Struct RowFilter 

pub struct RowFilter {
    pub kernels: Vec<String>,
    pub project_commits: Vec<String>,
    pub kernel_commits: Vec<String>,
    pub run_sources: Vec<String>,
    pub resolve_sources: Vec<String>,
    pub cpu_budgets: Vec<String>,
    pub schedulers: Vec<String>,
    pub topologies: Vec<String>,
    pub work_types: Vec<String>,
}

Re-export of the comparison-policy types so downstream crates using ktstr::cli as their public surface don’t need to reach into the internal ktstr::stats module (which is pub(crate) — see lib.rs — and therefore not a stable public path). The policy is the only item in stats that a CLI or external consumer constructs directly; every other item is internal plumbing reached via cli::compare_partitions. Typed-field filter set for narrowing GauntletRow sets in the cargo ktstr perf-delta pipeline. Every field is None / empty by default; populated fields are AND-combined ACROSS fields, with field-internal OR/AND semantics described per-field below. Applied via apply_row_filters in compare_partitions before the rows reach compare_rows_by.

Match semantics:

• scheduler / topology / work_type — STRICT EQUALITY against the row’s corresponding field. The sibling substring filter on compare_rows_by (-E) stays as the only fuzzy-match knob; typed fields are exact so a --scheduler scx_rusty filter does NOT spuriously match scx_rusty_alt.
• kernels — repeatable, OR-combined: a row matches iff its kernel_version equals ANY entry in kernels. Mirrors the --kernel flag on cargo ktstr test/coverage/llvm-cov so the same flag name carries the same multi-value semantic across every subcommand.
• project_commits — repeatable, OR-combined: a row matches iff its commit equals ANY entry in project_commits. Same multi-value semantic as kernels, applied to the ktstr project commit recorded by detect_project_commit at sidecar-write time. Surfaced as the --project-commit CLI flag.
• kernel_commits — repeatable, OR-combined: a row matches iff its kernel_commit equals ANY entry in kernel_commits. Same multi-value semantic as project_commits, applied to the kernel source-tree commit recorded by crate::test_support::sidecar::detect_kernel_commit at sidecar-write time. Filters on the kernel HEAD, NOT on the kernel release version (kernels is the version filter).
• run_sources — repeatable, OR-combined: a row matches iff its run_source equals ANY entry in run_sources. Same multi-value semantic as kernels / project_commits / kernel_commits, applied to the run-environment provenance tag ("local", "ci", "archive") recorded by crate::test_support::sidecar::detect_run_source at sidecar-write time, or rewritten to "archive" at load time when the consumer pulled the pool from a non-default --dir. Surfaced as the --run-source CLI flag.
• A kernels-populated filter against a row whose kernel_version is None ALWAYS fails (no wildcard semantic) — the operator wrote specific versions and a None-row would silently dilute the set. The same opt-in policy applies to project_commits against rows with commit == None, to kernel_commits against rows with kernel_commit == None, and to run_sources against rows with run_source == None.

Empty RowFilter (every field None/empty) is the no-op default and matches every row. Use RowFilter::default() to build it.

Fields

kernels: Vec<String>

Repeatable kernel-version filter, OR-combined: a row matches iff its GauntletRow::kernel_version equals ANY entry. Empty vec disables the filter (“do not filter on kernel”). A row whose kernel_version is itself None never matches a non-empty filter.

project_commits: Vec<String>

Repeatable project-commit filter, OR-combined: a row matches iff its GauntletRow::commit equals ANY entry. Empty vec disables the filter (“do not filter on commit”). A row whose commit is itself None never matches a non-empty filter — same opt-in semantic as kernels.

Field name project_commits (renamed from commits) disambiguates from the sibling kernel_commits field — both describe commit dimensions, so the prefix makes “which repository’s commit?” obvious at every call site.

kernel_commits: Vec<String>

Repeatable kernel-source-commit filter, OR-combined: a row matches iff its GauntletRow::kernel_commit equals ANY entry. Empty vec disables the filter (“do not filter on kernel commit”). A row whose kernel_commit is itself None never matches a non-empty filter — same opt-in semantic as project_commits.

Distinct from project_commits (the ktstr framework commit) and from kernels (the kernel release version): two runs with the same kernel_version but different kernel_commit values represent the same release rebuilt from different trees (e.g. WIP patches on top, a different remote ref).

run_sources: Vec<String>

Repeatable run-environment-source filter, OR-combined: a row matches iff its GauntletRow::run_source equals ANY entry. Empty vec disables the filter (“do not filter on run_source”). A row whose run_source is itself None (sidecar pre-dates the field) never matches a non-empty filter — same opt-in semantic as kernels / project_commits / kernel_commits. Typical values: "local", "ci", "archive". The schema is open: any string is acceptable so a future producer can introduce a new tag without a version bump.

Field name run_sources (renamed from sources) disambiguates from KernelMetadata.source / crate::cache::KernelSource — those describe the kernel build’s input, this describes the run-environment provenance.

resolve_sources: Vec<String>

Repeatable scheduler-resolution-source filter, OR-combined: a row matches iff its GauntletRow::resolve_source equals ANY entry. Empty vec disables the filter. A row whose resolve_source is None (sidecar pre-dates the field, or a skip resolved no binary) never matches a non-empty filter — same opt-in semantic as run_sources. Values are the crate::test_support::ResolveSource::as_str tags ("auto_built", "target_debug", "path", …). Distinct from run_sources (the run ENVIRONMENT): this is HOW the scheduler binary was found. Backs the Dimension::ResolveSource filter + pairing dim (--resolve-source).

cpu_budgets: Vec<String>

Repeatable cpu-budget filter, OR-combined: a row matches iff its GauntletRow::cpu_budget (the effective host-CPU budget, as a decimal string) equals ANY entry. Empty vec disables the filter. Rows with cpu_budget == None (skips) are dropped when this filter is non-empty, mirroring kernels / run_sources. Backs the Dimension::CpuBudget filter + pairing dim (--cpu-budget).

schedulers: Vec<String>

Repeatable scheduler-name filter, OR-combined: a row matches iff its GauntletRow::scheduler equals ANY entry. Empty vec disables the filter (“do not filter on scheduler”). Strict equality on each entry — the substring -E filter is the only fuzzy-match knob; typed flags exact-match. Mirrors the shape of kernels / project_commits / kernel_commits / run_sources so every typed dimension supports the same repeatable OR-combined idiom.

topologies: Vec<String>

Repeatable topology filter, OR-combined: a row matches iff its GauntletRow::topology equals ANY entry. The filter values are the rendered form (e.g. "1n2l4c2t") that Topology::Display emits and cargo ktstr stats list shows. Empty vec disables the filter.

work_types: Vec<String>

Repeatable work-type filter, OR-combined: a row matches iff its GauntletRow::work_type equals ANY entry. Valid names are the PascalCase variants of WorkType::ALL_NAMES. Empty vec disables the filter.

Implementations

impl RowFilter

pub fn matches(&self, row: &GauntletRow) -> bool

Returns true when every populated filter field matches the row. The empty RowFilter (default) returns true for every row — it’s the identity filter.

Trait Implementations

impl Clone for RowFilter

fn clone(&self) -> RowFilter

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for RowFilter

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

Formats the value using the given formatter.

impl Default for RowFilter

fn default() -> RowFilter

Returns the “default value” for a type.