docstring.py

"""
    sphinx.ext.napoleon.docstring
    ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~


    Classes for docstring parsing and formatting.


    :copyright: Copyright 2007-2021 by the Sphinx team, see AUTHORS.
    :license: BSD, see LICENSE for details.
"""

import collections
import inspect
import re
from functools import partial
from typing import Any, Callable, Dict, Iterable, List, Tuple, Type, Union

from sphinx.application import Sphinx
from sphinx.config import Config as SphinxConfig
from sphinx.locale import _, __
from sphinx.util import logging
from sphinx.util.inspect import stringify_annotation
from sphinx.util.typing import get_type_hints

logger = logging.getLogger(__name__)

_directive_regex = re.compile(r'\.\. \S+::')
_google_section_regex = re.compile(r'^(\s|\w)+:\s*$')
_google_typed_arg_regex = re.compile(r'(.+?)\(\s*(.*[^\s]+)\s*\)')
_numpy_section_regex = re.compile(r'^[=\-`:\'"~^_*+#<>]{2,}\s*$')
_single_colon_regex = re.compile(r'(?<!:):(?!:)')
_xref_or_code_regex = re.compile(
    r'((?::(?:[a-zA-Z0-9]+[\-_+:.])*[a-zA-Z0-9]+:`.+?`)|'
    r'(?:``.+?``))')
_xref_regex = re.compile(
    r'(?:(?::(?:[a-zA-Z0-9]+[\-_+:.])*[a-zA-Z0-9]+:)?`.+?`)'
)
_bullet_list_regex = re.compile(r'^(\*|\+|\-)(\s+\S|\s*$)')
_enumerated_list_regex = re.compile(
    r'^(?P<paren>\()?'
    r'(\d+|#|[ivxlcdm]+|[IVXLCDM]+|[a-zA-Z])'
    r'(?(paren)\)|\.)(\s+\S|\s*$)')
_token_regex = re.compile(
    r"(,\sor\s|\sor\s|\sof\s|:\s|\sto\s|,\sand\s|\sand\s|,\s"
    r"|[{]|[}]"
    r'|"(?:\\"|[^"])*"'
    r"|'(?:\\'|[^'])*')"
)
_default_regex = re.compile(
    r"^default[^_0-9A-Za-z].*$",
)
_SINGLETONS = ("None", "True", "False", "Ellipsis")


class _Stack:
    """A stack of strs, represented intenally as a python list."""

    sentinel = object()

    def __init__(self, items: Iterable[str]):
        """Initialize the stack from an Iterable of items."""
        self._items = list(items)[::-1]

    def __bool__(self) -> bool:
        """Return whether the stack is empty."""
        return bool(self._items)

    def get(self, n: int) -> Any:
        """
        Return the nth element of the stack, or ``self.sentinel`` if n is
        greater than the stack size.
        """
        return (self._items[len(self._items) - 1 - n]
                if n < len(self._items) else self.sentinel)

    def pop(self) -> str:
        """Pop and return the topmost stack element."""
        return self._items.pop()


def _convert_type_spec(_type: str, translations: Dict[str, str] = {}) -> str:
    """Convert type specification to reference in reST."""
    if _type in translations:
        return translations[_type]
    else:
        if _type == 'None':
            return ':obj:`None`'
        else:
            return ':class:`%s`' % _type

    return _type


class GoogleDocstring:
    """Convert Google style docstrings to reStructuredText.

    Parameters
    ----------
    docstring : :obj:`str` or :obj:`list` of :obj:`str`
        The docstring to parse, given either as a string or split into
        individual lines.
    config: :obj:`sphinx.ext.napoleon.Config` or :obj:`sphinx.config.Config`
        The configuration settings to use. If not given, defaults to the
        config object on `app`; or if `app` is not given defaults to the
        a new :class:`sphinx.ext.napoleon.Config` object.


    Other Parameters
    ----------------
    app : :class:`sphinx.application.Sphinx`, optional
        Application object representing the Sphinx process.
    what : :obj:`str`, optional
        A string specifying the type of the object to which the docstring
        belongs. Valid values: "module", "class", "exception", "function",
        "method", "attribute".
    name : :obj:`str`, optional
        The fully qualified name of the object.
    obj : module, class, exception, function, method, or attribute
        The object to which the docstring belongs.
    options : :class:`sphinx.ext.autodoc.Options`, optional
        The options given to the directive: an object with attributes
        inherited_members, undoc_members, show_inheritance and noindex that
        are True if the flag option of same name was given to the auto
        directive.


