Module export 

Export query results or whole tables to files.

All row-oriented formats (CSV, Parquet, Arrow IPC, Iceberg) go through hyperd’s native COPY (query) TO 'path' WITH (format => '…') writer. The MCP only issues a single SQL statement — no Rust-side type inference, no JSON intermediate, no in-memory buffering of the output. Hyperd writes the file (or directory) directly to disk and reports the written row count back as the statement’s affected-rows.

Supported formats:

• CSV — format => 'csv', header => true.
• Parquet — format => 'parquet'. Preserves NUMERIC precision, DATE/TIMESTAMP types, and column nullability. Order of magnitude faster than the previous JSON-mediated path.
• Arrow IPC Stream — format => 'arrowstream'. Bit-identical to what Hyper speaks on the wire for its binary Arrow protocol.
• Iceberg — format => 'iceberg'. Destination is a directory (the Iceberg table root with metadata/ and data/ subdirs); hyperd creates it. Round-trips cleanly with load_iceberg.
• Hyper — new .hyper file populated via CREATE DATABASE + ATTACH DATABASE + CREATE TABLE AS SELECT, openable directly in Tableau Desktop. (Cannot use plain std::fs::copy because on Windows hyperd holds an exclusive lock on the workspace file.)

Structs

ExportOptions

Specifies what to export and where.

ExportResult

Returned by export_to_file with the exported row count and telemetry.

Functions

export_to_file

Top-level export dispatcher. Resolves the source SQL, then delegates to the format-specific exporter.