Arrow Flows

Arrow is the most nuanced data-plane area currently documented in the Go SDK.

The SDK already supports:

• Arrow row append
• direct arrow.Record append
• Arrow log fetch and decode
• projection-aware fetch on ARROW log tables

Bootstrapped Arrow Writes

The dev harness uses BootstrapRowWriter(...) to pin table metadata, schema ID, compression settings, and optional writer state:

bootstrap, err := arrowClient.BootstrapRowWriter(ctx, client.WriterBootstrapOptions{
	InitWriter: true,
})
if err != nil {
	log.Fatal(err)
}

Then it appends Arrow rows:

result, err := bootstrap.AppendArrowRows(ctx, 0, []client.Row{row1, row2})

Direct arrow.Record Append

The SDK also supports direct record append through:

• AppendArrowRecord(...)
• AppendArrowRecordAuto(...)

This path is useful when your application already builds Arrow records directly.

Fetch and Decode

After fetch, the current public decode helper is:

rows, err := client.DecodeArrowLogBatchRows(logSchema, fetched[0].Records)

Projection Caveat

Projection support is aligned with upstream/server behavior:

• projection is supported for ARROW log format
• projection should not be documented as a universal log-table feature across all encodings

That caveat should remain explicit anywhere Arrow projection is shown.