# $Id: references.py 10303 2026-03-27 08:41:10Z milde $
# Author: David Goodger <goodger@python.org>
# Copyright: This module has been placed in the public domain.

"""
Transforms for resolving references.
"""

from __future__ import annotations

__docformat__ = 'reStructuredText'

from docutils import nodes, utils
from docutils.transforms import Transform


class SectionIDs(Transform):
    """
    Add identifiers to sections.

    If the "legacy_ids" configuration setting is False, the rST parser
    does not generate identifiers for implicit targets (e.g. sections)
    in order to give explicit targets preferential access to identifiers
    matching their reference name.

    However, the `parts.Contents` transform and most writers
    expect sections to have an identifier, so this transform adds them.
    """
    default_priority = 240

    def apply(self) -> None:
        if getattr(self.document.settings, "legacy_ids", True):
            return
        for node in self.document.findall(nodes.section):
            self.document.set_id(node)


class PropagateTargets(Transform):
    """
    Propagate empty internal targets to the next element.

    Given the following nodes::

        <target ids="internal1" names="internal1">
        <target anonymous="1" ids="id1">
        <target ids="internal2" names="internal2">
        <paragraph>
            This is a test.

    `PropagateTargets` propagates the ids and names of the internal
    targets preceding the paragraph to the paragraph itself::

        <target refid="internal1">
        <target anonymous="1" refid="id1">
        <target refid="internal2">
        <paragraph ids="internal2 id1 internal1" names="internal2 internal1">
            This is a test.
    """

    default_priority = 260

    def apply(self) -> None:
        for target in self.document.findall(nodes.target):
            # Only block-level targets without reference (like ".. _target:"):
            if (isinstance(target.parent, nodes.TextElement)
                or (target.hasattr('refid') or target.hasattr('refuri')
                    or target.hasattr('refname'))):
                continue
            assert len(target) == 0, 'error: block-level target has children'
            next_node = target.next_node(ascend=True)
            # skip system messages (may be removed by universal.FilterMessages)
            while isinstance(next_node, nodes.system_message):
                next_node = next_node.next_node(ascend=True, descend=False)
            # Do not move names and ids into Invisibles (we'd lose the
            # attributes) or different Targetables (e.g. footnotes).
            if (next_node is None
                or isinstance(next_node, (nodes.Invisible, nodes.Targetable))
                and not isinstance(next_node, nodes.target)):
                continue
            next_node['ids'].extend(target['ids'])
            next_node['names'].extend(target['names'])
            # Set defaults for next_node.expect_referenced_by_name/id.
            if not hasattr(next_node, 'expect_referenced_by_name'):
                next_node.expect_referenced_by_name = {}
            if not hasattr(next_node, 'expect_referenced_by_id'):
                next_node.expect_referenced_by_id = {}
            for id in target['ids']:
                # Update IDs to node mapping.
                self.document.ids[id] = next_node
                # If next_node is referenced by id ``id``, this
                # target shall be marked as referenced.
                next_node.expect_referenced_by_id[id] = target
            for name in target['names']:
                next_node.expect_referenced_by_name[name] = target
            # If there are any expect_referenced_by_... attributes
            # in target set, copy them to next_node.
            next_node.expect_referenced_by_name.update(
                getattr(target, 'expect_referenced_by_name', {}))
            next_node.expect_referenced_by_id.update(
                getattr(target, 'expect_referenced_by_id', {}))
            # Remove target node from places where it is invalid.
            # TODO: always remove target?
            # +1 It did complete its mission and is currently ignored.
            #    (except for the Sphinx LaTeX writer)
            # -1 It may help a future rST writer.
            if isinstance(target.parent, nodes.figure) and isinstance(
                    next_node, nodes.caption):
                target.parent.remove(target)
                continue
            # Set refid to point to the first former ID of target
            # which is now an ID of next_node.
            target['refid'] = target['ids'][0]
            # Clear ids and names; they have been moved to next_node.
            target['ids'] = []
            target['names'] = []
            self.document.note_refid(target)


class AnonymousHyperlinks(Transform):

    """
    Link anonymous references to targets.  Given::

