.. currentmodule:: matplotlib.pyplot

This legend guide is an extension of the documentation available at :func:`~matplotlib.pyplot.legend` - please ensure you are familiar with contents of that documentation before proceeding with this guide.

This guide makes use of some common terms, which are documented here for clarity:

.. glossary::

    legend entry
        A legend is made up of one or more legend entries. An entry is made up of
        exactly one key and one label.

    legend key
        The colored/patterned marker to the left of each legend label.

    legend label
        The text which describes the handle represented by the key.

    legend handle
        The original object which is used to generate an appropriate entry in
        the legend.

Calling :func:`legend` with no arguments automatically fetches the legend handles and their associated labels. This functionality is equivalent to:

handles, labels = ax.get_legend_handles_labels()
ax.legend(handles, labels)

The :meth:`~matplotlib.axes.Axes.get_legend_handles_labels` function returns a list of handles/artists which exist on the Axes which can be used to generate entries for the resulting legend - it is worth noting however that not all artists can be added to a legend, at which point a "proxy" will have to be created (see :ref:`proxy_legend_handles` for further details).

For full control of what is being added to the legend, it is common to pass the appropriate handles directly to :func:`legend`:

line_up, = plt.plot([1,2,3], label='Line 2')
line_down, = plt.plot([3,2,1], label='Line 1')
plt.legend(handles=[line_up, line_down])

In some cases, it is not possible to set the label of the handle, so it is possible to pass through the list of labels to :func:`legend`:

line_up, = plt.plot([1,2,3], label='Line 2')
line_down, = plt.plot([3,2,1], label='Line 1')
plt.legend([line_up, line_down], ['Line Up', 'Line Down'])

Not all handles can be turned into legend entries automatically, so it is often necessary to create an artist which can. Legend handles don't have to exists on the Figure or Axes in order to be used.

Suppose we wanted to create a legend which has an entry for some data which is represented by a red color:

.. plot::
    :include-source:

    import matplotlib.patches as mpatches
    import matplotlib.pyplot as plt

    red_patch = mpatches.Patch(color='red', label='The red data')
    plt.legend(handles=[red_patch])

    plt.show()

There are many supported legend handles, instead of creating a patch of color we could have created a line with a marker:

.. plot::
    :include-source:

    import matplotlib.lines as mlines
    import matplotlib.pyplot as plt

    blue_line = mlines.Line2D([], [], color='blue', marker='*',
                              markersize=15, label='Blue stars')
    plt.legend(handles=[blue_line])

    plt.show()

The location of the legend can be specified by the keyword argument loc. Please see the documentation at :func:`legend` for more details.

The bbox_to_anchor keyword gives a great degree of control for manual legend placement. For example, if you want your axes legend located at the figure's top right-hand corner instead of the axes' corner, simply specify the corner's location, and the coordinate system of that location:

plt.legend(bbox_to_anchor=(1, 1),
           bbox_transform=plt.gcf().transFigure)

More examples of custom legend placement:

.. plot:: users/plotting/examples/simple_legend01.py
   :include-source:

Sometimes it is more clear to split legend entries across multiple legends. Whilst the instinctive approach to doing this might be to call the :func:`legend` function multiple times, you will find that only one legend ever exists on the Axes. This has been done so that it is possible to call :func:`legend` repeatedly to update the legend to the latest handles on the Axes, so to persist old legend instances, we must add them manually to the Axes:

.. plot:: users/plotting/examples/simple_legend02.py
   :include-source:

In order to create legend entries, handles are given as an argument to an appropriate :class:`~matplotlib.legend_handler.HandlerBase` subclass. The choice of handler subclass is determined by the following rules:

1. Update :func:`~matplotlib.legend.Legend.get_legend_handler_map` with the value in the handler_map keyword.
2. Check if the handle is in the newly created handler_map.
3. Check if the type of handle is in the newly created handler_map.
4. Check if any of the types in the handle's mro is in the newly created handler_map.

For completeness, this logic is mostly implemented in :func:`~matplotlib.legend.Legend.get_legend_handler`.

All of this flexibility means that we have the necessary hooks to implement custom handlers for our own type of legend key.

The simplest example of using custom handlers is to instantiate one of the existing :class:`~matplotlib.legend_handler.HandlerBase` subclasses. For the sake of simplicity, let's choose :class:`matplotlib.legend_handler.HandlerLine2D` which accepts a numpoints argument (note numpoints is a keyword on the :func:`legend` function for convenience). We can then pass the mapping of instance to Handler as a keyword to legend.

