Struct HostView 

#[non_exhaustive]
pub struct HostView<'a> { /* private fields */ }

Borrowed view over the per-sample per-CPU PerCpuTimeStats data that the host capture pipeline populates into each FailureDumpReport::per_cpu_time. Returned by SampleSeries::host; exposes a per-CPU timeline (rows sorted ascending by elapsed-ms, stable on ties) plus a closure-based projector that emits a SeriesField<u64> compatible with the temporal-assertion patterns in crate::assert::temporal.

Orthogonal to super::MonitorView: this view is the per-sample per-CPU TIMELINE source; MonitorView exposes the per-VM-run cross-CPU AGGREGATE. The two draw from different fields on the captured reports (FailureDumpReport::per_cpu_time here vs MonitorReport.summary for the monitor view) and never overlap.

Placeholder samples (the freeze rendezvous timed out, the capture pipeline otherwise failed) carry an empty per_cpu_time slice and naturally drop out of every per-CPU timeline without an explicit filter — temporal-assertion patterns see the surrounding non-placeholder samples in order.

Implementations

impl<'a> HostView<'a>

pub fn cpus(&self) -> Vec<u32>

Discover every CPU id that appears in at least one sample’s per_cpu_time slice. Returned in ascending order, deduped. Useful for “fan-out over every captured CPU” assertion loops: for cpu in host.cpus() { ... }.

pub fn per_cpu_time_timeline(&self, cpu: u32) -> Vec<(u64, &'a PerCpuTimeStats)>

Per-CPU timeline: every sample that captured cpu, sorted ascending by elapsed_ms. Ties retain insertion order (stable sort). Samples whose per_cpu_time slice didn’t include cpu (placeholder reports, or a kernel without per-CPU stats) are absent from the returned timeline rather than producing a default-zero row that would silently advance counter-style assertions.

Returns an empty Vec when cpu was not captured in any sample. Test authors that need explicit per-sample coverage discrimination iterate via SampleSeries::iter_samples and consult crate::scenario::snapshot::Snapshot::per_cpu_time_at per sample.

Inherits the first-match-wins contract for duplicate-cpu entries from crate::scenario::snapshot::Snapshot::per_cpu_time_at: production walker (collect_per_cpu_time) enforces one entry per cpu per sample, but the lookup leaves the contract first-match for graceful degradation on a malformed report. Samples whose elapsed_ms is None (the bridge recorded no timestamp) are EXCLUDED: a timestamp-less sample has no position on a time-ordered axis, and placing it at a fabricated 0 would corrupt the ascending-by-time contract.

pub fn per_cpu_field_u64( &self, cpu: u32, label: impl Into<String>, project: impl Fn(&PerCpuTimeStats) -> u64, ) -> SeriesField<u64>

Project a single u64 field out of each per-sample PerCpuTimeStats row for cpu into a SeriesField<u64> suitable for the temporal-assertion patterns (nondecreasing, rate_within, steady_within, converges_to, etc.) in crate::assert::temporal. Mirrors the shape of SampleSeries::bpf so identical assertion pipelines compose against either axis.

Samples whose per_cpu_time slice didn’t include cpu surface as a per-sample SnapshotError::HostFieldUnavailable slot — gap-tolerant temporal patterns skip with a rendered Note, strict patterns fail the assertion so coverage gaps can never silently slip past the call site.

Trait Implementations

impl<'a> Clone for HostView<'a>

fn clone(&self) -> HostView<'a>

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl<'a> Debug for HostView<'a>

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

Formats the value using the given formatter.

impl<'a> Copy for HostView<'a>