Struct HDF5Dataset 

pub struct HDF5Dataset { /* private fields */ }

An HDF5 dataset — a typed, shaped, chunked (or contiguous) array of data.

Provides synchronous access to parsed metadata (shape, dtype, chunk shape, filters, fill value) and async access to the chunk index (byte offsets of all chunks in the file).

Implementations

impl HDF5Dataset

pub fn new( name: String, header: ObjectHeader, reader: Arc<dyn AsyncFileReader>, raw_reader: Arc<dyn AsyncFileReader>, superblock: Arc<Superblock>, ) -> Result<Self>

Create a new dataset by parsing metadata messages from its object header.

pub fn name(&self) -> &str

The dataset’s name (not the full path).

pub fn shape(&self) -> &[u64]

Dataset shape (current dimensions).

pub fn ndim(&self) -> usize

Number of dimensions.

pub fn dtype(&self) -> &DataType

Data type.

pub fn element_size(&self) -> u32

Element size in bytes.

pub fn chunk_shape(&self) -> Option<&[u64]>

Chunk shape (None for contiguous or compact storage).

pub fn filters(&self) -> &FilterPipeline

Filter pipeline.

pub fn fill_value(&self) -> Option<&[u8]>

Fill value bytes (interpretation depends on dtype).

pub fn layout(&self) -> &StorageLayout

Storage layout.

pub fn is_null_dataspace(&self) -> bool

Whether this dataset has a null dataspace (no data, type 2).

pub fn has_external_storage(&self) -> bool

Whether this dataset uses external data files (msg type 0x0007).

pub fn header(&self) -> &ObjectHeader

Access the object header.

pub async fn attributes(&self) -> Vec<Attribute>

Get all attributes attached to this dataset, resolving vlen data.

pub async fn attribute(&self, name: &str) -> Option<Attribute>

Get a single attribute by name.

pub async fn chunk_index(&self) -> Result<&ChunkIndex>

Extract the chunk index, caching the result after first resolution.

For chunked datasets, traverses the B-tree to enumerate all chunks. For contiguous datasets, returns a single-entry index. For compact datasets, returns an empty index (data is inline in the header).

pub async fn batch_get_chunks( &self, chunk_indices: &[Vec<u64>], ) -> Result<Vec<Option<Bytes>>>

Fetch multiple chunks in a single batched I/O call.

Looks up byte ranges from the chunk index and fetches them all via raw_reader.get_byte_ranges(), bypassing the BlockCache.

Returns one entry per input index, in the same order. Chunks that are not present in the index (unallocated) are returned as None.

pub async fn batch_fetch_ranges( &self, ranges: &[(u64, u64)], ) -> Result<Vec<Bytes>>

Fetch multiple byte ranges in a single batched I/O call.

Unlike batch_get_chunks, this takes pre-resolved (offset, length) pairs — no chunk index lookup is performed. Use this when the caller has already resolved the chunk index and wants to avoid re-parsing.

Trait Implementations

impl Debug for HDF5Dataset

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

Formats the value using the given formatter.