Robin Wilson: Easily hiding items from the legend in matplotlib

When producing some graphs for a client recently, I wanted to hide some labels from a legend in matplotlib. I started investigating complex arguments to the plt.legend function, but it turned out that there was a really simple way to do it…

If you start your label for a plot item with an underscore (_) then that item will be hidden from the legend.

For example:

plt.plot(np.random.rand(20), label='Random 1')
plt.plot(np.random.rand(20), label='Random 2')
plt.plot(np.random.rand(20), label='_Hidden label')
plt.legend()

produces a plot like this:

You can see that the third line is hidden from the legend – just because we started its label with an underscore.

I found this particularly useful when I wanted to plot a load of lines in the same colour to show all the data for something, and then highlight a few lines that meant specific things. For example:

for i in range(20):
    plt.plot(np.random.rand(20), label='_Hidden', color='gray', alpha=0.3)
plt.plot(np.random.rand(20), label='Important Line 1')
plt.plot(np.random.rand(20), label='Important Line 2')
plt.legend()

My next step was to do this when plotting from pandas. In this case I had a dataframe that had a column for each line I wanted to plot in the ‘background’, and then a separate dataframe with each of the ‘special’ lines to highlight.

This code will create a couple of example dataframes:

df = pd.DataFrame()

for i in range(20):
    df[f'Data{i}'] = np.random.rand(20)

special = pd.Series(data=np.random.rand(20))

Plotting this produces a legend with all the individual lines showing:

df.plot(color='gray', alpha=0.3)

However, just by changing the column names to start with an underscore you can hide all the entries in the legend. In this example, I actually set one of the columns to a name without an underscore, so that column can be used as a label to represent all of these lines:

cols = ["_" + col for col in df.columns]
cols[0] = 'All other data'
df.columns = cols

Plotting again using exactly the same command as above gives us this – along with some warnings saying that a load of legend items are going to be ignored (in case we accidentally had pandas columns starting with _)

Putting it all together, we can plot both dataframes, with a sensible legend:

ax = df.plot(color='gray', alpha=0.3)
special.plot(ax=ax, label='Special data')
plt.legend()

from Planet Python
via read more