[Python] Allow creating RecordBatch from Python dict

[asfimport]

There's already Table.to_pydict() and Table.from_pydict() methods , but only RecordBatch.to_pydict(). Perhaps, we should also add RecordBatch.from_pydict() to make the API consistent.

 

Reporter: Andrei Nesterov
Assignee: Kharoly / @kharoc

PRs and other links:

• GitHub Pull Request #10854

_{Note: This issue was originally created as ARROW-13089. Please see the migration documentation for further details.}

[asfimport]

Joris Van den Bossche / @jorisvandenbossche:
Thanks for the issue, agreed that it would be good to add a RecordBatch.from_pydict as well for consistency.

(contributions always welcome!)

[asfimport]

Andrei Nesterov:
@jorisvandenbossche What the reason to have both of these classes? They seem to have identical interfaces. Wouldn't it better to depreciate one of them?

[asfimport]

Joris Van den Bossche / @jorisvandenbossche:
RecordBatch is the more fundamental building block which is defined in the specification and used in the IPC format (https://arrow.apache.org/docs/format/Columnar.html#serialization-and-interprocess-communication-ipc), and so the python RecordBatch class reflects that data structure.

The Table on the other hand is more a convenience class to handle chunked arrays / multiple record batches in a single object (so the field arrays in a RecordBatch are single, contiguous arrays, while in a Table those are chunked arrays)

[asfimport]

Ben Kietzman / @bkietz:
Issue resolved by pull request 10854
#10854