Struct Comparison 

pub struct Comparison {
    pub baseline_suite: String,
    pub current_suite: String,
    pub headline: Headline,
    pub excluded: Vec<ExcludedItem>,
    pub changes: Vec<Change>,
}

What changed between two runs of the same suite.

The field order is the reading order, and it is deliberate: the headline (or the refusal to produce one) and the exclusions come before the changes, in the machine-readable form as much as in Comparison::summary.

Fields

baseline_suite: String

The suite the baseline reported on.

current_suite: String

The suite the current run reported on.

headline: Headline

The figures, or the reason there are none.

excluded: Vec<ExcludedItem>

Items that appeared in both runs and could not be paired, with why.

changes: Vec<Change>

Every change on the items that stayed paired, in current-run item order followed by removals.

Implementations

impl Comparison

pub fn is_unchanged(&self) -> bool

Returns true when nothing changed and nothing was excluded.

pub fn deterministic_regressions(&self) -> Vec<&Change>

The deterministic regressions — behaviour that got worse.

pub fn signal_regressions(&self) -> Vec<&Change>

The deterministic regressions that a control run says are larger than the harness’s own variation.

Without a noise floor this is every regression: an Unmeasured movement is not known to be noise, and treating the unknown as harmless is how a regression ships.

pub fn judge_drifts(&self) -> Vec<&Change>

The judge drifts — the same behaviour, graded differently.

pub fn derived_input_changes(&self) -> Vec<&Change>

The items whose declared-derived parts changed: the intended effect of the change being measured, and not a broken pairing.

pub fn has_deterministic_regression(&self) -> bool

Returns true when at least one item’s deterministic assertions got worse.

pub fn has_signal_regression(&self) -> bool

Returns true when at least one regression exceeds the measured noise floor. This is the question a merge gate should ask.

pub fn has_judge_drift(&self) -> bool

Returns true when at least one judge score moved beyond the tolerance. This is a question about the measurement, not about the agent.

pub fn corpus_defects(&self) -> Vec<&ExcludedItem>

The excluded items whose recorded parts changed.

These are not a result about the model and must never be read as one: something regenerated testimony about what the system actually emitted, which is a defect in the corpus or in the tooling that touched it. They are excluded from every figure, and this is where they are named.

pub fn has_corpus_defect(&self) -> bool

Returns true when a recorded part changed anywhere in the corpus.

pub fn excluded_share(&self) -> f64

The fraction of the paired corpus that could not be compared.

pub fn is_withheld(&self) -> bool

Returns true when there is no headline figure to read, because too much of the corpus went unpaired.

pub fn summary(&self) -> String

A readable rendering.

The first line is the pairing: how much of the corpus was compared and how much was dropped. The second is the headline, or the refusal. Only then come the individual changes, regressions first.

Trait Implementations

impl Clone for Comparison

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 Comparison

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for Comparison

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl PartialEq for Comparison

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

Equality operator ==.

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

Inequality operator !=.

impl Serialize for Comparison

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.

impl StructuralPartialEq for Comparison