无需Try Except：Pandas列类型转换报错排查及字符串检测问询

Absolutely! You don’t need try-except blocks to detect non-date strings in your object column or handle datetime conversion errors in Pandas. Here are a few straightforward, efficient approaches:

Pandas built-in pd.to_datetime has an errors='coerce' parameter that turns unconvertible values into NaT (Not a Time) instead of throwing errors. This lets you easily spot non-date strings and handle conversion in one go:

import pandas as pd

# Sample DataFrame with mixed dates and non-date strings
df = pd.DataFrame({
    'transaction_date': ['2023-11-05', '2023/11/06', 'invalid_text', '2023-11-07', '11-08-2023']
})

# Step 1: Identify rows with non-date strings
invalid_mask = pd.to_datetime(df['transaction_date'], errors='coerce').isna()
invalid_rows = df[invalid_mask]
print("Rows containing non-date strings:")
print(invalid_rows)

# Step 2: Convert valid entries to datetime, keep NaT for invalid ones
df['transaction_date_parsed'] = pd.to_datetime(df['transaction_date'], errors='coerce')

This method works for most common date formats (ISO, US-style, European-style) without writing custom logic.

If you know your dates follow a strict format (e.g., YYYY-MM-DD), you can use regex to filter out non-matching strings. Note: This only works for your specific format, so it’s less flexible than the pd.to_datetime approach.

import re

# Regex pattern for YYYY-MM-DD format
date_regex = r'^\d{4}-\d{2}-\d{2}$'

# Flag rows that don't match the pattern
df['is_invalid'] = ~df['transaction_date'].str.match(date_regex, na=False)

# Filter invalid rows
invalid_rows = df[df['is_invalid']]
print("Non-date strings (regex check):")
print(invalid_rows)

The na=False argument ensures missing values (if any) are flagged as invalid too.

If you want to keep the original string for invalid entries instead of NaT, you can use numpy.where to conditionally set values:

import numpy as np

# Create parsed column: datetime for valid entries, original string for invalid
df['date_or_original'] = np.where(
    pd.to_datetime(df['transaction_date'], errors='coerce').notna(),
    pd.to_datetime(df['transaction_date']),
    df['transaction_date']
)

This way you retain both valid dates (as datetime type) and the original invalid strings for further inspection.

内容的提问来源于stack exchange，提问作者uckocaman