from itertools import product
import io
import platform

import matplotlib as mpl
import matplotlib.pyplot as plt
import matplotlib.ticker as mticker
from matplotlib import cbook
from matplotlib.backend_bases import MouseEvent
from matplotlib.colors import LogNorm
from matplotlib.patches import Circle, Ellipse
from matplotlib.transforms import Bbox, TransformedBbox
from matplotlib.testing.decorators import (
    check_figures_equal, image_comparison, remove_ticks_and_titles)

from mpl_toolkits.axes_grid1 import (
    axes_size as Size,
    host_subplot, make_axes_locatable,
    Grid, AxesGrid, ImageGrid)
from mpl_toolkits.axes_grid1.anchored_artists import (
    AnchoredAuxTransformBox, AnchoredDrawingArea,
    AnchoredDirectionArrows, AnchoredSizeBar)
from mpl_toolkits.axes_grid1.axes_divider import (
    Divider, HBoxDivider, make_axes_area_auto_adjustable, SubplotDivider,
    VBoxDivider)
from mpl_toolkits.axes_grid1.axes_rgb import RGBAxes
from mpl_toolkits.axes_grid1.inset_locator import (
    zoomed_inset_axes, mark_inset, inset_axes, BboxConnectorPatch)
import mpl_toolkits.axes_grid1.mpl_axes
import pytest

import numpy as np
from numpy.testing import assert_array_equal, assert_array_almost_equal


def test_divider_append_axes():
    fig, ax = plt.subplots()
    divider = make_axes_locatable(ax)
    axs = {
        "main": ax,
        "top": divider.append_axes("top", 1.2, pad=0.1, sharex=ax),
        "bottom": divider.append_axes("bottom", 1.2, pad=0.1, sharex=ax),
        "left": divider.append_axes("left", 1.2, pad=0.1, sharey=ax),
        "right": divider.append_axes("right", 1.2, pad=0.1, sharey=ax),
    }
    fig.canvas.draw()
    bboxes = {k: axs[k].get_window_extent() for k in axs}
    dpi = fig.dpi
    assert bboxes["top"].height == pytest.approx(1.2 * dpi)
    assert bboxes["bottom"].height == pytest.approx(1.2 * dpi)
    assert bboxes["left"].width == pytest.approx(1.2 * dpi)
    assert bboxes["right"].width == pytest.approx(1.2 * dpi)
    assert bboxes["top"].y0 - bboxes["main"].y1 == pytest.approx(0.1 * dpi)
    assert bboxes["main"].y0 - bboxes["bottom"].y1 == pytest.approx(0.1 * dpi)
    assert bboxes["main"].x0 - bboxes["left"].x1 == pytest.approx(0.1 * dpi)
    assert bboxes["right"].x0 - bboxes["main"].x1 == pytest.approx(0.1 * dpi)
    assert bboxes["left"].y0 == bboxes["main"].y0 == bboxes["right"].y0
    assert bboxes["left"].y1 == bboxes["main"].y1 == bboxes["right"].y1
    assert bboxes["top"].x0 == bboxes["main"].x0 == bboxes["bottom"].x0
    assert bboxes["top"].x1 == bboxes["main"].x1 == bboxes["bottom"].x1


# Update style when regenerating the test image
@image_comparison(['twin_axes_empty_and_removed'], extensions=["png"], tol=1,
                  style=('classic', '_classic_test_patch'))
def test_twin_axes_empty_and_removed():
    # Purely cosmetic font changes (avoid overlap)
    mpl.rcParams.update(
        {"font.size": 8, "xtick.labelsize": 8, "ytick.labelsize": 8})
    generators = ["twinx", "twiny", "twin"]
    modifiers = ["", "host invisible", "twin removed", "twin invisible",
                 "twin removed\nhost invisible"]
    # Unmodified host subplot at the beginning for reference
    h = host_subplot(len(modifiers)+1, len(generators), 2)
    h.text(0.5, 0.5, "host_subplot",
           horizontalalignment="center", verticalalignment="center")
    # Host subplots with various modifications (twin*, visibility) applied
    for i, (mod, gen) in enumerate(product(modifiers, generators),
                                   len(generators) + 1):
        h = host_subplot(len(modifiers)+1, len(generators), i)
        t = getattr(h, gen)()
        if "twin invisible" in mod:
            t.axis[:].set_visible(False)
        if "twin removed" in mod:
            t.remove()
        if "host invisible" in mod:
            h.axis[:].set_visible(False)
        h.text(0.5, 0.5, gen + ("\n" + mod if mod else ""),
               horizontalalignment="center", verticalalignment="center")
    plt.subplots_adjust(wspace=0.5, hspace=1)


