trellis-testing

Companion testing support for Trellis graph invariants.

trellis-testing helps applications test the parts of Trellis that matter most: deterministic transaction traces, replayable canonical input scripts, scoped resource lifecycle, materialized output coherence, host status classification, audit explanations, and full-recompute equivalence.

It is intentionally narrow. It is not a general Rust testing framework, async runtime, network simulator, UI harness, or mock library.

Install

[dev-dependencies]
trellis-testing = "0.1"

Optional helpers are feature-gated:

[dev-dependencies]
trellis-testing = { version = "0.1", features = ["proptest"] }

Available optional features:

• proptest: model sequence strategy helpers;
• insta: snapshot-friendly debug output integration points;
• trybuild: compile-fail gate marker;
• fuzz: shared helper boundary for fuzz targets;
• serde: enables trellis-core/serde.

What It Provides

• TransactionScript for replayable canonical input scripts.
• TrellisHarness for committing named transaction steps against an application-owned graph builder.
• ResourceLedger for scoped resource lifecycle assertions.
• OutputLedger for revisioned frame and rebaseline assertions.
• FakeHost and host status helpers for success, failure, duplicate, stale, future, and late status classification.
• Audit assertions for explaining resource commands and output frames.
• Conformance support levels for downstream application graph tests.
• Full-recompute oracle assertion helpers.

Example Shape

Most applications wrap Trellis in an app-owned graph builder that returns both the graph and stable typed handles. A test then records canonical input changes and replays them against a fresh graph:

use trellis_testing::{TransactionScript, TrellisHarness};

let app = build_app_graph();
let handles = app.handles();

let mut script = TransactionScript::new();
script
    .step("select workspace")
    .input(handles.active_workspace, workspace_id)
    .commit();

let first = TrellisHarness::replay(build_app_graph, &script)?;
let second = TrellisHarness::replay(build_app_graph, &script)?;

first.assert_replay_matches(&second)?;

Design Boundary

Applications own canonical truth and graph construction. trellis-testing owns the reusable Trellis-specific test machinery: scripts, traces, ledgers, replay, audit assertions, and conformance reporting.

The crate does not hide graph propagation behind callbacks, execute real host resources, or replace ordinary Rust test tools.

Documentation

• Testing guide: https://github.com/pablof7z/trellis/blob/master/docs/TESTING.md
• Release-candidate gate: https://github.com/pablof7z/trellis/blob/master/docs/RELEASE_CANDIDATE.md
• Invariants: https://github.com/pablof7z/trellis/blob/master/docs/INVARIANTS.md
• Core crate: https://crates.io/crates/trellis-core

License

Licensed under MIT OR Apache-2.0.