figure.py

"""
`matplotlib.figure` implements the following classes:

`Figure`
    Top level `~matplotlib.artist.Artist`, which holds all plot elements.
    Many methods are implemented in `FigureBase`.

`SubFigure`
    A logical figure inside a figure, usually added to a figure (or parent
    `SubFigure`) with `Figure.add_subfigure` or `Figure.subfigures` methods
    (provisional API v3.4).

Figures are typically created using pyplot methods `~.pyplot.figure`,
`~.pyplot.subplots`, and `~.pyplot.subplot_mosaic`.

.. plot::
    :include-source:

    fig, ax = plt.subplots(figsize=(2, 2), facecolor='lightskyblue',
                           layout='constrained')
    fig.suptitle('Figure')
    ax.set_title('Axes', loc='left', fontstyle='oblique', fontsize='medium')

Some situations call for directly instantiating a `~.figure.Figure` class,
usually inside an application of some sort (see :ref:`user_interfaces` for a
list of examples) .  More information about Figures can be found at
:ref:`figure-intro`.
"""

from contextlib import ExitStack
import inspect
import itertools
import logging
from numbers import Integral
import threading

import numpy as np

import matplotlib as mpl
from matplotlib import _blocking_input, backend_bases, _docstring, projections
from matplotlib.artist import (
    Artist, allow_rasterization, _finalize_rasterization)
from matplotlib.backend_bases import (
    DrawEvent, FigureCanvasBase, NonGuiException, MouseButton, _get_renderer)
import matplotlib._api as _api
import matplotlib.cbook as cbook
import matplotlib.colorbar as cbar
import matplotlib.image as mimage

from matplotlib.axes import Axes
from matplotlib.gridspec import GridSpec, SubplotParams
from matplotlib.layout_engine import (
    ConstrainedLayoutEngine, TightLayoutEngine, LayoutEngine,
    PlaceHolderLayoutEngine
)
import matplotlib.legend as mlegend
from matplotlib.patches import Rectangle
from matplotlib.text import Text
from matplotlib.transforms import (Affine2D, Bbox, BboxTransformTo,
                                   TransformedBbox)

_log = logging.getLogger(__name__)


def _stale_figure_callback(self, val):
    if self.figure:
        self.figure.stale = val


class _AxesStack:
    """
    Helper class to track Axes in a figure.

    Axes are tracked both in the order in which they have been added
    (``self._axes`` insertion/iteration order) and in the separate "gca" stack
    (which is the index to which they map in the ``self._axes`` dict).
    """

    def __init__(self):
        self._axes = {}  # Mapping of Axes to "gca" order.
        self._counter = itertools.count()

    def as_list(self):
        """List the Axes that have been added to the figure."""
        return [*self._axes]  # This relies on dict preserving order.

    def remove(self, a):
        """Remove the Axes from the stack."""
        self._axes.pop(a)

    def bubble(self, a):
        """Move an Axes, which must already exist in the stack, to the top."""
        if a not in self._axes:
            raise ValueError("Axes has not been added yet")
        self._axes[a] = next(self._counter)

    def add(self, a):
        """Add an Axes to the stack, ignoring it if already present."""
        if a not in self._axes:
            self._axes[a] = next(self._counter)

    def current(self):
        """Return the active Axes, or None if the stack is empty."""
        return max(self._axes, key=self._axes.__getitem__, default=None)

    def __getstate__(self):
        return {
            **vars(self),
            "_counter": max(self._axes.values(), default=0)
        }

    def __setstate__(self, state):
        next_counter = state.pop('_counter')
        vars(self).update(state)
        self._counter = itertools.count(next_counter)


class FigureBase(Artist):
    """
    Base class for `.Figure` and `.SubFigure` containing the methods that add
    artists to the figure or subfigure, create Axes, etc.
    """
    def __init__(self, **kwargs):
        super().__init__()
        # remove the non-figure artist _axes property
        # as it makes no sense for a figure to be _in_ an Axes
        # this is used by the property methods in the artist base class
        # which are over-ridden in this class
        del self._axes

        self._suptitle = None
        self._supxlabel = None
        self._supylabel = None

        # groupers to keep track of x, y labels and title we want to align.
        # see self.align_xlabels, self.align_ylabels,
        # self.align_titles, and axis._get_tick_boxes_siblings
        self._align_label_groups = {
            "x": cbook.Grouper(),
            "y": cbook.Grouper(),
            "title": cbook.Grouper()
        }

        self._localaxes = []  # track all Axes
        self.artists = []
        self.lines = []
        self.patches = []
        self.texts = []
        self.images = []
        self.legends = []
        self.subfigs = []
        self.stale = True
        self.suppressComposite = None
        self.set(**kwargs)

    def _get_draw_artists(self, renderer):
        """Also runs apply_aspect"""
        artists = self.get_children()

        artists.remove(self.patch)
        artists = sorted(
            (artist for artist in artists if not artist.get_animated()),
            key=lambda artist: artist.get_zorder())
        for ax in self._localaxes:
            locator = ax.get_axes_locator()
            ax.apply_aspect(locator(ax, renderer) if locator else None)

            for child in ax.get_children():
                if hasattr(child, 'apply_aspect'):
                    locator = child.get_axes_locator()
                    child.apply_aspect(
                        locator(child, renderer) if locator else None)
        return artists

    def autofmt_xdate(
            self, bottom=0.2, rotation=30, ha='right', which='major'):
        """
        Date ticklabels often overlap, so it is useful to rotate them
        and right align them.  Also, a common use case is a number of
        subplots with shared x-axis where the x-axis is date data.  The
        ticklabels are often long, and it helps to rotate them on the
        bottom subplot and turn them off on other subplots, as well as
        turn off xlabels.

        Parameters
        ----------
        bottom : float, default: 0.2
            The bottom of the subplots for `subplots_adjust`.
        rotation : float, default: 30 degrees
            The rotation angle of the xtick labels in degrees.
        ha : {'left', 'center', 'right'}, default: 'right'
            The horizontal alignment of the xticklabels.
        which : {'major', 'minor', 'both'}, default: 'major'
            Selects which ticklabels to rotate.
        """
        _api.check_in_list(['major', 'minor', 'both'], which=which)
        allsubplots = all(ax.get_subplotspec() for ax in self.axes)
        if len(self.axes) == 1:
            for label in self.axes[0].get_xticklabels(which=which):
                label.set_ha(ha)
                label.set_rotation(rotation)
        else:
            if allsubplots:
                for ax in self.get_axes():
                    if ax.get_subplotspec().is_last_row():
                        for label in ax.get_xticklabels(which=which):
                            label.set_ha(ha)
                            label.set_rotation(rotation)
                    else:
                        for label in ax.get_xticklabels(which=which):
                            label.set_visible(False)
                        ax.set_xlabel('')

        if allsubplots:
            self.subplots_adjust(bottom=bottom)
        self.stale = True

    def get_children(self):
        """Get a list of artists contained in the figure."""
        return [self.patch,
                *self.artists,
                *self._localaxes,
                *self.lines,
                *self.patches,
                *self.texts,
                *self.images,
                *self.legends,
                *self.subfigs]

    def contains(self, mouseevent):
        """
        Test whether the mouse event occurred on the figure.

        Returns
        -------
            bool, {}
        """
        if self._different_canvas(mouseevent):
            return False, {}
        inside = self.bbox.contains(mouseevent.x, mouseevent.y)
        return inside, {}

    def get_window_extent(self, renderer=None):
        # docstring inherited
        return self.bbox

    def _suplabels(self, t, info, **kwargs):
        """
        Add a centered %(name)s to the figure.

        Parameters
        ----------
        t : str
            The %(name)s text.
        x : float, default: %(x0)s
            The x location of the text in figure coordinates.
        y : float, default: %(y0)s
            The y location of the text in figure coordinates.
        horizontalalignment, ha : {'center', 'left', 'right'}, default: %(ha)s
            The horizontal alignment of the text relative to (*x*, *y*).
        verticalalignment, va : {'top', 'center', 'bottom', 'baseline'}, \
default: %(va)s
            The vertical alignment of the text relative to (*x*, *y*).
        fontsize, size : default: :rc:`figure.%(rc)ssize`
            The font size of the text. See `.Text.set_size` for possible
            values.
        fontweight, weight : default: :rc:`figure.%(rc)sweight`
            The font weight of the text. See `.Text.set_weight` for possible
            values.

        Returns
        -------
        text
            The `.Text` instance of the %(name)s.

        Other Parameters
        ----------------
        fontproperties : None or dict, optional
            A dict of font properties. If *fontproperties* is given the
            default values for font size and weight are taken from the
            `.FontProperties` defaults. :rc:`figure.%(rc)ssize` and
            :rc:`figure.%(rc)sweight` are ignored in this case.

        **kwargs
            Additional kwargs are `matplotlib.text.Text` properties.
        """

        x = kwargs.pop('x', None)
        y = kwargs.pop('y', None)
        if info['name'] in ['_supxlabel', '_suptitle']:
            autopos = y is None
        elif info['name'] == '_supylabel':
            autopos = x is None
        if x is None:
            x = info['x0']
        if y is None:
            y = info['y0']

        kwargs = cbook.normalize_kwargs(kwargs, Text)
        kwargs.setdefault('horizontalalignment', info['ha'])
        kwargs.setdefault('verticalalignment', info['va'])
        kwargs.setdefault('rotation', info['rotation'])

        if 'fontproperties' not in kwargs:
            kwargs.setdefault('fontsize', mpl.rcParams[info['size']])
            kwargs.setdefault('fontweight', mpl.rcParams[info['weight']])

        suplab = getattr(self, info['name'])
        if suplab is not None:
            suplab.set_text(t)
            suplab.set_position((x, y))
            suplab.set(**kwargs)
        else:
            suplab = self.text(x, y, t, **kwargs)
            setattr(self, info['name'], suplab)
        suplab._autopos = autopos
        self.stale = True
        return suplab

    @_docstring.Substitution(x0=0.5, y0=0.98, name='suptitle', ha='center',
                             va='top', rc='title')
    @_docstring.copy(_suplabels)
    def suptitle(self, t, **kwargs):
        # docstring from _suplabels...
        info = {'name': '_suptitle', 'x0': 0.5, 'y0': 0.98,
                'ha': 'center', 'va': 'top', 'rotation': 0,
                'size': 'figure.titlesize', 'weight': 'figure.titleweight'}
        return self._suplabels(t, info, **kwargs)

    def get_suptitle(self):
        """Return the suptitle as string or an empty string if not set."""
        text_obj = self._suptitle
        return "" if text_obj is None else text_obj.get_text()

    @_docstring.Substitution(x0=0.5, y0=0.01, name='supxlabel', ha='center',
                             va='bottom', rc='label')
    @_docstring.copy(_suplabels)
    def supxlabel(self, t, **kwargs):
        # docstring from _suplabels...
        info = {'name': '_supxlabel', 'x0': 0.5, 'y0': 0.01,
                'ha': 'center', 'va': 'bottom', 'rotation': 0,
                'size': 'figure.labelsize', 'weight': 'figure.labelweight'}
        return self._suplabels(t, info, **kwargs)

    def get_supxlabel(self):
        """Return the supxlabel as string or an empty string if not set."""
        text_obj = self._supxlabel
        return "" if text_obj is None else text_obj.get_text()

    @_docstring.Substitution(x0=0.02, y0=0.5, name='supylabel', ha='left',
                             va='center', rc='label')
    @_docstring.copy(_suplabels)
    def supylabel(self, t, **kwargs):
        # docstring from _suplabels...
        info = {'name': '_supylabel', 'x0': 0.02, 'y0': 0.5,
                'ha': 'left', 'va': 'center', 'rotation': 'vertical',
                'rotation_mode': 'anchor', 'size': 'figure.labelsize',
                'weight': 'figure.labelweight'}
        return self._suplabels(t, info, **kwargs)

    def get_supylabel(self):
        """Return the supylabel as string or an empty string if not set."""
        text_obj = self._supylabel
        return "" if text_obj is None else text_obj.get_text()

    def get_edgecolor(self):
        """Get the edge color of the Figure rectangle."""
        return self.patch.get_edgecolor()

    def get_facecolor(self):
        """Get the face color of the Figure rectangle."""
        return self.patch.get_facecolor()

    def get_frameon(self):
        """
        Return the figure's background patch visibility, i.e.
        whether the figure background will be drawn. Equivalent to
        ``Figure.patch.get_visible()``.
        """
        return self.patch.get_visible()

    def set_linewidth(self, linewidth):
        """
        Set the line width of the Figure rectangle.

        Parameters
        ----------
        linewidth : number
        """
        self.patch.set_linewidth(linewidth)

    def get_linewidth(self):
        """
        Get the line width of the Figure rectangle.
        """
        return self.patch.get_linewidth()

    def set_edgecolor(self, color):
        """
        Set the edge color of the Figure rectangle.

        Parameters
        ----------
        color : :mpltype:`color`
        """
        self.patch.set_edgecolor(color)

    def set_facecolor(self, color):
        """
        Set the face color of the Figure rectangle.

        Parameters
        ----------
        color : :mpltype:`color`
        """
        self.patch.set_facecolor(color)

    def set_frameon(self, b):
        """
        Set the figure's background patch visibility, i.e.
        whether the figure background will be drawn. Equivalent to
        ``Figure.patch.set_visible()``.

        Parameters
        ----------
        b : bool
        """
        self.patch.set_visible(b)
        self.stale = True

    frameon = property(get_frameon, set_frameon)

    def add_artist(self, artist, clip=False):
        """
        Add an `.Artist` to the figure.

        Usually artists are added to `~.axes.Axes` objects using
        `.Axes.add_artist`; this method can be used in the rare cases where
        one needs to add artists directly to the figure instead.

        Parameters
        ----------
        artist : `~matplotlib.artist.Artist`
            The artist to add to the figure. If the added artist has no
            transform previously set, its transform will be set to
            ``figure.transSubfigure``.
        clip : bool, default: False
            Whether the added artist should be clipped by the figure patch.

        Returns
        -------
        `~matplotlib.artist.Artist`
            The added artist.
        """
        artist.set_figure(self)
        self.artists.append(artist)
        artist._remove_method = self.artists.remove

        if not artist.is_transform_set():
            artist.set_transform(self.transSubfigure)

        if clip and artist.get_clip_path() is None:
            artist.set_clip_path(self.patch)

        self.stale = True
        return artist

    @_docstring.dedent_interpd
    def add_axes(self, *args, **kwargs):
        """
        Add an `~.axes.Axes` to the figure.

