# query-lang - Roadmap
> Path from scaffold to a stable 1.0. Hard parts are front-loaded; each phase has hard exit criteria.
> Master plan: ../../_strategy/LANG_COLLECTION.md
>
> **Anti-deferral rule:** no listed hard task moves to a later phase unless this file records the move and the reason.
## v0.1.0 - Scaffold (DONE)
Compiles, CI green, structure correct, no domain logic.
- [x] Manifest, README, CHANGELOG, REPS, dual license, CI, deny, clippy, rustfmt.
## v0.2.0 - Core (THE HARD PART, NOT DEFERRED) (DONE)
A query-based incremental compilation framework (salsa/rust-analyzer model).
Dependencies (none) are wired here, when first used.
Exit criteria:
- [x] Every public item has rustdoc + a runnable example.
- [x] Core invariants property-tested (full DIRECTIVES + API authored at this stage).
Delivered: `System` trait, `Database<S>` engine (revision-based validation with
early cutoff), `Revision`, `Stats`, `QueryError`. No first-party dependency wired
(a generic caching/invalidation engine, like the sibling generic crates); the
reasoning is recorded here under the anti-deferral rule. `no_std` on `alloc`,
`#![forbid(unsafe_code)]`, keys in a `BTreeMap` (`Key: Ord`, no hashing dep).
## v1.0.0 - API freeze (DONE)
Public surface stable and frozen until 2.0.
- [x] docs/API.md marked stable; SemVer promise recorded.
- [x] Full test + benchmark suite green on all three platforms.
Shipped 2026-07-11. No code changes from 0.2.0; the surface (`System`, `Database`,
`Revision`, `Stats`, `QueryError`) is frozen, and the resolution semantics
(unchanged-input no-op, early cutoff) plus the `serde` representations and MSRV
1.85 are recorded as part of the contract in `docs/API.md#stability`.