Module codec 

The COPY FROM STDIN codec framework.

A codec binds one COPY data format to its implementation. Codecs are stateless strategies: they validate the options they accept at statement-parse time, describe which column types they can ingest, and create the stateful RecordParser per stream.

Formats are not limited to PostgreSQL’s own — GreptimeDB-specific formats (a line protocol, a JSON-lines dialect, …) register exactly like csv does, and may define their own options through the open WITH (...) clause.

Parsing and conversion

A RecordParser consumes raw CopyData bytes and emits complete, insertable rows: one ParsedRow per record, in the order of the targeted columns. Parsers validate record shape (field counts) and convert field values themselves — csv delegates both splitting and typed conversion to arrow-csv’s incremental decoder (the engine behind COPY ... FROM 'file.csv') and turns decoded batches into rows with the shared api::helper::vectors_to_rows. A future format that carries values as text (e.g. a line protocol) would reintroduce shared field-conversion helpers here; see the removed text codec in git history for the shape they had.

Options

The WITH (...) clause is open: every KEY value entry the client writes is collected into a [CopyOptionSet] (see options) regardless of whether any format understands it — the statement is parsed textually (see parse) precisely so options beyond sqlparser’s COPY grammar can reach the codecs. A codec builder takes the options it understands (take_string, take_single_byte, take_flag, or raw take_value for codec-specific shapes) and rejects the rest with reject_leftovers, so a future format like COPY t FROM STDIN WITH (FORMAT influxdb_line, precision 'ns') only needs its own codec and registry entry.

To add a format:

1. implement CopyInCodec for a config struct holding the format’s validated options (see csv::CsvCodec for the example),
2. write a builder fn(CopyOptionSet) -> PgWireResult<Arc<dyn CopyInCodec>> that applies defaults and rejects options the format does not accept,
3. register the builder under every accepted FORMAT name (including aliases) in REGISTERED_FORMATS in the parent module.

Traits

CopyInCodec 🔒

A COPY FROM STDIN data format implementation.

RecordParser 🔒

Incremental parser of one COPY FROM STDIN stream.

Type Aliases

ParsedRow 🔒

A parsed, insertable row: one gRPC value per targeted column, in COPY field order. Every parser emits these directly, whatever mechanism it used to produce them.