        <paragraph>
            <reference anonymous="1">
                internal
            <reference anonymous="1">
                external
        <target anonymous="1" ids="id1">
        <target anonymous="1" ids="id2" refuri="http://external">

    Corresponding references are linked via "refid" or resolved via "refuri"::

        <paragraph>
            <reference anonymous="1" refid="id1">
                text
            <reference anonymous="1" refuri="http://external">
                external
        <target anonymous="1" ids="id1">
        <target anonymous="1" ids="id2" refuri="http://external">
    """

    default_priority = 440

    def apply(self) -> None:
        anonymous_refs = [
            node for node in self.document.findall(nodes.reference)
            if node.get('anonymous')]
        anonymous_targets = [
            node for node in self.document.findall(nodes.target)
            if node.get('anonymous')]
        if len(anonymous_refs) != len(anonymous_targets):
            msg = self.document.reporter.error(
                  'Anonymous hyperlink mismatch: %s references but %s '
                  'targets.\nSee "backrefs" attribute for IDs.'
                  % (len(anonymous_refs), len(anonymous_targets)))
            msgid = self.document.set_id(msg)
            for ref in anonymous_refs:
                prb = nodes.problematic(
                      ref.rawsource, ref.rawsource, refid=msgid)
                prbid = self.document.set_id(prb)
                msg.add_backref(prbid)
                ref.replace_self(prb)
            return
        for ref, target in zip(anonymous_refs, anonymous_targets):
            if ref.hasattr('refid') or ref.hasattr('refuri'):
                continue
            target.referenced = True
            while True:
                if target.hasattr('refuri'):
                    ref['refuri'] = target['refuri']
                    ref.resolved = True
                    break
                else:
                    if not target['ids']:
                        # Propagated target.
                        target = self.document.ids[target['refid']]
                        continue
                    ref['refid'] = target['ids'][0]
                    self.document.note_refid(ref)
                    break


class IndirectHyperlinks(Transform):

    """
    a) Indirect external references::

           <paragraph>
               <reference refname="indirect-external">
                   indirect external
           <target ids="id1" names="direct-external"
               refuri="http://indirect">
           <target ids="id2" names="indirect-external"
               refname="direct-external">

       The "refuri" attribute is migrated back to all indirect targets
       from the final direct target (i.e. a target not referring to
       another indirect target)::

           <paragraph>
               <reference refname="indirect-external">
                   indirect external
           <target ids="id1" names="direct-external"
               refuri="http://indirect">
           <target ids="id2" names="indirect-external"
               refuri="http://indirect">

       Once the attribute is migrated, the preexisting "refname" attribute
       is dropped.

    b) Indirect internal references::

           <target ids="id1" names="final-target">
           <paragraph>
               <reference refname="indirect-internal">
                   indirect internal
           <target ids="id2" names="indirect-internal-2"
               refname="final target">
           <target ids="id3" names="indirect-internal"
               refname="indirect-internal-2">

       Targets which indirectly refer to an internal target become one-hop
       indirect (their "refid" attributes are directly set to the internal
       target's "id"). References which indirectly refer to an internal
       target become direct internal references::

           <target ids="id1" names="final-target">
           <paragraph>
               <reference refid="id1">
                   indirect internal
           <target ids="id2" names="indirect-internal-2" refid="id1">
           <target ids="id3" names="indirect-internal" refid="id1">
    """

    default_priority = 460

    def apply(self) -> None:
        for target in self.document.indirect_targets:
            if not target.resolved:
                self.resolve_indirect_target(target)
            self.resolve_indirect_references(target)

