echidna: Client.res AffineScript-TEA port — missing primitives (Http/Promise/Json/Dict) [meta]

[hyperpolymath]

Context

Porting src/rescript/src/api/Client.res to AffineScript-TEA (migration roadmap commit b942b9d, [migration-roadmap.rescript-to-affinescript]) is gated on four stdlib primitives that AffineScript does not yet expose. Each was filed separately; this parent formalises the cohort.

Blocker

Typed surfaces wait on hyperpolymath/affinescript#228 (module-qualified type paths).

Children

• AffineScript primitive needed for Client.res: Http.fetch #61 — Http.fetch (Http.get / Http.post / Response accessors)
• AffineScript primitive needed for Client.res: Promise / Async #62 — Promise / Async (Async[T], bind, catch)
• AffineScript primitive needed for Client.res: Json (encode / decode) #63 — Json (encode / decode, codec-style acceptable)
• AffineScript primitive needed for Client.res: Dict (mutable key-value map) #64 — Dict (mutable or affine-immutable key-value map)

Source

Split from a single porting context across four separate primitive-need issues. Formalising as a parent so the cohort can be tracked together and closed in lockstep with the Client.res port (migration step 1 of 10).

Downstream

All UI components that call Client.* functions (Store.res + 6 UI components) are blocked behind this cohort.

[hyperpolymath]

Status update on this cohort (2026-05-30, post-#447 merge):

Resolution

All four child issues closed as already-resolved-upstream:

• AffineScript primitive needed for Client.res: Http.fetch #61 — Http.fetch / get / post shipped in affinescript/stdlib/Http.affine (253 LOC, lowered to globalThis.fetch via Deno-ESM; WasmGC via CPS transform per ADR-013)
• AffineScript primitive needed for Client.res: Promise / Async #62 — Async is a reserved built-in effect; Promise.then equivalent via Http.thenableThen + CPS lowering
• AffineScript primitive needed for Client.res: Json (encode / decode) #63 — affinescript/stdlib/json.affine (335 LOC, v0.3) — full encode/decode + Zig-FFI parse + stringify
• AffineScript primitive needed for Client.res: Dict (mutable key-value map) #64 — affinescript/stdlib/dict.affine (122 LOC) — association-list assoc map; composes with AffineScript primitive needed for Client.res: Json (encode / decode) #63's decoded objects

Upstream blocker (#228) status

hyperpolymath/affinescript#447 (module-qualified type/effect path resolution, ADR-014) merged 2026-05-30T12:51:12Z.

Residual: lowercase stdlib module names (json, dict, option, prelude, …) still cannot be qualifier-referenced — head requires upper_ident. Workaround for the Client.res port: use use json::{Json, ...} style imports rather than json::Json qualified paths.

Remaining work

Client.res port itself is no longer blocked on primitives. The 6 downstream UI components (Store.res + 5 others) can also unblock once Client lands.

Suggested close-this-meta when the port lands as a PR.

[hyperpolymath]

Upstream blocker cleared

Status update on the cohort blocker affinescript#228 lowercase-stdlib-module qualified-path gap (issue #448 item 1):

• affinescript#523 (feat(parser): accept lower_ident head in qualified_type_name) — relaxes qualified_type_name to accept lower_ident head while keeping the tail segment uppercase. Menhir conflict count unchanged (21/1/68/7); 366 tests pass (was 363); regression guards in place for Tier-1-style false-positives and for tail-segment lowercase rejection.
• After it merges, x: json.Value, x: dict.Dict<…>, x: option.Option<…> etc. all parse cleanly at type/effect position with use json; / use dict; / use option; declarations.

This was the last grammar-level blocker noted on this meta-issue. All four primitive children (#61–#64) had already closed as resolved-upstream; with the parser side now matching, the Client.res AffineScript-TEA port is fully unblocked.

