Struct TruncationParams 

pub struct TruncationParams {
    pub rtol: Option<f64>,
    pub max_rank: Option<usize>,
    pub cutoff: Option<f64>,
}

Common truncation parameters.

This struct contains the core parameters used for rank truncation across various tensor decomposition and compression operations.

Semantics

This crate uses relative tolerance (rtol) semantics:

• Singular values are truncated when σ_i / σ_max < rtol

ITensorMPS.jl uses cutoff semantics:

• Singular values are truncated when σ_i² < cutoff

Conversion: For normalized tensors (where σ_max = 1):

• ITensorMPS.jl’s cutoff = tensor4all-rs’s rtol²
• To match ITensorMPS.jl behavior: use rtol = sqrt(cutoff)
• Example: ITensorMPS.jl cutoff=1e-10 ↔ tensor4all-rs rtol=1e-5

Fields

rtol: Option<f64>

Relative tolerance for truncation.

Singular values satisfying σ_i / σ_max < rtol are truncated, where σ_max is the largest singular value.

If None, uses the algorithm’s default tolerance.

max_rank: Option<usize>

Maximum rank (bond dimension).

If None, no rank limit is applied.

cutoff: Option<f64>

Cutoff value (ITensorMPS.jl convention).

When set via with_cutoff, rtol is automatically set to √cutoff. This field tracks the original cutoff value for inspection; rtol is always the authoritative tolerance.

If None, cutoff was not used.

Implementations

impl TruncationParams

pub fn new() -> Self

Create new truncation parameters with default values.

pub fn with_rtol(self, rtol: f64) -> Self

Set the relative tolerance.

Clears any previously set cutoff origin.

pub fn with_max_rank(self, max_rank: usize) -> Self

Set the maximum rank.

pub fn with_cutoff(self, cutoff: f64) -> Self

Set cutoff (ITensorMPS.jl convention).

Internally converted to rtol = √cutoff. Clears any previously set rtol origin so cutoff becomes the authoritative tolerance source.

pub fn with_maxdim(self, maxdim: usize) -> Self

Set maxdim (alias for with_max_rank).

This is provided for ITensorMPS.jl compatibility.

pub fn effective_rtol(&self, default: f64) -> f64

Get the effective rtol, using the provided default if not set.

pub fn effective_max_rank(&self) -> usize

Get the effective max_rank, using usize::MAX if not set.

pub fn merge(&self, other: &Self) -> Self

Merge with another set of parameters, preferring self’s values.

Trait Implementations

impl Clone for TruncationParams

fn clone(&self) -> TruncationParams

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for TruncationParams

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

Formats the value using the given formatter.

impl Default for TruncationParams

fn default() -> TruncationParams

Returns the “default value” for a type.

impl HasTruncationParams for TruncationParams

fn truncation_params(&self) -> &TruncationParams

Get a reference to the truncation parameters.

fn truncation_params_mut(&mut self) -> &mut TruncationParams

Get a mutable reference to the truncation parameters.

fn rtol(&self) -> Option<f64>

Get the rtol value.

fn max_rank(&self) -> Option<usize>

Get the max_rank value.

fn with_rtol(self, rtol: f64) -> Self

where Self: Sized,

Set the rtol value (builder pattern).

fn with_max_rank(self, max_rank: usize) -> Self

where Self: Sized,

Set the max_rank value (builder pattern).

fn with_cutoff(self, cutoff: f64) -> Self

where Self: Sized,

Set cutoff (ITensorMPS.jl convention, builder pattern).

fn with_maxdim(self, maxdim: usize) -> Self

where Self: Sized,

Set maxdim (alias for max_rank, builder pattern).

fn set_cutoff(&mut self, cutoff: f64)

Set cutoff via mutable reference.

fn set_maxdim(&mut self, maxdim: usize)

Set maxdim via mutable reference (alias for max_rank).

impl Copy for TruncationParams