        Call signatures::

            add_axes(rect, projection=None, polar=False, **kwargs)
            add_axes(ax)

        Parameters
        ----------
        rect : tuple (left, bottom, width, height)
            The dimensions (left, bottom, width, height) of the new
            `~.axes.Axes`. All quantities are in fractions of figure width and
            height.

        projection : {None, 'aitoff', 'hammer', 'lambert', 'mollweide', \
'polar', 'rectilinear', str}, optional
            The projection type of the `~.axes.Axes`. *str* is the name of
            a custom projection, see `~matplotlib.projections`. The default
            None results in a 'rectilinear' projection.

        polar : bool, default: False
            If True, equivalent to projection='polar'.

        axes_class : subclass type of `~.axes.Axes`, optional
            The `.axes.Axes` subclass that is instantiated.  This parameter
            is incompatible with *projection* and *polar*.  See
            :ref:`axisartist_users-guide-index` for examples.

        sharex, sharey : `~matplotlib.axes.Axes`, optional
            Share the x or y `~matplotlib.axis` with sharex and/or sharey.
            The axis will have the same limits, ticks, and scale as the axis
            of the shared Axes.

        label : str
            A label for the returned Axes.

        Returns
        -------
        `~.axes.Axes`, or a subclass of `~.axes.Axes`
            The returned Axes class depends on the projection used. It is
            `~.axes.Axes` if rectilinear projection is used and
            `.projections.polar.PolarAxes` if polar projection is used.

        Other Parameters
        ----------------
        **kwargs
            This method also takes the keyword arguments for
            the returned Axes class. The keyword arguments for the
            rectilinear Axes class `~.axes.Axes` can be found in
            the following table but there might also be other keyword
            arguments if another projection is used, see the actual Axes
            class.

            %(Axes:kwdoc)s

        Notes
        -----
        In rare circumstances, `.add_axes` may be called with a single
        argument, an Axes instance already created in the present figure but
        not in the figure's list of Axes.

        See Also
        --------
        .Figure.add_subplot
        .pyplot.subplot
        .pyplot.axes
        .Figure.subplots
        .pyplot.subplots

        Examples
        --------
        Some simple examples::

            rect = l, b, w, h
            fig = plt.figure()
            fig.add_axes(rect)
            fig.add_axes(rect, frameon=False, facecolor='g')
            fig.add_axes(rect, polar=True)
            ax = fig.add_axes(rect, projection='polar')
            fig.delaxes(ax)
            fig.add_axes(ax)
        """

        if not len(args) and 'rect' not in kwargs:
            raise TypeError(
                "add_axes() missing 1 required positional argument: 'rect'")
        elif 'rect' in kwargs:
            if len(args):
                raise TypeError(
                    "add_axes() got multiple values for argument 'rect'")
            args = (kwargs.pop('rect'), )

        if isinstance(args[0], Axes):
            a, *extra_args = args
            key = a._projection_init
            if a.get_figure() is not self:
                raise ValueError(
                    "The Axes must have been created in the present figure")
        else:
            rect, *extra_args = args
            if not np.isfinite(rect).all():
                raise ValueError(f'all entries in rect must be finite not {rect}')
            projection_class, pkw = self._process_projection_requirements(**kwargs)

            # create the new Axes using the Axes class given
            a = projection_class(self, rect, **pkw)
            key = (projection_class, pkw)

        if extra_args:
            _api.warn_deprecated(
                "3.8",
                name="Passing more than one positional argument to Figure.add_axes",
                addendum="Any additional positional arguments are currently ignored.")
        return self._add_axes_internal(a, key)

    @_docstring.dedent_interpd
    def add_subplot(self, *args, **kwargs):
        """
        Add an `~.axes.Axes` to the figure as part of a subplot arrangement.

        Call signatures::

           add_subplot(nrows, ncols, index, **kwargs)
           add_subplot(pos, **kwargs)
           add_subplot(ax)
           add_subplot()

        Parameters
        ----------
        *args : int, (int, int, *index*), or `.SubplotSpec`, default: (1, 1, 1)
            The position of the subplot described by one of

            - Three integers (*nrows*, *ncols*, *index*). The subplot will
              take the *index* position on a grid with *nrows* rows and
              *ncols* columns. *index* starts at 1 in the upper left corner
              and increases to the right.  *index* can also be a two-tuple
              specifying the (*first*, *last*) indices (1-based, and including
              *last*) of the subplot, e.g., ``fig.add_subplot(3, 1, (1, 2))``
              makes a subplot that spans the upper 2/3 of the figure.
            - A 3-digit integer. The digits are interpreted as if given
              separately as three single-digit integers, i.e.
              ``fig.add_subplot(235)`` is the same as
              ``fig.add_subplot(2, 3, 5)``. Note that this can only be used
              if there are no more than 9 subplots.
            - A `.SubplotSpec`.

            In rare circumstances, `.add_subplot` may be called with a single
            argument, a subplot Axes instance already created in the
            present figure but not in the figure's list of Axes.

        projection : {None, 'aitoff', 'hammer', 'lambert', 'mollweide', \
'polar', 'rectilinear', str}, optional
            The projection type of the subplot (`~.axes.Axes`). *str* is the
            name of a custom projection, see `~matplotlib.projections`. The
            default None results in a 'rectilinear' projection.

        polar : bool, default: False
            If True, equivalent to projection='polar'.

        axes_class : subclass type of `~.axes.Axes`, optional
            The `.axes.Axes` subclass that is instantiated.  This parameter
            is incompatible with *projection* and *polar*.  See
            :ref:`axisartist_users-guide-index` for examples.

        sharex, sharey : `~matplotlib.axes.Axes`, optional
            Share the x or y `~matplotlib.axis` with sharex and/or sharey.
            The axis will have the same limits, ticks, and scale as the axis
            of the shared Axes.

        label : str
            A label for the returned Axes.

        Returns
        -------
        `~.axes.Axes`

            The Axes of the subplot. The returned Axes can actually be an
            instance of a subclass, such as `.projections.polar.PolarAxes` for
            polar projections.

        Other Parameters
        ----------------
        **kwargs
            This method also takes the keyword arguments for the returned Axes
            base class; except for the *figure* argument. The keyword arguments
            for the rectilinear base class `~.axes.Axes` can be found in
            the following table but there might also be other keyword
            arguments if another projection is used.

            %(Axes:kwdoc)s

        See Also
        --------
        .Figure.add_axes
        .pyplot.subplot
        .pyplot.axes
        .Figure.subplots
        .pyplot.subplots

        Examples
        --------
        ::

            fig = plt.figure()

            fig.add_subplot(231)
            ax1 = fig.add_subplot(2, 3, 1)  # equivalent but more general

            fig.add_subplot(232, frameon=False)  # subplot with no frame
            fig.add_subplot(233, projection='polar')  # polar subplot
            fig.add_subplot(234, sharex=ax1)  # subplot sharing x-axis with ax1
            fig.add_subplot(235, facecolor="red")  # red subplot

            ax1.remove()  # delete ax1 from the figure
            fig.add_subplot(ax1)  # add ax1 back to the figure
        """
        if 'figure' in kwargs:
            # Axes itself allows for a 'figure' kwarg, but since we want to
            # bind the created Axes to self, it is not allowed here.
            raise _api.kwarg_error("add_subplot", "figure")

        if (len(args) == 1
                and isinstance(args[0], mpl.axes._base._AxesBase)
                and args[0].get_subplotspec()):
            ax = args[0]
            key = ax._projection_init
            if ax.get_figure() is not self:
                raise ValueError("The Axes must have been created in "
                                 "the present figure")
        else:
            if not args:
                args = (1, 1, 1)
            # Normalize correct ijk values to (i, j, k) here so that
            # add_subplot(211) == add_subplot(2, 1, 1).  Invalid values will
            # trigger errors later (via SubplotSpec._from_subplot_args).
            if (len(args) == 1 and isinstance(args[0], Integral)
                    and 100 <= args[0] <= 999):
                args = tuple(map(int, str(args[0])))
            projection_class, pkw = self._process_projection_requirements(**kwargs)
            ax = projection_class(self, *args, **pkw)
            key = (projection_class, pkw)
        return self._add_axes_internal(ax, key)

    def _add_axes_internal(self, ax, key):
        """Private helper for `add_axes` and `add_subplot`."""
        self._axstack.add(ax)
        if ax not in self._localaxes:
            self._localaxes.append(ax)
        self.sca(ax)
        ax._remove_method = self.delaxes
        # this is to support plt.subplot's re-selection logic
        ax._projection_init = key
        self.stale = True
        ax.stale_callback = _stale_figure_callback
        return ax

    def subplots(self, nrows=1, ncols=1, *, sharex=False, sharey=False,
                 squeeze=True, width_ratios=None, height_ratios=None,
                 subplot_kw=None, gridspec_kw=None):
        """
        Add a set of subplots to this figure.

        This utility wrapper makes it convenient to create common layouts of
        subplots in a single call.

        Parameters
        ----------
        nrows, ncols : int, default: 1
            Number of rows/columns of the subplot grid.

        sharex, sharey : bool or {'none', 'all', 'row', 'col'}, default: False
            Controls sharing of x-axis (*sharex*) or y-axis (*sharey*):

            - True or 'all': x- or y-axis will be shared among all subplots.
            - False or 'none': each subplot x- or y-axis will be independent.
            - 'row': each subplot row will share an x- or y-axis.
            - 'col': each subplot column will share an x- or y-axis.

            When subplots have a shared x-axis along a column, only the x tick
            labels of the bottom subplot are created. Similarly, when subplots
            have a shared y-axis along a row, only the y tick labels of the
            first column subplot are created. To later turn other subplots'
            ticklabels on, use `~matplotlib.axes.Axes.tick_params`.

            When subplots have a shared axis that has units, calling
            `.Axis.set_units` will update each axis with the new units.

        squeeze : bool, default: True
            - If True, extra dimensions are squeezed out from the returned
              array of Axes:

              - if only one subplot is constructed (nrows=ncols=1), the
                resulting single Axes object is returned as a scalar.
              - for Nx1 or 1xM subplots, the returned object is a 1D numpy
                object array of Axes objects.
              - for NxM, subplots with N>1 and M>1 are returned as a 2D array.

            - If False, no squeezing at all is done: the returned Axes object
              is always a 2D array containing Axes instances, even if it ends
              up being 1x1.

        width_ratios : array-like of length *ncols*, optional
            Defines the relative widths of the columns. Each column gets a
            relative width of ``width_ratios[i] / sum(width_ratios)``.
            If not given, all columns will have the same width.  Equivalent
            to ``gridspec_kw={'width_ratios': [...]}``.

        height_ratios : array-like of length *nrows*, optional
            Defines the relative heights of the rows. Each row gets a
            relative height of ``height_ratios[i] / sum(height_ratios)``.
            If not given, all rows will have the same height. Equivalent
            to ``gridspec_kw={'height_ratios': [...]}``.

        subplot_kw : dict, optional
            Dict with keywords passed to the `.Figure.add_subplot` call used to
            create each subplot.

        gridspec_kw : dict, optional
            Dict with keywords passed to the
            `~matplotlib.gridspec.GridSpec` constructor used to create
            the grid the subplots are placed on.

        Returns
        -------
        `~.axes.Axes` or array of Axes
            Either a single `~matplotlib.axes.Axes` object or an array of Axes
            objects if more than one subplot was created. The dimensions of the
            resulting array can be controlled with the *squeeze* keyword, see
            above.

        See Also
        --------
        .pyplot.subplots
        .Figure.add_subplot
        .pyplot.subplot

        Examples
        --------
        ::

            # First create some toy data:
            x = np.linspace(0, 2*np.pi, 400)
            y = np.sin(x**2)

            # Create a figure
            fig = plt.figure()

            # Create a subplot
            ax = fig.subplots()
            ax.plot(x, y)
            ax.set_title('Simple plot')

            # Create two subplots and unpack the output array immediately
            ax1, ax2 = fig.subplots(1, 2, sharey=True)
            ax1.plot(x, y)
            ax1.set_title('Sharing Y axis')
            ax2.scatter(x, y)

            # Create four polar Axes and access them through the returned array
            axes = fig.subplots(2, 2, subplot_kw=dict(projection='polar'))
            axes[0, 0].plot(x, y)
            axes[1, 1].scatter(x, y)

            # Share an X-axis with each column of subplots
            fig.subplots(2, 2, sharex='col')

            # Share a Y-axis with each row of subplots
            fig.subplots(2, 2, sharey='row')

            # Share both X- and Y-axes with all subplots
            fig.subplots(2, 2, sharex='all', sharey='all')

            # Note that this is the same as
            fig.subplots(2, 2, sharex=True, sharey=True)
        """
        gridspec_kw = dict(gridspec_kw or {})
        if height_ratios is not None:
            if 'height_ratios' in gridspec_kw:
                raise ValueError("'height_ratios' must not be defined both as "
                                 "parameter and as key in 'gridspec_kw'")
            gridspec_kw['height_ratios'] = height_ratios
        if width_ratios is not None:
            if 'width_ratios' in gridspec_kw:
                raise ValueError("'width_ratios' must not be defined both as "
                                 "parameter and as key in 'gridspec_kw'")
            gridspec_kw['width_ratios'] = width_ratios

        gs = self.add_gridspec(nrows, ncols, figure=self, **gridspec_kw)
        axs = gs.subplots(sharex=sharex, sharey=sharey, squeeze=squeeze,
                          subplot_kw=subplot_kw)
        return axs

    def delaxes(self, ax):
        """
        Remove the `~.axes.Axes` *ax* from the figure; update the current Axes.
        """
        self._remove_axes(ax, owners=[self._axstack, self._localaxes])

    def _remove_axes(self, ax, owners):
        """
        Common helper for removal of standard Axes (via delaxes) and of child Axes.

        Parameters
        ----------
        ax : `~.AxesBase`
            The Axes to remove.
        owners
            List of objects (list or _AxesStack) "owning" the Axes, from which the Axes
            will be remove()d.
        """
        for owner in owners:
            owner.remove(ax)

        self._axobservers.process("_axes_change_event", self)
        self.stale = True
        self.canvas.release_mouse(ax)

        for name in ax._axis_names:  # Break link between any shared Axes
            grouper = ax._shared_axes[name]
            siblings = [other for other in grouper.get_siblings(ax) if other is not ax]
            if not siblings:  # Axes was not shared along this axis; we're done.
                continue
            grouper.remove(ax)
            # Formatters and locators may previously have been associated with the now
            # removed axis.  Update them to point to an axis still there (we can pick
            # any of them, and use the first sibling).
            remaining_axis = siblings[0]._axis_map[name]
            remaining_axis.get_major_formatter().set_axis(remaining_axis)
            remaining_axis.get_major_locator().set_axis(remaining_axis)
            remaining_axis.get_minor_formatter().set_axis(remaining_axis)
            remaining_axis.get_minor_locator().set_axis(remaining_axis)

        ax._twinned_axes.remove(ax)  # Break link between any twinned Axes.

    def clear(self, keep_observers=False):
        """
        Clear the figure.