    def resolve_indirect_target(self, target) -> None:
        # indirect targets have either a refname or refid attribute
        refname = target.get('refname')
        refid = target.get('refid')
        if refid:
            reftarget = self.document.ids.get(refid)
        else:
            reftarget = self.document.names.get(refname)
            refid = self.document.nameids.get(refname)
            if reftarget and not refid:
                refid = self.document.set_id(reftarget)
        if not reftarget:
            # Check the unknown_reference_resolvers
            for resolver_function in \
                self.document.transformer.unknown_reference_resolvers:
                if resolver_function(target):
                    break
            else:
                self.nonexistent_indirect_target(target)
            return
        reftarget.note_referenced_by(id=refid)
        if (isinstance(reftarget, nodes.target)
            and not reftarget.resolved
            and reftarget.hasattr('refname')):
            if hasattr(target, 'multiply_indirect'):
                self.circular_indirect_reference(target)
                return
            target.multiply_indirect = 1
            self.resolve_indirect_target(reftarget)  # multiply indirect
            del target.multiply_indirect
        if reftarget.hasattr('refuri'):
            target['refuri'] = reftarget['refuri']
            if 'refid' in target:
                del target['refid']
        elif reftarget.hasattr('refid'):
            target['refid'] = reftarget['refid']
            self.document.note_refid(target)
        else:
            if reftarget['ids']:
                target['refid'] = refid
                self.document.note_refid(target)
            else:
                self.nonexistent_indirect_target(target)
                return
        if refname is not None:
            del target['refname']
        target.resolved = True

    def nonexistent_indirect_target(self, target) -> None:
        if self.document.names.get(target['refname'], '') is None:
            self.indirect_target_error(target, 'which is a duplicate, and '
                                       'cannot be used as a unique reference')
        else:
            self.indirect_target_error(target, 'which does not exist')

    def circular_indirect_reference(self, target) -> None:
        self.indirect_target_error(target, 'forming a circular reference')

    def indirect_target_error(self, target, explanation) -> None:
        naming = ''
        reflist = []
        if target['names']:
            naming = '"%s" ' % target['names'][0]
        for name in target['names']:
            reflist.extend(self.document.refnames.get(name, []))
        for id in target['ids']:
            reflist.extend(self.document.refids.get(id, []))
        if target['ids']:
            naming += '(id="%s")' % target['ids'][0]
        msg = self.document.reporter.error(
              'Indirect hyperlink target %s refers to target "%s", %s.'
              % (naming, target['refname'], explanation), base_node=target)
        msgid = self.document.set_id(msg)
        for ref in utils.uniq(reflist):
            prb = nodes.problematic(
                  ref.rawsource, ref.rawsource, refid=msgid)
            prbid = self.document.set_id(prb)
            msg.add_backref(prbid)
            ref.replace_self(prb)
        target.resolved = True

    def resolve_indirect_references(self, target) -> None:
        if target.hasattr('refid'):
            attname = 'refid'
            call_method = self.document.note_refid
        elif target.hasattr('refuri'):
            attname = 'refuri'
            call_method = None
        else:
            return
        attval = target[attname]
        for name in target['names']:
            reflist = self.document.refnames.get(name, [])
            if reflist:
                target.note_referenced_by(name=name)
            for ref in reflist:
                if ref.resolved:
                    continue
                del ref['refname']
                ref[attname] = attval
                if call_method:
                    call_method(ref)
                ref.resolved = True
                if isinstance(ref, nodes.target):
                    self.resolve_indirect_references(ref)
        for id in target['ids']:
            reflist = self.document.refids.get(id, [])
            if reflist:
                target.note_referenced_by(id=id)
            for ref in reflist:
                if ref.resolved:
                    continue
                del ref['refid']
                ref[attname] = attval
                if call_method:
                    call_method(ref)
                ref.resolved = True
                if isinstance(ref, nodes.target):
                    self.resolve_indirect_references(ref)


class ExternalTargets(Transform):

    """
    Given::

        <paragraph>
            <reference refname="direct-external">
                direct external
        <target ids="id1" names="direct-external" refuri="http://direct">

    The "refname" attribute is replaced by the direct "refuri" attribute::

        <paragraph>
            <reference refuri="http://direct">
                direct external
        <target ids="id1" names="direct-external" refuri="http://direct">
    """

    default_priority = 640

    def apply(self) -> None:
        for target in self.document.findall(nodes.target):
            if target.hasattr('refuri'):
                refuri = target['refuri']
                for name in target['names']:
                    reflist = self.document.refnames.get(name, [])
                    if reflist:
                        target.note_referenced_by(name=name)
                    for ref in reflist:
                        if ref.resolved:
                            continue
                        del ref['refname']
                        ref['refuri'] = refuri
                        ref.resolved = True


class InternalTargets(Transform):

    default_priority = 660

    def apply(self) -> None:
        for target in self.document.findall(nodes.target):
            if not target.hasattr('refuri') and not target.hasattr('refid'):
                self.resolve_reference_ids(target)

    def resolve_reference_ids(self, target) -> None:
        """
        Given::