    Example
    -------
    >>> from sphinx.ext.napoleon import Config
    >>> config = Config(napoleon_use_param=True, napoleon_use_rtype=True)
    >>> docstring = '''One line summary.
    ...
    ... Extended description.
    ...
    ... Args:
    ...   arg1(int): Description of `arg1`
    ...   arg2(str): Description of `arg2`
    ... Returns:
    ...   str: Description of return value.
    ... '''
    >>> print(GoogleDocstring(docstring, config))
    One line summary.
    <BLANKLINE>
    Extended description.
    <BLANKLINE>
    :param arg1: Description of `arg1`
    :type arg1: int
    :param arg2: Description of `arg2`
    :type arg2: str
    <BLANKLINE>
    :returns: Description of return value.
    :rtype: str
    <BLANKLINE>

    """

    _name_rgx = re.compile(r"^\s*((?::(?P<role>\S+):)?`(?P<name>~?[a-zA-Z0-9_.-]+)`|"
                           r" (?P<name2>~?[a-zA-Z0-9_.-]+))\s*", re.X)

    def __init__(self, docstring: Union[str, List[str]], config: SphinxConfig = None,
                 app: Sphinx = None, what: str = '', name: str = '',
                 obj: Any = None, options: Any = None) -> None:
        self._config = config
        self._app = app

        if not self._config:
            from sphinx.ext.napoleon import Config
            self._config = self._app.config if self._app else Config()  # type: ignore

        if not what:
            if inspect.isclass(obj):
                what = 'class'
            elif inspect.ismodule(obj):
                what = 'module'
            elif callable(obj):
                what = 'function'
            else:
                what = 'object'

        self._what = what
        self._name = name
        self._obj = obj
        self._opt = options
        if isinstance(docstring, str):
            lines = docstring.splitlines()
        else:
            lines = docstring
        self._lines = _Stack(map(str.rstrip, lines))
        self._parsed_lines: List[str] = []
        self._is_in_section = False
        self._section_indent = 0
        if not hasattr(self, '_directive_sections'):
            self._directive_sections: List[str] = []
        if not hasattr(self, '_sections'):
            self._sections: Dict[str, Callable] = {
                'args': self._parse_parameters_section,
                'arguments': self._parse_parameters_section,
                'attention': partial(self._parse_admonition, 'attention'),
                'attributes': self._parse_attributes_section,
                'caution': partial(self._parse_admonition, 'caution'),
                'danger': partial(self._parse_admonition, 'danger'),
                'error': partial(self._parse_admonition, 'error'),
                'example': self._parse_examples_section,
                'examples': self._parse_examples_section,
                'hint': partial(self._parse_admonition, 'hint'),
                'important': partial(self._parse_admonition, 'important'),
                'keyword args': self._parse_keyword_arguments_section,
                'keyword arguments': self._parse_keyword_arguments_section,
                'methods': self._parse_methods_section,
                'note': partial(self._parse_admonition, 'note'),
                'notes': self._parse_notes_section,
                'other parameters': self._parse_other_parameters_section,
                'parameters': self._parse_parameters_section,
                'receive': self._parse_receives_section,
                'receives': self._parse_receives_section,
                'return': self._parse_returns_section,
                'returns': self._parse_returns_section,
                'raise': self._parse_raises_section,
                'raises': self._parse_raises_section,
                'references': self._parse_references_section,
                'see also': self._parse_see_also_section,
                'tip': partial(self._parse_admonition, 'tip'),
                'todo': partial(self._parse_admonition, 'todo'),
                'warning': partial(self._parse_admonition, 'warning'),
                'warnings': partial(self._parse_admonition, 'warning'),
                'warn': self._parse_warns_section,
                'warns': self._parse_warns_section,
                'yield': self._parse_yields_section,
                'yields': self._parse_yields_section,
            }

        self._load_custom_sections()

        self._parse()

    def __str__(self) -> str:
        """Return the parsed docstring in reStructuredText format.

        Returns
        -------
        unicode
            Unicode version of the docstring.

        """
        return '\n'.join(self.lines())

    def lines(self) -> List[str]:
        """Return the parsed lines of the docstring in reStructuredText format.

        Returns
        -------
        list(str)
            The lines of the docstring in a list.

        """
        return self._parsed_lines

    def _consume_indented_block(self, indent: int = 1) -> List[str]:
        lines = []
        line = self._lines.get(0)
        while(not self._is_section_break() and
              (not line or self._is_indented(line, indent))):
            lines.append(self._lines.pop())
            line = self._lines.get(0)
        return lines

    def _consume_contiguous(self) -> List[str]:
        lines = []
        while (self._lines and
               self._lines.get(0) and
               not self._is_section_header()):
            lines.append(self._lines.pop())
        return lines

    def _consume_empty(self) -> List[str]:
        lines = []
        line = self._lines.get(0)
        while self._lines and not line:
            lines.append(self._lines.pop())
            line = self._lines.get(0)
        return lines

    def _consume_field(self, parse_type: bool = True, prefer_type: bool = False
                       ) -> Tuple[str, str, List[str]]:
        line = self._lines.pop()

        before, colon, after = self._partition_field_on_colon(line)
        _name, _type, _desc = before, '', after

        if parse_type:
            match = _google_typed_arg_regex.match(before)
            if match:
                _name = match.group(1).strip()
                _type = match.group(2)

