quoted! covers all three recording vocabularies

[0-jake-0]

Stacked on #765 → #764 → #763 → #762 → #761 → #760 → #759. Base is quoted-macro; this diff is only the top commit.

What this changes

quoted! previously handled a plain closure. That left the interesting case — a per-instrument parameter — needing the manual three-line func! + .with_src() form, which is exactly backwards: the shorthand worked for the easy case and not for the one the generator exists to serve.

One rule now covers everything, marked in the argument list:

Write	Records
map(|x| ..)	the closure
map([fee] move |x| ..)	the closure and the captured value
ticker(cfg period)	the data config
fold(cfg 0u64, |a, v| ..)	both
join(&other, |x, y| ..)	the closure; &other is an edge, left alone

So a per-instrument leg reads without leaving the macro:

let ticks = quoted!(g => ticker(cfg inst.tick)).count();
let px    = quoted!(ticks => map(|n: &u64| *n as f64));
let net   = quoted!(px => map([fee] move |p: &f64| p - fee));

cfg also removes a redundancy worth naming on its own: ticker(p).with_cfg(&p) wrote the value twice, so it could be forgotten in one place or — worse — changed in one place.

How it was verified

• cargo fmt --all
• cargo lint ✅ — cargo lint-all not run: --all-features cannot build here (aeron needs CMake ≥3.30, box has 3.28; etcd needs protoc). Documented prerequisites, unrelated to this diff; CI is the first real check on that feature set.
• --all-features blocked as above. Full default suite plus bench,async,csv,cache,augurs,market,ws,redis,postgres,kafka,web,kdb,prometheus,fix,dynamic-graph,tracing — green. Doctests green (16 passed).
• New behaviour covered by tests

Twelve tests, one per arm plus the composite. Two assert the shorthand records byte-identically to the manual form, so it stays a convenience rather than a second mechanism. The last one builds a two-instrument desk and asserts no node with a closure config is left unrecorded — which is the property a user actually cares about, rather than any individual arm working.

Notes for the reviewer

Arm order is load-bearing, and the reason is a macro_rules! sharp edge. Each arm is discriminated by a literal token (cfg, [) before any fragment is parsed, and arms run most-specific first. macro_rules! falls through cleanly on a literal mismatch, but a failed $x:expr parse is a hard error, not a fallthrough — so an arm that could swallow another's input has to come second. The comment in the macro says this, because the ordering looks arbitrary otherwise and would be "tidied" into breakage.

A cfg-marked argument is cloned into the call so the original survives to be rendered. Every EmitLiteral type is Clone, and for the common Copy ones (Duration, numbers) the clone compiles away.

Still not detected: an undeclared capture. map(move |p| p - fee) without [fee] records a body that will not resolve in an artifact. It surfaces as a generator refusal — or, if the name happens to resolve to something else at the splice site, as a pass-2 compile error. That is the one remaining hole in the quotation story, and closing it needs the fn-pointer coercion §3 specifies, which cannot be written arity-generically.

One doctest caught me: the markdown pipe-escaping (\|) needed for the table leaked into the adjacent code block, so the example was invalid Rust until the doctest run caught it. Worth knowing if you add rows to that table.

Stack: nine PRs, none reviewed, eight chained on #759.

---

Generated by Claude Code

[0-jake-0]

Superseded by #769, and like #765 not carried forward — quoted! was retired rather than extended, for the reason given there.

Closing.

---

Generated by Claude Code