def test_twin_axes_both_with_units():
    host = host_subplot(111)
    with pytest.warns(mpl.MatplotlibDeprecationWarning):
        host.plot_date([0, 1, 2], [0, 1, 2], xdate=False, ydate=True)
    twin = host.twinx()
    twin.plot(["a", "b", "c"])
    assert host.get_yticklabels()[0].get_text() == "00:00:00"
    assert twin.get_yticklabels()[0].get_text() == "a"


def test_axesgrid_colorbar_log_smoketest():
    fig = plt.figure()
    grid = AxesGrid(fig, 111,  # modified to be only subplot
                    nrows_ncols=(1, 1),
                    ngrids=1,
                    label_mode="L",
                    cbar_location="top",
                    cbar_mode="single",
                    )

    Z = 10000 * np.random.rand(10, 10)
    im = grid[0].imshow(Z, interpolation="nearest", norm=LogNorm())

    grid.cbar_axes[0].colorbar(im)


def test_inset_colorbar_tight_layout_smoketest():
    fig, ax = plt.subplots(1, 1)
    pts = ax.scatter([0, 1], [0, 1], c=[1, 5])

    cax = inset_axes(ax, width="3%", height="70%")
    plt.colorbar(pts, cax=cax)

    with pytest.warns(UserWarning, match="This figure includes Axes"):
        # Will warn, but not raise an error
        plt.tight_layout()


@image_comparison(['inset_locator.png'], style='default', remove_text=True)
def test_inset_locator():
    fig, ax = plt.subplots(figsize=[5, 4])

    # prepare the demo image
    # Z is a 15x15 array
    Z = cbook.get_sample_data("axes_grid/bivariate_normal.npy")
    extent = (-3, 4, -4, 3)
    Z2 = np.zeros((150, 150))
    ny, nx = Z.shape
    Z2[30:30+ny, 30:30+nx] = Z

    ax.imshow(Z2, extent=extent, interpolation="nearest",
              origin="lower")

    axins = zoomed_inset_axes(ax, zoom=6, loc='upper right')
    axins.imshow(Z2, extent=extent, interpolation="nearest",
                 origin="lower")
    axins.yaxis.get_major_locator().set_params(nbins=7)
    axins.xaxis.get_major_locator().set_params(nbins=7)
    # sub region of the original image
    x1, x2, y1, y2 = -1.5, -0.9, -2.5, -1.9
    axins.set_xlim(x1, x2)
    axins.set_ylim(y1, y2)

    plt.xticks(visible=False)
    plt.yticks(visible=False)

    # draw a bbox of the region of the inset axes in the parent axes and
    # connecting lines between the bbox and the inset axes area
    mark_inset(ax, axins, loc1=2, loc2=4, fc="none", ec="0.5")

    asb = AnchoredSizeBar(ax.transData,
                          0.5,
                          '0.5',
                          loc='lower center',
                          pad=0.1, borderpad=0.5, sep=5,
                          frameon=False)
    ax.add_artist(asb)


@image_comparison(['inset_axes.png'], style='default', remove_text=True)
def test_inset_axes():
    fig, ax = plt.subplots(figsize=[5, 4])

    # prepare the demo image
    # Z is a 15x15 array
    Z = cbook.get_sample_data("axes_grid/bivariate_normal.npy")
    extent = (-3, 4, -4, 3)
    Z2 = np.zeros((150, 150))
    ny, nx = Z.shape
    Z2[30:30+ny, 30:30+nx] = Z

    ax.imshow(Z2, extent=extent, interpolation="nearest",
              origin="lower")

    # creating our inset axes with a bbox_transform parameter
    axins = inset_axes(ax, width=1., height=1., bbox_to_anchor=(1, 1),
                       bbox_transform=ax.transAxes)

    axins.imshow(Z2, extent=extent, interpolation="nearest",
                 origin="lower")
    axins.yaxis.get_major_locator().set_params(nbins=7)
    axins.xaxis.get_major_locator().set_params(nbins=7)
    # sub region of the original image
    x1, x2, y1, y2 = -1.5, -0.9, -2.5, -1.9
    axins.set_xlim(x1, x2)
    axins.set_ylim(y1, y2)

    plt.xticks(visible=False)
    plt.yticks(visible=False)