            <paragraph>
                <reference refname="direct-internal">
                    direct internal
            <target ids="id1" names="direct-internal">

        The "refname" attribute is replaced by "refid" linking to the target's
        "id"::

            <paragraph>
                <reference refid="id1">
                    direct internal
            <target ids="id1" names="direct-internal">
        """
        for name in target['names']:
            refid = self.document.nameids.get(name)
            reflist = self.document.refnames.get(name, [])
            if reflist:
                target.note_referenced_by(name=name)
            for ref in reflist:
                if ref.resolved:
                    continue
                if refid:
                    del ref['refname']
                    ref['refid'] = refid
                ref.resolved = True


class Footnotes(Transform):

    """
    Assign numbers to autonumbered footnotes, and resolve links to footnotes,
    citations, and their references.

    Given the following ``document`` as input::

        <document>
            <paragraph>
                A labeled autonumbered footnote reference:
                <footnote_reference auto="1" ids="id1" refname="footnote">
            <paragraph>
                An unlabeled autonumbered footnote reference:
                <footnote_reference auto="1" ids="id2">
            <footnote auto="1" ids="id3">
                <paragraph>
                    Unlabeled autonumbered footnote.
            <footnote auto="1" ids="footnote" names="footnote">
                <paragraph>
                    Labeled autonumbered footnote.

    Auto-numbered footnotes have attribute ``auto="1"`` and no label.
    Auto-numbered footnote_references have no reference text (they're
    empty elements). When resolving the numbering, a ``label`` element
    is added to the beginning of the ``footnote``, and reference text
    to the ``footnote_reference``.

    The transformed result will be::

        <document>
            <paragraph>
                A labeled autonumbered footnote reference:
                <footnote_reference auto="1" ids="id1" refid="footnote">
                    2
            <paragraph>
                An unlabeled autonumbered footnote reference:
                <footnote_reference auto="1" ids="id2" refid="id3">
                    1
            <footnote auto="1" ids="id3" backrefs="id2">
                <label>
                    1
                <paragraph>
                    Unlabeled autonumbered footnote.
            <footnote auto="1" ids="footnote" names="footnote" backrefs="id1">
                <label>
                    2
                <paragraph>
                    Labeled autonumbered footnote.

    Note that the footnotes are not in the same order as the references.

    The labels and reference text are added to the auto-numbered ``footnote``
    and ``footnote_reference`` elements.  Footnote elements are backlinked to
    their references via "refids" attributes.  References are assigned "id"
    and "refid" attributes.

    After adding labels and reference text, the "auto" attributes can be
    ignored.
    """

    default_priority = 620

    autofootnote_labels = None
    """Keep track of unlabeled autonumbered footnotes."""

    symbols = [
          # Entries 1-4 and 6 below are from section 12.51 of
          # The Chicago Manual of Style, 14th edition.
          '*',                          # asterisk/star
          '\u2020',                    # † &dagger; dagger
          '\u2021',                    # ‡ &Dagger; double dagger
          '\u00A7',                    # § &sect; section mark
          '\u00B6',                    # ¶ &para; paragraph mark (pilcrow)
                                        # (parallels ['||'] in CMoS)
          '#',                          # number sign
          # The entries below were chosen arbitrarily.
          '\u2660',                    # ♠ &spades; spade suit
          '\u2665',                    # ♡ &hearts; heart suit
          '\u2666',                    # ♢ &diams; diamond suit
          '\u2663',                    # ♣ &clubs; club suit
          ]

    def apply(self) -> None:
        self.autofootnote_labels = []
        startnum = self.document.autofootnote_start
        self.document.autofootnote_start = self.number_footnotes(startnum)
        self.number_footnote_references(startnum)
        self.symbolize_footnotes()
        self.resolve_footnotes_and_citations()

    def number_footnotes(self, startnum):
        """
        Assign numbers to autonumbered footnotes.

