如何拆分DataFrame中含字符串的列？附CSV导入的DataFrame示例

Hey there! It looks like your event_params column is storing JSON-structured strings, and you want to break that into individual columns. Let's walk through how to do this cleanly with pandas.

Step 1: Parse the JSON Strings #

First, we need to convert those JSON strings into Python dictionaries, then expand those dictionaries into separate columns. Here are two reliable approaches:

Approach 1: Using apply() + json.loads()

This method is straightforward for simple JSON structures:

import pandas as pd
import json

# Assume your DataFrame is named df
# Parse each JSON string in event_params into a dictionary, then convert to a DataFrame
parsed_params = df['event_params'].apply(json.loads).apply(pd.Series)

# Merge the parsed columns back into the original DataFrame (dropping the old event_params column)
df_expanded = df.drop('event_params', axis=1).join(parsed_params)

Approach 2: Using pd.json_normalize()

This is pandas' built-in function for flattening JSON data, great if you might have nested structures later:

import pandas as pd
import json

# Convert the JSON strings to a list of dictionaries
params_dicts = df['event_params'].apply(json.loads).tolist()

# Normalize the dictionaries into columns and merge with original data
df_expanded = df.drop('event_params', axis=1).join(pd.json_normalize(params_dicts))

Handling Malformed JSON #

If some rows have truncated or invalid JSON (like the last row in your sample with "maximuma..."), add a safe parsing function to avoid errors:

def safe_parse_json(s):
    try:
        return json.loads(s)
    except json.JSONDecodeError:
        # Return an empty dict if parsing fails to keep the process running
        return {}

# Use the safe function instead of direct json.loads
parsed_params = df['event_params'].apply(safe_parse_json).apply(pd.Series)
df_expanded = df.drop('event_params', axis=1).join(parsed_params)

What You'll Get #

After running either approach, your DataFrame will have new columns for type, maximumangle, and duration (matching the keys in your JSON), all merged with your original columns like ts, employee_id, etc.

内容的提问来源于stack exchange，提问作者ashlock