[ARROW-3829] [Python] Support protocols to extract Arrow objects from third-party classes

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 0.15.0
• Component/s: Python
• Labels: pull-request-available
• External issue URL: https://github.com/apache/arrow/issues/20293
• Assignee: Joris Van den Bossche
• Reporter: Uwe Korn
• Created: 17/Nov/18 18:42
• Updated: 11/Jan/23 07:30
• Resolved: 27/Aug/19 23:00

Description

In the style of NumPy's _array_, we should be able to ask inputs to pa.array, pa.Table.from_X, ... whether they can convert themselves to Arrow objects. This would allow for example to turn objects that hold an Arrow object internally to expose them directly instead of going a conversion path.

My current use case involves Pandas ExtensionArray instances that internally have Arrow objects and should be reused when we pass the whole DataFrame to pa.Table.from_pandas.

Issue Links

• blocks: ARROW-5379 [Python] support pandas' nullable Integer type in from_pandas
• duplicates: ARROW-5271 [Python] Interface for converting pandas ExtensionArray / other custom array objects to pyarrow Array
• links to: GitHub Pull Request #5106