        For labeled autonumbered footnotes, copy the number over to
        corresponding footnote references.
        """
        for footnote in self.document.autofootnotes:
            while True:
                label = str(startnum)
                startnum += 1
                if label not in self.document.nameids:
                    break
            footnote.insert(0, nodes.label('', label))
            for name in footnote['names']:
                for ref in self.document.footnote_refs.get(name, []):
                    ref += nodes.Text(label)
                    ref.delattr('refname')
                    assert len(footnote['ids']) == len(ref['ids']) == 1
                    ref['refid'] = footnote['ids'][0]
                    footnote.add_backref(ref['ids'][0])
                    self.document.note_refid(ref)
                    ref.resolved = True
            if not footnote['names'] and not footnote['dupnames']:
                footnote['names'].append(label)
                self.document.note_explicit_target(footnote, footnote)
                self.autofootnote_labels.append(label)
        return startnum

    def number_footnote_references(self, startnum) -> None:
        """Assign numbers to autonumbered footnote references."""
        i = 0
        for ref in self.document.autofootnote_refs:
            if ref.resolved or ref.hasattr('refid'):
                continue
            try:
                label = self.autofootnote_labels[i]
            except IndexError:
                n = len(self.autofootnote_labels)
                s = 's' if n > 1 else ''
                msg = self.document.reporter.error(
                          'Too many autonumbered footnote references: '
                          f'only {n} corresponding footnote{s} available.',
                          base_node=ref)
                msgid = self.document.set_id(msg)
                for ref in self.document.autofootnote_refs[i:]:
                    if ref.resolved or ref.hasattr('refname'):
                        continue
                    prb = nodes.problematic(
                          ref.rawsource, ref.rawsource, refid=msgid)
                    prbid = self.document.set_id(prb)
                    msg.add_backref(prbid)
                    ref.replace_self(prb)
                break
            ref += nodes.Text(label)
            id = self.document.nameids[label]
            footnote = self.document.ids[id]
            ref['refid'] = id
            self.document.note_refid(ref)
            assert len(ref['ids']) == 1
            footnote.add_backref(ref['ids'][0])
            ref.resolved = True
            i += 1

    def symbolize_footnotes(self) -> None:
        """Add symbols indexes to "[*]"-style footnotes and references."""
        labels = []
        for footnote in self.document.symbol_footnotes:
            reps, index = divmod(self.document.symbol_footnote_start,
                                 len(self.symbols))
            labeltext = self.symbols[index] * (reps + 1)
            labels.append(labeltext)
            footnote.insert(0, nodes.label('', labeltext))
            self.document.symbol_footnote_start += 1
            self.document.set_id(footnote)
        i = 0
        for ref in self.document.symbol_footnote_refs:
            try:
                ref += nodes.Text(labels[i])
            except IndexError:
                msg = self.document.reporter.error(
                      'Too many symbol footnote references: only %s '
                      'corresponding footnotes available.' % len(labels),
                      base_node=ref)
                msgid = self.document.set_id(msg)
                for ref in self.document.symbol_footnote_refs[i:]:
                    if ref.resolved or ref.hasattr('refid'):
                        continue
                    prb = nodes.problematic(
                          ref.rawsource, ref.rawsource, refid=msgid)
                    prbid = self.document.set_id(prb)
                    msg.add_backref(prbid)
                    ref.replace_self(prb)
                break
            footnote = self.document.symbol_footnotes[i]
            assert len(footnote['ids']) == 1
            ref['refid'] = footnote['ids'][0]
            self.document.note_refid(ref)
            footnote.add_backref(ref['ids'][0])
            i += 1

    def resolve_footnotes_and_citations(self) -> None:
        """
        Link manually-labeled footnotes and citations to/from their
        references.
        """
        for footnote in self.document.footnotes:
            for label in footnote['names']:
                if label in self.document.footnote_refs:
                    reflist = self.document.footnote_refs[label]
                    self.resolve_references(footnote, reflist)
        for citation in self.document.citations:
            for label in citation['names']:
                if label in self.document.citation_refs:
                    reflist = self.document.citation_refs[label]
                    self.resolve_references(citation, reflist)

    def resolve_references(self, note, reflist) -> None:
        assert len(note['ids']) == 1
        id = note['ids'][0]
        for ref in reflist:
            if ref.resolved:
                continue
            ref.delattr('refname')
            ref['refid'] = id
            assert len(ref['ids']) == 1
            note.add_backref(ref['ids'][0])
            ref.resolved = True
        note.resolved = True


class CircularSubstitutionDefinitionError(Exception):
    pass


class Substitutions(Transform):