    # draw a bbox of the region of the inset axes in the parent axes and
    # connecting lines between the bbox and the inset axes area
    mark_inset(ax, axins, loc1=2, loc2=4, fc="none", ec="0.5")

    asb = AnchoredSizeBar(ax.transData,
                          0.5,
                          '0.5',
                          loc='lower center',
                          pad=0.1, borderpad=0.5, sep=5,
                          frameon=False)
    ax.add_artist(asb)


def test_inset_axes_complete():
    dpi = 100
    figsize = (6, 5)
    fig, ax = plt.subplots(figsize=figsize, dpi=dpi)
    fig.subplots_adjust(.1, .1, .9, .9)

    ins = inset_axes(ax, width=2., height=2., borderpad=0)
    fig.canvas.draw()
    assert_array_almost_equal(
        ins.get_position().extents,
        [(0.9*figsize[0]-2.)/figsize[0], (0.9*figsize[1]-2.)/figsize[1],
         0.9, 0.9])

    ins = inset_axes(ax, width="40%", height="30%", borderpad=0)
    fig.canvas.draw()
    assert_array_almost_equal(
        ins.get_position().extents, [.9-.8*.4, .9-.8*.3, 0.9, 0.9])

    ins = inset_axes(ax, width=1., height=1.2, bbox_to_anchor=(200, 100),
                     loc=3, borderpad=0)
    fig.canvas.draw()
    assert_array_almost_equal(
        ins.get_position().extents,
        [200/dpi/figsize[0], 100/dpi/figsize[1],
         (200/dpi+1)/figsize[0], (100/dpi+1.2)/figsize[1]])

    ins1 = inset_axes(ax, width="35%", height="60%", loc=3, borderpad=1)
    ins2 = inset_axes(ax, width="100%", height="100%",
                      bbox_to_anchor=(0, 0, .35, .60),
                      bbox_transform=ax.transAxes, loc=3, borderpad=1)
    fig.canvas.draw()
    assert_array_equal(ins1.get_position().extents,
                       ins2.get_position().extents)

    with pytest.raises(ValueError):
        ins = inset_axes(ax, width="40%", height="30%",
                         bbox_to_anchor=(0.4, 0.5))

    with pytest.warns(UserWarning):
        ins = inset_axes(ax, width="40%", height="30%",
                         bbox_transform=ax.transAxes)


def test_inset_axes_tight():
    # gh-26287 found that inset_axes raised with bbox_inches=tight
    fig, ax = plt.subplots()
    inset_axes(ax, width=1.3, height=0.9)

    f = io.BytesIO()
    fig.savefig(f, bbox_inches="tight")


@image_comparison(['fill_facecolor.png'], remove_text=True, style='mpl20')
def test_fill_facecolor():
    fig, ax = plt.subplots(1, 5)
    fig.set_size_inches(5, 5)
    for i in range(1, 4):
        ax[i].yaxis.set_visible(False)
    ax[4].yaxis.tick_right()
    bbox = Bbox.from_extents(0, 0.4, 1, 0.6)

    # fill with blue by setting 'fc' field
    bbox1 = TransformedBbox(bbox, ax[0].transData)
    bbox2 = TransformedBbox(bbox, ax[1].transData)
    # set color to BboxConnectorPatch
    p = BboxConnectorPatch(
        bbox1, bbox2, loc1a=1, loc2a=2, loc1b=4, loc2b=3,
        ec="r", fc="b")
    p.set_clip_on(False)
    ax[0].add_patch(p)
    # set color to marked area
    axins = zoomed_inset_axes(ax[0], 1, loc='upper right')
    axins.set_xlim(0, 0.2)
    axins.set_ylim(0, 0.2)
    plt.gca().axes.xaxis.set_ticks([])
    plt.gca().axes.yaxis.set_ticks([])
    mark_inset(ax[0], axins, loc1=2, loc2=4, fc="b", ec="0.5")