        Parameters
        ----------
        keep_observers : bool, default: False
            Set *keep_observers* to True if, for example,
            a gui widget is tracking the Axes in the figure.
        """
        self.suppressComposite = None

        # first clear the Axes in any subfigures
        for subfig in self.subfigs:
            subfig.clear(keep_observers=keep_observers)
        self.subfigs = []

        for ax in tuple(self.axes):  # Iterate over the copy.
            ax.clear()
            self.delaxes(ax)  # Remove ax from self._axstack.

        self.artists = []
        self.lines = []
        self.patches = []
        self.texts = []
        self.images = []
        self.legends = []
        if not keep_observers:
            self._axobservers = cbook.CallbackRegistry()
        self._suptitle = None
        self._supxlabel = None
        self._supylabel = None

        self.stale = True

    # synonym for `clear`.
    def clf(self, keep_observers=False):
        """
        [*Discouraged*] Alias for the `clear()` method.

        .. admonition:: Discouraged

            The use of ``clf()`` is discouraged. Use ``clear()`` instead.

        Parameters
        ----------
        keep_observers : bool, default: False
            Set *keep_observers* to True if, for example,
            a gui widget is tracking the Axes in the figure.
        """
        return self.clear(keep_observers=keep_observers)

    # Note: the docstring below is modified with replace for the pyplot
    # version of this function because the method name differs (plt.figlegend)
    # the replacements are:
    #    " legend(" -> " figlegend(" for the signatures
    #    "fig.legend(" -> "plt.figlegend" for the code examples
    #    "ax.plot" -> "plt.plot" for consistency in using pyplot when able
    @_docstring.dedent_interpd
    def legend(self, *args, **kwargs):
        """
        Place a legend on the figure.

        Call signatures::

            legend()
            legend(handles, labels)
            legend(handles=handles)
            legend(labels)

        The call signatures correspond to the following different ways to use
        this method:

        **1. Automatic detection of elements to be shown in the legend**

        The elements to be added to the legend are automatically determined,
        when you do not pass in any extra arguments.

        In this case, the labels are taken from the artist. You can specify
        them either at artist creation or by calling the
        :meth:`~.Artist.set_label` method on the artist::

            ax.plot([1, 2, 3], label='Inline label')
            fig.legend()

        or::

            line, = ax.plot([1, 2, 3])
            line.set_label('Label via method')
            fig.legend()

        Specific lines can be excluded from the automatic legend element
        selection by defining a label starting with an underscore.
        This is default for all artists, so calling `.Figure.legend` without
        any arguments and without setting the labels manually will result in
        no legend being drawn.


        **2. Explicitly listing the artists and labels in the legend**

        For full control of which artists have a legend entry, it is possible
        to pass an iterable of legend artists followed by an iterable of
        legend labels respectively::

            fig.legend([line1, line2, line3], ['label1', 'label2', 'label3'])


        **3. Explicitly listing the artists in the legend**

        This is similar to 2, but the labels are taken from the artists'
        label properties. Example::

            line1, = ax1.plot([1, 2, 3], label='label1')
            line2, = ax2.plot([1, 2, 3], label='label2')
            fig.legend(handles=[line1, line2])


        **4. Labeling existing plot elements**

        .. admonition:: Discouraged

            This call signature is discouraged, because the relation between
            plot elements and labels is only implicit by their order and can
            easily be mixed up.

        To make a legend for all artists on all Axes, call this function with
        an iterable of strings, one for each legend item. For example::

            fig, (ax1, ax2) = plt.subplots(1, 2)
            ax1.plot([1, 3, 5], color='blue')
            ax2.plot([2, 4, 6], color='red')
            fig.legend(['the blues', 'the reds'])


        Parameters
        ----------
        handles : list of `.Artist`, optional
            A list of Artists (lines, patches) to be added to the legend.
            Use this together with *labels*, if you need full control on what
            is shown in the legend and the automatic mechanism described above
            is not sufficient.

            The length of handles and labels should be the same in this
            case. If they are not, they are truncated to the smaller length.

        labels : list of str, optional
            A list of labels to show next to the artists.
            Use this together with *handles*, if you need full control on what
            is shown in the legend and the automatic mechanism described above
            is not sufficient.

        Returns
        -------
        `~matplotlib.legend.Legend`

        Other Parameters
        ----------------
        %(_legend_kw_figure)s

        See Also
        --------
        .Axes.legend

        Notes
        -----
        Some artists are not supported by this function.  See
        :ref:`legend_guide` for details.
        """

        handles, labels, kwargs = mlegend._parse_legend_args(self.axes, *args, **kwargs)
        # explicitly set the bbox transform if the user hasn't.
        kwargs.setdefault("bbox_transform", self.transSubfigure)
        l = mlegend.Legend(self, handles, labels, **kwargs)
        self.legends.append(l)
        l._remove_method = self.legends.remove
        self.stale = True
        return l

    @_docstring.dedent_interpd
    def text(self, x, y, s, fontdict=None, **kwargs):
        """
        Add text to figure.

        Parameters
        ----------
        x, y : float
            The position to place the text. By default, this is in figure
            coordinates, floats in [0, 1]. The coordinate system can be changed
            using the *transform* keyword.

        s : str
            The text string.

        fontdict : dict, optional
            A dictionary to override the default text properties. If not given,
            the defaults are determined by :rc:`font.*`. Properties passed as
            *kwargs* override the corresponding ones given in *fontdict*.

        Returns
        -------
        `~.text.Text`

        Other Parameters
        ----------------
        **kwargs : `~matplotlib.text.Text` properties
            Other miscellaneous text parameters.

            %(Text:kwdoc)s

        See Also
        --------
        .Axes.text
        .pyplot.text
        """
        effective_kwargs = {
            'transform': self.transSubfigure,
            **(fontdict if fontdict is not None else {}),
            **kwargs,
        }
        text = Text(x=x, y=y, text=s, **effective_kwargs)
        text.set_figure(self)
        text.stale_callback = _stale_figure_callback

        self.texts.append(text)
        text._remove_method = self.texts.remove
        self.stale = True
        return text

    @_docstring.dedent_interpd
    def colorbar(
            self, mappable, cax=None, ax=None, use_gridspec=True, **kwargs):
        """
        Add a colorbar to a plot.

        Parameters
        ----------
        mappable
            The `matplotlib.cm.ScalarMappable` (i.e., `.AxesImage`,
            `.ContourSet`, etc.) described by this colorbar.  This argument is
            mandatory for the `.Figure.colorbar` method but optional for the
            `.pyplot.colorbar` function, which sets the default to the current
            image.

            Note that one can create a `.ScalarMappable` "on-the-fly" to
            generate colorbars not attached to a previously drawn artist, e.g.
            ::

                fig.colorbar(cm.ScalarMappable(norm=norm, cmap=cmap), ax=ax)

        cax : `~matplotlib.axes.Axes`, optional
            Axes into which the colorbar will be drawn.  If `None`, then a new
            Axes is created and the space for it will be stolen from the Axes(s)
            specified in *ax*.

        ax : `~matplotlib.axes.Axes` or iterable or `numpy.ndarray` of Axes, optional
            The one or more parent Axes from which space for a new colorbar Axes
            will be stolen. This parameter is only used if *cax* is not set.

            Defaults to the Axes that contains the mappable used to create the
            colorbar.

        use_gridspec : bool, optional
            If *cax* is ``None``, a new *cax* is created as an instance of
            Axes.  If *ax* is positioned with a subplotspec and *use_gridspec*
            is ``True``, then *cax* is also positioned with a subplotspec.

        Returns
        -------
        colorbar : `~matplotlib.colorbar.Colorbar`

        Other Parameters
        ----------------
        %(_make_axes_kw_doc)s
        %(_colormap_kw_doc)s

        Notes
        -----
        If *mappable* is a `~.contour.ContourSet`, its *extend* kwarg is
        included automatically.

        The *shrink* kwarg provides a simple way to scale the colorbar with
        respect to the Axes. Note that if *cax* is specified, it determines the
        size of the colorbar, and *shrink* and *aspect* are ignored.

        For more precise control, you can manually specify the positions of the
        axes objects in which the mappable and the colorbar are drawn.  In this
        case, do not use any of the Axes properties kwargs.

        It is known that some vector graphics viewers (svg and pdf) render
        white gaps between segments of the colorbar.  This is due to bugs in
        the viewers, not Matplotlib.  As a workaround, the colorbar can be
        rendered with overlapping segments::

            cbar = colorbar()
            cbar.solids.set_edgecolor("face")
            draw()

        However, this has negative consequences in other circumstances, e.g.
        with semi-transparent images (alpha < 1) and colorbar extensions;
        therefore, this workaround is not used by default (see issue #1188).

        """

        if ax is None:
            ax = getattr(mappable, "axes", None)

        if cax is None:
            if ax is None:
                raise ValueError(
                    'Unable to determine Axes to steal space for Colorbar. '
                    'Either provide the *cax* argument to use as the Axes for '
                    'the Colorbar, provide the *ax* argument to steal space '
                    'from it, or add *mappable* to an Axes.')
            fig = (  # Figure of first Axes; logic copied from make_axes.
                [*ax.flat] if isinstance(ax, np.ndarray)
                else [*ax] if np.iterable(ax)
                else [ax])[0].figure
            current_ax = fig.gca()
            if (fig.get_layout_engine() is not None and
                    not fig.get_layout_engine().colorbar_gridspec):
                use_gridspec = False
            if (use_gridspec
                    and isinstance(ax, mpl.axes._base._AxesBase)
                    and ax.get_subplotspec()):
                cax, kwargs = cbar.make_axes_gridspec(ax, **kwargs)
            else:
                cax, kwargs = cbar.make_axes(ax, **kwargs)
            # make_axes calls add_{axes,subplot} which changes gca; undo that.
            fig.sca(current_ax)
            cax.grid(visible=False, which='both', axis='both')

        if hasattr(mappable, "figure") and mappable.figure is not None:
            # Get top level artists
            mappable_host_fig = mappable.figure
            if isinstance(mappable_host_fig, mpl.figure.SubFigure):
                mappable_host_fig = mappable_host_fig.figure
            # Warn in case of mismatch
            if mappable_host_fig is not self.figure:
                _api.warn_external(
                        f'Adding colorbar to a different Figure '
                        f'{repr(mappable.figure)} than '
                        f'{repr(self.figure)} which '
                        f'fig.colorbar is called on.')

        NON_COLORBAR_KEYS = [  # remove kws that cannot be passed to Colorbar
            'fraction', 'pad', 'shrink', 'aspect', 'anchor', 'panchor']
        cb = cbar.Colorbar(cax, mappable, **{
            k: v for k, v in kwargs.items() if k not in NON_COLORBAR_KEYS})
        cax.figure.stale = True
        return cb

    def subplots_adjust(self, left=None, bottom=None, right=None, top=None,
                        wspace=None, hspace=None):
        """
        Adjust the subplot layout parameters.

        Unset parameters are left unmodified; initial values are given by
        :rc:`figure.subplot.[name]`.

        Parameters
        ----------
        left : float, optional
            The position of the left edge of the subplots,
            as a fraction of the figure width.
        right : float, optional
            The position of the right edge of the subplots,
            as a fraction of the figure width.
        bottom : float, optional
            The position of the bottom edge of the subplots,
            as a fraction of the figure height.
        top : float, optional
            The position of the top edge of the subplots,
            as a fraction of the figure height.
        wspace : float, optional
            The width of the padding between subplots,
            as a fraction of the average Axes width.
        hspace : float, optional
            The height of the padding between subplots,
            as a fraction of the average Axes height.
        """
        if (self.get_layout_engine() is not None and
                not self.get_layout_engine().adjust_compatible):
            _api.warn_external(
                "This figure was using a layout engine that is "
                "incompatible with subplots_adjust and/or tight_layout; "
                "not calling subplots_adjust.")
            return
        self.subplotpars.update(left, bottom, right, top, wspace, hspace)
        for ax in self.axes:
            if ax.get_subplotspec() is not None:
                ax._set_position(ax.get_subplotspec().get_position(self))
        self.stale = True

    def align_xlabels(self, axs=None):
        """
        Align the xlabels of subplots in the same subplot row if label
        alignment is being done automatically (i.e. the label position is
        not manually set).

        Alignment persists for draw events after this is called.

        If a label is on the bottom, it is aligned with labels on Axes that
        also have their label on the bottom and that have the same
        bottom-most subplot row.  If the label is on the top,
        it is aligned with labels on Axes with the same top-most row.

        Parameters
        ----------
        axs : list of `~matplotlib.axes.Axes`
            Optional list of (or `~numpy.ndarray`) `~matplotlib.axes.Axes`
            to align the xlabels.
            Default is to align all Axes on the figure.

        See Also
        --------
        matplotlib.figure.Figure.align_ylabels
        matplotlib.figure.Figure.align_titles
        matplotlib.figure.Figure.align_labels

        Notes
        -----
        This assumes that ``axs`` are from the same `.GridSpec`, so that
        their `.SubplotSpec` positions correspond to figure positions.

        Examples
        --------
        Example with rotated xtick labels::

            fig, axs = plt.subplots(1, 2)
            for tick in axs[0].get_xticklabels():
                tick.set_rotation(55)
            axs[0].set_xlabel('XLabel 0')
            axs[1].set_xlabel('XLabel 1')
            fig.align_xlabels()
        """
        if axs is None:
            axs = self.axes
        axs = [ax for ax in np.ravel(axs) if ax.get_subplotspec() is not None]
        for ax in axs:
            _log.debug(' Working on: %s', ax.get_xlabel())
            rowspan = ax.get_subplotspec().rowspan
            pos = ax.xaxis.get_label_position()  # top or bottom
            # Search through other Axes for label positions that are same as
            # this one and that share the appropriate row number.
            # Add to a grouper associated with each Axes of siblings.
            # This list is inspected in `axis.draw` by
            # `axis._update_label_position`.
            for axc in axs:
                if axc.xaxis.get_label_position() == pos:
                    rowspanc = axc.get_subplotspec().rowspan
                    if (pos == 'top' and rowspan.start == rowspanc.start or
                            pos == 'bottom' and rowspan.stop == rowspanc.stop):
                        # grouper for groups of xlabels to align
                        self._align_label_groups['x'].join(ax, axc)

    def align_ylabels(self, axs=None):
        """
        Align the ylabels of subplots in the same subplot column if label
        alignment is being done automatically (i.e. the label position is
        not manually set).

