Coverage
How much of each engine its tests reach. Generated on every docs build, so this page describes the commit it was built from rather than a number somebody typed once.
| Engine | Lines | Branches | Functions | Tests | Measured with |
|---|---|---|---|---|---|
| SwiftUI | 93.42% | 89.77% (regions) | 85.83% | 112 | llvm-cov |
| Compose | 92.18% | 81.03% | 89.90% | 113 | Kover |
| TypeScript | 97.66% | 94.08% | 98.31% | 158 | node –experimental-test-coverage |
| React binding | 98.10% | 85.48% | 93.33% | 19 | node –experimental-test-coverage |
| CLI | 98.85% | 93.59% | 97.97% | 101 | node –experimental-test-coverage |
Read these as a floor, not a score. Every engine passes the same 1620 conformance vectors, which is the definition of correct here; coverage only says how much of each implementation those vectors plus its own tests happen to execute. A line nobody runs is a line nobody has checked, which is why the numbers are worth publishing, but a high number is not evidence of conformance and a lower one is not evidence of a defect.
Swift is measured with llvm-cov, which counts regions rather than branches; the column is labelled accordingly. Kotlin is measured with Kover, and the three TypeScript packages with Node’s built-in coverage, which also enforces a floor on every test run so a regression fails CI rather than showing up here.
The uncovered remainder is mostly defensive: branches that exist because a type system cannot prove they are unreachable, and error paths for inputs the gate has already rejected.