Feature Request: facet-specific legends

After faceting by rows and columns for two variables, and hueing for a third one, some facets might contain only some hue values. It makes little sense then to have one overall FacetGrid legend and diving through it to find the keys for any specific facet. It may be more practical to have the legend of each facet only to show the hue values that appear in that specific facet. This was discussed, and resolved, in a stackoverflow post but it may be nice to make it a feature, possibly to be chosen with an argument of .add_legend() to be passed to the FacetGrid object.

[mwaskom]

I think this statement is incorrect:

It makes little sense then to have one overall FacetGrid legend and diving through it to find the keys for any specific facet.

It makes the same amount of sense as with other uses of FacetGrid: the figure represents the entire dataset, and the figure legend shows the hue values. Putting a legend in each facet also leads to a very cluttered appearance.

Here is another way to get what you want.

g = sns.FacetGrid(tips, row="sex", col="day")
for (row_var, col_var), facet_df in tips.groupby(["sex", "day"]):
    ax = g.axes[g.row_names.index(row_var), g.col_names.index(col_var)]
    sns.swarmplot(x="time", y="total_bill", hue="size", data=facet_df, ax=ax)
    ax.set(xlabel="", ylabel="")
g.set_axis_labels("time", "total_bill")
g.set_titles()

I am going to close this issue, but thanks for the suggestion.

[massimopinto]

Thank you @mwaskom, I understand that there is more than one solution to this (the answer by @ParfaitG to my original question on Overflow was also updated to work in a more general case). With respect to #2046, I first need to figure out how to deploy seaborn after pulling from a branch different from master (such as facet_grid_convenience) without upsetting my conda installation (now at 0.10.1). Using pip after git checkout facet_grid_convenience perhaps?