source-academy / py-slang / 29980398089

86%

push

github

pynter-wasm: real npm dependency, opcode fix, and test coverage (#267)

* pynter-wasm.ts: use null (not undefined) for cwrap's void return type

Same fix as source-academy/pynter#18 (Gemini's review there): Emscripten's
cwrap convention (and the @types/emscripten typings many consumers
reference) expects null for a void return type, not undefined — runtime
behavior is identical either way, but null matches the documented
convention.

Applying the literal fix here surfaced a real type error: pynterwasm.d.ts
(vendored, hand-copied from an old Emscripten build) typed cwrap's
returnType as `string | undefined`, not accepting null at all — unlike
ccall's own type just above it in the same file, which already allows
`string | null`. This is a genuine asymmetry in Emscripten's own JSDoc
(cwrap's `@param {string=}` vs ccall's `@param {string|null=}`), not
something introduced by staleness on our side — confirmed by generating
a fresh .d.ts from current pynter source via --emit-tsd, which reproduces
the same narrower cwrap signature.

Since null and undefined are handled identically at cwrap's actual JS
runtime (both take the same "not a string" branch), widening the type to
`(string | null) | undefined` reflects real behavior correctly rather
than just casting around a self-inflicted type error.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

* Consume the real @sourceacademy/pynter-wasm npm package, drop vendored files

source-academy/pynter#18 published pynterwasm.js/pynterwasm.wasm as a real,
versioned npm package for the first time — until now the only way to get
these artifacts was building Emscripten output locally and copying it by
hand, which is exactly what happened once, months ago, and never again:
the vendored copy predated pynter#14's siwasm_run fix (the WASM build was
completely non-functional before that landed), let alone every
Python-specific feature since (for-loop support, //, %, ** fixes,
True/False/None display).

- package.json... (continued)

4386 of 5462 branches covered (80.3%)

Branch coverage included in aggregate %.

9458 of 10555 relevant lines covered (89.61%)

180272.04 hits per line