Got it, dealing with a 2000-column dataset where IPython truncates your output is super frustrating. Let’s break down exactly how to get the counts you need and identify non-empty columns clearly:

1. 快速统计空列与非空列的数量 #

First, let’s calculate the number of fully empty columns (where every row is null) and non-empty columns in one go:

import pandas as pd

# 先计算每列的空值总数
null_counts = df.isnull().sum()

# 总行数
total_rows = len(df)

# 统计完全空的列数（空值数等于总行数）
empty_columns_count = (null_counts == total_rows).sum()
# 非空列数就是总列数减去空列数
non_empty_columns_count = len(df.columns) - empty_columns_count

print(f"完全空的列数量: {empty_columns_count}")
print(f"非空列数量: {non_empty_columns_count}")

2. 获取所有非空列的列表（方便后续处理） #

If you want to actually get the names of non-empty columns (or create a new dataset with only those columns), use this:

# 筛选出空值数小于总行数的列（也就是至少有一个非空值的列）
non_empty_columns = null_counts[null_counts < total_rows].index.tolist()

# 直接生成只包含非空列的数据集
filtered_df = df[non_empty_columns]

3. 解决IPython输出截断问题（查看完整空值统计） #

If you just want to see the full df.isnull().sum() output without truncation, tweak Pandas’ display settings:

# 取消行数和列数的显示限制
pd.set_option('display.max_rows', None)
pd.set_option('display.max_columns', None)

# 现在打印就能看到所有2000列的空值统计了
print(null_counts)

This should cover everything you need—whether you just want the counts, the actual column names, or to view the full output without truncation.

内容的提问来源于stack exchange，提问作者CathyQian