        _name = self._escape_args_and_kwargs(_name)

        if prefer_type and not _type:
            _type, _name = _name, _type

        if _type and self._config.napoleon_preprocess_types:
            _type = _convert_type_spec(_type, self._config.napoleon_type_aliases or {})

        indent = self._get_indent(line) + 1
        _descs = [_desc] + self._dedent(self._consume_indented_block(indent))
        _descs = self.__class__(_descs, self._config).lines()
        return _name, _type, _descs

    def _consume_fields(self, parse_type: bool = True, prefer_type: bool = False,
                        multiple: bool = False) -> List[Tuple[str, str, List[str]]]:
        self._consume_empty()
        fields = []
        while not self._is_section_break():
            _name, _type, _desc = self._consume_field(parse_type, prefer_type)
            if multiple and _name:
                for name in _name.split(","):
                    fields.append((name.strip(), _type, _desc))
            elif _name or _type or _desc:
                fields.append((_name, _type, _desc,))
        return fields

    def _consume_inline_attribute(self) -> Tuple[str, List[str]]:
        line = self._lines.pop()
        _type, colon, _desc = self._partition_field_on_colon(line)
        if not colon or not _desc:
            _type, _desc = _desc, _type
            _desc += colon
        _descs = [_desc] + self._dedent(self._consume_to_end())
        _descs = self.__class__(_descs, self._config).lines()
        return _type, _descs

    def _consume_returns_section(self, preprocess_types: bool = False
                                 ) -> List[Tuple[str, str, List[str]]]:
        lines = self._dedent(self._consume_to_next_section())
        if lines:
            before, colon, after = self._partition_field_on_colon(lines[0])
            _name, _type, _desc = '', '', lines

            if colon:
                if after:
                    _desc = [after] + lines[1:]
                else:
                    _desc = lines[1:]

                _type = before

            if (_type and preprocess_types and
                    self._config.napoleon_preprocess_types):
                _type = _convert_type_spec(_type, self._config.napoleon_type_aliases or {})

            _desc = self.__class__(_desc, self._config).lines()
            return [(_name, _type, _desc,)]
        else:
            return []

    def _consume_usage_section(self) -> List[str]:
        lines = self._dedent(self._consume_to_next_section())
        return lines

    def _consume_section_header(self) -> str:
        section = self._lines.pop()
        stripped_section = section.strip(':')
        if stripped_section.lower() in self._sections:
            section = stripped_section
        return section

    def _consume_to_end(self) -> List[str]:
        lines = []
        while self._lines:
            lines.append(self._lines.pop())
        return lines

    def _consume_to_next_section(self) -> List[str]:
        self._consume_empty()
        lines = []
        while not self._is_section_break():
            lines.append(self._lines.pop())
        return lines + self._consume_empty()

    def _dedent(self, lines: List[str], full: bool = False) -> List[str]:
        if full:
            return [line.lstrip() for line in lines]
        else:
            min_indent = self._get_min_indent(lines)
            return [line[min_indent:] for line in lines]

    def _escape_args_and_kwargs(self, name: str) -> str:
        if name.endswith('_') and getattr(self._config, 'strip_signature_backslash', False):
            name = name[:-1] + r'\_'

        if name[:2] == '**':
            return r'\*\*' + name[2:]
        elif name[:1] == '*':
            return r'\*' + name[1:]
        else:
            return name

    def _fix_field_desc(self, desc: List[str]) -> List[str]:
        if self._is_list(desc):
            desc = [''] + desc
        elif desc[0].endswith('::'):
            desc_block = desc[1:]
            indent = self._get_indent(desc[0])
            block_indent = self._get_initial_indent(desc_block)
            if block_indent > indent:
                desc = [''] + desc
            else:
                desc = ['', desc[0]] + self._indent(desc_block, 4)
        return desc

    def _format_admonition(self, admonition: str, lines: List[str]) -> List[str]:
        lines = self._strip_empty(lines)
        if len(lines) == 1:
            return ['.. %s:: %s' % (admonition, lines[0].strip()), '']
        elif lines:
            lines = self._indent(self._dedent(lines), 3)
            return ['.. %s::' % admonition, ''] + lines + ['']
        else:
            return ['.. %s::' % admonition, '']

    def _format_block(self, prefix: str, lines: List[str], padding: str = None) -> List[str]:
        if lines:
            if padding is None:
                padding = ' ' * len(prefix)
            result_lines = []
            for i, line in enumerate(lines):
                if i == 0:
                    result_lines.append((prefix + line).rstrip())
                elif line:
                    result_lines.append(padding + line)
                else:
                    result_lines.append('')
            return result_lines
        else:
            return [prefix]

    def _format_docutils_params(self, fields: List[Tuple[str, str, List[str]]],
                                field_role: str = 'param', type_role: str = 'type'
                                ) -> List[str]:
        lines = []
        for _name, _type, _desc in fields:
            _desc = self._strip_empty(_desc)
            if any(_desc):
                _desc = self._fix_field_desc(_desc)
                field = ':%s %s: ' % (field_role, _name)
                lines.extend(self._format_block(field, _desc))
            else:
                lines.append(':%s %s:' % (field_role, _name))

            if _type:
                lines.append(':%s %s: %s' % (type_role, _name, _type))
        return lines + ['']

    def _format_field(self, _name: str, _type: str, _desc: List[str]) -> List[str]:
        _desc = self._strip_empty(_desc)
        has_desc = any(_desc)
        separator = ' -- ' if has_desc else ''
        if _name:
            if _type:
                if '`' in _type:
                    field = '**%s** (%s)%s' % (_name, _type, separator)
                else:
                    field = '**%s** (*%s*)%s' % (_name, _type, separator)
            else:
                field = '**%s**%s' % (_name, separator)
        elif _type:
            if '`' in _type:
                field = '%s%s' % (_type, separator)
            else:
                field = '*%s*%s' % (_type, separator)
        else:
            field = ''

