WASM: demo notebook and ecosystem coordination with QuantStack

[mmcky]

Note

Updated 2026-08-21. The second item said quantecon was already on emscripten-forge; it is not (no recipe exists — that is #932) and it reaches the browser from conda-forge noarch. Added prerequisites and the code paths a demo notebook must avoid for now, based on the Phase 0/1 findings recorded in #927 and #944. The original text is preserved in the edit history.

Part of #925 (Phase 4 — ecosystem, ongoing).

The announcement closes with an explicit invitation to "validate more packages from the wider Numba ecosystem", and it already demos EconForge's stack (interpolation.py, Dolo.py). Answering that invitation is cheap and high-signal.

Prerequisites

• The WASM: JupyterLite proof-of-concept deployment and browser smoke suite #928 results table — the smoke suite from WASM: JupyterLite environment config and browser smoke suite #938 is on main and the browser runner is tracked in WASM: add a WebAssembly/JupyterLite smoke-test job to CI #933; its results are the evidence base for what a demo can show
• A release containing the gini_coefficient rewrite in PR Rewrite Gini coefficient in O(n log n) #937 (WASM: remove @njit(parallel=True) from gini_coefficient via an O(n log n) rewrite #926), so the demo runs against a published version rather than main
• For live %mamba install quantecon inside the notebook: the emscripten-forge recipe, WASM: contribute a quantecon recipe to emscripten-forge #932. Not a hard blocker — quantecon (and requests) already install at build time from conda-forge noarch alongside the emscripten-forge-4x channel, as in WASM: JupyterLite proof-of-concept deployment and browser smoke suite #928's environment

Paths the demo must avoid for now

• gini_coefficient until a release containing PR Rewrite Gini coefficient in O(n log n) #937 ships — on main it still uses @njit(parallel=True) (WASM: remove @njit(parallel=True) from gini_coefficient via an O(n log n) rewrite #926), which runs on Emscripten only via the temporary numba patch 0010 (Fallback Numba parallel JIT to serial on Emscripten emscripten-forge/recipes#6293)
• Any jitted generator: support_enumeration and vertex_enumeration hang on Emscripten (WASM: support_enumeration and vertex_enumeration hang in the browser — root cause is jitted generators (numba.np.linalg._LAPACK verified working) #927) — use lemke_howson for the game-theory example
• simplex_grid, num_compositions_jit and k_array_rank_jit — cache=True cache-save failure on warm sessions (WASM: cache=True functions that link a cache-restored callee fail with "no compiled object yet" (simplex_grid → num_compositions_jit → comb_jit) #944; upstream Numba: RuntimeError: no compiled object yet emscripten-forge/recipes#6309)

Items

• A QuantEcon demo notebook on notebook.link (Markov chains, LQ control, a game-theory solve via lemke_howson) alongside the existing @anutosh491/numba-ecosystem demos; coordinate with QuantStack
• Longer term: QuantEcon lectures that execute without Binder or a kernel server — the rest of the lecture stack (numpy, scipy, matplotlib, pandas, numba) is on emscripten-forge; quantecon (and requests) currently resolve from conda-forge noarch alongside it (see WASM: JupyterLite proof-of-concept deployment and browser smoke suite #928) until the emscripten-forge recipe in WASM: contribute a quantecon recipe to emscripten-forge #932 lands
• Performance exploration: the emscripten-forge recipe tests show @njit(fastmath=True) autovectorises to WASM SIMD (f32x4 + simd128) — a possible pass over hot kernels, with the usual fastmath reproducibility caveats