Skip to content

TScert v0.6

Choose a tag to compare

@cristianoc cristianoc released this 02 Nov 21:26
· 677 commits to main since this release

v0.6 is TScert's second tagged release, at commit 4291f9ce: a research verifier
for type safety of TypeScript functions. TScert accepts a function when its
certificate passes a checker whose soundness is proved in Lean.

The version number marks a coverage threshold: v0.N requires at least N×10% of
the named top-level function declarations with bodies in tools/, TScert's own
front end, to verify in the census on main. v0.6 meets the 60% threshold.

The census of 4291f9ce, measured on 2026-11-01 with a 300 s stage limit,
verifies 411 of 644 functions (63.8%), against 351 of 637 (55.1%) at v0.5.
Within those function bodies, 4,441 of 15,235 counted statements are covered
(29.1%), against 21.5% at v0.5; the
coverage definition
states what is counted.

What acceptance guarantees

For inputs meeting their types and a heap meeting its storage contracts, an
accepted function's translation into the calculus does not get stuck or return a
value outside its result type. It may throw or run forever. The acceptance
theorem,
SourceClaimJson.checkJson_safe,
establishes this guarantee for successful checking.

The guarantee relies on source translation and annotation interpretation,
library and host contracts, conforming inputs, and the trusted tools and runtime
assumptions listed in the
trust surface.
Since v0.5, the native contracts of join, some, filter, flatMap and
JSON.stringify admit arrays with holes, under the existing assumptions of
stable standard objects and ordinary array construction, and the module-loading
host boundary states its deferred exports operation.

Scope

The supported subset includes control flow, objects, arrays, classes, closures
and generic functions; the
README
outlines it. The
fragment boundary
describes source restrictions, and the
work ledger
tracks remaining verification work.

License

TScert is licensed under the
NON-AI MIT License:
the MIT terms, with the code excluded from use in training or improving
machine-learning systems and from datasets for that training, also for its
derivatives.

Try it

Follow the
setup instructions,
then check the included example:

make check FILE=pipeline/examples/array-last-element.ts