如何在Pandas中实现groupby分组内排序并保留聚合结果，同时按store实现facet分面可视化？

Got it, let's break down how to build those store-specific facet plots with your already aggregated and sorted data. First, we need to tweak the data structure a bit since your df_2 has multi-level columns/indexes, which can throw off plotting libraries.

Step 1: Clean Up the Data Format #

First, let's flatten that multi-level column and convert the index into regular columns—this makes plotting way easier:

# Flatten the multi-level column name ('sales', 'sum') to a simple 'sales_sum'
df_2.columns = ['sales_sum']
# Turn the multi-level index into regular columns
df_plot = df_2.reset_index()

Step 2: Use Seaborn FacetGrid for Easy Faceting #

Seaborn's FacetGrid is perfect for this kind of per-group visualization. It'll automatically create a subplot for each store, and we can keep your pre-sorted sales order intact:

import seaborn as sns
import matplotlib.pyplot as plt

# Set a clean plot style
sns.set_style("whitegrid")

# Create the facet grid—one subplot per store, stacked vertically
g = sns.FacetGrid(df_plot, col="store", col_wrap=1, height=4, aspect=2)

# Map a barplot to each subplot, using our pre-sorted data
g.map(sns.barplot, "product", "sales_sum")

# Add labels and titles to make it readable
g.set_axis_labels("Product", "Total Sales")
g.set_titles("Store: {col_name}")

# Adjust layout so nothing gets cut off
plt.tight_layout()
plt.show()

Pro Tip: Enforce Group-Specific Sorting #

If you want to make 100% sure each store's products are sorted by sales descending (in case the global product order doesn't match each store's local order), use map_dataframe with a custom plotting function:

g = sns.FacetGrid(df_plot, col="store", col_wrap=1, height=4, aspect=2)

def plot_sorted_bars(data, **kwargs):
    # Sort the store's data again (just to be safe)
    sorted_data = data.sort_values('sales_sum', ascending=False)
    sns.barplot(x="product", y="sales_sum", data=sorted_data, **kwargs)

g.map_dataframe(plot_sorted_bars)

# Add labels/titles as before
g.set_axis_labels("Product", "Total Sales")
g.set_titles("Store: {col_name}")
plt.tight_layout()
plt.show()

Alternative: Manual Plotting with Matplotlib #

If you prefer full control with Matplotlib, you can loop through each store and create subplots manually:

stores = df_plot['store'].unique()
# Create a grid of subplots—one row per store
fig, axes = plt.subplots(nrows=len(stores), ncols=1, figsize=(10, 4*len(stores)))

for ax, store in zip(axes, stores):
    # Grab data for the current store and sort it
    store_data = df_plot[df_plot['store'] == store].sort_values('sales_sum', ascending=False)
    # Plot the bar chart
    ax.bar(store_data['product'], store_data['sales_sum'])
    # Add titles and labels
    ax.set_title(f"Store: {store}")
    ax.set_xlabel("Product")
    ax.set_ylabel("Total Sales")

plt.tight_layout()
plt.show()

Either approach will give you the facet plots you're looking for—each store gets its own subplot showing total sales per product, sorted from highest to lowest.

内容的提问来源于stack exchange，提问作者Napier