        Alignment persists for draw events after this is called.

        If a label is on the left, it is aligned with labels on Axes that
        also have their label on the left and that have the same
        left-most subplot column.  If the label is on the right,
        it is aligned with labels on Axes with the same right-most column.

        Parameters
        ----------
        axs : list of `~matplotlib.axes.Axes`
            Optional list (or `~numpy.ndarray`) of `~matplotlib.axes.Axes`
            to align the ylabels.
            Default is to align all Axes on the figure.

        See Also
        --------
        matplotlib.figure.Figure.align_xlabels
        matplotlib.figure.Figure.align_titles
        matplotlib.figure.Figure.align_labels

        Notes
        -----
        This assumes that ``axs`` are from the same `.GridSpec`, so that
        their `.SubplotSpec` positions correspond to figure positions.

        Examples
        --------
        Example with large yticks labels::

            fig, axs = plt.subplots(2, 1)
            axs[0].plot(np.arange(0, 1000, 50))
            axs[0].set_ylabel('YLabel 0')
            axs[1].set_ylabel('YLabel 1')
            fig.align_ylabels()
        """
        if axs is None:
            axs = self.axes
        axs = [ax for ax in np.ravel(axs) if ax.get_subplotspec() is not None]
        for ax in axs:
            _log.debug(' Working on: %s', ax.get_ylabel())
            colspan = ax.get_subplotspec().colspan
            pos = ax.yaxis.get_label_position()  # left or right
            # Search through other Axes for label positions that are same as
            # this one and that share the appropriate column number.
            # Add to a list associated with each Axes of siblings.
            # This list is inspected in `axis.draw` by
            # `axis._update_label_position`.
            for axc in axs:
                if axc.yaxis.get_label_position() == pos:
                    colspanc = axc.get_subplotspec().colspan
                    if (pos == 'left' and colspan.start == colspanc.start or
                            pos == 'right' and colspan.stop == colspanc.stop):
                        # grouper for groups of ylabels to align
                        self._align_label_groups['y'].join(ax, axc)

    def align_titles(self, axs=None):
        """
        Align the titles of subplots in the same subplot row if title
        alignment is being done automatically (i.e. the title position is
        not manually set).

        Alignment persists for draw events after this is called.

        Parameters
        ----------
        axs : list of `~matplotlib.axes.Axes`
            Optional list of (or ndarray) `~matplotlib.axes.Axes`
            to align the titles.
            Default is to align all Axes on the figure.

        See Also
        --------
        matplotlib.figure.Figure.align_xlabels
        matplotlib.figure.Figure.align_ylabels
        matplotlib.figure.Figure.align_labels

        Notes
        -----
        This assumes that ``axs`` are from the same `.GridSpec`, so that
        their `.SubplotSpec` positions correspond to figure positions.

        Examples
        --------
        Example with titles::

            fig, axs = plt.subplots(1, 2)
            axs[0].set_aspect('equal')
            axs[0].set_title('Title 0')
            axs[1].set_title('Title 1')
            fig.align_titles()
        """
        if axs is None:
            axs = self.axes
        axs = [ax for ax in np.ravel(axs) if ax.get_subplotspec() is not None]
        for ax in axs:
            _log.debug(' Working on: %s', ax.get_title())
            rowspan = ax.get_subplotspec().rowspan
            for axc in axs:
                rowspanc = axc.get_subplotspec().rowspan
                if (rowspan.start == rowspanc.start):
                    self._align_label_groups['title'].join(ax, axc)

    def align_labels(self, axs=None):
        """
        Align the xlabels and ylabels of subplots with the same subplots
        row or column (respectively) if label alignment is being
        done automatically (i.e. the label position is not manually set).

        Alignment persists for draw events after this is called.

        Parameters
        ----------
        axs : list of `~matplotlib.axes.Axes`
            Optional list (or `~numpy.ndarray`) of `~matplotlib.axes.Axes`
            to align the labels.
            Default is to align all Axes on the figure.

        See Also
        --------
        matplotlib.figure.Figure.align_xlabels
        matplotlib.figure.Figure.align_ylabels
        matplotlib.figure.Figure.align_titles
        """
        self.align_xlabels(axs=axs)
        self.align_ylabels(axs=axs)

    def add_gridspec(self, nrows=1, ncols=1, **kwargs):
        """
        Low-level API for creating a `.GridSpec` that has this figure as a parent.

        This is a low-level API, allowing you to create a gridspec and
        subsequently add subplots based on the gridspec. Most users do
        not need that freedom and should use the higher-level methods
        `~.Figure.subplots` or `~.Figure.subplot_mosaic`.

        Parameters
        ----------
        nrows : int, default: 1
            Number of rows in grid.

        ncols : int, default: 1
            Number of columns in grid.

        Returns
        -------
        `.GridSpec`

        Other Parameters
        ----------------
        **kwargs
            Keyword arguments are passed to `.GridSpec`.

        See Also
        --------
        matplotlib.pyplot.subplots

        Examples
        --------
        Adding a subplot that spans two rows::

            fig = plt.figure()
            gs = fig.add_gridspec(2, 2)
            ax1 = fig.add_subplot(gs[0, 0])
            ax2 = fig.add_subplot(gs[1, 0])
            # spans two rows:
            ax3 = fig.add_subplot(gs[:, 1])

        """

        _ = kwargs.pop('figure', None)  # pop in case user has added this...
        gs = GridSpec(nrows=nrows, ncols=ncols, figure=self, **kwargs)
        return gs

    def subfigures(self, nrows=1, ncols=1, squeeze=True,
                   wspace=None, hspace=None,
                   width_ratios=None, height_ratios=None,
                   **kwargs):
        """
        Add a set of subfigures to this figure or subfigure.

        A subfigure has the same artist methods as a figure, and is logically
        the same as a figure, but cannot print itself.
        See :doc:`/gallery/subplots_axes_and_figures/subfigures`.

        .. note::
            The *subfigure* concept is new in v3.4, and the API is still provisional.

        Parameters
        ----------
        nrows, ncols : int, default: 1
            Number of rows/columns of the subfigure grid.

        squeeze : bool, default: True
            If True, extra dimensions are squeezed out from the returned
            array of subfigures.

        wspace, hspace : float, default: None
            The amount of width/height reserved for space between subfigures,
            expressed as a fraction of the average subfigure width/height.
            If not given, the values will be inferred from rcParams if using
            constrained layout (see `~.ConstrainedLayoutEngine`), or zero if
            not using a layout engine.

        width_ratios : array-like of length *ncols*, optional
            Defines the relative widths of the columns. Each column gets a
            relative width of ``width_ratios[i] / sum(width_ratios)``.
            If not given, all columns will have the same width.

        height_ratios : array-like of length *nrows*, optional
            Defines the relative heights of the rows. Each row gets a
            relative height of ``height_ratios[i] / sum(height_ratios)``.
            If not given, all rows will have the same height.
        """
        gs = GridSpec(nrows=nrows, ncols=ncols, figure=self,
                      wspace=wspace, hspace=hspace,
                      width_ratios=width_ratios,
                      height_ratios=height_ratios,
                      left=0, right=1, bottom=0, top=1)

        sfarr = np.empty((nrows, ncols), dtype=object)
        for i in range(ncols):
            for j in range(nrows):
                sfarr[j, i] = self.add_subfigure(gs[j, i], **kwargs)

        if self.get_layout_engine() is None and (wspace is not None or
                                                 hspace is not None):
            # Gridspec wspace and hspace is ignored on subfigure instantiation,
            # and no space is left.  So need to account for it here if required.
            bottoms, tops, lefts, rights = gs.get_grid_positions(self)
            for sfrow, bottom, top in zip(sfarr, bottoms, tops):
                for sf, left, right in zip(sfrow, lefts, rights):
                    bbox = Bbox.from_extents(left, bottom, right, top)
                    sf._redo_transform_rel_fig(bbox=bbox)

        if squeeze:
            # Discarding unneeded dimensions that equal 1.  If we only have one
            # subfigure, just return it instead of a 1-element array.
            return sfarr.item() if sfarr.size == 1 else sfarr.squeeze()
        else:
            # Returned axis array will be always 2-d, even if nrows=ncols=1.
            return sfarr

    def add_subfigure(self, subplotspec, **kwargs):
        """
        Add a `.SubFigure` to the figure as part of a subplot arrangement.

        Parameters
        ----------
        subplotspec : `.gridspec.SubplotSpec`
            Defines the region in a parent gridspec where the subfigure will
            be placed.

        Returns
        -------
        `.SubFigure`

        Other Parameters
        ----------------
        **kwargs
            Are passed to the `.SubFigure` object.

        See Also
        --------
        .Figure.subfigures
        """
        sf = SubFigure(self, subplotspec, **kwargs)
        self.subfigs += [sf]
        sf._remove_method = self.subfigs.remove
        return sf

    def sca(self, a):
        """Set the current Axes to be *a* and return *a*."""
        self._axstack.bubble(a)
        self._axobservers.process("_axes_change_event", self)
        return a

    def gca(self):
        """
        Get the current Axes.

        If there is currently no Axes on this Figure, a new one is created
        using `.Figure.add_subplot`.  (To test whether there is currently an
        Axes on a Figure, check whether ``figure.axes`` is empty.  To test
        whether there is currently a Figure on the pyplot figure stack, check
        whether `.pyplot.get_fignums()` is empty.)
        """
        ax = self._axstack.current()
        return ax if ax is not None else self.add_subplot()

    def _gci(self):
        # Helper for `~matplotlib.pyplot.gci`.  Do not use elsewhere.
        """
        Get the current colorable artist.

        Specifically, returns the current `.ScalarMappable` instance (`.Image`
        created by `imshow` or `figimage`, `.Collection` created by `pcolor` or
        `scatter`, etc.), or *None* if no such instance has been defined.

        The current image is an attribute of the current Axes, or the nearest
        earlier Axes in the current figure that contains an image.

        Notes
        -----
        Historically, the only colorable artists were images; hence the name
        ``gci`` (get current image).
        """
        # Look first for an image in the current Axes.
        ax = self._axstack.current()
        if ax is None:
            return None
        im = ax._gci()
        if im is not None:
            return im
        # If there is no image in the current Axes, search for
        # one in a previously created Axes.  Whether this makes
        # sense is debatable, but it is the documented behavior.
        for ax in reversed(self.axes):
            im = ax._gci()
            if im is not None:
                return im
        return None

    def _process_projection_requirements(self, *, axes_class=None, polar=False,
                                         projection=None, **kwargs):
        """
        Handle the args/kwargs to add_axes/add_subplot/gca, returning::

            (axes_proj_class, proj_class_kwargs)

        which can be used for new Axes initialization/identification.
        """
        if axes_class is not None:
            if polar or projection is not None:
                raise ValueError(
                    "Cannot combine 'axes_class' and 'projection' or 'polar'")
            projection_class = axes_class
        else:

            if polar:
                if projection is not None and projection != 'polar':
                    raise ValueError(
                        f"polar={polar}, yet projection={projection!r}. "
                        "Only one of these arguments should be supplied."
                    )
                projection = 'polar'

            if isinstance(projection, str) or projection is None:
                projection_class = projections.get_projection_class(projection)
            elif hasattr(projection, '_as_mpl_axes'):
                projection_class, extra_kwargs = projection._as_mpl_axes()
                kwargs.update(**extra_kwargs)
            else:
                raise TypeError(
                    f"projection must be a string, None or implement a "
                    f"_as_mpl_axes method, not {projection!r}")
        return projection_class, kwargs

    def get_default_bbox_extra_artists(self):
        """
        Return a list of Artists typically used in `.Figure.get_tightbbox`.
        """
        bbox_artists = [artist for artist in self.get_children()
                        if (artist.get_visible() and artist.get_in_layout())]
        for ax in self.axes:
            if ax.get_visible():
                bbox_artists.extend(ax.get_default_bbox_extra_artists())
        return bbox_artists

    @_api.make_keyword_only("3.8", "bbox_extra_artists")
    def get_tightbbox(self, renderer=None, bbox_extra_artists=None):
        """
        Return a (tight) bounding box of the figure *in inches*.

        Note that `.FigureBase` differs from all other artists, which return
        their `.Bbox` in pixels.

        Artists that have ``artist.set_in_layout(False)`` are not included
        in the bbox.

        Parameters
        ----------
        renderer : `.RendererBase` subclass
            Renderer that will be used to draw the figures (i.e.
            ``fig.canvas.get_renderer()``)

        bbox_extra_artists : list of `.Artist` or ``None``
            List of artists to include in the tight bounding box.  If
            ``None`` (default), then all artist children of each Axes are
            included in the tight bounding box.

        Returns
        -------
        `.BboxBase`
            containing the bounding box (in figure inches).
        """

        if renderer is None:
            renderer = self.figure._get_renderer()

        bb = []
        if bbox_extra_artists is None:
            artists = [artist for artist in self.get_children()
                       if (artist not in self.axes and artist.get_visible()
                           and artist.get_in_layout())]
        else:
            artists = bbox_extra_artists

        for a in artists:
            bbox = a.get_tightbbox(renderer)
            if bbox is not None:
                bb.append(bbox)

        for ax in self.axes:
            if ax.get_visible():
                # some Axes don't take the bbox_extra_artists kwarg so we
                # need this conditional....
                try:
                    bbox = ax.get_tightbbox(
                        renderer, bbox_extra_artists=bbox_extra_artists)
                except TypeError:
                    bbox = ax.get_tightbbox(renderer)
                bb.append(bbox)
        bb = [b for b in bb
              if (np.isfinite(b.width) and np.isfinite(b.height)
                  and (b.width != 0 or b.height != 0))]

        isfigure = hasattr(self, 'bbox_inches')
        if len(bb) == 0:
            if isfigure:
                return self.bbox_inches
            else:
                # subfigures do not have bbox_inches, but do have a bbox
                bb = [self.bbox]

        _bbox = Bbox.union(bb)

        if isfigure:
            # transform from pixels to inches...
            _bbox = TransformedBbox(_bbox, self.dpi_scale_trans.inverted())

        return _bbox

    @staticmethod
    def _norm_per_subplot_kw(per_subplot_kw):
        expanded = {}
        for k, v in per_subplot_kw.items():
            if isinstance(k, tuple):
                for sub_key in k:
                    if sub_key in expanded:
                        raise ValueError(f'The key {sub_key!r} appears multiple times.')
                    expanded[sub_key] = v
            else:
                if k in expanded:
                    raise ValueError(f'The key {k!r} appears multiple times.')
                expanded[k] = v
        return expanded

    @staticmethod
    def _normalize_grid_string(layout):
        if '\n' not in layout:
            # single-line string
            return [list(ln) for ln in layout.split(';')]
        else:
            # multi-line string
            layout = inspect.cleandoc(layout)
            return [list(ln) for ln in layout.strip('\n').split('\n')]

    def subplot_mosaic(self, mosaic, *, sharex=False, sharey=False,
                       width_ratios=None, height_ratios=None,
                       empty_sentinel='.',
                       subplot_kw=None, per_subplot_kw=None, gridspec_kw=None):
        """
        Build a layout of Axes based on ASCII art or nested lists.

