Classes
struct	RecordPrefix
	Fixed-prefix wire header for a serialization record. More...

Enumerations
enum class	Status { Ok , Eof }
	Outcome of a wire-format read operation. More...

Functions
void	encode_record_prefix (std::uint8_t *out, const RecordPrefix &p)
	Encode a `RecordPrefix` into the `out` buffer using field-by-field little-endian encoding.

void	decode_record_prefix (const std::uint8_t *in, RecordPrefix &p)
	Decode a `RecordPrefix` from the `in` buffer.

void	write_record_prefix (std::ostream &out, const RecordPrefix &p)
	Write a `RecordPrefix` to `out` by writing each field through `byte_io`'s stream adapters.

Status	read_record_prefix (std::istream &in, RecordPrefix &p)
	Read a `RecordPrefix` from `in` by reading each field through `byte_io`'s stream adapters.

Enumeration Type Documentation

◆ Status

enum class models::bmi::protocols::serialization::wire_format::Status

strong

Outcome of a wire-format read operation.

read_record_prefix returns this directly: Ok if a full prefix arrived, Eof if the stream ended (cleanly or torn) before the prefix completed. There is no third "error" state at this layer because the wire-format primitive does not validate the bytes — magic / wire_version / payload-cap checks live at the validation layer in serialization_record.hpp, which returns expected<Status, std::string> and reserves the error arm for malformed records.

Sub-byte primitives (byte_io::read_*) still return bool — they are an implementation detail of the wire format with no meaningful three-state distinction.

Enumerator
Ok
Eof

Function Documentation

◆ decode_record_prefix()

void models::bmi::protocols::serialization::wire_format::decode_record_prefix	(	const std::uint8_t *	in,
		RecordPrefix &	p )

inline

Decode a RecordPrefix from the in buffer.

The buffer MUST be at least RecordPrefix::PREFIX_BYTES bytes long.

Magic and wire-version validation are the caller's responsibility — this function only decodes the fields.

◆ encode_record_prefix()

void models::bmi::protocols::serialization::wire_format::encode_record_prefix	(	std::uint8_t *	out,
		const RecordPrefix &	p )

inline

Encode a RecordPrefix into the out buffer using field-by-field little-endian encoding.

The buffer MUST be at least RecordPrefix::PREFIX_BYTES bytes long; exactly PREFIX_BYTES bytes are written.

This is the source of truth for the on-disk prefix layout. Stream and direct-fd writers both go through it, so the encoded bytes are byte-identical regardless of how the caller delivers them to the underlying medium.

References models::bmi::protocols::serialization::wire_format::RecordPrefix::checkpoint_epoch, models::bmi::protocols::serialization::wire_format::RecordPrefix::id_length, models::bmi::protocols::serialization::wire_format::RecordPrefix::magic, models::bmi::protocols::serialization::wire_format::RecordPrefix::payload_length, models::bmi::protocols::serialization::wire_format::RecordPrefix::simulation_timestamp, models::bmi::protocols::serialization::byte_io::store_little_u8(), models::bmi::protocols::serialization::wire_format::RecordPrefix::time_step, and models::bmi::protocols::serialization::wire_format::RecordPrefix::wire_version.

Referenced by ngen::serialization::FileBackend::Writer::write().

◆ read_record_prefix()

Status models::bmi::protocols::serialization::wire_format::read_record_prefix	(	std::istream &	in,
		RecordPrefix &	p )

inline

Read a RecordPrefix from in by reading each field through byte_io's stream adapters.

Returns Status::Ok on a full read; Status::Eof if the stream ended before the prefix completed (EOF at start, or short read mid-prefix).

Magic and wire-version validation are the caller's responsibility — this function reports only "did a full prefix arrive." A reader that needs to distinguish "torn file at the prefix boundary" from "this is not a record" inspects magic and wire_version after a successful read.

Referenced by models::bmi::protocols::read_validated_prefix().

◆ write_record_prefix()

void models::bmi::protocols::serialization::wire_format::write_record_prefix	(	std::ostream &	out,
		const RecordPrefix &	p )

inline

Write a RecordPrefix to out by writing each field through byte_io's stream adapters.

Exactly PREFIX_BYTES bytes are emitted.

The caller is responsible for following the prefix on the stream with id_length bytes of id and payload_length bytes of payload — the prefix carries the lengths but does not carry the bodies.

Referenced by models::bmi::protocols::write_record(), and models::bmi::protocols::write_record().

Classes

Enumerations

Functions

Enumeration Type Documentation

◆ Status

Function Documentation

◆ decode_record_prefix()

◆ encode_record_prefix()

◆ read_record_prefix()

◆ write_record_prefix()