    """
    Given the following ``document`` as input::

        <document>
            <paragraph>
                The
                <substitution_reference refname="biohazard">
                    biohazard
                 symbol is deservedly scary-looking.
            <substitution_definition names="biohazard">
                <image alt="biohazard" uri="biohazard.png">

    The ``substitution_reference`` will simply be replaced by the
    contents of the corresponding ``substitution_definition``.

    The transformed result will be::

        <document>
            <paragraph>
                The
                <image alt="biohazard" uri="biohazard.png">
                 symbol is deservedly scary-looking.
            <substitution_definition names="biohazard">
                <image alt="biohazard" uri="biohazard.png">
    """

    default_priority = 220
    """The Substitutions transform has to be applied very early, before
    `docutils.transforms.frontmatter.DocTitle` and others."""

    def apply(self):
        defs = self.document.substitution_defs
        normed = self.document.substitution_names
        nested = {}
        line_length_limit = getattr(self.document.settings,
                                    "line_length_limit", 10000)

        subreflist = list(self.document.findall(nodes.substitution_reference))
        for ref in subreflist:
            msg = ''
            refname = ref['refname']
            if refname in defs:
                key = refname
            else:
                normed_name = refname.lower()
                key = normed.get(normed_name, None)
            if key is None:
                msg = self.document.reporter.error(
                      'Undefined substitution referenced: "%s".'
                      % refname, base_node=ref)
            else:
                subdef = defs[key]
                if len(subdef.astext()) > line_length_limit:
                    msg = self.document.reporter.error(
                            'Substitution definition "%s" exceeds the'
                            ' line-length-limit.' % key)
            if msg:
                msgid = self.document.set_id(msg)
                prb = nodes.problematic(
                      ref.rawsource, ref.rawsource, refid=msgid)
                prbid = self.document.set_id(prb)
                msg.add_backref(prbid)
                ref.replace_self(prb)
                continue

            parent = ref.parent
            index = parent.index(ref)
            if ('ltrim' in subdef.attributes
                or 'trim' in subdef.attributes):
                if index > 0 and isinstance(parent[index - 1],
                                            nodes.Text):
                    parent[index - 1] = parent[index - 1].rstrip()
            if ('rtrim' in subdef.attributes
                or 'trim' in subdef.attributes):
                if (len(parent) > index + 1
                    and isinstance(parent[index + 1], nodes.Text)):
                    parent[index + 1] = parent[index + 1].lstrip()
            subdef_copy = subdef.deepcopy()
            try:
                # Take care of nested substitution references:
                for nested_ref in subdef_copy.findall(
                        nodes.substitution_reference):
                    nested_name = normed[nested_ref['refname'].lower()]
                    if nested_name in nested.setdefault(nested_name, []):
                        raise CircularSubstitutionDefinitionError
                    nested[nested_name].append(key)
                    nested_ref['ref-origin'] = ref
                    subreflist.append(nested_ref)
            except CircularSubstitutionDefinitionError:
                parent = ref.parent
                if isinstance(parent, nodes.substitution_definition):
                    msg = self.document.reporter.error(
                        'Circular substitution definition detected:',
                        nodes.literal_block(parent.rawsource,
                                            parent.rawsource),
                        line=parent.line, base_node=parent)
                    parent.replace_self(msg)
                else:
                    # find original ref substitution which caused this error
                    ref_origin = ref
                    while ref_origin.hasattr('ref-origin'):
                        ref_origin = ref_origin['ref-origin']
                    msg = self.document.reporter.error(
                        'Circular substitution definition referenced: '
                        '"%s".' % refname, base_node=ref_origin)
                    msgid = self.document.set_id(msg)
                    prb = nodes.problematic(
                        ref.rawsource, ref.rawsource, refid=msgid)
                    prbid = self.document.set_id(prb)
                    msg.add_backref(prbid)
                    ref.replace_self(prb)
                continue
            ref.replace_self(subdef_copy.children)
            # register refname of the replacement node(s)
            # (needed for resolution of references)
            for node in subdef_copy.children:
                if isinstance(node, nodes.Referential):
                    # HACK: verify refname attribute exists.
                    # Test with docs/dev/todo.rst, see. |donate|
                    if 'refname' in node:
                        self.document.note_refname(node)


