buildIpcBytes
fn
ubyte[] buildIpcBytes(ColumnBatch batch,
IpcCompression compression = IpcCompression.NONE) @trustedSerializes a frame to the Arrow IPC file format (.arrows, Feather V2): one schema message, one dictionary batch per categorical column, one record-batch message per aligned chunk row (a frame whose columns share k chunk boundaries writes k batches; otherwise columns rechunk to one), and the footer. Uncompressed by default; LZ4_FRAME bodies as the explicit option. The bytes are standard: pyarrow/DuckDB/pandas read them.
Parameters
batch | the columns to write (one logical image, however chunked; the alignment policy is the codec's) |
compression | the body compression (default none) |
Returns
the file bytes.
Throws
DataFrameException on unsupported columns (STRUCT, and
the shapes the reader rejects) naming them.