fix(syntax): implement Carp's reader macros and literals

[takeokunn]

docs/LanguageGuide.md:170-173 defines & and @ under a literal "Reader Macros" heading — &x ;; same as (ref x), @x ;; same as (copy x) — but Carp routed through classify_legacy, which implements neither.

So @(f x) split into a bare @ atom plus a sibling, inflating the enclosing call's arity in 116 of 248 files (47%), 1493 atoms in all. Byte spans survived, so round-trips were lossless; what broke was structure, which makes any argument-counting analysis unsound for Carp.

Also fixed: @"…" silently split string literals (@ glued to the next token and swallowed the opening quote); character literals unrecognized, so \{ \} \[ \] \( \) \" were read as real delimiters; #"…" Pattern literals unrecognized.

Three more the brief missed, all real: ~ (deref) is also a reader macro (76 glued atoms, 15 files); $[…] static arrays (55 bare $ atoms, same arity-inflating defect); and , is whitespace in Carp, not unquote — 39 sites where the legacy reader gave max and val phantom Unquote prefixes in [min, max, val].

What src/Parsing.hs settles

readerMacro is string macroStr then expr <- sexpr — recursing into sexpr, not atom — so prefixes bind any form, @@x stacks, and @"…" is a copy of a genuine string literal. validCharacters excludes & ~ @ % $ #, so a sigil can never be inside a symbol. aChar is \ then exactly one character with no terminator, so \{ and \ (space) are valid and \ab is char a then symbol b.

Differential over all 248 files

parse failures        6 → 0     (exactly the six the README names)
trees changed              171
child slots removed       1623   (reconciled token-by-token)
files that GAINED atoms      0

Deliberately deferred

%/%@ — Carp's real unquote — needs two new ReaderPrefix variants threaded through 40+ sites and adds findings inside macro templates, so it wants its own FP audit. Leaving it keeps templates inert, which is the suppressing direction. r"…" has zero corpus occurrences.

Ref/Copy/Deref/StaticArray are new variants rather than reuses of Function, whose as_source() is #' and which feeds three real re-emission paths — reusing it would have been latent corruption in place of the old kind.

Honest limitation

No carp binary was obtainable — it was removed from nixpkgs on 2026-02-05 as broken for over a year, and there's no ghc/stack/cabal. So the semantics are grammar-derived from src/Parsing.hs, not oracle-checked: 20 hand-verified cases covering prefix-on-list, stacking, @"a b", all seven delimiter char literals, \space, #"a\"b", $[…], comma, and refusals. Every other dialect fix in this series had a real oracle; this one does not, and that's stated rather than implied.

Rebase: resolved arm by arm

Crossed #96, #98, #100 and #102. The classify_reader_macro table is the trap — ours had Unknown | Hy | Carp with Lfe pulled out by #98; theirs had Unknown | Lfe | Hy with Carp pulled out. Each side's group line still named the other's dialect, so taking either would have silently deleted a dialect's reader support.

Verified by building a pristine origin/main control binary and diffing it against the rebased one over 36 repo Lisp files × 10 dialects plus a 28-case snippet battery: 444 comparisons, 0 differences. Nothing outside Carp changed at all.

The agent also caught that the source worktree's HEAD was b61fe5d, not the commit I named — so a naive git diff HEAD would have swept in all of #95 as spurious content.

Three updates to lint-carp-idiom (#95), one a real test failure

That package was authored before this fix and documents these defects as current behaviour. the_correct_corpus_exercises_the_readers_arity_inflation pinned the bug outright with assert!(bare > 0, "…which the reader splits into a bare sigil atom").

Inverted rather than deleted: it now asserts bare == 0 and prefixed_lists == 8, keeping the corpus-coverage pin it existed for — with the 8 derived from the corpus text (7 &(/@( paren forms plus one &[…]) rather than read off the parser, which then independently agreed.

edit format — confirmed fixed by #100

@(unsafe-nth world i) survives in examples/game_of_life.carp. That was the original agent's top-ranked follow-up; #100's carries_reader_prefix fix is dialect-agnostic, so Carp's new variants were covered automatically. Corpus-wide: format exits 0, output reparses, @ & ~ $ counts unchanged, idempotent.

Verification

cargo build --workspace, cargo test --workspace (13915 passed), cargo test --test cli (3085 passed), cargo fmt --all --check, cargo clippy --all-targets --all-features -- -D warnings — all exit 0. No golden or pinned count moved.

Note for sequencing: #103 (Hy) is open and will pull Hy out of Unknown | Hy in that same dispatch table. Whichever merges second needs the same arm-by-arm treatment.