.. plot::
    :include-source:

    import matplotlib.pyplot as plt
    from matplotlib.legend_handler import HandlerLine2D

    line1, = plt.plot([3,2,1], marker='o', label='Line 1')
    line2, = plt.plot([1,2,3], marker='o', label='Line 2')

    plt.legend(handler_map={line1: HandlerLine2D(numpoints=4)})

As you can see, "Line 1" now has 4 marker points, where "Line 2" has 2 (the default). Try the above code, only change the map's key from line1 to type(line1). Notice how now both :class:`~matplotlib.lines.Line2D` instances get 4 markers.

Along with handlers for complex plot types such as errorbars, stem plots and histograms, the default handler_map has a special tuple handler (:class:`~matplotlib.legend_handler.HandlerTuple`) which simply plots the handles on top of one another for each item in the given tuple. The following example demonstrates combining two legend keys on top of one another:

.. plot::
    :include-source:

    import matplotlib.pyplot as plt
    from numpy.random import randn

    z = randn(10)

    red_dot, = plt.plot(z, "ro", markersize=15)
    # Put a white cross over some of the data.
    white_cross, = plt.plot(z[:5], "w+", markeredgewidth=3, markersize=15)

    plt.legend([red_dot, (red_dot, white_cross)], ["Attr A", "Attr A+B"])

The :class:`~matplotlib.legend_handler.HandlerTuple` class can also be used to assign several legend keys to the same entry:

.. plot::
    :include-source:

    import matplotlib.pyplot as plt
    from matplotlib.legend_handler import HandlerLine2D, HandlerTuple

    p1, = plt.plot([1, 2.5, 3], 'r-d')
    p2, = plt.plot([3, 2, 1], 'k-o')

    l = plt.legend([(p1, p2)], ['Two keys'], numpoints=1,
                   handler_map={tuple: HandlerTuple(ndivide=None)})

A custom handler can be implemented to turn any handle into a legend key (handles don't necessarily need to be matplotlib artists). The handler must implement a "legend_artist" method which returns a single artist for the legend to use. Signature details about the "legend_artist" are documented at :meth:`~matplotlib.legend_handler.HandlerBase.legend_artist`.

.. plot::
    :include-source:

    import matplotlib.pyplot as plt
    import matplotlib.patches as mpatches

    class AnyObject(object):
        pass

    class AnyObjectHandler(object):
        def legend_artist(self, legend, orig_handle, fontsize, handlebox):
            x0, y0 = handlebox.xdescent, handlebox.ydescent
            width, height = handlebox.width, handlebox.height
            patch = mpatches.Rectangle([x0, y0], width, height, facecolor='red',
                                       edgecolor='black', hatch='xx', lw=3,
                                       transform=handlebox.get_transform())
            handlebox.add_artist(patch)
            return patch

    plt.legend([AnyObject()], ['My first handler'],
               handler_map={AnyObject: AnyObjectHandler()})

Alternatively, had we wanted to globally accept AnyObject instances without needing to manually set the handler_map keyword all the time, we could have registered the new handler with:

from matplotlib.legend import Legend
Legend.update_default_handler_map({AnyObject: AnyObjectHandler()})

Whilst the power here is clear, remember that there are already many handlers implemented and what you want to achieve may already be easily possible with existing classes. For example, to produce elliptical legend keys, rather than rectangular ones:

.. plot::
    :include-source:

    from matplotlib.legend_handler import HandlerPatch
    import matplotlib.pyplot as plt
    import matplotlib.patches as mpatches


    class HandlerEllipse(HandlerPatch):
        def create_artists(self, legend, orig_handle,
                           xdescent, ydescent, width, height, fontsize, trans):
            center = 0.5 * width - 0.5 * xdescent, 0.5 * height - 0.5 * ydescent
            p = mpatches.Ellipse(xy=center, width=width + xdescent,
                                 height=height + ydescent)
            self.update_prop(p, orig_handle, legend)
            p.set_transform(trans)
            return [p]


    c = mpatches.Circle((0.5, 0.5), 0.25, facecolor="green",
                        edgecolor="red", linewidth=3)
    plt.gca().add_patch(c)

    plt.legend([c], ["An ellipse, not a rectangle"],
               handler_map={mpatches.Circle: HandlerEllipse()})

Here is a non-exhaustive list of the examples available involving legend being used in various ways:

• :ref:`lines_bars_and_markers-scatter_with_legend`
• :ref:`api-legend_demo`
• :ref:`pylab_examples-contourf_hatching`
• :ref:`pylab_examples-figlegend_demo`
• :ref:`pylab_examples-finance_work2`
• :ref:`pylab_examples-scatter_symbol`