Crate nornir_testmatrix 

nornir-testmatrix

The portable core of nornir’s multi-aspect test matrix — extracted from nornir (EPIC L) so any leaf repo (znippy, skade, lbzip2-rs …) can run the same matrix with a tiny dep set: std + serde/serde_json + anyhow. No iceberg, arrow, eframe, tantivy, or skade.

What it does

1. Runner (runner) — wrap a repo’s native Rust test framework (cargo nextest run --message-format libtest-json if nextest is on PATH, else plain cargo test), parse each test’s pass/fail + duration, enforce a stall watchdog. Pure subprocess driving — no test registration.
2. Multi-aspect engine (aspect) — beyond unit tests, run MANY aspects of a repo’s health (Aspect): build, doctest, clippy, fmt, audit, bench-smoke, coverage, feature-powerset, msrv, examples. Each shells the right cargo subcommand, parses its result, and emits TestResultRows carrying an aspect tag + a numeric metric (coverage %, warning count, advisory count …). A missing tool is a skip, never a hard fail.
3. Model (model) — the pure row/summary types (TestResultRow, RunSummary, TestSelector, the status tags) + the renderers (render_matrix, summarize_runs, rows_to_json).
4. Sinks (sink) — the TestSink trait + built-in JsonFileSink and NullSink so a leaf repo with no warehouse can still persist its matrix. nornir implements TestSink over its Iceberg warehouse.
5. Discovery (discover) — the autonom (AUT2) anti-drift core: pure enumerators that build the testable Surface from data (facett components, viz tabs × {thin,fat}, CLI subcommands, MCP tools, and the unreached-function set from symbol_facts − test-reachable(call_edges)), then compute_gap differences it against coverage so the gate can enforce Gap == ∅. No warehouse here — the caller feeds the rows.
6. Functional-status mode (functional) — a component reports whether it actually works (functional_status(component, check, ok, detail)) from its own headless-render / self-test path. The Aspect::Functional runner drains those into rows so a broken render is a RED matrix row WITHOUT eyeballing a GUI. Gated behind the testmatrix cargo feature: release builds (feature off) strip the emit to a no-op.

How a leaf repo uses it

use nornir_testmatrix::{run_full_matrix, Aspect, JsonFileSink, TestSink};
use std::path::Path;

let aspects = [Aspect::Build, Aspect::Unit, Aspect::Doctest, Aspect::Clippy, Aspect::Fmt];
let rows = run_full_matrix(Path::new("."), &aspects);
let sink = JsonFileSink::new("target/nornir-testmatrix.json");
sink.append(&rows).unwrap();

Re-exports

pub use aspect::aspect_label;

pub use aspect::parse_aspect;

pub use aspect::parse_funnel_decompose;

pub use aspect::run_aspect;

pub use aspect::run_full_matrix;

pub use aspect::Aspect;

pub use aspect::AspectOutcome;

pub use coverage::rows_for;

pub use coverage::seed_allowlist;

pub use coverage::stale_allowlist_entries;

pub use coverage::AllowEntry;

pub use coverage::Allowlist;

pub use coverage::CoverageRow;

pub use coverage::CoverageSummary;

pub use coverage::GateReport;

pub use coverage::Verdict;

pub use discover::cli_commands;

pub use discover::compute_gap;

pub use discover::facett_components;

pub use discover::mcp_tools;

pub use discover::test_reachable;

pub use discover::unreached_functions;

pub use discover::viz_tabs;

pub use discover::CallEdge;

pub use discover::FacetRow;

pub use discover::Gap;

pub use discover::Mode;

pub use discover::Surface;

pub use discover::SurfaceKind;

pub use discover::SurfaceNode;

pub use discover::SymbolRow;

pub use functional::drain_functional_rows;

pub use functional::functional_row;

pub use functional::functional_status;

pub use model::listed_rows;

pub use model::new_run_id;

pub use model::parse_cargo_test_list;

pub use model::parse_nextest_list;

pub use model::render_matrix;

pub use model::rows_to_json;

pub use model::short_run;

pub use model::status;

pub use model::summarize_runs;

pub use model::RunSummary;

pub use model::TestResultRow;

pub use model::TestSelector;

pub use runner::detect_runner;

pub use runner::list_tests;

pub use runner::run_matrix;

pub use runner::stall_secs;

pub use runner::MatrixRun;

pub use runner::Runner;

pub use runner::TestCase;

pub use runner::DEFAULT_STALL_SECS;

pub use sink::JsonFileSink;

pub use sink::NullSink;

pub use sink::TestSink;

Modules

aspect

The multi-aspect engine — test MANY aspects of a repo’s health, not just its unit tests. Each Aspect shells the right cargo subcommand, parses its result into an AspectOutcome, and emits TestResultRows carrying the aspect tag + a numeric metric (coverage %, warning count, advisory count …).

coverage

autonom coverage gate — the PURE verdict model (AUT2 / n-005)

discover

autonom self-discovery — the anti-drift core of the completeness gate

functional

Functional-status mode — a component reports whether it actually works.

model

The pure test-matrix model: rows, summaries, selectors, status tags, and the human/JSON renderers. Zero heavy deps — std only (no chrono, no uuid): a tiny RFC3339 formatter + a getrandom-free UUIDv4 keep the crate lean.

runner

Standardized test-matrix runner — wrap a repo’s native Rust test framework, parse per-test pass/fail + duration, and hand the rows back.

sink

Where a matrix run’s rows go. The TestSink trait is the durable seam: nornir implements it over its Iceberg warehouse; a leaf repo with no warehouse uses the built-in JsonFileSink (append-only JSON history file) or NullSink (drop, for dry-runs / tests).