    # fill with yellow by setting 'facecolor' field
    bbox3 = TransformedBbox(bbox, ax[1].transData)
    bbox4 = TransformedBbox(bbox, ax[2].transData)
    # set color to BboxConnectorPatch
    p = BboxConnectorPatch(
        bbox3, bbox4, loc1a=1, loc2a=2, loc1b=4, loc2b=3,
        ec="r", facecolor="y")
    p.set_clip_on(False)
    ax[1].add_patch(p)
    # set color to marked area
    axins = zoomed_inset_axes(ax[1], 1, loc='upper right')
    axins.set_xlim(0, 0.2)
    axins.set_ylim(0, 0.2)
    plt.gca().axes.xaxis.set_ticks([])
    plt.gca().axes.yaxis.set_ticks([])
    mark_inset(ax[1], axins, loc1=2, loc2=4, facecolor="y", ec="0.5")

    # fill with green by setting 'color' field
    bbox5 = TransformedBbox(bbox, ax[2].transData)
    bbox6 = TransformedBbox(bbox, ax[3].transData)
    # set color to BboxConnectorPatch
    p = BboxConnectorPatch(
        bbox5, bbox6, loc1a=1, loc2a=2, loc1b=4, loc2b=3,
        ec="r", color="g")
    p.set_clip_on(False)
    ax[2].add_patch(p)
    # set color to marked area
    axins = zoomed_inset_axes(ax[2], 1, loc='upper right')
    axins.set_xlim(0, 0.2)
    axins.set_ylim(0, 0.2)
    plt.gca().axes.xaxis.set_ticks([])
    plt.gca().axes.yaxis.set_ticks([])
    mark_inset(ax[2], axins, loc1=2, loc2=4, color="g", ec="0.5")

    # fill with green but color won't show if set fill to False
    bbox7 = TransformedBbox(bbox, ax[3].transData)
    bbox8 = TransformedBbox(bbox, ax[4].transData)
    # BboxConnectorPatch won't show green
    p = BboxConnectorPatch(
        bbox7, bbox8, loc1a=1, loc2a=2, loc1b=4, loc2b=3,
        ec="r", fc="g", fill=False)
    p.set_clip_on(False)
    ax[3].add_patch(p)
    # marked area won't show green
    axins = zoomed_inset_axes(ax[3], 1, loc='upper right')
    axins.set_xlim(0, 0.2)
    axins.set_ylim(0, 0.2)
    axins.xaxis.set_ticks([])
    axins.yaxis.set_ticks([])
    mark_inset(ax[3], axins, loc1=2, loc2=4, fc="g", ec="0.5", fill=False)


# Update style when regenerating the test image
@image_comparison(['zoomed_axes.png', 'inverted_zoomed_axes.png'],
                  style=('classic', '_classic_test_patch'),
                  tol=0 if platform.machine() == 'x86_64' else 0.02)
def test_zooming_with_inverted_axes():
    fig, ax = plt.subplots()
    ax.plot([1, 2, 3], [1, 2, 3])
    ax.axis([1, 3, 1, 3])
    inset_ax = zoomed_inset_axes(ax, zoom=2.5, loc='lower right')
    inset_ax.axis([1.1, 1.4, 1.1, 1.4])

    fig, ax = plt.subplots()
    ax.plot([1, 2, 3], [1, 2, 3])
    ax.axis([3, 1, 3, 1])
    inset_ax = zoomed_inset_axes(ax, zoom=2.5, loc='lower right')
    inset_ax.axis([1.4, 1.1, 1.4, 1.1])


# Update style when regenerating the test image
@image_comparison(['anchored_direction_arrows.png'],
                  tol=0 if platform.machine() == 'x86_64' else 0.01,
                  style=('classic', '_classic_test_patch'))
def test_anchored_direction_arrows():
    fig, ax = plt.subplots()
    ax.imshow(np.zeros((10, 10)), interpolation='nearest')

    simple_arrow = AnchoredDirectionArrows(ax.transAxes, 'X', 'Y')
    ax.add_artist(simple_arrow)