        This is a helper function to build complex GridSpec layouts visually.

        See :ref:`mosaic`
        for an example and full API documentation

        Parameters
        ----------
        mosaic : list of list of {hashable or nested} or str

            A visual layout of how you want your Axes to be arranged
            labeled as strings.  For example ::

               x = [['A panel', 'A panel', 'edge'],
                    ['C panel', '.',       'edge']]

            produces 4 Axes:

            - 'A panel' which is 1 row high and spans the first two columns
            - 'edge' which is 2 rows high and is on the right edge
            - 'C panel' which in 1 row and 1 column wide in the bottom left
            - a blank space 1 row and 1 column wide in the bottom center

            Any of the entries in the layout can be a list of lists
            of the same form to create nested layouts.

            If input is a str, then it can either be a multi-line string of
            the form ::

              '''
              AAE
              C.E
              '''

            where each character is a column and each line is a row. Or it
            can be a single-line string where rows are separated by ``;``::

              'AB;CC'

            The string notation allows only single character Axes labels and
            does not support nesting but is very terse.

            The Axes identifiers may be `str` or a non-iterable hashable
            object (e.g. `tuple` s may not be used).

        sharex, sharey : bool, default: False
            If True, the x-axis (*sharex*) or y-axis (*sharey*) will be shared
            among all subplots.  In that case, tick label visibility and axis
            units behave as for `subplots`.  If False, each subplot's x- or
            y-axis will be independent.

        width_ratios : array-like of length *ncols*, optional
            Defines the relative widths of the columns. Each column gets a
            relative width of ``width_ratios[i] / sum(width_ratios)``.
            If not given, all columns will have the same width.  Equivalent
            to ``gridspec_kw={'width_ratios': [...]}``. In the case of nested
            layouts, this argument applies only to the outer layout.

        height_ratios : array-like of length *nrows*, optional
            Defines the relative heights of the rows. Each row gets a
            relative height of ``height_ratios[i] / sum(height_ratios)``.
            If not given, all rows will have the same height. Equivalent
            to ``gridspec_kw={'height_ratios': [...]}``. In the case of nested
            layouts, this argument applies only to the outer layout.

        subplot_kw : dict, optional
            Dictionary with keywords passed to the `.Figure.add_subplot` call
            used to create each subplot.  These values may be overridden by
            values in *per_subplot_kw*.

        per_subplot_kw : dict, optional
            A dictionary mapping the Axes identifiers or tuples of identifiers
            to a dictionary of keyword arguments to be passed to the
            `.Figure.add_subplot` call used to create each subplot.  The values
            in these dictionaries have precedence over the values in
            *subplot_kw*.

            If *mosaic* is a string, and thus all keys are single characters,
            it is possible to use a single string instead of a tuple as keys;
            i.e. ``"AB"`` is equivalent to ``("A", "B")``.

            .. versionadded:: 3.7

        gridspec_kw : dict, optional
            Dictionary with keywords passed to the `.GridSpec` constructor used
            to create the grid the subplots are placed on. In the case of
            nested layouts, this argument applies only to the outer layout.
            For more complex layouts, users should use `.Figure.subfigures`
            to create the nesting.

        empty_sentinel : object, optional
            Entry in the layout to mean "leave this space empty".  Defaults
            to ``'.'``. Note, if *layout* is a string, it is processed via
            `inspect.cleandoc` to remove leading white space, which may
            interfere with using white-space as the empty sentinel.

        Returns
        -------
        dict[label, Axes]
           A dictionary mapping the labels to the Axes objects.  The order of
           the Axes is left-to-right and top-to-bottom of their position in the
           total layout.

        """
        subplot_kw = subplot_kw or {}
        gridspec_kw = dict(gridspec_kw or {})
        per_subplot_kw = per_subplot_kw or {}

        if height_ratios is not None:
            if 'height_ratios' in gridspec_kw:
                raise ValueError("'height_ratios' must not be defined both as "
                                 "parameter and as key in 'gridspec_kw'")
            gridspec_kw['height_ratios'] = height_ratios
        if width_ratios is not None:
            if 'width_ratios' in gridspec_kw:
                raise ValueError("'width_ratios' must not be defined both as "
                                 "parameter and as key in 'gridspec_kw'")
            gridspec_kw['width_ratios'] = width_ratios

        # special-case string input
        if isinstance(mosaic, str):
            mosaic = self._normalize_grid_string(mosaic)
            per_subplot_kw = {
                tuple(k): v for k, v in per_subplot_kw.items()
            }

        per_subplot_kw = self._norm_per_subplot_kw(per_subplot_kw)

        # Only accept strict bools to allow a possible future API expansion.
        _api.check_isinstance(bool, sharex=sharex, sharey=sharey)

        def _make_array(inp):
            """
            Convert input into 2D array

            We need to have this internal function rather than
            ``np.asarray(..., dtype=object)`` so that a list of lists
            of lists does not get converted to an array of dimension > 2.

            Returns
            -------
            2D object array
            """
            r0, *rest = inp
            if isinstance(r0, str):
                raise ValueError('List mosaic specification must be 2D')
            for j, r in enumerate(rest, start=1):
                if isinstance(r, str):
                    raise ValueError('List mosaic specification must be 2D')
                if len(r0) != len(r):
                    raise ValueError(
                        "All of the rows must be the same length, however "
                        f"the first row ({r0!r}) has length {len(r0)} "
                        f"and row {j} ({r!r}) has length {len(r)}."
                    )
            out = np.zeros((len(inp), len(r0)), dtype=object)
            for j, r in enumerate(inp):
                for k, v in enumerate(r):
                    out[j, k] = v
            return out

        def _identify_keys_and_nested(mosaic):
            """
            Given a 2D object array, identify unique IDs and nested mosaics

            Parameters
            ----------
            mosaic : 2D object array

            Returns
            -------
            unique_ids : tuple
                The unique non-sub mosaic entries in this mosaic
            nested : dict[tuple[int, int], 2D object array]
            """
            # make sure we preserve the user supplied order
            unique_ids = cbook._OrderedSet()
            nested = {}
            for j, row in enumerate(mosaic):
                for k, v in enumerate(row):
                    if v == empty_sentinel:
                        continue
                    elif not cbook.is_scalar_or_string(v):
                        nested[(j, k)] = _make_array(v)
                    else:
                        unique_ids.add(v)

            return tuple(unique_ids), nested

        def _do_layout(gs, mosaic, unique_ids, nested):
            """
            Recursively do the mosaic.

            Parameters
            ----------
            gs : GridSpec
            mosaic : 2D object array
                The input converted to a 2D array for this level.
            unique_ids : tuple
                The identified scalar labels at this level of nesting.
            nested : dict[tuple[int, int]], 2D object array
                The identified nested mosaics, if any.

            Returns
            -------
            dict[label, Axes]
                A flat dict of all of the Axes created.
            """
            output = dict()

            # we need to merge together the Axes at this level and the Axes
            # in the (recursively) nested sub-mosaics so that we can add
            # them to the figure in the "natural" order if you were to
            # ravel in c-order all of the Axes that will be created
            #
            # This will stash the upper left index of each object (axes or
            # nested mosaic) at this level
            this_level = dict()

            # go through the unique keys,
            for name in unique_ids:
                # sort out where each axes starts/ends
                indx = np.argwhere(mosaic == name)
                start_row, start_col = np.min(indx, axis=0)
                end_row, end_col = np.max(indx, axis=0) + 1
                # and construct the slice object
                slc = (slice(start_row, end_row), slice(start_col, end_col))
                # some light error checking
                if (mosaic[slc] != name).any():
                    raise ValueError(
                        f"While trying to layout\n{mosaic!r}\n"
                        f"we found that the label {name!r} specifies a "
                        "non-rectangular or non-contiguous area.")
                # and stash this slice for later
                this_level[(start_row, start_col)] = (name, slc, 'axes')

            # do the same thing for the nested mosaics (simpler because these
            # cannot be spans yet!)
            for (j, k), nested_mosaic in nested.items():
                this_level[(j, k)] = (None, nested_mosaic, 'nested')

            # now go through the things in this level and add them
            # in order left-to-right top-to-bottom
            for key in sorted(this_level):
                name, arg, method = this_level[key]
                # we are doing some hokey function dispatch here based
                # on the 'method' string stashed above to sort out if this
                # element is an Axes or a nested mosaic.
                if method == 'axes':
                    slc = arg
                    # add a single Axes
                    if name in output:
                        raise ValueError(f"There are duplicate keys {name} "
                                         f"in the layout\n{mosaic!r}")
                    ax = self.add_subplot(
                        gs[slc], **{
                            'label': str(name),
                            **subplot_kw,
                            **per_subplot_kw.get(name, {})
                        }
                    )
                    output[name] = ax
                elif method == 'nested':
                    nested_mosaic = arg
                    j, k = key
                    # recursively add the nested mosaic
                    rows, cols = nested_mosaic.shape
                    nested_output = _do_layout(
                        gs[j, k].subgridspec(rows, cols),
                        nested_mosaic,
                        *_identify_keys_and_nested(nested_mosaic)
                    )
                    overlap = set(output) & set(nested_output)
                    if overlap:
                        raise ValueError(
                            f"There are duplicate keys {overlap} "
                            f"between the outer layout\n{mosaic!r}\n"
                            f"and the nested layout\n{nested_mosaic}"
                        )
                    output.update(nested_output)
                else:
                    raise RuntimeError("This should never happen")
            return output

        mosaic = _make_array(mosaic)
        rows, cols = mosaic.shape
        gs = self.add_gridspec(rows, cols, **gridspec_kw)
        ret = _do_layout(gs, mosaic, *_identify_keys_and_nested(mosaic))
        ax0 = next(iter(ret.values()))
        for ax in ret.values():
            if sharex:
                ax.sharex(ax0)
                ax._label_outer_xaxis(skip_non_rectangular_axes=True)
            if sharey:
                ax.sharey(ax0)
                ax._label_outer_yaxis(skip_non_rectangular_axes=True)
        if extra := set(per_subplot_kw) - set(ret):
            raise ValueError(
                f"The keys {extra} are in *per_subplot_kw* "
                "but not in the mosaic."
            )
        return ret

    def _set_artist_props(self, a):
        if a != self:
            a.set_figure(self)
        a.stale_callback = _stale_figure_callback
        a.set_transform(self.transSubfigure)


@_docstring.interpd
class SubFigure(FigureBase):
    """
    Logical figure that can be placed inside a figure.

    See :ref:`figure-api-subfigure` for an index of methods on this class.
    Typically instantiated using `.Figure.add_subfigure` or
    `.SubFigure.add_subfigure`, or `.SubFigure.subfigures`.  A subfigure has
    the same methods as a figure except for those particularly tied to the size
    or dpi of the figure, and is confined to a prescribed region of the figure.
    For example the following puts two subfigures side-by-side::

        fig = plt.figure()
        sfigs = fig.subfigures(1, 2)
        axsL = sfigs[0].subplots(1, 2)
        axsR = sfigs[1].subplots(2, 1)

    See :doc:`/gallery/subplots_axes_and_figures/subfigures`

    .. note::
        The *subfigure* concept is new in v3.4, and the API is still provisional.
    """

    def __init__(self, parent, subplotspec, *,
                 facecolor=None,
                 edgecolor=None,
                 linewidth=0.0,
                 frameon=None,
                 **kwargs):
        """
        Parameters
        ----------
        parent : `.Figure` or `.SubFigure`
            Figure or subfigure that contains the SubFigure.  SubFigures
            can be nested.

        subplotspec : `.gridspec.SubplotSpec`
            Defines the region in a parent gridspec where the subfigure will
            be placed.

        facecolor : default: ``"none"``
            The figure patch face color; transparent by default.

        edgecolor : default: :rc:`figure.edgecolor`
            The figure patch edge color.

        linewidth : float
            The linewidth of the frame (i.e. the edge linewidth of the figure
            patch).

        frameon : bool, default: :rc:`figure.frameon`
            If ``False``, suppress drawing the figure background patch.

        Other Parameters
        ----------------
        **kwargs : `.SubFigure` properties, optional

            %(SubFigure:kwdoc)s
        """
        super().__init__(**kwargs)
        if facecolor is None:
            facecolor = "none"
        if edgecolor is None:
            edgecolor = mpl.rcParams['figure.edgecolor']
        if frameon is None:
            frameon = mpl.rcParams['figure.frameon']

        self._subplotspec = subplotspec
        self._parent = parent
        self.figure = parent.figure

        # subfigures use the parent axstack
        self._axstack = parent._axstack
        self.subplotpars = parent.subplotpars
        self.dpi_scale_trans = parent.dpi_scale_trans
        self._axobservers = parent._axobservers
        self.canvas = parent.canvas
        self.transFigure = parent.transFigure
        self.bbox_relative = Bbox.null()
        self._redo_transform_rel_fig()
        self.figbbox = self._parent.figbbox
        self.bbox = TransformedBbox(self.bbox_relative,
                                    self._parent.transSubfigure)
        self.transSubfigure = BboxTransformTo(self.bbox)

        self.patch = Rectangle(
            xy=(0, 0), width=1, height=1, visible=frameon,
            facecolor=facecolor, edgecolor=edgecolor, linewidth=linewidth,
            # Don't let the figure patch influence bbox calculation.
            in_layout=False, transform=self.transSubfigure)
        self._set_artist_props(self.patch)
        self.patch.set_antialiased(False)

    @property
    def dpi(self):
        return self._parent.dpi

    @dpi.setter
    def dpi(self, value):
        self._parent.dpi = value

    def get_dpi(self):
        """
        Return the resolution of the parent figure in dots-per-inch as a float.
        """
        return self._parent.dpi

    def set_dpi(self, val):
        """
        Set the resolution of parent figure in dots-per-inch.

        Parameters
        ----------
        val : float
        """
        self._parent.dpi = val
        self.stale = True

    def _get_renderer(self):
        return self._parent._get_renderer()

    def _redo_transform_rel_fig(self, bbox=None):
        """
        Make the transSubfigure bbox relative to Figure transform.