Suggested next-step shape

Translating the 360-LOC Client.res is substantive engineering with several design choices that benefit from owner direction:

1. Mutable session state (currentSessionId: ref<option<string>>). AS has no global ref idiom; canonical options are (a) a TEA Model field, (b) a Ref-style mutable cell in stdlib, (c) thread the session id through every function call. (a) is the AS-TEA pattern; recommend that unless a global is intended.
2. Belt.Option.flatMap + Belt.Array.keepMap. AS option.affine ships the flatMap equivalent (PR affinescript#509 / #505 surface); the keepMap shape is arr |> List.filterMap style. Decision: do the port use the |> pipe consistently, or stick with the -> chain shape?
3. Promise.then → AS Async. The reserved built-in effect Async is in stdlib; the canonical pattern in affinescript/stdlib/Http.affine uses CPS lowering per ADR-013. Recommend mirroring Http.affine's shape so the port doesn't introduce a divergent idiom.
4. Webapi.Fetch types. Response.t, RequestInit, BodyInit, HeadersInit, HeadersInit.make({"Content-Type": …}). The AS Http stdlib (PR affinescript#447 era) exposes Http.fetch, Http.get, Http.post directly; recommend collapsing the per-call init-record construction down to one of those three.
5. Dict.t<JSON.t> → AS dict.Dict<json.Value>. The qualified-path syntax is now legal at type position; for the value-side dict::make() pattern, see affinescript#178 lowering work.

If those design choices are settled, the port shape is roughly:

• Client.affine mirroring the 8 API functions
• Client.affine exports a module Decode { … } and module Encode { … } matching the ReScript layout
• Tests in src/rescript/test/ or wherever TEA tests land
• The 6 downstream UI components (Store.res + 5 others) are sequenced after Client.affine lands

Happy to take the port forward as a draft PR once the design choices above are decided — or owner may prefer to drive it directly given the architectural touchpoints.

Refs hyperpolymath/affinescript#523, refs #61, refs #62, refs #63, refs #64.

[hyperpolymath]

Status update — AffineScript stdlib codegen has landed (2026-06-01) 📌

Surfacing the AS-side baseline shift for the parking context here:

• affinescript#522 (merged 2026-06-01) — Sqlite.affine codegen foundation; first stdlib extern with full Deno-ESM lowering + tests end-to-end.
• affinescript#504 (merged 2026-06-01) — closes 6 of 10 STEP-3 Deno-ESM gaps; 6 new stdlib externs (stat helpers, byte ops, importMetaUrl) + wildcard-let codegen fix that was breaking strict-mode ESM.
• affinescript#463 (merged 2026-05-30) — non-ASCII string literals emit correct \uXXXX/\u{X} Unicode escapes in Deno-ESM (unblocks any source with emoji/CJK/accented chars).
• affinescript#421 — stdlib/json v0.3 rewired to hpm-json-rsr Zig FFI (11 externs).

What now works that didn't before: stdlib externs (Sqlite, Deno, json) lower to runnable Deno-ESM via lib/codegen_deno.ml; the lowering pattern (declare in stdlib/*.affine → wire in deno_builtins table + prelude → smoke-test in tests/codegen-deno/*) is repeatable and tested.

What is NOT yet resolved:

• User-module codegen — only stdlib has codegen so far; user-defined pub extern fn in app code is not yet wired.
• Stdlib primitive gaps still open: affinescript#160 (Http.fetch), chapel: real-corpus speedup bench (10-30s prover invocations) #161 (portable Json — partially closed by stdlib/json v0.3), chapel: bench_mrr telemetry — per-prover preempted/timeout/success breakdown #162 (Dict/Map).
• Migration parser affinescript#57; module-qualified type paths affinescript#228; effect-row v1 affinescript#59.

Action for this ticket: No new sweep being kicked off here from this status post — surfacing so anyone landing here can see the current AS baseline. Re-evaluate when user-module codegen + the named stdlib primitives land.

🤖 Filed as part of a cross-estate AS-status sweep on 2026-06-01.