如何在水平 Seaborn 条形图上注释文本？

Question

I have the problem that as soon as I want to annotate my horizontal bars there will be some error message:

posx and posy should be finite values

Than I looked into the code and surprisingly I got some nan values which only appear when the hue parameter is used.

Code:

ax = sns.barplot(x="Points", y="Characters", hue="Average Speeds", data=albion_dataset, palette="Set1", dodge=False)
for p in ax.patches:
    width = p.get_width()
    print(width)

Output:
nan
nan
2.57562
nan
nan
nan
nan
1.526325
nan
...

But when I remove the hue option than there is no nan and the annotation works flawless. The dataframe itself has no nan values. How can this be fixed, so that I can use the hue function. The dtypes are floats for x and hue and object for y.

UPDATE: Found a way to annotate the bars, but now the last bar has no annotation text.

i = 0
for p in ax.patches:
    ax.annotate("%.4f" % albion_dataset["Average Speeds"][i], (p.get_x() + p.get_width(), p.get_y() + 1.2),
            xytext=(5, 10), textcoords='offset points')
    print(i)
    i += 1

Furthermore, how can I add the text from the hue legend to the bars because the above code is not considering the order of the hue values. Thus I get wrong values on the bars.

Answer 1

One option is to rely on the width of the patch itself, instead of trying to match the bar to the dataframe:

tips = sns.load_dataset("tips")
tips.loc[(tips.day=="Thur")&(tips.sex=='Female')] = np.nan
tips.loc[(tips.day=="Sat")&(tips.sex=='Male')] = np.nan
ax = sns.barplot(y="day", x="total_bill", hue="sex", data=tips, ci=None)

for p in ax.patches:
    ax.annotate("%.4f" % p.get_width(), xy=(p.get_width(), p.get_y()+p.get_height()/2),
            xytext=(5, 0), textcoords='offset points', ha="left", va="center")

If you need to know the value of the hue (or access the sub-dataframe corresponding to each bar), then I would recommend you explicitly pass an order= and a hue_order= argument, so you know in what order the bars are drawn.

import itertools
tips = sns.load_dataset("tips")
tips.loc[(tips.day=="Thur")&(tips.sex=='Female')] = np.nan
tips.loc[(tips.day=="Sat")&(tips.sex=='Male')] = np.nan

group_col = 'day'
hue_col = 'sex'
order=['Sat','Sun','Thur','Fri']
hue_order = ['Female','Male']

ax = sns.barplot(y=group_col, x="total_bill", hue=hue_col, order=order, hue_order=hue_order, data=tips, ci=None)

for p,(cur_hue, cur_y) in zip(ax.patches,itertools.product(hue_order,order)):
    temp_df = tips.loc[(tips[group_col]==cur_y)&(tips[hue_col]==cur_hue)]
    # temp_df is the sub-dataframe that corresponds to the current bar `p`. It can contain 0 or more rows
    pos = p.get_width() if p.get_width()>0 else 0
    ax.annotate(cur_hue, xy=(pos, p.get_y()+p.get_height()/2),
                xytext=(5, 0), textcoords='offset points', ha="left", va="center")