Pandas: When Does Selecting from a DataFrame Create a View vs. a Copy?

Pandas: Understanding the Rules for Generating Views vs Copies

Confusion in Identifying Views and Copies

Determining whether a selection from a Pandas DataFrame results in a view or a copy can be puzzling. This confusion arises primarily from the variety of indexing operations and their varying behaviors.

General Rules

• All DataFrame operations inherently create copies.
• The inplace=True option, when available, allows for in-place modification.
• Indexers that set values (e.g., .loc, .iloc, .iat, .at) typically perform operations in-place.
• Getting values from a single data type object using an indexer often results in a view (although memory layout can affect this).
• Getting values from a multiple data type object using an indexer always produces a copy.

Specific Examples

• Assigning values to a comparison:

  <code class="python">df[df.C <= df.B] = 7654321</code>

  Copy after login

  This assignment modifies the original DataFrame df because the indexer .loc is used to set values in-place.

• Chained indexing:

  <code class="python">df[df.C <= df.B].loc[:, 'B':'E']</code>

  Copy after login

  This operation is discouraged as it may not be reliable. To avoid potential issues, use the following syntax instead:

  <code class="python">df.loc[df.C <= df.B, 'B':'E']</code>

  Copy after login

Modifying Values Based on a Query

To modify all values in a DataFrame that meet a specific query condition, use the loc indexer with the query condition as an argument. For example:

<code class="python">df.loc[df.C <= df.B, 'E'] = 40</code>

This assignment will only change the values in column 'E' for rows where df.C is less than or equal to df.B.

The above is the detailed content of Pandas: When Does Selecting from a DataFrame Create a View vs. a Copy?. For more information, please follow other related articles on the PHP Chinese website!