list

list previews the exact data set an analyze pass would consume, without running the analysis, letting you confirm the commit range and discriminant sets first. It takes a bare positional subject — runs, discriminants, or blessings. A bare list with no subject is an error that names the three.

The preview includes ignored benchmarks and ghosts: these are exclusions from analysis, not from stored-data inspection. list runs and examine show raw measurements independently of blessing records. Their output remains available when a blessing payload cannot be decoded; analysis and blessing inspection still validate the acceptance records they use.

# Per discriminant set, the run / series / per-commit counts of the selected runs.
cargo bench-history list runs --local=./bench-history

# A discovery catalog of the discriminant sets present in storage (needs no repository).
cargo bench-history list discriminants --local=./bench-history

# Audit blessings.
cargo bench-history list blessings --local=./bench-history

Subjects

• runs — mirrors analyze’s data-set-selection parameters exactly through the shared selection pipeline, and reports, per discriminant set, the run / series / per-commit counts of the selected runs (each commit’s clean/dirty split), oldest-first by topology.
• discriminants — a different view: a discovery catalog of the sets present in storage. It requires no repository and so ignores the timeline and data-filtering groups. With no discriminant filters it lists every stored partition, so you can find triples and machine keys you do not already know.
• blessings — audits blessings (see bless / unbless): records at --context (default HEAD), with their actual persisted blessing scopes, even without measurements. A scope is shown when it intersects the query, but is never narrowed in the output. Pass --engine all --target-triple all --machine-key all to audit every scope at that anchor. JSON’s discriminant_scope contains the accepted values per axis; an empty array is unrestricted. Add --all to show the most recent effective blessing per measured benchmark and partition across the analysis window, using the same history or branch evidence line as analyze. The window view describes existing series, not the stored scope’s coverage of future partitions.

Blessing JSON views

The top-level JSON scope identifies the view: head means records at the single selected context commit, and window means effective acceptance across the analysis window. The commit field identifies the selected context; it need not be the checked-out HEAD. Each row’s discriminant_scope separately describes which engines, targets and machine keys the blessing accepts.