class TargetNotes(Transform):

    """
    Creates a footnote for each external target in the text, and corresponding
    footnote references after each reference.
    """

    default_priority = 540
    """The TargetNotes transform has to be applied after `IndirectHyperlinks`
    but before `Footnotes`."""

    def __init__(self, document, startnode) -> None:
        Transform.__init__(self, document, startnode=startnode)

        self.classes = startnode.details.get('class', [])

    def apply(self) -> None:
        notes = {}
        nodelist = []
        for target in self.document.findall(nodes.target):
            # Only external targets.
            if not target.hasattr('refuri'):
                continue
            names = target['names']
            refs = []
            for name in names:
                refs.extend(self.document.refnames.get(name, []))
            if not refs:
                continue
            footnote = self.make_target_footnote(target['refuri'], refs,
                                                 notes)
            if target['refuri'] not in notes:
                notes[target['refuri']] = footnote
                nodelist.append(footnote)
        # Take care of anonymous references.
        for ref in self.document.findall(nodes.reference):
            if not ref.get('anonymous'):
                continue
            if ref.hasattr('refuri'):
                footnote = self.make_target_footnote(ref['refuri'], [ref],
                                                     notes)
                if ref['refuri'] not in notes:
                    notes[ref['refuri']] = footnote
                    nodelist.append(footnote)
        self.startnode.replace_self(nodelist)

    def make_target_footnote(self, refuri, refs, notes):
        if refuri in notes:  # duplicate?
            footnote = notes[refuri]
            assert len(footnote['names']) == 1
            footnote_name = footnote['names'][0]
        else:                           # original
            footnote = nodes.footnote()
            footnote_id = self.document.set_id(footnote)
            # Use uppercase letters and a colon; they can't be
            # produced inside names by the parser.
            footnote_name = 'TARGET_NOTE: ' + footnote_id
            footnote['auto'] = 1
            footnote['names'] = [footnote_name]
            footnote_paragraph = nodes.paragraph()
            footnote_paragraph += nodes.reference('', refuri, refuri=refuri)
            footnote += footnote_paragraph
            self.document.note_autofootnote(footnote)
            self.document.note_explicit_target(footnote, footnote)
        for ref in refs:
            if isinstance(ref, nodes.target):
                continue
            refnode = nodes.footnote_reference(refname=footnote_name, auto=1)
            refnode['classes'] += self.classes
            self.document.note_autofootnote_ref(refnode)
            self.document.note_footnote_ref(refnode)
            index = ref.parent.index(ref) + 1
            reflist = [refnode]
            if not utils.get_trim_footnote_ref_space(self.document.settings):
                if self.classes:
                    reflist.insert(
                        0, nodes.inline(text=' ', Classes=self.classes))
                else:
                    reflist.insert(0, nodes.Text(' '))
            ref.parent.insert(index, reflist)
        return footnote


class CitationReferences(Transform):
    """Resolve <citation_references>.

    The 'use_bibtex'__ configuration setting indicates that citation entries
    are fetched from a BibTeX database by the backend (LaTeX).

