Python seaborn facetgrid

Question

I have a count of green/orange/red by salesman for each customer : Something like :

    sellor  customer    red green   orange
0   73  c1  5   96  15
1   77  c1  88  18  79
2   97  c1  58  59  71

Which I can construct with :

df = pd.DataFrame()
df["sellor"]   = np.random.randint(0,100, 20)
df["customer"]   = ["c1"]*5 + ["C2"]*5 + ["C3"]*5 + ["c4"]*5
df["red"] = np.random.randint(0,100, 20)
df["green"] = np.random.randint(0,100, 20)
df["orange"] = np.random.randint(0,100, 20)

df.sellor = df.sellor.astype("category")
df.customer = df.customer.astype("category")

Now I want to represent the data as graphs : For now I do :

for customer in df.customer.unique():
    df[df.customer==customer].plot.bar(title=customer)

Which gives me 4 images like the folowing one :

I think I could do the same with seaborn facetgrid, but didn't really find a way. I tried :

sns.barplot(data=df, x="sellor", y="red", hue="customer")

but it gives me only one plot (no subdivision by customer) + It doesn't gives me the 3 bars by sellors :(

How would I use a facetgrid to get the same result as my loop ?

Answer 1

Without using seaborn you may plot your barplots to different subplots of the same figure,

import matplotlib.pyplot as plt

u = df.customer.unique()
fig, axes = plt.subplots(ncols=len(u), figsize=(10,3))

for customer, ax in zip(u, axes):
    df[df.customer==customer].plot.bar(x="sellor", title=customer, ax =ax )

plt.tight_layout()    
plt.show()

You may achieve something similar with seaborn's FacetGrid via

import seaborn as sns
g = sns.FacetGrid(data=df, col="customer", col_order=["c1", "C2", "C3", "c4"])

def plot(*args,**kwargs):
    kwargs["data"].plot(kind="bar", x="sellor", ax=plt.gca())
g.map_dataframe(plot)
plt.show()

Finally, to change the colors used here to match those of the dataframe names you may set the color cycle, e.g.

plt.rcParams["axes.prop_cycle"] = plt.cycler("color", ["red", "green", "orange"])