Add Arrow IPC stream iterator for direct access to Arrow buffer

[jadewang-db]

Summary

This PR adds support for accessing raw Arrow IPC (Inter-Process Communication) streams
directly, providing an alternative to the existing parsed Arrow record interface. This
allows users to work with the Arrow binary format without the overhead of parsing,
enabling use cases like streaming data to other systems or custom processing pipelines.

Changes

• Refactored internal batch iterator to expose IPC streams:

  • Renamed BatchIterator → IPCStreamIterator (returns io.Reader)
  • Updated implementations to return raw IPC streams instead of parsed batches
  • Created backward-compatible BatchIterator wrapper

• Added public API in rows package:

  • New GetArrowIPCStreams() method on Rows interface
  • New ArrowIPCStreamIterator interface with methods:
    • Next() (io.Reader, error) - returns raw IPC stream
    • HasNext() bool - Close() - SchemaBytes() ([]byte, error)

• Updated all row scanner implementations to support IPC streams

• Added example demonstrating IPC stream usage

Benefits

• Performance: Avoid parsing overhead when forwarding Arrow data
• Flexibility: Direct access to Arrow binary format for custom processing
• Compatibility: Easier integration with other Arrow-based systems
• Memory efficiency: Process streams without loading all records into memory

Testing

• All existing tests pass
• Backward compatibility maintained through wrapper pattern
• Example provided in examples/ipcstreams/

Usage Example

rows, err := db.QueryContext(ctx, "SELECT * FROM table") ipcStreams, err := rows.(dbsqlrows.Rows).GetArrowIPCStreams(ctx) defer ipcStreams.Close() for ipcStreams.HasNext() { reader, err := ipcStreams.Next() // Process raw Arrow IPC stream }

[jadewang-db]

jenkins merge