        if has_desc:
            _desc = self._fix_field_desc(_desc)
            if _desc[0]:
                return [field + _desc[0]] + _desc[1:]
            else:
                return [field] + _desc
        else:
            return [field]

    def _format_fields(self, field_type: str, fields: List[Tuple[str, str, List[str]]]
                       ) -> List[str]:
        field_type = ':%s:' % field_type.strip()
        padding = ' ' * len(field_type)
        multi = len(fields) > 1
        lines: List[str] = []
        for _name, _type, _desc in fields:
            field = self._format_field(_name, _type, _desc)
            if multi:
                if lines:
                    lines.extend(self._format_block(padding + ' * ', field))
                else:
                    lines.extend(self._format_block(field_type + ' * ', field))
            else:
                lines.extend(self._format_block(field_type + ' ', field))
        if lines and lines[-1]:
            lines.append('')
        return lines

    def _get_current_indent(self, peek_ahead: int = 0) -> int:
        line = self._lines.get(peek_ahead)
        while line is not self._lines.sentinel:
            if line:
                return self._get_indent(line)
            peek_ahead += 1
            line = self._lines.get(peek_ahead)
        return 0

    def _get_indent(self, line: str) -> int:
        for i, s in enumerate(line):
            if not s.isspace():
                return i
        return len(line)

    def _get_initial_indent(self, lines: List[str]) -> int:
        for line in lines:
            if line:
                return self._get_indent(line)
        return 0

    def _get_min_indent(self, lines: List[str]) -> int:
        min_indent = None
        for line in lines:
            if line:
                indent = self._get_indent(line)
                if min_indent is None:
                    min_indent = indent
                elif indent < min_indent:
                    min_indent = indent
        return min_indent or 0

    def _indent(self, lines: List[str], n: int = 4) -> List[str]:
        return [(' ' * n) + line for line in lines]

    def _is_indented(self, line: str, indent: int = 1) -> bool:
        for i, s in enumerate(line):
            if i >= indent:
                return True
            elif not s.isspace():
                return False
        return False

    def _is_list(self, lines: List[str]) -> bool:
        if not lines:
            return False
        if _bullet_list_regex.match(lines[0]):
            return True
        if _enumerated_list_regex.match(lines[0]):
            return True
        if len(lines) < 2 or lines[0].endswith('::'):
            return False
        indent = self._get_indent(lines[0])
        next_indent = indent
        for line in lines[1:]:
            if line:
                next_indent = self._get_indent(line)
                break
        return next_indent > indent

    def _is_section_header(self) -> bool:
        section = self._lines.get(0).lower()
        match = _google_section_regex.match(section)
        if match and section.strip(':') in self._sections:
            header_indent = self._get_indent(section)
            section_indent = self._get_current_indent(peek_ahead=1)
            return section_indent > header_indent
        elif self._directive_sections:
            if _directive_regex.match(section):
                for directive_section in self._directive_sections:
                    if section.startswith(directive_section):
                        return True
        return False

    def _is_section_break(self) -> bool:
        line = self._lines.get(0)
        return (not self._lines or
                self._is_section_header() or
                (self._is_in_section and
                    line and
                    not self._is_indented(line, self._section_indent)))

    def _load_custom_sections(self) -> None:
        if self._config.napoleon_custom_sections is not None:
            for entry in self._config.napoleon_custom_sections:
                if isinstance(entry, str):
                    # if entry is just a label, add to sections list,
                    # using generic section logic.
                    self._sections[entry.lower()] = self._parse_custom_generic_section
                else:
                    # otherwise, assume entry is container;
                    if entry[1] == "params_style":
                        self._sections[entry[0].lower()] = \
                            self._parse_custom_params_style_section
                    elif entry[1] == "returns_style":
                        self._sections[entry[0].lower()] = \
                            self._parse_custom_returns_style_section
                    else:
                        # [0] is new section, [1] is the section to alias.
                        # in the case of key mismatch, just handle as generic section.
                        self._sections[entry[0].lower()] = \
                            self._sections.get(entry[1].lower(),
                                               self._parse_custom_generic_section)

    def _parse(self) -> None:
        self._parsed_lines = self._consume_empty()

        if self._name and self._what in ('attribute', 'data', 'property'):
            # Implicit stop using StopIteration no longer allowed in
            # Python 3.7; see PEP 479
            res: List[str] = []
            try:
                res = self._parse_attribute_docstring()
            except StopIteration:
                pass
            self._parsed_lines.extend(res)
            return

        while self._lines:
            if self._is_section_header():
                try:
                    section = self._consume_section_header()
                    self._is_in_section = True
                    self._section_indent = self._get_current_indent()
                    if _directive_regex.match(section):
                        lines = [section] + self._consume_to_next_section()
                    else:
                        lines = self._sections[section.lower()](section)
                finally:
                    self._is_in_section = False
                    self._section_indent = 0
            else:
                if not self._parsed_lines:
                    lines = self._consume_contiguous() + self._consume_empty()
                else:
                    lines = self._consume_to_next_section()
            self._parsed_lines.extend(lines)

    def _parse_admonition(self, admonition: str, section: str) -> List[str]:
        # type (str, str) -> List[str]
        lines = self._consume_to_next_section()
        return self._format_admonition(admonition, lines)

