Struct ChunkCacheConfig 

pub struct ChunkCacheConfig { /* private fields */ }

Configuration for a per-dataset chunk cache.

The byte and slot limits are the hdf5-pure counterpart of the rdcc_nbytes and rdcc_nslots raw-data chunk-cache settings from HDF5’s H5Pset_cache. They apply to decompressed raw chunk data. The optional chunk-index cache controls whether hdf5-pure retains the parsed chunk address index between reads of the same crate::Dataset. Disabling the index cache lowers retained metadata memory at the cost of re-scanning the on-disk chunk index for repeated reads.

Implementations

impl ChunkCacheConfig

pub const fn new() -> Self

Create a config matching the historical defaults: 1 MiB of decompressed chunks, 16 slots, and retained parsed chunk indexes.

pub const fn from_h5p_cache(rdcc_nslots: usize, rdcc_nbytes: usize) -> Self

Create a config from HDF5 H5Pset_cache raw data chunk-cache values.

rdcc_nslots maps to the maximum retained chunk slots and rdcc_nbytes maps to the maximum retained decompressed chunk bytes. Modern HDF5 ignores H5Pset_cache’s mdc_nelmts; use crate::MetadataCacheConfig for the metadata-cache budget.

rdcc_w0 has nothing to decide here

C’s third field is a head start, in entries, for a rule that preempts only chunks the caller consumed in full before the unqualified rule joins the walk (H5D__chunk_cache_prune, discriminating on H5D_rdcc_ent_t::rd_count / wr_count). Nothing here answers to it. This cache holds decompressed data read from the file, so there is no dirty chunk for wr_count to describe; and it records only whether a chunk is held, never how much of it a caller took, so the rd_count class it would sort by does not exist to be sorted.

What w0 exists to prevent — a scan evicting chunks for chunks it will not ask for again — is handled here without a knob: a whole read fills the cache and stops offering, so it has no prune for a preemption policy to order. The field-by-field argument, including why a row window is the one path that does evict and why recency already keeps what it needs, is in the property-support reference.

pub const fn disabled() -> Self

Disable retained decompressed chunks and parsed chunk indexes.

pub const fn with_max_bytes(self, max_bytes: usize) -> Self

Set the maximum decompressed chunk bytes retained per dataset.

pub const fn with_max_slots(self, max_slots: usize) -> Self

Set the maximum number of decompressed chunk slots retained per dataset.

pub const fn with_index_cache(self, enabled: bool) -> Self

Enable or disable retaining the parsed chunk index between reads.

pub const fn max_bytes(&self) -> usize

Return the maximum decompressed chunk bytes retained per dataset.

pub const fn max_slots(&self) -> usize

Return the maximum decompressed chunk slots retained per dataset.

pub const fn index_cache_enabled(&self) -> bool

Return whether parsed chunk indexes are retained between reads.

Trait Implementations

impl Clone for ChunkCacheConfig

fn clone(&self) -> ChunkCacheConfig

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Copy for ChunkCacheConfig

impl Debug for ChunkCacheConfig

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

Formats the value using the given formatter.

impl Default for ChunkCacheConfig

fn default() -> Self

Returns the “default value” for a type.

impl Eq for ChunkCacheConfig

impl PartialEq for ChunkCacheConfig

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

Equality operator ==.

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

Inequality operator !=.

impl StructuralPartialEq for ChunkCacheConfig