Enum RecordsPayload 

pub enum RecordsPayload {
    V2(Vec<RecordBatch>),
    Raw(Bytes),
    Legacy(Bytes),
    FileRegions(Vec<FileRegion>),
}

Owned form of a records-field payload.

Variants

V2(Vec<RecordBatch>)

Zero or more parsed v2 batches. The records field is a sequence.

Raw(Bytes)

Verbatim wire-format v2 bytes, one or more batches, forwarded without parsing. The fetch pass-through path produces this variant.

Legacy(Bytes)

Opaque pre-v2 bytes, a v0 or v1 MessageSet. Decode them with krabka_records_legacy::decode_message_set.

FileRegions(Vec<FileRegion>)

Zero-copy fetch (Increments D + E): the records run stays in segment .log files. The broker sendfile(2)s it straight to a plaintext socket and never materializes it in userspace. There is one crate::records::FileRegion per contributing segment. encode_to falls back to pread + put_slice on TLS and non-sendfile platforms, and for encoded_len agreement. This variant is gated on the SENDFILE alias (Linux + Apple + FreeBSD/DragonFly) because it only exists to feed the sendfile drainer. On Windows the fallback pread path runs.

Implementations

impl RecordsPayload

pub fn from_bytes(bytes: Bytes) -> Result<Self, RecordsError>

Constructs a payload from raw records-field bytes.

When the bytes look like v2, this function decodes every batch in the field. If they do not, it keeps the bytes as opaque legacy bytes.

Errors

Returns the underlying protocol error when input is truncated, contains an invalid length or tag, or cannot be encoded for the selected version.

pub fn from_bytes_with_policy( bytes: Bytes, policy: RecordDecompressionPolicy, ) -> Result<Self, RecordsError>

Constructs a payload from raw records bytes with explicit decompression limits.

Errors

Returns the underlying records error for malformed, truncated, corrupt, or over-limit v2 input.

pub fn payload_len(&self) -> usize

Wire size of the records-field bytes (no outer length prefix).

pub fn encode_to<B: BufMut>(&self, buf: &mut B) -> Result<(), RecordsError>

Writes the payload bytes into buf.

The caller owns the outer framing.

For FileRegions this is the fallback path, used on TLS, on non-Linux platforms, and for encoded_len agreement. It preads each region out of the segment file and copies it into buf. The broker’s zero-copy sendfile path never calls this method. That path consumes the FileRegions directly.

Errors

Returns the underlying protocol error when input is truncated, contains an invalid length or tag, or cannot be encoded for the selected version.

pub fn as_v2(&self) -> Option<&[RecordBatch]>

Borrows the parsed v2 batches, if this is a parsed V2 payload.

Returns None for Raw, which is intentionally unparsed, for Legacy, and for FileRegions, which is deliberately never materialized.

pub fn as_legacy(&self) -> Option<&Bytes>

Borrows the raw legacy bytes, if this payload is Legacy.

pub fn from_fetch_bytes(bytes: Bytes) -> Result<Self, RecordsError>

Decodes a response-side records field and tolerates a truncated trailing batch.

Kafka returns a partial final RecordBatch when a partition hits its fetch byte budget mid-batch. The JVM consumer stops at the first incomplete batch and re-fetches it from the next offset. This function does the same: it decodes every complete batch, then stops at the first HeaderTooShort or BodyTooShort and drops the remainder. A complete batch that is corrupt in its CRC, magic byte, or content still returns an error, because the leniency forgives truncation only. Produce-request validation keeps the strict from_bytes.

This function treats only HeaderTooShort and BodyTooShort as truncation. An invalid batch_length gives RecordParse, which is corruption and still returns an error. Kafka truncation always keeps a valid batch_length prefix, so truncation can appear only as the two too-short variants.

Errors

Returns the underlying protocol error when input is truncated, contains an invalid length or tag, or cannot be encoded for the selected version.

pub fn decode_lenient<B: Buf>( buf: &mut B, _version: i16, ) -> Result<Self, ProtocolError>

Lenient Decode-shaped entry point for records fields in response messages.

The generated codec calls this method. It consumes the whole sliced field buffer, which the caller has already framed, and parses the bytes leniently through from_fetch_bytes.

Errors

Returns the underlying protocol error when input is truncated, contains an invalid length or tag, or cannot be encoded for the selected version.

Trait Implementations

impl Clone for RecordsPayload

fn clone(&self) -> Self

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for RecordsPayload

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Decode<'_> for RecordsPayload

fn decode<B: Buf>(buf: &mut B, _version: i16) -> Result<Self, ProtocolError>

Errors

impl Default for RecordsPayload

fn default() -> Self

Returns the “default value” for a type.

impl Encode for RecordsPayload

fn encode<B: BufMut>( &self, buf: &mut B, _version: i16, ) -> Result<(), ProtocolError>

Errors

fn encoded_len(&self, _version: i16) -> usize

Size in bytes that encode will write.

impl Eq for RecordsPayload

impl From<RecordBatch> for RecordsPayload

fn from(rb: RecordBatch) -> Self

Converts to this type from the input type.

impl From<Vec<RecordBatch>> for RecordsPayload

fn from(v: Vec<RecordBatch>) -> Self

Converts to this type from the input type.

impl PartialEq for RecordsPayload

fn eq(&self, other: &Self) -> bool

Equality operator ==.

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=.

impl StructuralPartialEq for RecordsPayload