    def _parse_attribute_docstring(self) -> List[str]:
        _type, _desc = self._consume_inline_attribute()
        lines = self._format_field('', '', _desc)
        if _type:
            lines.extend(['', ':type: %s' % _type])
        return lines

    def _parse_attributes_section(self, section: str) -> List[str]:
        lines = []
        for _name, _type, _desc in self._consume_fields():
            if not _type:
                _type = self._lookup_annotation(_name)
            if self._config.napoleon_use_ivar:
                _name = self._qualify_name(_name, self._obj)
                field = ':ivar %s: ' % _name
                lines.extend(self._format_block(field, _desc))
                if _type:
                    lines.append(':vartype %s: %s' % (_name, _type))
            else:
                lines.append('.. attribute:: ' + _name)
                if self._opt and 'noindex' in self._opt:
                    lines.append('   :noindex:')
                lines.append('')

                fields = self._format_field('', '', _desc)
                lines.extend(self._indent(fields, 3))
                if _type:
                    lines.append('')
                    lines.extend(self._indent([':type: %s' % _type], 3))
                lines.append('')
        if self._config.napoleon_use_ivar:
            lines.append('')
        return lines

    def _parse_examples_section(self, section: str) -> List[str]:
        labels = {
            'example': _('Example'),
            'examples': _('Examples'),
        }
        use_admonition = self._config.napoleon_use_admonition_for_examples
        label = labels.get(section.lower(), section)
        return self._parse_generic_section(label, use_admonition)

    def _parse_custom_generic_section(self, section: str) -> List[str]:
        # for now, no admonition for simple custom sections
        return self._parse_generic_section(section, False)

    def _parse_custom_params_style_section(self, section: str) -> List[str]:
        return self._format_fields(section, self._consume_fields())

    def _parse_custom_returns_style_section(self, section: str) -> List[str]:
        fields = self._consume_returns_section(preprocess_types=True)
        return self._format_fields(section, fields)

    def _parse_usage_section(self, section: str) -> List[str]:
        header = ['.. rubric:: Usage:', '']
        block = ['.. code-block:: python', '']
        lines = self._consume_usage_section()
        lines = self._indent(lines, 3)
        return header + block + lines + ['']

    def _parse_generic_section(self, section: str, use_admonition: bool) -> List[str]:
        lines = self._strip_empty(self._consume_to_next_section())
        lines = self._dedent(lines)
        if use_admonition:
            header = '.. admonition:: %s' % section
            lines = self._indent(lines, 3)
        else:
            header = '.. rubric:: %s' % section
        if lines:
            return [header, ''] + lines + ['']
        else:
            return [header, '']

    def _parse_keyword_arguments_section(self, section: str) -> List[str]:
        fields = self._consume_fields()
        if self._config.napoleon_use_keyword:
            return self._format_docutils_params(
                fields,
                field_role="keyword",
                type_role="kwtype")
        else:
            return self._format_fields(_('Keyword Arguments'), fields)

    def _parse_methods_section(self, section: str) -> List[str]:
        lines: List[str] = []
        for _name, _type, _desc in self._consume_fields(parse_type=False):
            lines.append('.. method:: %s' % _name)
            if self._opt and 'noindex' in self._opt:
                lines.append('   :noindex:')
            if _desc:
                lines.extend([''] + self._indent(_desc, 3))
            lines.append('')
        return lines

    def _parse_notes_section(self, section: str) -> List[str]:
        use_admonition = self._config.napoleon_use_admonition_for_notes
        return self._parse_generic_section(_('Notes'), use_admonition)

    def _parse_other_parameters_section(self, section: str) -> List[str]:
        if self._config.napoleon_use_param:
            # Allow to declare multiple parameters at once (ex: x, y: int)
            fields = self._consume_fields(multiple=True)
            return self._format_docutils_params(fields)
        else:
            fields = self._consume_fields()
            return self._format_fields(_('Other Parameters'), fields)

    def _parse_parameters_section(self, section: str) -> List[str]:
        if self._config.napoleon_use_param:
            # Allow to declare multiple parameters at once (ex: x, y: int)
            fields = self._consume_fields(multiple=True)
            return self._format_docutils_params(fields)
        else:
            fields = self._consume_fields()
            return self._format_fields(_('Parameters'), fields)

    def _parse_raises_section(self, section: str) -> List[str]:
        fields = self._consume_fields(parse_type=False, prefer_type=True)
        lines: List[str] = []
        for _name, _type, _desc in fields:
            m = self._name_rgx.match(_type)
            if m and m.group('name'):
                _type = m.group('name')
            elif _xref_regex.match(_type):
                pos = _type.find('`')
                _type = _type[pos + 1:-1]
            _type = ' ' + _type if _type else ''
            _desc = self._strip_empty(_desc)
            _descs = ' ' + '\n    '.join(_desc) if any(_desc) else ''
            lines.append(':raises%s:%s' % (_type, _descs))
        if lines:
            lines.append('')
        return lines

    def _parse_receives_section(self, section: str) -> List[str]:
        if self._config.napoleon_use_param:
            # Allow to declare multiple parameters at once (ex: x, y: int)
            fields = self._consume_fields(multiple=True)
            return self._format_docutils_params(fields)
        else:
            fields = self._consume_fields()
            return self._format_fields(_('Receives'), fields)

