Enum CheckpointPolicy 

pub enum CheckpointPolicy {
    Flush,
    Fsync,
    FsyncEveryN(u32),
}

Persistence policy a CheckpointSink requests from the apply driver.

After a checkpoint is recorded, the apply driver inspects this value to decide how aggressively to push pending writes through the operating system. Cheap sinks (in-memory test capture) ask for Self::Flush; durability-sensitive sinks (persist-to-disk so a crash recovers cleanly) ask for Self::Fsync or Self::FsyncEveryN.

The driver calls ApplySession::sync_all when the policy demands it, which both flushes every cached BufWriter and calls File::sync_all on the underlying handle. Honouring this on every record would gut throughput on patches with millions of regions — hence Self::FsyncEveryN for the typical “fsync every N records” cadence downstream consumers want.

Mid-block checkpoints — the per-DEFLATE-block emissions inside crate::chunk::sqpk::SqpkFile AddFile — never flush and never fsync regardless of policy. Those emissions fire often enough on a multi-GB file that interleaving a sync syscall would gut throughput. The driver guarantees the next chunk-boundary checkpoint flushes the bytes the mid-block run accumulated in its BufWriter, so a resume from an in-flight checkpoint can never miss data that a later chunk-boundary checkpoint already covered.

Variants

Flush

Flush BufWriter buffers to the OS only; no fsync. Survives a process crash but not an OS crash or power loss between checkpoint and recovery.

Fsync

Flush and fsync every cached file handle on every recorded checkpoint. Strongest durability; pay the syscall cost on every record.

FsyncEveryN(u32)

Flush every record; fsync once every N records. N == 0 is rejected at sink-installation time (crate::ApplyConfig::with_checkpoint_sink and crate::IndexApplier::with_checkpoint_sink both panic) — use Self::Fsync for “fsync every record” instead.

In-flight mid-block checkpoints (the per-DEFLATE-block emissions inside crate::chunk::sqpk::SqpkFile AddFile) never fsync regardless of policy: those emissions are too frequent to interleave with a sync syscall, and the apply driver guarantees that a resume from an in-flight checkpoint can never miss data that a later chunk-boundary checkpoint already covered.

Trait Implementations

impl Clone for CheckpointPolicy

fn clone(&self) -> CheckpointPolicy

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for CheckpointPolicy

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for CheckpointPolicy

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Hash for CheckpointPolicy

fn hash<__H: Hasher>(&self, state: &mut __H)

Feeds this value into the given Hasher.

fn hash_slice<H>(data: &[Self], state: &mut H)

where H: Hasher, Self: Sized,

Feeds a slice of this type into the given Hasher.

impl PartialEq for CheckpointPolicy

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

Tests for self and other values to be equal, and is used by ==.

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

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl Serialize for CheckpointPolicy

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.

impl Copy for CheckpointPolicy

impl Eq for CheckpointPolicy

impl StructuralPartialEq for CheckpointPolicy