        Parameters
        ----------
        bbox : bbox or None
            If not None, then the bbox is used for relative bounding box.
            Otherwise, it is calculated from the subplotspec.
        """
        if bbox is not None:
            self.bbox_relative.p0 = bbox.p0
            self.bbox_relative.p1 = bbox.p1
            return
        # need to figure out *where* this subplotspec is.
        gs = self._subplotspec.get_gridspec()
        wr = np.asarray(gs.get_width_ratios())
        hr = np.asarray(gs.get_height_ratios())
        dx = wr[self._subplotspec.colspan].sum() / wr.sum()
        dy = hr[self._subplotspec.rowspan].sum() / hr.sum()
        x0 = wr[:self._subplotspec.colspan.start].sum() / wr.sum()
        y0 = 1 - hr[:self._subplotspec.rowspan.stop].sum() / hr.sum()
        self.bbox_relative.p0 = (x0, y0)
        self.bbox_relative.p1 = (x0 + dx, y0 + dy)

    def get_constrained_layout(self):
        """
        Return whether constrained layout is being used.

        See :ref:`constrainedlayout_guide`.
        """
        return self._parent.get_constrained_layout()

    def get_constrained_layout_pads(self, relative=False):
        """
        Get padding for ``constrained_layout``.

        Returns a list of ``w_pad, h_pad`` in inches and
        ``wspace`` and ``hspace`` as fractions of the subplot.

        See :ref:`constrainedlayout_guide`.

        Parameters
        ----------
        relative : bool
            If `True`, then convert from inches to figure relative.
        """
        return self._parent.get_constrained_layout_pads(relative=relative)

    def get_layout_engine(self):
        return self._parent.get_layout_engine()

    @property
    def axes(self):
        """
        List of Axes in the SubFigure.  You can access and modify the Axes
        in the SubFigure through this list.

        Modifying this list has no effect. Instead, use `~.SubFigure.add_axes`,
        `~.SubFigure.add_subplot` or `~.SubFigure.delaxes` to add or remove an
        Axes.

        Note: The `.SubFigure.axes` property and `~.SubFigure.get_axes` method
        are equivalent.
        """
        return self._localaxes[:]

    get_axes = axes.fget

    def draw(self, renderer):
        # docstring inherited

        # draw the figure bounding box, perhaps none for white figure
        if not self.get_visible():
            return

        artists = self._get_draw_artists(renderer)

        try:
            renderer.open_group('subfigure', gid=self.get_gid())
            self.patch.draw(renderer)
            mimage._draw_list_compositing_images(
                renderer, self, artists, self.figure.suppressComposite)
            renderer.close_group('subfigure')

        finally:
            self.stale = False


@_docstring.interpd
class Figure(FigureBase):
    """
    The top level container for all the plot elements.

    See `matplotlib.figure` for an index of class methods.

    Attributes
    ----------
    patch
        The `.Rectangle` instance representing the figure background patch.

    suppressComposite
        For multiple images, the figure will make composite images
        depending on the renderer option_image_nocomposite function.  If
        *suppressComposite* is a boolean, this will override the renderer.
    """

    # we want to cache the fonts and mathtext at a global level so that when
    # multiple figures are created we can reuse them.  This helps with a bug on
    # windows where the creation of too many figures leads to too many open
    # file handles and improves the performance of parsing mathtext.  However,
    # these global caches are not thread safe.  The solution here is to let the
    # Figure acquire a shared lock at the start of the draw, and release it when it
    # is done.  This allows multiple renderers to share the cached fonts and
    # parsed text, but only one figure can draw at a time and so the font cache
    # and mathtext cache are used by only one renderer at a time.

    _render_lock = threading.RLock()

    def __str__(self):
        return "Figure(%gx%g)" % tuple(self.bbox.size)

    def __repr__(self):
        return "<{clsname} size {h:g}x{w:g} with {naxes} Axes>".format(
            clsname=self.__class__.__name__,
            h=self.bbox.size[0], w=self.bbox.size[1],
            naxes=len(self.axes),
        )

    def __init__(self,
                 figsize=None,
                 dpi=None,
                 *,
                 facecolor=None,
                 edgecolor=None,
                 linewidth=0.0,
                 frameon=None,
                 subplotpars=None,  # rc figure.subplot.*
                 tight_layout=None,  # rc figure.autolayout
                 constrained_layout=None,  # rc figure.constrained_layout.use
                 layout=None,
                 **kwargs
                 ):
        """
        Parameters
        ----------
        figsize : 2-tuple of floats, default: :rc:`figure.figsize`
            Figure dimension ``(width, height)`` in inches.

        dpi : float, default: :rc:`figure.dpi`
            Dots per inch.

        facecolor : default: :rc:`figure.facecolor`
            The figure patch facecolor.

        edgecolor : default: :rc:`figure.edgecolor`
            The figure patch edge color.

        linewidth : float
            The linewidth of the frame (i.e. the edge linewidth of the figure
            patch).

        frameon : bool, default: :rc:`figure.frameon`
            If ``False``, suppress drawing the figure background patch.

        subplotpars : `~matplotlib.gridspec.SubplotParams`
            Subplot parameters. If not given, the default subplot
            parameters :rc:`figure.subplot.*` are used.

        tight_layout : bool or dict, default: :rc:`figure.autolayout`
            Whether to use the tight layout mechanism. See `.set_tight_layout`.

            .. admonition:: Discouraged

                The use of this parameter is discouraged. Please use
                ``layout='tight'`` instead for the common case of
                ``tight_layout=True`` and use `.set_tight_layout` otherwise.

        constrained_layout : bool, default: :rc:`figure.constrained_layout.use`
            This is equal to ``layout='constrained'``.

            .. admonition:: Discouraged

                The use of this parameter is discouraged. Please use
                ``layout='constrained'`` instead.

        layout : {'constrained', 'compressed', 'tight', 'none', `.LayoutEngine`, \
None}, default: None
            The layout mechanism for positioning of plot elements to avoid
            overlapping Axes decorations (labels, ticks, etc). Note that
            layout managers can have significant performance penalties.

            - 'constrained': The constrained layout solver adjusts Axes sizes
              to avoid overlapping Axes decorations.  Can handle complex plot
              layouts and colorbars, and is thus recommended.

              See :ref:`constrainedlayout_guide` for examples.

            - 'compressed': uses the same algorithm as 'constrained', but
              removes extra space between fixed-aspect-ratio Axes.  Best for
              simple grids of Axes.

            - 'tight': Use the tight layout mechanism. This is a relatively
              simple algorithm that adjusts the subplot parameters so that
              decorations do not overlap.

              See :ref:`tight_layout_guide` for examples.

            - 'none': Do not use a layout engine.

            - A `.LayoutEngine` instance. Builtin layout classes are
              `.ConstrainedLayoutEngine` and `.TightLayoutEngine`, more easily
              accessible by 'constrained' and 'tight'.  Passing an instance
              allows third parties to provide their own layout engine.

            If not given, fall back to using the parameters *tight_layout* and
            *constrained_layout*, including their config defaults
            :rc:`figure.autolayout` and :rc:`figure.constrained_layout.use`.

        Other Parameters
        ----------------
        **kwargs : `.Figure` properties, optional

            %(Figure:kwdoc)s
        """
        super().__init__(**kwargs)
        self.figure = self
        self._layout_engine = None

        if layout is not None:
            if (tight_layout is not None):
                _api.warn_external(
                    "The Figure parameters 'layout' and 'tight_layout' cannot "
                    "be used together. Please use 'layout' only.")
            if (constrained_layout is not None):
                _api.warn_external(
                    "The Figure parameters 'layout' and 'constrained_layout' "
                    "cannot be used together. Please use 'layout' only.")
            self.set_layout_engine(layout=layout)
        elif tight_layout is not None:
            if constrained_layout is not None:
                _api.warn_external(
                    "The Figure parameters 'tight_layout' and "
                    "'constrained_layout' cannot be used together. Please use "
                    "'layout' parameter")
            self.set_layout_engine(layout='tight')
            if isinstance(tight_layout, dict):
                self.get_layout_engine().set(**tight_layout)
        elif constrained_layout is not None:
            if isinstance(constrained_layout, dict):
                self.set_layout_engine(layout='constrained')
                self.get_layout_engine().set(**constrained_layout)
            elif constrained_layout:
                self.set_layout_engine(layout='constrained')

        else:
            # everything is None, so use default:
            self.set_layout_engine(layout=layout)

        # Callbacks traditionally associated with the canvas (and exposed with
        # a proxy property), but that actually need to be on the figure for
        # pickling.
        self._canvas_callbacks = cbook.CallbackRegistry(
            signals=FigureCanvasBase.events)
        connect = self._canvas_callbacks._connect_picklable
        self._mouse_key_ids = [
            connect('key_press_event', backend_bases._key_handler),
            connect('key_release_event', backend_bases._key_handler),
            connect('key_release_event', backend_bases._key_handler),
            connect('button_press_event', backend_bases._mouse_handler),
            connect('button_release_event', backend_bases._mouse_handler),
            connect('scroll_event', backend_bases._mouse_handler),
            connect('motion_notify_event', backend_bases._mouse_handler),
        ]
        self._button_pick_id = connect('button_press_event', self.pick)
        self._scroll_pick_id = connect('scroll_event', self.pick)

        if figsize is None:
            figsize = mpl.rcParams['figure.figsize']
        if dpi is None:
            dpi = mpl.rcParams['figure.dpi']
        if facecolor is None:
            facecolor = mpl.rcParams['figure.facecolor']
        if edgecolor is None:
            edgecolor = mpl.rcParams['figure.edgecolor']
        if frameon is None:
            frameon = mpl.rcParams['figure.frameon']

        if not np.isfinite(figsize).all() or (np.array(figsize) < 0).any():
            raise ValueError('figure size must be positive finite not '
                             f'{figsize}')
        self.bbox_inches = Bbox.from_bounds(0, 0, *figsize)

        self.dpi_scale_trans = Affine2D().scale(dpi)
        # do not use property as it will trigger
        self._dpi = dpi
        self.bbox = TransformedBbox(self.bbox_inches, self.dpi_scale_trans)
        self.figbbox = self.bbox
        self.transFigure = BboxTransformTo(self.bbox)
        self.transSubfigure = self.transFigure

        self.patch = Rectangle(
            xy=(0, 0), width=1, height=1, visible=frameon,
            facecolor=facecolor, edgecolor=edgecolor, linewidth=linewidth,
            # Don't let the figure patch influence bbox calculation.
            in_layout=False)
        self._set_artist_props(self.patch)
        self.patch.set_antialiased(False)

        FigureCanvasBase(self)  # Set self.canvas.

        if subplotpars is None:
            subplotpars = SubplotParams()

        self.subplotpars = subplotpars

        self._axstack = _AxesStack()  # track all figure Axes and current Axes
        self.clear()

    def pick(self, mouseevent):
        if not self.canvas.widgetlock.locked():
            super().pick(mouseevent)

    def _check_layout_engines_compat(self, old, new):
        """
        Helper for set_layout engine

        If the figure has used the old engine and added a colorbar then the
        value of colorbar_gridspec must be the same on the new engine.
        """
        if old is None or new is None:
            return True
        if old.colorbar_gridspec == new.colorbar_gridspec:
            return True
        # colorbar layout different, so check if any colorbars are on the
        # figure...
        for ax in self.axes:
            if hasattr(ax, '_colorbar'):
                # colorbars list themselves as a colorbar.
                return False
        return True

    def set_layout_engine(self, layout=None, **kwargs):
        """
        Set the layout engine for this figure.

        Parameters
        ----------
        layout : {'constrained', 'compressed', 'tight', 'none', `.LayoutEngine`, None}

            - 'constrained' will use `~.ConstrainedLayoutEngine`
            - 'compressed' will also use `~.ConstrainedLayoutEngine`, but with
              a correction that attempts to make a good layout for fixed-aspect
              ratio Axes.
            - 'tight' uses `~.TightLayoutEngine`
            - 'none' removes layout engine.

            If a `.LayoutEngine` instance, that instance will be used.

            If `None`, the behavior is controlled by :rc:`figure.autolayout`
            (which if `True` behaves as if 'tight' was passed) and
            :rc:`figure.constrained_layout.use` (which if `True` behaves as if
            'constrained' was passed).  If both are `True`,
            :rc:`figure.autolayout` takes priority.

            Users and libraries can define their own layout engines and pass
            the instance directly as well.

        **kwargs
            The keyword arguments are passed to the layout engine to set things
            like padding and margin sizes.  Only used if *layout* is a string.

        """
        if layout is None:
            if mpl.rcParams['figure.autolayout']:
                layout = 'tight'
            elif mpl.rcParams['figure.constrained_layout.use']:
                layout = 'constrained'
            else:
                self._layout_engine = None
                return
        if layout == 'tight':
            new_layout_engine = TightLayoutEngine(**kwargs)
        elif layout == 'constrained':
            new_layout_engine = ConstrainedLayoutEngine(**kwargs)
        elif layout == 'compressed':
            new_layout_engine = ConstrainedLayoutEngine(compress=True,
                                                        **kwargs)
        elif layout == 'none':
            if self._layout_engine is not None:
                new_layout_engine = PlaceHolderLayoutEngine(
                    self._layout_engine.adjust_compatible,
                    self._layout_engine.colorbar_gridspec
                )
            else:
                new_layout_engine = None
        elif isinstance(layout, LayoutEngine):
            new_layout_engine = layout
        else:
            raise ValueError(f"Invalid value for 'layout': {layout!r}")

        if self._check_layout_engines_compat(self._layout_engine,
                                             new_layout_engine):
            self._layout_engine = new_layout_engine
        else:
            raise RuntimeError('Colorbar layout of new layout engine not '
                               'compatible with old engine, and a colorbar '
                               'has been created.  Engine not changed.')

    def get_layout_engine(self):
        return self._layout_engine

    # TODO: I'd like to dynamically add the _repr_html_ method
    # to the figure in the right context, but then IPython doesn't
    # use it, for some reason.

    def _repr_html_(self):
        # We can't use "isinstance" here, because then we'd end up importing
        # webagg unconditionally.
        if 'WebAgg' in type(self.canvas).__name__:
            from matplotlib.backends import backend_webagg
            return backend_webagg.ipython_inline_display(self)

    def show(self, warn=True):
        """
        If using a GUI backend with pyplot, display the figure window.

        If the figure was not created using `~.pyplot.figure`, it will lack
        a `~.backend_bases.FigureManagerBase`, and this method will raise an
        AttributeError.

