Struct CompactOptions 

pub struct CompactOptions {
    pub max_live_ratio: f64,
    pub max_pass_bytes: u64,
    pub runtime: Option<HyperFileRuntimeConfig>,
    pub relocate_chunk_bytes: usize,
    pub dry_run: bool,
    pub allow_consecutive: bool,
    pub allow_foreign_keys: bool,
}

Which objects compaction may rewrite, and how much it may do at once.

Fields

max_live_ratio: f64

Rewrite an object only if at most this much of it is still live. At 1.0 every live object is a candidate, which rewrites the whole container.

max_pass_bytes: u64

The most live bytes one pass rewrites. Objects are taken whole and cheapest first until the next would pass this; what is left waits for the next pass. Unlimited by default.

On a parted container with partial segments enabled — the default for both — this is only a size: fs_relocate writes the pass out as partials once it crosses the dirty-data limits, so memory is bounded by those limits whatever the pass. Where no partial can be written, everything the pass relocates stays dirty until its one flush, so the pass is also capped at UNBOUNDED_PASS_BYTES; and a container that is not parted writes a checkpoint as one object, so its pass is capped at what one object holds.

runtime: Option<HyperFileRuntimeConfig>

The runtime configuration of the handle that does the rewrite: the size of the partials a pass is written out as (data_cache_dirty_min_bytes_to_part and the dirty-data limits), segment_buffer_size, the read cache, multipart uploads. Give the one the container’s writer uses, and what a compaction writes is shaped like what the writer writes.

None uses compaction_runtime_default. Either way the cleaner overrides one field, flush_conflict_policy, to FailFast: a compactor writes bytes it read earlier, and retrying on a conflict would republish them over another writer’s newer ones.

relocate_chunk_bytes: usize

How many bytes of blocks to hand fs_relocate at once. It holds what it reads in a buffer of its own before the dirty cache takes it, so this bounds that second copy.

dry_run: bool

Select and report, write nothing. No file is opened.

allow_consecutive: bool

Run even though the newest checkpoint is already a compaction. The flag that keeps content_cno still holds for one checkpoint only, so a second pass in a row is seen by consumers as a change of content. Set this when that is acceptable, for example to finish a container larger than one pass.

allow_foreign_keys: bool

Run even though the prefix holds keys this crate does not recognize. Only set this having established that they are not a WAL.

Trait Implementations

impl Clone for CompactOptions

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 CompactOptions

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

Formats the value using the given formatter.

impl Default for CompactOptions

fn default() -> Self

Returns the “default value” for a type.