Jan 14, 2019 · Dec 25, 2018 · Dec 25, 2018 · Dec 25, 2018 · Dec 25, 2018 · Dec 29, 2018
diff --git a/lib/matplotlib/contour.py b/lib/matplotlib/contour.py
            are not given explicitly via *levels*.
            Defaults to `~.ticker.MaxNLocator`.

        extend : {'neither', 'both', 'min', 'max'}, optional
            Unless this is 'neither', contour levels are automatically
            added to one or both ends of the range so that all data
            are included. These added ranges are then mapped to the
            special colormap values which default to the ends of the
            colormap range, but can be set via
            :meth:`matplotlib.colors.Colormap.set_under` and
            :meth:`matplotlib.colors.Colormap.set_over` methods.
        extend : {'neither', 'both', 'min', 'max'}, optional, default: \
 'neither'
            Determines the ``contourf``-coloring of values that are outside the
            *levels* range.

            If 'neither', values outside the *levels* range are not colored.
            If 'min', 'max' or 'both', color the values below, above or below
            and above the *levels* range.

            Values below ``min(levels)`` and above ``max(levels)`` are mapped
            to the under/over values of the `.Colormap`. Note, that most
            colormaps do not have dedicated colors for these by default, so
            that the over and under values are the edge values of the colormap.
            You may want to set these values explicitly using
            `.Colormap.set_under` and `.Colormap.set_over`.

            .. note::

                An exising `.QuadContourSet` does not get notified if
                properties of its colormap are changed. Therefore, an explicit
                call `.QuadContourSet.changed()` is needed after modifying the
                colormap. The explicit call can be left out, if a colorbar is
                assigned to the `.QuadContourSet` because it interally calls
                `.QuadContourSet.changed()`.

            Example::

                x = np.arange(1, 10)
                y = x.reshape(-1, 1)
                h = x * y

                cs = plt.contourf(h, levels=[10, 30, 50],
                    colors=['#808080', '#A0A0A0', '#C0C0C0'], extend='both')
                cs.cmap.set_over('red')
                cs.cmap.set_under('blue')
                cs.changed()

        xunits, yunits : registered units, optional
            Override axis units by specifying an instance of a
Original file line number	Diff line number	Diff line change
Expand Up		@@ -1766,14 +1766,42 @@ def _initialize_x_y(self, z):
		are not given explicitly via levels.
		Defaults to `~.ticker.MaxNLocator`.

		extend : {'neither', 'both', 'min', 'max'}, optional
		Unless this is 'neither', contour levels are automatically
		added to one or both ends of the range so that all data
		are included. These added ranges are then mapped to the
		special colormap values which default to the ends of the
		colormap range, but can be set via
		:meth:`matplotlib.colors.Colormap.set_under` and
		:meth:`matplotlib.colors.Colormap.set_over` methods.
		extend : {'neither', 'both', 'min', 'max'}, optional, default: \
		'neither'
		Determines the ``contourf``-coloring of values that are outside the
		levels range.

		If 'neither', values outside the levels range are not colored.
		If 'min', 'max' or 'both', color the values below, above or below
		and above the levels range.

		Values below ``min(levels)`` and above ``max(levels)`` are mapped
		to the under/over values of the `.Colormap`. Note, that most
		colormaps do not have dedicated colors for these by default, so
		that the over and under values are the edge values of the colormap.
		You may want to set these values explicitly using
		`.Colormap.set_under` and `.Colormap.set_over`.

		.. note::

		An exising `.QuadContourSet` does not get notified if
		properties of its colormap are changed. Therefore, an explicit
		call `.QuadContourSet.changed()` is needed after modifying the
		colormap. The explicit call can be left out, if a colorbar is
		assigned to the `.QuadContourSet` because it interally calls
		`.QuadContourSet.changed()`.

		Example::

		x = np.arange(1, 10)
		y = x.reshape(-1, 1)
		h = x * y

		cs = plt.contourf(h, levels=[10, 30, 50],
		colors=['#808080', '#A0A0A0', '#C0C0C0'], extend='both')
		cs.cmap.set_over('red')
Copy link Member tacaswellDec 25, 2018 Choose a reason for hiding this comment The reason will be displayed to describe this comment to others.Learn more. It is probably better to show updating the color map before creating the contour and then passing that in to`contourf`. I think that is a bit for idiomatic and does not expose this updating edge-case in the examples. Copy link Member tacaswellDec 25, 2018 Choose a reason for hiding this comment The reason will be displayed to describe this comment to others.Learn more. Ah, the point is we are creating the colormap on-the-fly so can not do that. We have not noticed the lack of update on images because we unconditionally consult the colormap every time we draw. Copy link MemberAuthor timhoffmDec 25, 2018• edited Loading Choose a reason for hiding this comment The reason will be displayed to describe this comment to others.Learn more. Updating a`QuadContourSet` on each draw (to get automatic updates and the same behavior as with images) would essentially mean that we have to run the full code of`changed()`. Not quite sure if we want to do this. I'm just being defensive here and clearly document the current behavior. Copy link Member tacaswellDec 29, 2018 Choose a reason for hiding this comment The reason will be displayed to describe this comment to others.Learn more. I agree this is the conservative path here.
		cs.cmap.set_under('blue')
		cs.changed()

		xunits, yunits : registered units, optional
		Override axis units by specifying an instance of a
Expand Down