    def _parse_references_section(self, section: str) -> List[str]:
        use_admonition = self._config.napoleon_use_admonition_for_references
        return self._parse_generic_section(_('References'), use_admonition)

    def _parse_returns_section(self, section: str) -> List[str]:
        fields = self._consume_returns_section()
        multi = len(fields) > 1
        if multi:
            use_rtype = False
        else:
            use_rtype = self._config.napoleon_use_rtype

        lines: List[str] = []
        for _name, _type, _desc in fields:
            if use_rtype:
                field = self._format_field(_name, '', _desc)
            else:
                field = self._format_field(_name, _type, _desc)

            if multi:
                if lines:
                    lines.extend(self._format_block('          * ', field))
                else:
                    lines.extend(self._format_block(':returns: * ', field))
            else:
                lines.extend(self._format_block(':returns: ', field))
                if _type and use_rtype:
                    lines.extend([':rtype: %s' % _type, ''])
        if lines and lines[-1]:
            lines.append('')
        return lines

    def _parse_see_also_section(self, section: str) -> List[str]:
        return self._parse_admonition('seealso', section)

    def _parse_warns_section(self, section: str) -> List[str]:
        return self._format_fields(_('Warns'), self._consume_fields())

    def _parse_yields_section(self, section: str) -> List[str]:
        fields = self._consume_returns_section(preprocess_types=True)
        return self._format_fields(_('Yields'), fields)

    def _partition_field_on_colon(self, line: str) -> Tuple[str, str, str]:
        before_colon = []
        after_colon = []
        colon = ''
        found_colon = False
        for i, source in enumerate(_xref_or_code_regex.split(line)):
            if found_colon:
                after_colon.append(source)
            else:
                m = _single_colon_regex.search(source)
                if (i % 2) == 0 and m:
                    found_colon = True
                    colon = source[m.start(): m.end()]
                    before_colon.append(source[:m.start()])
                    after_colon.append(source[m.end():])
                else:
                    before_colon.append(source)

        return ("".join(before_colon).strip(),
                colon,
                "".join(after_colon).strip())

    def _qualify_name(self, attr_name: str, klass: Type) -> str:
        if klass and '.' not in attr_name:
            if attr_name.startswith('~'):
                attr_name = attr_name[1:]
            try:
                q = klass.__qualname__
            except AttributeError:
                q = klass.__name__
            return '~%s.%s' % (q, attr_name)
        return attr_name

    def _strip_empty(self, lines: List[str]) -> List[str]:
        if lines:
            start = -1
            for i, line in enumerate(lines):
                if line:
                    start = i
                    break
            if start == -1:
                lines = []
            end = -1
            for i in reversed(range(len(lines))):
                line = lines[i]
                if line:
                    end = i
                    break
            if start > 0 or end + 1 < len(lines):
                lines = lines[start:end + 1]
        return lines

    def _lookup_annotation(self, _name: str) -> str:
        if self._config.napoleon_attr_annotations:
            if self._what in ("module", "class", "exception") and self._obj:
                # cache the class annotations
                if not hasattr(self, "_annotations"):
                    localns = getattr(self._config, "autodoc_type_aliases", {})
                    localns.update(getattr(
                                   self._config, "napoleon_type_aliases", {}
                                   ) or {})
                    self._annotations = get_type_hints(self._obj, None, localns)
                if _name in self._annotations:
                    return stringify_annotation(self._annotations[_name])
        # No annotation found
        return ""


def _recombine_set_tokens(tokens: List[str]) -> List[str]:
    token_queue = collections.deque(tokens)
    keywords = ("optional", "default")

    def takewhile_set(tokens):
        open_braces = 0
        previous_token = None
        while True:
            try:
                token = tokens.popleft()
            except IndexError:
                break

            if token == ", ":
                previous_token = token
                continue

            if not token.strip():
                continue

            if token in keywords:
                tokens.appendleft(token)
                if previous_token is not None:
                    tokens.appendleft(previous_token)
                break

            if previous_token is not None:
                yield previous_token
                previous_token = None

            if token == "{":
                open_braces += 1
            elif token == "}":
                open_braces -= 1

            yield token

            if open_braces == 0:
                break

    def combine_set(tokens):
        while True:
            try:
                token = tokens.popleft()
            except IndexError:
                break

            if token == "{":
                tokens.appendleft("{")
                yield "".join(takewhile_set(tokens))
            else:
                yield token

    return list(combine_set(token_queue))


def _tokenize_type_spec(spec: str) -> List[str]:
    def postprocess(item):
        if _default_regex.match(item):
            default = item[:7]
            # can't be separated by anything other than a single space
            # for now
            other = item[8:]

            return [default, " ", other]
        else:
            return [item]

    tokens = list(
        item
        for raw_token in _token_regex.split(spec)
        for item in postprocess(raw_token)
        if item
    )
    return tokens


def _token_type(token: str, location: str = None) -> str:
    def is_numeric(token):
        try:
            # use complex to make sure every numeric value is detected as literal
            complex(token)
        except ValueError:
            return False
        else:
            return True

