Hey there! Let's tackle your two Pandas questions one by one:

The most reliable way to do this is using regular expressions with str.extract, which lets you target the exact word and capture everything that follows it. Here's a concrete example:

Suppose you have a DataFrame with a column containing strings like "text: your desired content" and you want to pull out everything after "text:":

import pandas as pd

# Sample DataFrame
df = pd.DataFrame({
    'content': [
        "text: extract this part",
        "random leading text text: another example",
        "no matching keyword here"
    ]
})

# Extract text after "text:" (ignores any spaces right after the keyword)
df['extracted_text'] = df['content'].str.extract(r'text:\s*(.*)', expand=False)

Breakdown of the regex: #

• text:\s* matches the exact keyword "text:" plus any number of spaces that follow it
• (.*) captures all characters after that point (the () creates a capture group that str.extract will return)
• expand=False ensures we get a Series back, which fits perfectly into a new DataFrame column

If you need to target a different word, just replace "text:" in the regex with your desired keyword (e.g., r'keyword:\s*(.*)').

Your error happens because some values in the fields.description.content column are missing values (NaN). In Pandas, NaNs are stored as float objects, and you can't use dictionary-style indexing (x['text']) on a float.

Here are two simple fixes:

Option 1: Use str.get() (cleanest approach) #

Pandas has a built-in str.get() method that safely pulls values from dictionary-like objects, and returns NaN for non-dictionary values (like missing data):

issues_df['new_column'] = issues_df['fields.description.content'].str.get('text')

Option 2: Add a type check in your lambda #

If you prefer using apply, add a check to make sure x is a dictionary before accessing its 'text' key:

import pandas as pd

issues_df['new_column'] = issues_df['fields.description.content'].apply(
    lambda x: x['text'] if isinstance(x, dict) else pd.NA
)

Both approaches will handle missing values gracefully without throwing errors. The first option is more concise and idiomatic for Pandas, so it's my go-to recommendation.

内容的提问来源于stack exchange，提问作者Junior P