    __ https://docutils.sourceforge.io/docs/user/config.html#use-bibtex
    """
    # TODO: Bibliography database support for other output formats.

    default_priority = 770
    # Apply between `InternalTargets` (660) and `DanglingReferences` (850)

    def apply(self) -> None:
        if not getattr(self.document.settings, 'use_bibtex', []):
            return
        for node in self.document.findall(nodes.citation_reference):
            # Skip nodes that are resolved or have a matching target
            # and will be resolved by `DanglingReferences`:
            # TODO: drop the second condition when `DanglingReferences` is
            #       replaced by two separate transitions.
            if node.resolved or self.document.nameids.get(node.get('refname')):
                continue
            if node.astext():  # ensure text content (becomes the BibTeX key)
                node.delattr('refname')
                node.resolved = True


class DanglingReferences(Transform):

    """
    Check for dangling references (incl. footnote & citation) and for
    unreferenced targets.

    Provisional : pending deprecation
      Docutils readers will add separate transforms for resolving
      refnames to refids and for reporting unresolved references
      instead of this transform (to make space for reference-resolving
      transforms added by extensions or applications) in Docutils 1.0.
      This transform will be removed in Docutils 2.0.
    """

    default_priority = 850

    def apply(self) -> None:
        visitor = DanglingReferencesVisitor(
            self.document,
            self.document.transformer.unknown_reference_resolvers)
        self.document.walk(visitor)
        # *After* resolving all references, check for unreferenced
        # targets:
        for target in self.document.findall(nodes.target):
            if not target.referenced:
                if target.get('anonymous'):
                    # If we have unreferenced anonymous targets, there
                    # is already an error message about anonymous
                    # hyperlink mismatch; no need to generate another
                    # message.
                    continue
                if target['names']:
                    naming = target['names'][0]
                elif target['ids']:
                    naming = target['ids'][0]
                else:
                    # Hack: Propagated targets always have their refid
                    # attribute set.
                    naming = target['refid']
                self.document.reporter.info(
                    'Hyperlink target "%s" is not referenced.'
                    % naming, base_node=target)


class DanglingReferencesVisitor(nodes.SparseNodeVisitor):
    """Provisional : pending deprecation

    This auxiliary class is used by the `DanglingReferences` transform
    which will no longer be used in Docutils 1.0.
    It will be removed in Docutils 2.0.
    """

    def __init__(self, document, unknown_reference_resolvers) -> None:
        nodes.SparseNodeVisitor.__init__(self, document)
        self.document = document
        self.unknown_reference_resolvers = unknown_reference_resolvers

    def unknown_visit(self, node) -> None:
        pass

    def visit_reference(self, node) -> None:
        if node.resolved or not node.hasattr('refname'):
            return
        refname = node['refname']
        id = self.document.nameids.get(refname, '')
        if id:
            # target found, set refid
            del node['refname']
            node['refid'] = id
            self.document.ids[id].note_referenced_by(id=id)
            node.resolved = True
            return
        # Apply component-specific resolving functions (cf. TransformSpec):
        # (will be removed in Docutils 1.0)
        for resolver_function in self.unknown_reference_resolvers:
            if resolver_function(node):
                return
        # Report unresolved references:
        if id is None:
            msg = self.document.reporter.error(
                'Duplicate target name, cannot be used as a unique '
                f'reference: "{refname}".', base_node=node)
        else:
            if '<' in refname or '>' in refname:
                hint = 'Did you want to embed a URI or alias?'
                if '<' not in refname:
                    hint += '\nOpening bracket missing.'
                elif ' <' not in refname:
                    hint += ('\nThe embedded reference must be preceded'
                             ' by whitespace.')
                if '>' not in refname:
                    hint += '\nClosing bracket missing.'
                elif not refname.endswith('>'):
                    hint += ('\nThe embedded reference must be the last text'
                             ' before the end string.')
                if '< ' in refname or ' >' in refname:
                    hint += ('\nWhitespace around the embedded reference'
                             ' is not allowed.')
                details = [nodes.paragraph('', hint)]
            else:
                details = []
            msg = self.document.reporter.error(
                      f'Unknown target name: "{refname}".',
                      *details, base_node=node)
        msgid = self.document.set_id(msg)
        prb = nodes.problematic(node.rawsource, node.rawsource, refid=msgid)
        try:
            prbid = node['ids'][0]
        except IndexError:
            prbid = self.document.set_id(prb)
        msg.add_backref(prbid)
        node.replace_self(prb)

    visit_footnote_reference = visit_citation_reference = visit_reference