Pandas数据处理：当分组结果元素数大于1时删除对应零值行

Got it, let's tackle this problem step by step. You want to filter your original DataFrame based on grouped results: when a group (grouped by id_client and date) has multiple count values (forming a list after your apply(list) command), you need to remove rows where count is 0; if the group only has a single count value, keep that row no matter if it's 0 or not.

Method 1: Direct Group Filtering (Most Efficient) #

You don't even need to generate the list_count intermediate DataFrame to achieve this. We can use groupby().filter() to handle the logic directly on the original data:

First, let's recreate your sample DataFrame for testing:

import pandas as pd

# Original sample data
data = {
    'id_client': [908, 908, 907, 907, 909, 910],
    'date': ['01/2020'] * 6,
    'count': [0, 35, 0, 37, 50, 0]
}
df = pd.DataFrame(data)

Now apply the filtering logic:

# Filter groups based on their size and count values
filtered_df = df.groupby(['id_client', 'date']).filter(
    lambda group: True if len(group) == 1 else group['count'] != 0
)

# Reset index (optional but cleaner)
filtered_df = filtered_df.reset_index(drop=True)
print(filtered_df)

Explanation:

• groupby(['id_client', 'date']).filter() applies the lambda function to each group. Rows where the function returns True are kept.
• For groups with only 1 row (equivalent to your single-element grouping result), we keep all rows (True).
• For groups with multiple rows (equivalent to your list-based grouping result), we only keep rows where count != 0.

Expected Output:

id_client     date  count
0        908  01/2020     35
1        907  01/2020     37
2        909  01/2020     50
3        910  01/2020      0

Method 2: Using Your Intermediate list_count DataFrame #

If you want to work with the list_count DataFrame you generated, you can merge it back to the original data and apply filtering:

First, generate your list_count as you did:

list_count = df.groupby(['id_client', 'date'])['count'].apply(list).reset_index()

Then merge and filter:

# Merge group info into original DataFrame
df_with_groups = df.merge(
    list_count.rename(columns={'count': 'group_values'}),
    on=['id_client', 'date']
)

# Apply filtering logic
filtered_df = df_with_groups[
    # Keep rows where group is a single value OR count is not 0
    (df_with_groups['group_values'].apply(lambda x: not isinstance(x, list) or len(x) == 1))
    | (df_with_groups['count'] != 0)
].drop(columns=['group_values']).reset_index(drop=True)

This will give you the same expected output as Method 1.

内容的提问来源于stack exchange，提问作者tamo007