# Update style when regenerating the test image
@image_comparison(['anchored_direction_arrows_many_args.png'],
                  style=('classic', '_classic_test_patch'))
def test_anchored_direction_arrows_many_args():
    fig, ax = plt.subplots()
    ax.imshow(np.ones((10, 10)))

    direction_arrows = AnchoredDirectionArrows(
            ax.transAxes, 'A', 'B', loc='upper right', color='red',
            aspect_ratio=-0.5, pad=0.6, borderpad=2, frameon=True, alpha=0.7,
            sep_x=-0.06, sep_y=-0.08, back_length=0.1, head_width=9,
            head_length=10, tail_width=5)
    ax.add_artist(direction_arrows)


def test_axes_locatable_position():
    fig, ax = plt.subplots()
    divider = make_axes_locatable(ax)
    with mpl.rc_context({"figure.subplot.wspace": 0.02}):
        cax = divider.append_axes('right', size='5%')
    fig.canvas.draw()
    assert np.isclose(cax.get_position(original=False).width,
                      0.03621495327102808)


@image_comparison(['image_grid_each_left_label_mode_all.png'], style='mpl20',
                  savefig_kwarg={'bbox_inches': 'tight'})
def test_image_grid_each_left_label_mode_all():
    imdata = np.arange(100).reshape((10, 10))

    fig = plt.figure(1, (3, 3))
    grid = ImageGrid(fig, (1, 1, 1), nrows_ncols=(3, 2), axes_pad=(0.5, 0.3),
                     cbar_mode="each", cbar_location="left", cbar_size="15%",
                     label_mode="all")
    # 3-tuple rect => SubplotDivider
    assert isinstance(grid.get_divider(), SubplotDivider)
    assert grid.get_axes_pad() == (0.5, 0.3)
    assert grid.get_aspect()  # True by default for ImageGrid
    for ax, cax in zip(grid, grid.cbar_axes):
        im = ax.imshow(imdata, interpolation='none')
        cax.colorbar(im)


@image_comparison(['image_grid_single_bottom_label_mode_1.png'], style='mpl20',
                  savefig_kwarg={'bbox_inches': 'tight'})
def test_image_grid_single_bottom():
    imdata = np.arange(100).reshape((10, 10))

    fig = plt.figure(1, (2.5, 1.5))
    grid = ImageGrid(fig, (0, 0, 1, 1), nrows_ncols=(1, 3),
                     axes_pad=(0.2, 0.15), cbar_mode="single", cbar_pad=0.3,
                     cbar_location="bottom", cbar_size="10%", label_mode="1")
    # 4-tuple rect => Divider, isinstance will give True for SubplotDivider
    assert type(grid.get_divider()) is Divider
    for i in range(3):
        im = grid[i].imshow(imdata, interpolation='none')
    grid.cbar_axes[0].colorbar(im)


def test_image_grid_label_mode_invalid():
    fig = plt.figure()
    with pytest.raises(ValueError, match="'foo' is not a valid value for mode"):
        ImageGrid(fig, (0, 0, 1, 1), (2, 1), label_mode="foo")


@image_comparison(['image_grid.png'],
                  remove_text=True, style='mpl20',
                  savefig_kwarg={'bbox_inches': 'tight'})
def test_image_grid():
    # test that image grid works with bbox_inches=tight.
    im = np.arange(100).reshape((10, 10))

    fig = plt.figure(1, (4, 4))
    grid = ImageGrid(fig, 111, nrows_ncols=(2, 2), axes_pad=0.1)
    assert grid.get_axes_pad() == (0.1, 0.1)
    for i in range(4):
        grid[i].imshow(im, interpolation='nearest')


def test_gettightbbox():
    fig, ax = plt.subplots(figsize=(8, 6))

    l, = ax.plot([1, 2, 3], [0, 1, 0])

    ax_zoom = zoomed_inset_axes(ax, 4)
    ax_zoom.plot([1, 2, 3], [0, 1, 0])

    mark_inset(ax, ax_zoom, loc1=1, loc2=3, fc="none", ec='0.3')

    remove_ticks_and_titles(fig)
    bbox = fig.get_tightbbox(fig.canvas.get_renderer())
    np.testing.assert_array_almost_equal(bbox.extents,
                                         [-17.7, -13.9, 7.2, 5.4])