    if token.startswith(" ") or token.endswith(" "):
        type_ = "delimiter"
    elif (
            is_numeric(token) or
            (token.startswith("{") and token.endswith("}")) or
            (token.startswith('"') and token.endswith('"')) or
            (token.startswith("'") and token.endswith("'"))
    ):
        type_ = "literal"
    elif token.startswith("{"):
        logger.warning(
            __("invalid value set (missing closing brace): %s"),
            token,
            location=location,
        )
        type_ = "literal"
    elif token.endswith("}"):
        logger.warning(
            __("invalid value set (missing opening brace): %s"),
            token,
            location=location,
        )
        type_ = "literal"
    elif token.startswith("'") or token.startswith('"'):
        logger.warning(
            __("malformed string literal (missing closing quote): %s"),
            token,
            location=location,
        )
        type_ = "literal"
    elif token.endswith("'") or token.endswith('"'):
        logger.warning(
            __("malformed string literal (missing opening quote): %s"),
            token,
            location=location,
        )
        type_ = "literal"
    elif token in ("optional", "default"):
        # default is not a official keyword (yet) but supported by the
        # reference implementation (numpydoc) and widely used
        type_ = "control"
    elif _xref_regex.match(token):
        type_ = "reference"
    else:
        type_ = "obj"

    return type_


def _convert_numpy_type_spec(_type: str, location: str = None, translations: dict = {}) -> str:
    def convert_obj(obj, translations, default_translation):
        translation = translations.get(obj, obj)

        # use :class: (the default) only if obj is not a standard singleton
        if translation in _SINGLETONS and default_translation == ":class:`%s`":
            default_translation = ":obj:`%s`"
        elif translation == "..." and default_translation == ":class:`%s`":
            # allow referencing the builtin ...
            default_translation = ":obj:`%s <Ellipsis>`"

        if _xref_regex.match(translation) is None:
            translation = default_translation % translation

        return translation

    tokens = _tokenize_type_spec(_type)
    combined_tokens = _recombine_set_tokens(tokens)
    types = [
        (token, _token_type(token, location))
        for token in combined_tokens
    ]

    converters = {
        "literal": lambda x: "``%s``" % x,
        "obj": lambda x: convert_obj(x, translations, ":class:`%s`"),
        "control": lambda x: "*%s*" % x,
        "delimiter": lambda x: x,
        "reference": lambda x: x,
    }

    converted = "".join(converters.get(type_)(token) for token, type_ in types)

    return converted


class NumpyDocstring(GoogleDocstring):
    """Convert NumPy style docstrings to reStructuredText.

    Parameters
    ----------
    docstring : :obj:`str` or :obj:`list` of :obj:`str`
        The docstring to parse, given either as a string or split into
        individual lines.
    config: :obj:`sphinx.ext.napoleon.Config` or :obj:`sphinx.config.Config`
        The configuration settings to use. If not given, defaults to the
        config object on `app`; or if `app` is not given defaults to the
        a new :class:`sphinx.ext.napoleon.Config` object.


    Other Parameters
    ----------------
    app : :class:`sphinx.application.Sphinx`, optional
        Application object representing the Sphinx process.
    what : :obj:`str`, optional
        A string specifying the type of the object to which the docstring
        belongs. Valid values: "module", "class", "exception", "function",
        "method", "attribute".
    name : :obj:`str`, optional
        The fully qualified name of the object.
    obj : module, class, exception, function, method, or attribute
        The object to which the docstring belongs.
    options : :class:`sphinx.ext.autodoc.Options`, optional
        The options given to the directive: an object with attributes
        inherited_members, undoc_members, show_inheritance and noindex that
        are True if the flag option of same name was given to the auto
        directive.


    Example
    -------
    >>> from sphinx.ext.napoleon import Config
    >>> config = Config(napoleon_use_param=True, napoleon_use_rtype=True)
    >>> docstring = '''One line summary.
    ...
    ... Extended description.
    ...
    ... Parameters
    ... ----------
    ... arg1 : int
    ...     Description of `arg1`
    ... arg2 : str
    ...     Description of `arg2`
    ... Returns
    ... -------
    ... str
    ...     Description of return value.
    ... '''
    >>> print(NumpyDocstring(docstring, config))
    One line summary.
    <BLANKLINE>
    Extended description.
    <BLANKLINE>
    :param arg1: Description of `arg1`
    :type arg1: int
    :param arg2: Description of `arg2`
    :type arg2: str
    <BLANKLINE>
    :returns: Description of return value.
    :rtype: str
    <BLANKLINE>

    Methods
    -------
    __str__()
        Return the parsed docstring in reStructuredText format.

        Returns
        -------
        str
            UTF-8 encoded version of the docstring.

    __unicode__()
        Return the parsed docstring in reStructuredText format.

        Returns
        -------
        unicode
            Unicode version of the docstring.

    lines()
        Return the parsed lines of the docstring in reStructuredText format.

        Returns
        -------
        list(str)
            The lines of the docstring in a list.

