML-source.
    """
    xml = StringView(xml)
    PlainText = MockParser('', 'PlainText')
    mock_parsers = {':PlainText': PlainText}

    def parse_attributes