从R转Python用dfply模拟dplyr：异名列及多列连接问题求助

Handling Joins with Different Column Names in dfply #

Since dfply only supports single-column joins out of the box, the simplest workaround is to rename one of the columns to match the other before joining. Here's a concrete example:

Say you have two DataFrames with mismatched join columns:

from dfply import *
import pandas as pd

# Sample data
users = pd.DataFrame({"user_id": [1, 2, 3], "full_name": ["Alice Smith", "Bob Jones", "Charlie Brown"]})
orders = pd.DataFrame({"customer_id": [1, 2, 4], "order_total": [99.99, 49.99, 149.99]})

To join them on user_id (from users) and customer_id (from orders), just rename the column first in your dfply chain:

# Rename orders' customer_id to user_id, then do an inner join
joined_data = orders >> rename(user_id = X.customer_id) >> inner_join(users, by="user_id")

This gets the job done, even if it's an extra step.

Multi-Column Joins in dfply #

As you found in the docs, dfply doesn't support native multi-column joins. If you absolutely need to use dfply for this, you could create a composite key by combining the columns into a single new column (e.g., concatenating strings or integers), but this is clunky and not ideal for all data types.

For a smoother experience, it's worth switching to a tool that supports multi-column joins natively.

Top Dplyr-like Python Packages for Full Join Support #

If you want a seamless dplyr-esque workflow with full support for multi-column and mismatched-name joins, these are your best bets:

Siuba #

Siuba is the closest you'll get to dplyr in Python—it copies dplyr's syntax almost exactly and handles all join types with ease. For multi-column joins with different names:

from siuba import _, inner_join
import pandas as pd

# Sample data with two join columns
users = pd.DataFrame({"user_id": [1,2,3], "region": ["North", "South", "East"]})
orders = pd.DataFrame({"cust_id": [1,2,4], "region_code": ["North", "South", "West"]})

# Join on user_id/cust_id AND region/region_code
joined_df = inner_join(users, orders, left_on=["user_id", "region"], right_on=["cust_id", "region_code"])

Pandas (with Chained Pipes) #

You don't even need a third-party package—pandas has all the join functionality you need, and you can use .pipe() to replicate dplyr's chaining style. Multi-column joins are straightforward:

import pandas as pd

users = pd.DataFrame({"user_id": [1,2,3], "name": ["Alice", "Bob", "Charlie"]})
orders = pd.DataFrame({"cust_id": [1,2,4], "order_date": ["2024-01-01", "2024-01-02", "2024-01-03"]})

# Chain operations with parentheses, join on multiple columns
result = (users
          .pipe(lambda df: pd.merge(df, orders, left_on=["user_id"], right_on=["cust_id"], how="left")))

Dplython #

Another package built to mirror dplyr's API. It supports multi-column joins using left_on and right_on parameters with lists:

from dplython import DplyFrame, inner_join
import pandas as pd

users = DplyFrame(pd.DataFrame({"user_id": [1,2,3], "city": ["NYC", "LA", "Chicago"]}))
orders = DplyFrame(pd.DataFrame({"customer_id": [1,2,4], "city_name": ["NYC", "LA", "Houston"]}))

# Join on user_id/customer_id and city/city_name
joined_data = inner_join(users, orders, left_on=["user_id", "city"], right_on=["customer_id", "city_name"])

内容的提问来源于stack exchange，提问作者Murali