    """
    def __init__(self, docstring: Union[str, List[str]], config: SphinxConfig = None,
                 app: Sphinx = None, what: str = '', name: str = '',
                 obj: Any = None, options: Any = None) -> None:
        self._directive_sections = ['.. index::']
        super().__init__(docstring, config, app, what, name, obj, options)

    def _get_location(self) -> str:
        try:
            filepath = inspect.getfile(self._obj) if self._obj is not None else None
        except TypeError:
            filepath = None
        name = self._name

        if filepath is None and name is None:
            return None
        elif filepath is None:
            filepath = ""

        return ":".join([filepath, "docstring of %s" % name])

    def _escape_args_and_kwargs(self, name: str) -> str:
        func = super()._escape_args_and_kwargs

        if ", " in name:
            return ", ".join(func(param) for param in name.split(", "))
        else:
            return func(name)

    def _consume_field(self, parse_type: bool = True, prefer_type: bool = False
                       ) -> Tuple[str, str, List[str]]:
        line = self._lines.pop()
        if parse_type:
            _name, _, _type = self._partition_field_on_colon(line)
        else:
            _name, _type = line, ''
        _name, _type = _name.strip(), _type.strip()
        _name = self._escape_args_and_kwargs(_name)

        if parse_type and not _type:
            _type = self._lookup_annotation(_name)

        if prefer_type and not _type:
            _type, _name = _name, _type

        if self._config.napoleon_preprocess_types:
            _type = _convert_numpy_type_spec(
                _type,
                location=self._get_location(),
                translations=self._config.napoleon_type_aliases or {},
            )

        indent = self._get_indent(line) + 1
        _desc = self._dedent(self._consume_indented_block(indent))
        _desc = self.__class__(_desc, self._config).lines()
        return _name, _type, _desc

    def _consume_returns_section(self, preprocess_types: bool = False
                                 ) -> List[Tuple[str, str, List[str]]]:
        return self._consume_fields(prefer_type=True)

    def _consume_section_header(self) -> str:
        section = self._lines.pop()
        if not _directive_regex.match(section):
            # Consume the header underline
            self._lines.pop()
        return section

    def _is_section_break(self) -> bool:
        line1, line2 = self._lines.get(0), self._lines.get(1)
        return (not self._lines or
                self._is_section_header() or
                ['', ''] == [line1, line2] or
                (self._is_in_section and
                    line1 and
                    not self._is_indented(line1, self._section_indent)))

    def _is_section_header(self) -> bool:
        section, underline = self._lines.get(0), self._lines.get(1)
        section = section.lower()
        if section in self._sections and isinstance(underline, str):
            return bool(_numpy_section_regex.match(underline))
        elif self._directive_sections:
            if _directive_regex.match(section):
                for directive_section in self._directive_sections:
                    if section.startswith(directive_section):
                        return True
        return False

    def _parse_see_also_section(self, section: str) -> List[str]:
        lines = self._consume_to_next_section()
        try:
            return self._parse_numpydoc_see_also_section(lines)
        except ValueError:
            return self._format_admonition('seealso', lines)

    def _parse_numpydoc_see_also_section(self, content: List[str]) -> List[str]:
        """
        Derived from the NumpyDoc implementation of _parse_see_also.

        See Also
        --------
        func_name : Descriptive text
            continued text
        another_func_name : Descriptive text
        func_name1, func_name2, :meth:`func_name`, func_name3

        """
        items = []

        def parse_item_name(text: str) -> Tuple[str, str]:
            """Match ':role:`name`' or 'name'"""
            m = self._name_rgx.match(text)
            if m:
                g = m.groups()
                if g[1] is None:
                    return g[3], None
                else:
                    return g[2], g[1]
            raise ValueError("%s is not a item name" % text)

        def push_item(name: str, rest: List[str]) -> None:
            if not name:
                return
            name, role = parse_item_name(name)
            items.append((name, list(rest), role))
            del rest[:]

        def translate(func, description, role):
            translations = self._config.napoleon_type_aliases
            if role is not None or not translations:
                return func, description, role

            translated = translations.get(func, func)
            match = self._name_rgx.match(translated)
            if not match:
                return translated, description, role

            groups = match.groupdict()
            role = groups["role"]
            new_func = groups["name"] or groups["name2"]

            return new_func, description, role

        current_func = None
        rest: List[str] = []

        for line in content:
            if not line.strip():
                continue

            m = self._name_rgx.match(line)
            if m and line[m.end():].strip().startswith(':'):
                push_item(current_func, rest)
                current_func, line = line[:m.end()], line[m.end():]
                rest = [line.split(':', 1)[1].strip()]
                if not rest[0]:
                    rest = []
            elif not line.startswith(' '):
                push_item(current_func, rest)
                current_func = None
                if ',' in line:
                    for func in line.split(','):
                        if func.strip():
                            push_item(func, [])
                elif line.strip():
                    current_func = line
            elif current_func is not None:
                rest.append(line.strip())
        push_item(current_func, rest)

        if not items:
            return []

        # apply type aliases
        items = [
            translate(func, description, role)
            for func, description, role in items
        ]

        lines: List[str] = []
        last_had_desc = True
        for name, desc, role in items:
            if role:
                link = ':%s:`%s`' % (role, name)
            else:
                link = ':obj:`%s`' % name
            if desc or last_had_desc:
                lines += ['']
                lines += [link]
            else:
                lines[-1] += ", %s" % link
            if desc:
                lines += self._indent([' '.join(desc)])
                last_had_desc = True
            else:
                last_had_desc = False
        lines += ['']

        return self._format_admonition('seealso', lines)