Struct WriteOptions 

#[non_exhaustive]
pub struct WriteOptions {
    pub compression: Option<u8>,
    pub version: Option<String>,
    pub validate: bool,
}

Tuning knobs for write_with.

Construct from Default and adjust:

use nir_rs::io::WriteOptions;

let opts = WriteOptions::default().with_compression(None);
assert_eq!(opts.compression, None);

Fields (Non-exhaustive)

Non-exhaustive structs could have additional fields added in future. Therefore, non-exhaustive structs cannot be constructed in external crates using the traditional Struct { .. } syntax; cannot be matched against without a wildcard ..; and struct update syntax will not work.

compression: Option<u8>

Deflate (gzip) level for numeric array datasets, must be in 0..=9.

None writes arrays uncompressed. Values above 9 are rejected; use Self::with_compression to clamp automatically.

Three kinds of dataset are never compressed regardless of this setting:

• Scalars. HDF5 cannot chunk them, and chunking is a prerequisite for any filter.
• Empty arrays — any tensor with a zero-length axis. There are no bytes to compress, and a filter would still cost a chunked layout.
• String arrays — edges and MetadataValue::StringList. These are variable-length, so the dataset holds only heap descriptors and the characters live on HDF5’s global heap. A filter applies to the descriptors, not to the payload, so deflating them would add chunking overhead while compressing almost nothing.

version: Option<String>

Version string to write, overriding NirGraph::version.

validate: bool

Run NirGraph::validate_structure and lossless-representation checks before writing. Defaults to true.

Set this to false to rewrite a graph whose edges do not all resolve, or that contains values the wire format cannot preserve (nested graph versions, rank-0 metadata tensors). Such files exist in the wild — upstream’s own braille_noDelay_bias_zero_subgraph.nir has a subgraph edge naming a node that is not in that subgraph — and read loads them faithfully, so writing them back has to be possible. The resulting file will not load in Python nir.read with its default type checking, and may lose or change the unrepresentable values on readback.

Implementations

impl WriteOptions

pub fn with_compression(self, level: Option<u8>) -> Self

Set the deflate level for array datasets; None disables compression.

pub fn with_version(self, version: impl Into<String>) -> Self

Override the version string written to /version.

pub fn with_validation(self, validate: bool) -> Self

Enable or disable the pre-write structure check.

Trait Implementations

impl Clone for WriteOptions

fn clone(&self) -> WriteOptions

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for WriteOptions

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

Formats the value using the given formatter.

impl Default for WriteOptions

fn default() -> Self

Returns the “default value” for a type.

impl Eq for WriteOptions

impl PartialEq for WriteOptions

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

Equality operator ==.

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

Inequality operator !=.

impl StructuralPartialEq for WriteOptions