        .. warning::

            This does not manage an GUI event loop. Consequently, the figure
            may only be shown briefly or not shown at all if you or your
            environment are not managing an event loop.

            Use cases for `.Figure.show` include running this from a GUI
            application (where there is persistently an event loop running) or
            from a shell, like IPython, that install an input hook to allow the
            interactive shell to accept input while the figure is also being
            shown and interactive.  Some, but not all, GUI toolkits will
            register an input hook on import.  See :ref:`cp_integration` for
            more details.

            If you're in a shell without input hook integration or executing a
            python script, you should use `matplotlib.pyplot.show` with
            ``block=True`` instead, which takes care of starting and running
            the event loop for you.

        Parameters
        ----------
        warn : bool, default: True
            If ``True`` and we are not running headless (i.e. on Linux with an
            unset DISPLAY), issue warning when called on a non-GUI backend.

        """
        if self.canvas.manager is None:
            raise AttributeError(
                "Figure.show works only for figures managed by pyplot, "
                "normally created by pyplot.figure()")
        try:
            self.canvas.manager.show()
        except NonGuiException as exc:
            if warn:
                _api.warn_external(str(exc))

    @property
    def axes(self):
        """
        List of Axes in the Figure. You can access and modify the Axes in the
        Figure through this list.

        Do not modify the list itself. Instead, use `~Figure.add_axes`,
        `~.Figure.add_subplot` or `~.Figure.delaxes` to add or remove an Axes.

        Note: The `.Figure.axes` property and `~.Figure.get_axes` method are
        equivalent.
        """
        return self._axstack.as_list()

    get_axes = axes.fget

    def _get_renderer(self):
        if hasattr(self.canvas, 'get_renderer'):
            return self.canvas.get_renderer()
        else:
            return _get_renderer(self)

    def _get_dpi(self):
        return self._dpi

    def _set_dpi(self, dpi, forward=True):
        """
        Parameters
        ----------
        dpi : float

        forward : bool
            Passed on to `~.Figure.set_size_inches`
        """
        if dpi == self._dpi:
            # We don't want to cause undue events in backends.
            return
        self._dpi = dpi
        self.dpi_scale_trans.clear().scale(dpi)
        w, h = self.get_size_inches()
        self.set_size_inches(w, h, forward=forward)

    dpi = property(_get_dpi, _set_dpi, doc="The resolution in dots per inch.")

    def get_tight_layout(self):
        """Return whether `.Figure.tight_layout` is called when drawing."""
        return isinstance(self.get_layout_engine(), TightLayoutEngine)

    @_api.deprecated("3.6", alternative="set_layout_engine",
                     pending=True)
    def set_tight_layout(self, tight):
        """
        Set whether and how `.Figure.tight_layout` is called when drawing.

        Parameters
        ----------
        tight : bool or dict with keys "pad", "w_pad", "h_pad", "rect" or None
            If a bool, sets whether to call `.Figure.tight_layout` upon drawing.
            If ``None``, use :rc:`figure.autolayout` instead.
            If a dict, pass it as kwargs to `.Figure.tight_layout`, overriding the
            default paddings.
        """
        if tight is None:
            tight = mpl.rcParams['figure.autolayout']
        _tight = 'tight' if bool(tight) else 'none'
        _tight_parameters = tight if isinstance(tight, dict) else {}
        self.set_layout_engine(_tight, **_tight_parameters)
        self.stale = True

    def get_constrained_layout(self):
        """
        Return whether constrained layout is being used.

        See :ref:`constrainedlayout_guide`.
        """
        return isinstance(self.get_layout_engine(), ConstrainedLayoutEngine)

    @_api.deprecated("3.6", alternative="set_layout_engine('constrained')",
                     pending=True)
    def set_constrained_layout(self, constrained):
        """
        Set whether ``constrained_layout`` is used upon drawing.

        If None, :rc:`figure.constrained_layout.use` value will be used.

        When providing a dict containing the keys ``w_pad``, ``h_pad``
        the default ``constrained_layout`` paddings will be
        overridden.  These pads are in inches and default to 3.0/72.0.
        ``w_pad`` is the width padding and ``h_pad`` is the height padding.

        Parameters
        ----------
        constrained : bool or dict or None
        """
        if constrained is None:
            constrained = mpl.rcParams['figure.constrained_layout.use']
        _constrained = 'constrained' if bool(constrained) else 'none'
        _parameters = constrained if isinstance(constrained, dict) else {}
        self.set_layout_engine(_constrained, **_parameters)
        self.stale = True

    @_api.deprecated(
         "3.6", alternative="figure.get_layout_engine().set()",
         pending=True)
    def set_constrained_layout_pads(self, **kwargs):
        """
        Set padding for ``constrained_layout``.

        Tip: The parameters can be passed from a dictionary by using
        ``fig.set_constrained_layout(**pad_dict)``.

        See :ref:`constrainedlayout_guide`.

        Parameters
        ----------
        w_pad : float, default: :rc:`figure.constrained_layout.w_pad`
            Width padding in inches.  This is the pad around Axes
            and is meant to make sure there is enough room for fonts to
            look good.  Defaults to 3 pts = 0.04167 inches

        h_pad : float, default: :rc:`figure.constrained_layout.h_pad`
            Height padding in inches. Defaults to 3 pts.

        wspace : float, default: :rc:`figure.constrained_layout.wspace`
            Width padding between subplots, expressed as a fraction of the
            subplot width.  The total padding ends up being w_pad + wspace.

        hspace : float, default: :rc:`figure.constrained_layout.hspace`
            Height padding between subplots, expressed as a fraction of the
            subplot width. The total padding ends up being h_pad + hspace.

        """
        if isinstance(self.get_layout_engine(), ConstrainedLayoutEngine):
            self.get_layout_engine().set(**kwargs)

    @_api.deprecated("3.6", alternative="fig.get_layout_engine().get()",
                     pending=True)
    def get_constrained_layout_pads(self, relative=False):
        """
        Get padding for ``constrained_layout``.

        Returns a list of ``w_pad, h_pad`` in inches and
        ``wspace`` and ``hspace`` as fractions of the subplot.
        All values are None if ``constrained_layout`` is not used.

        See :ref:`constrainedlayout_guide`.

        Parameters
        ----------
        relative : bool
            If `True`, then convert from inches to figure relative.
        """
        if not isinstance(self.get_layout_engine(), ConstrainedLayoutEngine):
            return None, None, None, None
        info = self.get_layout_engine().get()
        w_pad = info['w_pad']
        h_pad = info['h_pad']
        wspace = info['wspace']
        hspace = info['hspace']

        if relative and (w_pad is not None or h_pad is not None):
            renderer = self._get_renderer()
            dpi = renderer.dpi
            w_pad = w_pad * dpi / renderer.width
            h_pad = h_pad * dpi / renderer.height

        return w_pad, h_pad, wspace, hspace

    def set_canvas(self, canvas):
        """
        Set the canvas that contains the figure

        Parameters
        ----------
        canvas : FigureCanvas
        """
        self.canvas = canvas

    @_docstring.interpd
    def figimage(self, X, xo=0, yo=0, alpha=None, norm=None, cmap=None,
                 vmin=None, vmax=None, origin=None, resize=False, **kwargs):
        """
        Add a non-resampled image to the figure.

        The image is attached to the lower or upper left corner depending on
        *origin*.

        Parameters
        ----------
        X
            The image data. This is an array of one of the following shapes:

            - (M, N): an image with scalar data.  Color-mapping is controlled
              by *cmap*, *norm*, *vmin*, and *vmax*.
            - (M, N, 3): an image with RGB values (0-1 float or 0-255 int).
            - (M, N, 4): an image with RGBA values (0-1 float or 0-255 int),
              i.e. including transparency.

        xo, yo : int
            The *x*/*y* image offset in pixels.

        alpha : None or float
            The alpha blending value.

        %(cmap_doc)s

            This parameter is ignored if *X* is RGB(A).

        %(norm_doc)s

            This parameter is ignored if *X* is RGB(A).

        %(vmin_vmax_doc)s

            This parameter is ignored if *X* is RGB(A).

        origin : {'upper', 'lower'}, default: :rc:`image.origin`
            Indicates where the [0, 0] index of the array is in the upper left
            or lower left corner of the Axes.

        resize : bool
            If *True*, resize the figure to match the given image size.

        Returns
        -------
        `matplotlib.image.FigureImage`

        Other Parameters
        ----------------
        **kwargs
            Additional kwargs are `.Artist` kwargs passed on to `.FigureImage`.

        Notes
        -----
        figimage complements the Axes image (`~matplotlib.axes.Axes.imshow`)
        which will be resampled to fit the current Axes.  If you want
        a resampled image to fill the entire figure, you can define an
        `~matplotlib.axes.Axes` with extent [0, 0, 1, 1].

        Examples
        --------
        ::

            f = plt.figure()
            nx = int(f.get_figwidth() * f.dpi)
            ny = int(f.get_figheight() * f.dpi)
            data = np.random.random((ny, nx))
            f.figimage(data)
            plt.show()
        """
        if resize:
            dpi = self.get_dpi()
            figsize = [x / dpi for x in (X.shape[1], X.shape[0])]
            self.set_size_inches(figsize, forward=True)

        im = mimage.FigureImage(self, cmap=cmap, norm=norm,
                                offsetx=xo, offsety=yo,
                                origin=origin, **kwargs)
        im.stale_callback = _stale_figure_callback

        im.set_array(X)
        im.set_alpha(alpha)
        if norm is None:
            im.set_clim(vmin, vmax)
        self.images.append(im)
        im._remove_method = self.images.remove
        self.stale = True
        return im

    def set_size_inches(self, w, h=None, forward=True):
        """
        Set the figure size in inches.

        Call signatures::

             fig.set_size_inches(w, h)  # OR
             fig.set_size_inches((w, h))

        Parameters
        ----------
        w : (float, float) or float
            Width and height in inches (if height not specified as a separate
            argument) or width.
        h : float
            Height in inches.
        forward : bool, default: True
            If ``True``, the canvas size is automatically updated, e.g.,
            you can resize the figure window from the shell.

        See Also
        --------
        matplotlib.figure.Figure.get_size_inches
        matplotlib.figure.Figure.set_figwidth
        matplotlib.figure.Figure.set_figheight

        Notes
        -----
        To transform from pixels to inches divide by `Figure.dpi`.
        """
        if h is None:  # Got called with a single pair as argument.
            w, h = w
        size = np.array([w, h])
        if not np.isfinite(size).all() or (size < 0).any():
            raise ValueError(f'figure size must be positive finite not {size}')
        self.bbox_inches.p1 = size
        if forward:
            manager = self.canvas.manager
            if manager is not None:
                manager.resize(*(size * self.dpi).astype(int))
        self.stale = True

    def get_size_inches(self):
        """
        Return the current size of the figure in inches.

        Returns
        -------
        ndarray
           The size (width, height) of the figure in inches.

        See Also
        --------
        matplotlib.figure.Figure.set_size_inches
        matplotlib.figure.Figure.get_figwidth
        matplotlib.figure.Figure.get_figheight

        Notes
        -----
        The size in pixels can be obtained by multiplying with `Figure.dpi`.
        """
        return np.array(self.bbox_inches.p1)

    def get_figwidth(self):
        """Return the figure width in inches."""
        return self.bbox_inches.width

    def get_figheight(self):
        """Return the figure height in inches."""
        return self.bbox_inches.height

    def get_dpi(self):
        """Return the resolution in dots per inch as a float."""
        return self.dpi

    def set_dpi(self, val):
        """
        Set the resolution of the figure in dots-per-inch.

        Parameters
        ----------
        val : float
        """
        self.dpi = val
        self.stale = True

    def set_figwidth(self, val, forward=True):
        """
        Set the width of the figure in inches.

        Parameters
        ----------
        val : float
        forward : bool
            See `set_size_inches`.

        See Also
        --------
        matplotlib.figure.Figure.set_figheight
        matplotlib.figure.Figure.set_size_inches
        """
        self.set_size_inches(val, self.get_figheight(), forward=forward)

    def set_figheight(self, val, forward=True):
        """
        Set the height of the figure in inches.

        Parameters
        ----------
        val : float
        forward : bool
            See `set_size_inches`.

        See Also
        --------
        matplotlib.figure.Figure.set_figwidth
        matplotlib.figure.Figure.set_size_inches
        """
        self.set_size_inches(self.get_figwidth(), val, forward=forward)

    def clear(self, keep_observers=False):
        # docstring inherited
        super().clear(keep_observers=keep_observers)
        # FigureBase.clear does not clear toolbars, as
        # only Figure can have toolbars
        toolbar = self.canvas.toolbar
        if toolbar is not None:
            toolbar.update()

    @_finalize_rasterization
    @allow_rasterization
    def draw(self, renderer):
        # docstring inherited
        if not self.get_visible():
            return

        with self._render_lock:

            artists = self._get_draw_artists(renderer)
            try:
                renderer.open_group('figure', gid=self.get_gid())
                if self.axes and self.get_layout_engine() is not None:
                    try:
                        self.get_layout_engine().execute(self)
                    except ValueError:
                        pass
                        # ValueError can occur when resizing a window.

                self.patch.draw(renderer)
                mimage._draw_list_compositing_images(
                    renderer, self, artists, self.suppressComposite)

                renderer.close_group('figure')
            finally:
                self.stale = False

            DrawEvent("draw_event", self.canvas, renderer)._process()

    def draw_without_rendering(self):
        """
        Draw the figure with no output.  Useful to get the final size of
        artists that require a draw before their size is known (e.g. text).
        """
        renderer = _get_renderer(self)
        with renderer._draw_disabled():
            self.draw(renderer)

    def draw_artist(self, a):
        """
        Draw `.Artist` *a* only.
        """
        a.draw(self.canvas.get_renderer())

    def __getstate__(self):
        state = super().__getstate__()

        # The canvas cannot currently be pickled, but this has the benefit
        # of meaning that a figure can be detached from one canvas, and
        # re-attached to another.
        state.pop("canvas")

        # discard any changes to the dpi due to pixel ratio changes
        state["_dpi"] = state.get('_original_dpi', state['_dpi'])

        # add version information to the state
        state['__mpl_version__'] = mpl.__version__

        # check whether the figure manager (if any) is registered with pyplot
        from matplotlib import _pylab_helpers
        if self.canvas.manager in _pylab_helpers.Gcf.figs.values():
            state['_restore_to_pylab'] = True
        return state

    def __setstate__(self, state):
        version = state.pop('__mpl_version__')
        restore_to_pylab = state.pop('_restore_to_pylab', False)

        if version != mpl.__version__:
            _api.warn_external(
                f"This figure was saved with matplotlib version {version} and "
                f"loaded with {mpl.__version__} so may not function correctly."
            )
        self.__dict__ = state

        # re-initialise some of the unstored state information
        FigureCanvasBase(self)  # Set self.canvas.

        if restore_to_pylab:
            # lazy import to avoid circularity
            import matplotlib.pyplot as plt
            import matplotlib._pylab_helpers as pylab_helpers
            allnums = plt.get_fignums()
            num = max(allnums) + 1 if allnums else 1
            backend = plt._get_backend_mod()
            mgr = backend.new_figure_manager_given_figure(num, self)
            pylab_helpers.Gcf._set_new_active_manager(mgr)
            plt.draw_if_interactive()

        self.stale = True

    def add_axobserver(self, func):
        """Whenever the Axes state change, ``func(self)`` will be called."""
        # Connect a wrapper lambda and not func itself, to avoid it being
        # weakref-collected.
        self._axobservers.connect("_axes_change_event", lambda arg: func(arg))

    def savefig(self, fname, *, transparent=None, **kwargs):
        """
        Save the current figure as an image or vector graphic to a file.