@pytest.mark.parametrize("click_on", ["big", "small"])
@pytest.mark.parametrize("big_on_axes,small_on_axes", [
    ("gca", "gca"),
    ("host", "host"),
    ("host", "parasite"),
    ("parasite", "host"),
    ("parasite", "parasite")
])
def test_picking_callbacks_overlap(big_on_axes, small_on_axes, click_on):
    """Test pick events on normal, host or parasite axes."""
    # Two rectangles are drawn and "clicked on", a small one and a big one
    # enclosing the small one. The axis on which they are drawn as well as the
    # rectangle that is clicked on are varied.
    # In each case we expect that both rectangles are picked if we click on the
    # small one and only the big one is picked if we click on the big one.
    # Also tests picking on normal axes ("gca") as a control.
    big = plt.Rectangle((0.25, 0.25), 0.5, 0.5, picker=5)
    small = plt.Rectangle((0.4, 0.4), 0.2, 0.2, facecolor="r", picker=5)
    # Machinery for "receiving" events
    received_events = []
    def on_pick(event):
        received_events.append(event)
    plt.gcf().canvas.mpl_connect('pick_event', on_pick)
    # Shortcut
    rectangles_on_axes = (big_on_axes, small_on_axes)
    # Axes setup
    axes = {"gca": None, "host": None, "parasite": None}
    if "gca" in rectangles_on_axes:
        axes["gca"] = plt.gca()
    if "host" in rectangles_on_axes or "parasite" in rectangles_on_axes:
        axes["host"] = host_subplot(111)
        axes["parasite"] = axes["host"].twin()
    # Add rectangles to axes
    axes[big_on_axes].add_patch(big)
    axes[small_on_axes].add_patch(small)
    # Simulate picking with click mouse event
    if click_on == "big":
        click_axes = axes[big_on_axes]
        axes_coords = (0.3, 0.3)
    else:
        click_axes = axes[small_on_axes]
        axes_coords = (0.5, 0.5)
    # In reality mouse events never happen on parasite axes, only host axes
    if click_axes is axes["parasite"]:
        click_axes = axes["host"]
    (x, y) = click_axes.transAxes.transform(axes_coords)
    m = MouseEvent("button_press_event", click_axes.get_figure(root=True).canvas, x, y,
                   button=1)
    click_axes.pick(m)
    # Checks
    expected_n_events = 2 if click_on == "small" else 1
    assert len(received_events) == expected_n_events
    event_rects = [event.artist for event in received_events]
    assert big in event_rects
    if click_on == "small":
        assert small in event_rects


@image_comparison(['anchored_artists.png'], remove_text=True, style='mpl20')
def test_anchored_artists():
    fig, ax = plt.subplots(figsize=(3, 3))
    ada = AnchoredDrawingArea(40, 20, 0, 0, loc='upper right', pad=0.,
                              frameon=False)
    p1 = Circle((10, 10), 10)
    ada.drawing_area.add_artist(p1)
    p2 = Circle((30, 10), 5, fc="r")
    ada.drawing_area.add_artist(p2)
    ax.add_artist(ada)

    box = AnchoredAuxTransformBox(ax.transData, loc='upper left')
    el = Ellipse((0, 0), width=0.1, height=0.4, angle=30, color='cyan')
    box.drawing_area.add_artist(el)
    ax.add_artist(box)

    # This block used to test the AnchoredEllipse class, but that was removed. The block
    # remains, though it duplicates the above ellipse, so that the test image doesn't
    # need to be regenerated.
    box = AnchoredAuxTransformBox(ax.transData, loc='lower left', frameon=True,
                                  pad=0.5, borderpad=0.4)
    el = Ellipse((0, 0), width=0.1, height=0.25, angle=-60)
    box.drawing_area.add_artist(el)
    ax.add_artist(box)

    asb = AnchoredSizeBar(ax.transData, 0.2, r"0.2 units", loc='lower right',
                          pad=0.3, borderpad=0.4, sep=4, fill_bar=True,
                          frameon=False, label_top=True, prop={'size': 20},
                          size_vertical=0.05, color='green')
    ax.add_artist(asb)


def test_hbox_divider():
    arr1 = np.arange(20).reshape((4, 5))
    arr2 = np.arange(20).reshape((5, 4))

    fig, (ax1, ax2) = plt.subplots(1, 2)
    ax1.imshow(arr1)
    ax2.imshow(arr2)

    pad = 0.5  # inches.
    divider = HBoxDivider(
        fig, 111,  # Position of combined axes.
        horizontal=[Size.AxesX(ax1), Size.Fixed(pad), Size.AxesX(ax2)],
        vertical=[Size.AxesY(ax1), Size.Scaled(1), Size.AxesY(ax2)])
    ax1.set_axes_locator(divider.new_locator(0))
    ax2.set_axes_locator(divider.new_locator(2))

