无需Pandas实现列表嵌套字典结构化及Pivot丢失name问题答疑

1. Yes, You Can Achieve Structured Output Without Pandas #

Let’s walk through a concrete example using basic Python tools like defaultdict and built-in data structures. First, let’s assume your input data looks something like this (adjust to match your actual structure):

data = [
    {'name': 'Alice', 'metric': 'score', 'value': 85},
    {'name': 'Alice', 'metric': 'age', 'value': 30},
    {'name': 'Bob', 'metric': 'score', 'value': 90},
    {'name': 'Charlie', 'metric': 'age', 'value': 25}
]

Here’s how to restructure it:

Step 1: Group Data by Name #

We’ll use collections.defaultdict to group each name’s metrics and values:

from collections import defaultdict

# Group entries by 'name'
grouped_data = defaultdict(dict)
for entry in data:
    name = entry.pop('name')
    metric = entry.pop('metric')
    grouped_data[name][metric] = entry['value']

Step 2: Collect All Unique Columns (Metrics) #

We need to know all possible metrics to ensure every row has the same columns:

# Get all unique metrics to use as columns
all_metrics = sorted({metric for name in grouped_data for metric in grouped_data[name].keys()})

Step 3: Build Structured Rows #

Now we’ll create a list of dictionaries where each entry represents a row with the name and all metrics (filling in None for missing values):

structured_output = []
for name, metrics in grouped_data.items():
    row = {'name': name}
    # Add each metric's value (or None if missing)
    for metric in all_metrics:
        row[metric] = metrics.get(metric, None)
    structured_output.append(row)

The result will look like this:

[
    {'name': 'Alice', 'age': 30, 'score': 85},
    {'name': 'Bob', 'age': None, 'score': 90},
    {'name': 'Charlie', 'age': 25, 'score': None}
]

This matches the structured format you’d get from Pandas, no external libraries required.

2. Why Pandas Pivot Isn’t Working for All Names #

Let’s break down the two issues you mentioned:

Issue 1: Losing the 'name' Field #

When you use df.pivot(index='name', columns='metric', values='value'), the name column becomes the index of the pivoted DataFrame, not a regular column. It’s not lost—just moved. To get it back as a column, add .reset_index() after pivoting:

pivoted_df = df.pivot(index='name', columns='metric', values='value').reset_index()

Issue 2: Not Generating Tables for All Names #

If you want a separate structured table for each individual name, pivot alone won’t do that. You need to group the DataFrame by name first, then apply pivot to each group:

# Group by name and create a pivot table for each group
for name, group in df.groupby('name'):
    # Pivot the group (drop index since each group only has one name)
    name_table = group.pivot(columns='metric', values='value').reset_index(drop=True)
    print(f"Table for {name}:")
    print(name_table)
    print("\n")

This will generate a unique table for every name in your dataset. If some names were missing in your original pivot output, it’s likely because you didn’t set index='name' in the pivot call—without that, Pandas won’t anchor the pivot to each individual name.

内容的提问来源于stack exchange，提问作者AmiB