        Call signature::

          savefig(fname, *, transparent=None, dpi='figure', format=None,
                  metadata=None, bbox_inches=None, pad_inches=0.1,
                  facecolor='auto', edgecolor='auto', backend=None,
                  **kwargs
                 )

        The available output formats depend on the backend being used.

        Parameters
        ----------
        fname : str or path-like or binary file-like
            A path, or a Python file-like object, or
            possibly some backend-dependent object such as
            `matplotlib.backends.backend_pdf.PdfPages`.

            If *format* is set, it determines the output format, and the file
            is saved as *fname*.  Note that *fname* is used verbatim, and there
            is no attempt to make the extension, if any, of *fname* match
            *format*, and no extension is appended.

            If *format* is not set, then the format is inferred from the
            extension of *fname*, if there is one.  If *format* is not
            set and *fname* has no extension, then the file is saved with
            :rc:`savefig.format` and the appropriate extension is appended to
            *fname*.

        Other Parameters
        ----------------
        transparent : bool, default: :rc:`savefig.transparent`
            If *True*, the Axes patches will all be transparent; the
            Figure patch will also be transparent unless *facecolor*
            and/or *edgecolor* are specified via kwargs.

            If *False* has no effect and the color of the Axes and
            Figure patches are unchanged (unless the Figure patch
            is specified via the *facecolor* and/or *edgecolor* keyword
            arguments in which case those colors are used).

            The transparency of these patches will be restored to their
            original values upon exit of this function.

            This is useful, for example, for displaying
            a plot on top of a colored background on a web page.

        dpi : float or 'figure', default: :rc:`savefig.dpi`
            The resolution in dots per inch.  If 'figure', use the figure's
            dpi value.

        format : str
            The file format, e.g. 'png', 'pdf', 'svg', ... The behavior when
            this is unset is documented under *fname*.

        metadata : dict, optional
            Key/value pairs to store in the image metadata. The supported keys
            and defaults depend on the image format and backend:

            - 'png' with Agg backend: See the parameter ``metadata`` of
              `~.FigureCanvasAgg.print_png`.
            - 'pdf' with pdf backend: See the parameter ``metadata`` of
              `~.backend_pdf.PdfPages`.
            - 'svg' with svg backend: See the parameter ``metadata`` of
              `~.FigureCanvasSVG.print_svg`.
            - 'eps' and 'ps' with PS backend: Only 'Creator' is supported.

            Not supported for 'pgf', 'raw', and 'rgba' as those formats do not support
            embedding metadata.
            Does not currently support 'jpg', 'tiff', or 'webp', but may include
            embedding EXIF metadata in the future.

        bbox_inches : str or `.Bbox`, default: :rc:`savefig.bbox`
            Bounding box in inches: only the given portion of the figure is
            saved.  If 'tight', try to figure out the tight bbox of the figure.

        pad_inches : float or 'layout', default: :rc:`savefig.pad_inches`
            Amount of padding in inches around the figure when bbox_inches is
            'tight'. If 'layout' use the padding from the constrained or
            compressed layout engine; ignored if one of those engines is not in
            use.

        facecolor : :mpltype:`color` or 'auto', default: :rc:`savefig.facecolor`
            The facecolor of the figure.  If 'auto', use the current figure
            facecolor.

        edgecolor : :mpltype:`color` or 'auto', default: :rc:`savefig.edgecolor`
            The edgecolor of the figure.  If 'auto', use the current figure
            edgecolor.

        backend : str, optional
            Use a non-default backend to render the file, e.g. to render a
            png file with the "cairo" backend rather than the default "agg",
            or a pdf file with the "pgf" backend rather than the default
            "pdf".  Note that the default backend is normally sufficient.  See
            :ref:`the-builtin-backends` for a list of valid backends for each
            file format.  Custom backends can be referenced as "module://...".

        orientation : {'landscape', 'portrait'}
            Currently only supported by the postscript backend.

        papertype : str
            One of 'letter', 'legal', 'executive', 'ledger', 'a0' through
            'a10', 'b0' through 'b10'. Only supported for postscript
            output.

        bbox_extra_artists : list of `~matplotlib.artist.Artist`, optional
            A list of extra artists that will be considered when the
            tight bbox is calculated.

        pil_kwargs : dict, optional
            Additional keyword arguments that are passed to
            `PIL.Image.Image.save` when saving the figure.

        """

        kwargs.setdefault('dpi', mpl.rcParams['savefig.dpi'])
        if transparent is None:
            transparent = mpl.rcParams['savefig.transparent']

        with ExitStack() as stack:
            if transparent:
                def _recursively_make_subfig_transparent(exit_stack, subfig):
                    exit_stack.enter_context(
                        subfig.patch._cm_set(
                            facecolor="none", edgecolor="none"))
                    for ax in subfig.axes:
                        exit_stack.enter_context(
                            ax.patch._cm_set(
                                facecolor="none", edgecolor="none"))
                    for sub_subfig in subfig.subfigs:
                        _recursively_make_subfig_transparent(
                            exit_stack, sub_subfig)

                def _recursively_make_axes_transparent(exit_stack, ax):
                    exit_stack.enter_context(
                        ax.patch._cm_set(facecolor="none", edgecolor="none"))
                    for child_ax in ax.child_axes:
                        exit_stack.enter_context(
                            child_ax.patch._cm_set(
                                facecolor="none", edgecolor="none"))
                    for child_childax in ax.child_axes:
                        _recursively_make_axes_transparent(
                            exit_stack, child_childax)

                kwargs.setdefault('facecolor', 'none')
                kwargs.setdefault('edgecolor', 'none')
                # set subfigure to appear transparent in printed image
                for subfig in self.subfigs:
                    _recursively_make_subfig_transparent(stack, subfig)
                # set Axes to be transparent
                for ax in self.axes:
                    _recursively_make_axes_transparent(stack, ax)
            self.canvas.print_figure(fname, **kwargs)

    def ginput(self, n=1, timeout=30, show_clicks=True,
               mouse_add=MouseButton.LEFT,
               mouse_pop=MouseButton.RIGHT,
               mouse_stop=MouseButton.MIDDLE):
        """
        Blocking call to interact with a figure.

        Wait until the user clicks *n* times on the figure, and return the
        coordinates of each click in a list.

        There are three possible interactions:

        - Add a point.
        - Remove the most recently added point.
        - Stop the interaction and return the points added so far.

        The actions are assigned to mouse buttons via the arguments
        *mouse_add*, *mouse_pop* and *mouse_stop*.

        Parameters
        ----------
        n : int, default: 1
            Number of mouse clicks to accumulate. If negative, accumulate
            clicks until the input is terminated manually.
        timeout : float, default: 30 seconds
            Number of seconds to wait before timing out. If zero or negative
            will never time out.
        show_clicks : bool, default: True
            If True, show a red cross at the location of each click.
        mouse_add : `.MouseButton` or None, default: `.MouseButton.LEFT`
            Mouse button used to add points.
        mouse_pop : `.MouseButton` or None, default: `.MouseButton.RIGHT`
            Mouse button used to remove the most recently added point.
        mouse_stop : `.MouseButton` or None, default: `.MouseButton.MIDDLE`
            Mouse button used to stop input.

        Returns
        -------
        list of tuples
            A list of the clicked (x, y) coordinates.

        Notes
        -----
        The keyboard can also be used to select points in case your mouse
        does not have one or more of the buttons.  The delete and backspace
        keys act like right-clicking (i.e., remove last point), the enter key
        terminates input and any other key (not already used by the window
        manager) selects a point.
        """
        clicks = []
        marks = []

        def handler(event):
            is_button = event.name == "button_press_event"
            is_key = event.name == "key_press_event"
            # Quit (even if not in infinite mode; this is consistent with
            # MATLAB and sometimes quite useful, but will require the user to
            # test how many points were actually returned before using data).
            if (is_button and event.button == mouse_stop
                    or is_key and event.key in ["escape", "enter"]):
                self.canvas.stop_event_loop()
            # Pop last click.
            elif (is_button and event.button == mouse_pop
                  or is_key and event.key in ["backspace", "delete"]):
                if clicks:
                    clicks.pop()
                    if show_clicks:
                        marks.pop().remove()
                        self.canvas.draw()
            # Add new click.
            elif (is_button and event.button == mouse_add
                  # On macOS/gtk, some keys return None.
                  or is_key and event.key is not None):
                if event.inaxes:
                    clicks.append((event.xdata, event.ydata))
                    _log.info("input %i: %f, %f",
                              len(clicks), event.xdata, event.ydata)
                    if show_clicks:
                        line = mpl.lines.Line2D([event.xdata], [event.ydata],
                                                marker="+", color="r")
                        event.inaxes.add_line(line)
                        marks.append(line)
                        self.canvas.draw()
            if len(clicks) == n and n > 0:
                self.canvas.stop_event_loop()

        _blocking_input.blocking_input_loop(
            self, ["button_press_event", "key_press_event"], timeout, handler)

        # Cleanup.
        for mark in marks:
            mark.remove()
        self.canvas.draw()

        return clicks

    def waitforbuttonpress(self, timeout=-1):
        """
        Blocking call to interact with the figure.

        Wait for user input and return True if a key was pressed, False if a
        mouse button was pressed and None if no input was given within
        *timeout* seconds.  Negative values deactivate *timeout*.
        """
        event = None

        def handler(ev):
            nonlocal event
            event = ev
            self.canvas.stop_event_loop()

        _blocking_input.blocking_input_loop(
            self, ["button_press_event", "key_press_event"], timeout, handler)

        return None if event is None else event.name == "key_press_event"

    def tight_layout(self, *, pad=1.08, h_pad=None, w_pad=None, rect=None):
        """
        Adjust the padding between and around subplots.

        To exclude an artist on the Axes from the bounding box calculation
        that determines the subplot parameters (i.e. legend, or annotation),
        set ``a.set_in_layout(False)`` for that artist.

        Parameters
        ----------
        pad : float, default: 1.08
            Padding between the figure edge and the edges of subplots,
            as a fraction of the font size.
        h_pad, w_pad : float, default: *pad*
            Padding (height/width) between edges of adjacent subplots,
            as a fraction of the font size.
        rect : tuple (left, bottom, right, top), default: (0, 0, 1, 1)
            A rectangle in normalized figure coordinates into which the whole
            subplots area (including labels) will fit.

        See Also
        --------
        .Figure.set_layout_engine
        .pyplot.tight_layout
        """
        # note that here we do not permanently set the figures engine to
        # tight_layout but rather just perform the layout in place and remove
        # any previous engines.
        engine = TightLayoutEngine(pad=pad, h_pad=h_pad, w_pad=w_pad, rect=rect)
        try:
            previous_engine = self.get_layout_engine()
            self.set_layout_engine(engine)
            engine.execute(self)
            if previous_engine is not None and not isinstance(
                previous_engine, (TightLayoutEngine, PlaceHolderLayoutEngine)
            ):
                _api.warn_external('The figure layout has changed to tight')
        finally:
            self.set_layout_engine('none')


def figaspect(arg):
    """
    Calculate the width and height for a figure with a specified aspect ratio.

    While the height is taken from :rc:`figure.figsize`, the width is
    adjusted to match the desired aspect ratio. Additionally, it is ensured
    that the width is in the range [4., 16.] and the height is in the range
    [2., 16.]. If necessary, the default height is adjusted to ensure this.

    Parameters
    ----------
    arg : float or 2D array
        If a float, this defines the aspect ratio (i.e. the ratio height /
        width).
        In case of an array the aspect ratio is number of rows / number of
        columns, so that the array could be fitted in the figure undistorted.

    Returns
    -------
    width, height : float
        The figure size in inches.

    Notes
    -----
    If you want to create an Axes within the figure, that still preserves the
    aspect ratio, be sure to create it with equal width and height. See
    examples below.

    Thanks to Fernando Perez for this function.

    Examples
    --------
    Make a figure twice as tall as it is wide::

        w, h = figaspect(2.)
        fig = Figure(figsize=(w, h))
        ax = fig.add_axes([0.1, 0.1, 0.8, 0.8])
        ax.imshow(A, **kwargs)

    Make a figure with the proper aspect for an array::

        A = rand(5, 3)
        w, h = figaspect(A)
        fig = Figure(figsize=(w, h))
        ax = fig.add_axes([0.1, 0.1, 0.8, 0.8])
        ax.imshow(A, **kwargs)
    """

    isarray = hasattr(arg, 'shape') and not np.isscalar(arg)

    # min/max sizes to respect when autoscaling.  If John likes the idea, they
    # could become rc parameters, for now they're hardwired.
    figsize_min = np.array((4.0, 2.0))  # min length for width/height
    figsize_max = np.array((16.0, 16.0))  # max length for width/height

    # Extract the aspect ratio of the array
    if isarray:
        nr, nc = arg.shape[:2]
        arr_ratio = nr / nc
    else:
        arr_ratio = arg

    # Height of user figure defaults
    fig_height = mpl.rcParams['figure.figsize'][1]

    # New size for the figure, keeping the aspect ratio of the caller
    newsize = np.array((fig_height / arr_ratio, fig_height))

    # Sanity checks, don't drop either dimension below figsize_min
    newsize /= min(1.0, *(newsize / figsize_min))

    # Avoid humongous windows as well
    newsize /= max(1.0, *(newsize / figsize_max))

    # Finally, if we have a really funky aspect ratio, break it but respect
    # the min/max dimensions (we don't want figures 10 feet tall!)
    newsize = np.clip(newsize, figsize_min, figsize_max)
    return newsize