    fig.canvas.draw()
    p1 = ax1.get_position()
    p2 = ax2.get_position()
    assert p1.height == p2.height
    assert p2.width / p1.width == pytest.approx((4 / 5) ** 2)


def test_vbox_divider():
    arr1 = np.arange(20).reshape((4, 5))
    arr2 = np.arange(20).reshape((5, 4))

    fig, (ax1, ax2) = plt.subplots(1, 2)
    ax1.imshow(arr1)
    ax2.imshow(arr2)

    pad = 0.5  # inches.
    divider = VBoxDivider(
        fig, 111,  # Position of combined axes.
        horizontal=[Size.AxesX(ax1), Size.Scaled(1), Size.AxesX(ax2)],
        vertical=[Size.AxesY(ax1), Size.Fixed(pad), Size.AxesY(ax2)])
    ax1.set_axes_locator(divider.new_locator(0))
    ax2.set_axes_locator(divider.new_locator(2))

    fig.canvas.draw()
    p1 = ax1.get_position()
    p2 = ax2.get_position()
    assert p1.width == p2.width
    assert p1.height / p2.height == pytest.approx((4 / 5) ** 2)


def test_axes_class_tuple():
    fig = plt.figure()
    axes_class = (mpl_toolkits.axes_grid1.mpl_axes.Axes, {})
    gr = AxesGrid(fig, 111, nrows_ncols=(1, 1), axes_class=axes_class)


def test_grid_axes_lists():
    """Test Grid axes_all, axes_row and axes_column relationship."""
    fig = plt.figure()
    grid = Grid(fig, 111, (2, 3), direction="row")
    assert_array_equal(grid, grid.axes_all)
    assert_array_equal(grid.axes_row, np.transpose(grid.axes_column))
    assert_array_equal(grid, np.ravel(grid.axes_row), "row")
    assert grid.get_geometry() == (2, 3)
    grid = Grid(fig, 111, (2, 3), direction="column")
    assert_array_equal(grid, np.ravel(grid.axes_column), "column")


@pytest.mark.parametrize('direction', ('row', 'column'))
def test_grid_axes_position(direction):
    """Test positioning of the axes in Grid."""
    fig = plt.figure()
    grid = Grid(fig, 111, (2, 2), direction=direction)
    loc = [ax.get_axes_locator() for ax in np.ravel(grid.axes_row)]
    # Test nx.
    assert loc[1].args[0] > loc[0].args[0]
    assert loc[0].args[0] == loc[2].args[0]
    assert loc[3].args[0] == loc[1].args[0]
    # Test ny.
    assert loc[2].args[1] < loc[0].args[1]
    assert loc[0].args[1] == loc[1].args[1]
    assert loc[3].args[1] == loc[2].args[1]


@pytest.mark.parametrize('rect, ngrids, error, message', (
    ((1, 1), None, TypeError, "Incorrect rect format"),
    (111, -1, ValueError, "ngrids must be positive"),
    (111, 7, ValueError, "ngrids must be positive"),
))
def test_grid_errors(rect, ngrids, error, message):
    fig = plt.figure()
    with pytest.raises(error, match=message):
        Grid(fig, rect, (2, 3), ngrids=ngrids)


@pytest.mark.parametrize('anchor, error, message', (
    (None, TypeError, "anchor must be str"),
    ("CC", ValueError, "'CC' is not a valid value for anchor"),
    ((1, 1, 1), TypeError, "anchor must be str"),
))
def test_divider_errors(anchor, error, message):
    fig = plt.figure()
    with pytest.raises(error, match=message):
        Divider(fig, [0, 0, 1, 1], [Size.Fixed(1)], [Size.Fixed(1)],
                anchor=anchor)


@check_figures_equal(extensions=["png"])
def test_mark_inset_unstales_viewlim(fig_test, fig_ref):
    inset, full = fig_test.subplots(1, 2)
    full.plot([0, 5], [0, 5])
    inset.set(xlim=(1, 2), ylim=(1, 2))
    # Check that mark_inset unstales full's viewLim before drawing the marks.
    mark_inset(full, inset, 1, 4)

    inset, full = fig_ref.subplots(1, 2)
    full.plot([0, 5], [0, 5])
    inset.set(xlim=(1, 2), ylim=(1, 2))
    mark_inset(full, inset, 1, 4)
    # Manually unstale the full's viewLim.
    fig_ref.canvas.draw()


def test_auto_adjustable():
    fig = plt.figure()
    ax = fig.add_axes([0, 0, 1, 1])
    pad = 0.1
    make_axes_area_auto_adjustable(ax, pad=pad)
    fig.canvas.draw()
    tbb = ax.get_tightbbox()
    assert tbb.x0 == pytest.approx(pad * fig.dpi)
    assert tbb.x1 == pytest.approx(fig.bbox.width - pad * fig.dpi)
    assert tbb.y0 == pytest.approx(pad * fig.dpi)
    assert tbb.y1 == pytest.approx(fig.bbox.height - pad * fig.dpi)


# Update style when regenerating the test image
@image_comparison(['rgb_axes.png'], remove_text=True,
                  style=('classic', '_classic_test_patch'))
def test_rgb_axes():
    fig = plt.figure()
    ax = RGBAxes(fig, (0.1, 0.1, 0.8, 0.8), pad=0.1)
    rng = np.random.default_rng(19680801)
    r = rng.random((5, 5))
    g = rng.random((5, 5))
    b = rng.random((5, 5))
    ax.imshow_rgb(r, g, b, interpolation='none')


# The original version of this test relied on mpl_toolkits's slightly different
# colorbar implementation; moving to matplotlib's own colorbar implementation
# caused the small image comparison error.
@image_comparison(['imagegrid_cbar_mode.png'],
                  remove_text=True, style='mpl20', tol=0.3)
def test_imagegrid_cbar_mode_edge():
    arr = np.arange(16).reshape((4, 4))

    fig = plt.figure(figsize=(18, 9))

    positions = (241, 242, 243, 244, 245, 246, 247, 248)
    directions = ['row']*4 + ['column']*4
    cbar_locations = ['left', 'right', 'top', 'bottom']*2

    for position, direction, location in zip(
            positions, directions, cbar_locations):
        grid = ImageGrid(fig, position,
                         nrows_ncols=(2, 2),
                         direction=direction,
                         cbar_location=location,
                         cbar_size='20%',
                         cbar_mode='edge')
        ax1, ax2, ax3, ax4 = grid

        ax1.imshow(arr, cmap='nipy_spectral')
        ax2.imshow(arr.T, cmap='hot')
        ax3.imshow(np.hypot(arr, arr.T), cmap='jet')
        ax4.imshow(np.arctan2(arr, arr.T), cmap='hsv')

        # In each row/column, the "first" colorbars must be overwritten by the
        # "second" ones.  To achieve this, clear out the axes first.
        for ax in grid:
            ax.cax.cla()
            cb = ax.cax.colorbar(ax.images[0])


def test_imagegrid():
    fig = plt.figure()
    grid = ImageGrid(fig, 111, nrows_ncols=(1, 1))
    ax = grid[0]
    im = ax.imshow([[1, 2]], norm=mpl.colors.LogNorm())
    cb = ax.cax.colorbar(im)
    assert isinstance(cb.locator, mticker.LogLocator)


def test_removal():
    import matplotlib.pyplot as plt
    import mpl_toolkits.axisartist as AA
    fig = plt.figure()
    ax = host_subplot(111, axes_class=AA.Axes, figure=fig)
    col = ax.fill_between(range(5), 0, range(5))
    fig.canvas.draw()
    col.remove()
    fig.canvas.draw()


@image_comparison(['anchored_locator_base_call.png'], style="mpl20")
def test_anchored_locator_base_call():
    fig = plt.figure(figsize=(3, 3))
    fig1, fig2 = fig.subfigures(nrows=2, ncols=1)

    ax = fig1.subplots()
    ax.set(aspect=1, xlim=(-15, 15), ylim=(-20, 5))
    ax.set(xticks=[], yticks=[])

    Z = cbook.get_sample_data("axes_grid/bivariate_normal.npy")
    extent = (-3, 4, -4, 3)

    axins = zoomed_inset_axes(ax, zoom=2, loc="upper left")
    axins.set(xticks=[], yticks=[])

    axins.imshow(Z, extent=extent, origin="lower")


def test_grid_with_axes_class_not_overriding_axis():
    Grid(plt.figure(), 111, (2, 2), axes_class=mpl.axes.Axes)
    RGBAxes(plt.figure(), 111, axes_class=mpl.axes.Axes)