Chez Phase 1 (increment 3c): multi-arity + variadic fn emission

emit-fn lowered multi-arity fns to a Scheme case-lambda and variadic fns to
a rest-arg lambda; the Scheme rest list is coerced to a jolt seq (nil when
empty, via list->cseq), and the named-let wrapper runs that coercion only on
first entry so recur carries the seq directly. Single fixed arity keeps the
plain-lambda fast path (fib untouched).

Also fixes a latent leak in the module-global known-procs: a throw mid-emit
(uncompilable body) left the fn's name registered, so a later corpus case
binding the same name to a keyword emitted a direct call to a non-procedure.
The cleanup now runs on the error path too. Only surfaced once the new arity
support let +24 cases compile further before hitting an uncompilable fn.

Gate: emit-test 81/81, subset probe 619/619 compiled (was 595), 0 divergences,
2036/2655 out of subset; full run-tests green (125 files).
This commit is contained in:
Yogthos 2026-06-17 16:08:27 -04:00
parent cb3cfaf0c2
commit 45208afff1
3 changed files with 78 additions and 20 deletions

View file

@ -152,28 +152,58 @@
(unless recur-target (error "emit: recur outside a loop/fn target"))
(string "(" recur-target " " (string/join (map emit (vv (get node :args))) " ") ")"))
(defn- emit-fn [node]
(def arities (map nn (vv (get node :arities))))
(when (not= 1 (length arities)) (error "emit: multi-arity fn not in this increment"))
(def a (first arities))
(when (get a :rest) (error "emit: variadic fn not in this increment"))
# One arity -> a Scheme lambda param-list + a named-let-wrapped body. The named
# let lets fn-level `recur` rebind this arity's params. A variadic arity takes a
# Scheme rest arg (proper list) and the let binding coerces it to a jolt seq
# (nil when empty — Clojure's rest semantics; list->cseq already does this); recur
# carries the rest seq directly, and the named let's init only runs on first
# entry, so the coercion isn't re-applied on a recur.
(defn- emit-arity-clause [a]
(def params (map munge (vv (get a :params))))
# wrap the body in a named let so fn-level `recur` rebinds the params
(def restp (when-let [r (get a :rest)] (munge r)))
(def label (fresh-label "fnrec"))
(def prev recur-target)
(set recur-target label)
# a named fn binds its own name as a known-procedure local in the body, so
# self-calls emit directly rather than via the jolt-invoke fallback.
(def body (emit (get a :body)))
(set recur-target prev)
(def paramlist
(cond
# only a rest param: Scheme formals are the bare symbol, not `( . xs)`
(and restp (empty? params)) restp
restp (string "(" (string/join params " ") " . " restp ")")
(string "(" (string/join params " ") ")")))
(def binds
(if restp
[;(map (fn [p] (string "(" p " " p ")")) params)
(string "(" restp " (list->cseq " restp "))")]
(map (fn [p] (string "(" p " " p ")")) params)))
[paramlist (string "(let " label " (" (string/join binds " ") ") " body ")")])
(defn- emit-fn [node]
(def arities (map nn (vv (get node :arities))))
# a named fn binds its own name as a known-procedure local across ALL arities,
# so self-calls (to any arity) emit directly rather than via jolt-invoke; the
# case-lambda value dispatches on argument count.
(def self (when-let [nm (get node :name)] (munge nm)))
(def had-self (and self (get known-procs self)))
(when self (put known-procs self true))
(def body (emit (get a :body)))
# Restore known-procs even when a body is uncompilable: a throw mid-emit must
# not leak this fn's name into the module global, or a LATER case binding the
# same name to a keyword/coll would emit a direct call to a non-procedure
# (runtime crash). The corpus probe shares one emit state across all cases, so
# this leak is order-dependent and otherwise invisible in single-case tests.
(def clauses
(try (map emit-arity-clause arities)
([err fib]
(unless had-self (when self (put known-procs self nil)))
(propagate err fib))))
(unless had-self (when self (put known-procs self nil)))
(set recur-target prev)
(def lambda
(string "(lambda (" (string/join params " ") ") "
"(let " label " (" (string/join (map (fn [p] (string "(" p " " p ")")) params) " ") ") "
body "))"))
(if (= 1 (length clauses))
(let [[pl body] (first clauses)] (string "(lambda " pl " " body ")"))
(string "(case-lambda "
(string/join (map (fn [c] (string "(" (get c 0) " " (get c 1) ")")) clauses) " ")
")")))
# A named fn (defn / (fn self [..])) references itself by name — the analyzer
# binds that name as a :local in the body. letrec makes the name visible to the
# lambda so self-calls resolve (recur stays a separate self-call to the arity).

View file

@ -48,13 +48,19 @@ compile-time signal) and are counted "out of subset", not as divergences.
JOLT_CHEZ_CORPUS=1 janet test/chez/run-corpus-chez.janet
Baseline after inc 3b (seq tier + dynamic IFn, jolt-5pso): **595/595 compiled
cases pass**, 0 divergences; 2060/2655 out of subset (await clojure.core on Chez).
The seq tier brought up a list/lazy-seq type with first/rest/next/seq/cons/list,
map/filter/reduce/into/remove, range/take/drop/concat/apply, keys/vals, and
nth/peek/pop over seqs; dynamic IFn dispatch (a keyword/vector/coll held in a
local and called as a fn) now routes through the `jolt-invoke` fallback, closing
the 3 ex-known divergences. The probe exits non-zero on any NEW divergence.
Baseline after inc 3c (multi-arity + variadic fns, jolt-gret): **619/619 compiled
cases pass**, 0 divergences; 2036/2655 out of subset (await clojure.core on Chez).
`emit-fn` now lowers multi-arity fns to a Scheme `case-lambda` and variadic fns to
a rest-arg lambda (the Scheme rest list is coerced to a jolt seq, nil when empty),
which is the gating lift for emitting the clojure.core tiers as a prelude.
Prior, inc 3b (seq tier + dynamic IFn, jolt-5pso): 595/595 compiled, 0 divergences,
2060/2655 out of subset. The seq tier brought up a list/lazy-seq type with
first/rest/next/seq/cons/list, map/filter/reduce/into/remove,
range/take/drop/concat/apply, keys/vals, and nth/peek/pop over seqs; dynamic IFn
dispatch (a keyword/vector/coll held in a local and called as a fn) routes through
the `jolt-invoke` fallback, closing the 3 ex-known divergences. The probe exits
non-zero on any NEW divergence.
(Prior, inc 3a: 433/436 compiled, 3 known IFn divergences, 2219 out of subset.
Inc 2: 182/182 compiled, 0 divergences, 2473 out of subset.)

View file

@ -185,6 +185,28 @@
(ok (string "seq=: " src) (and (= code 0) (= out "true"))
(string "chez=" out " | " err))))
# 3g) multi-arity + variadic fns (inc 3c): case-lambda dispatch, a Scheme rest
# arg collected into a jolt seq (nil when empty), recur within an arity and a
# self-call across arities. Value parity vs the CLI oracle.
(each src ["((fn ([x] (* x 2)) ([x y] (+ x y))) 5)"
"((fn ([x] (* x 2)) ([x y] (+ x y))) 3 4)"
"(defn g ([x] x) ([x y] (+ x y))) (g 10)"
"(defn g ([x] x) ([x y] (+ x y))) (g 10 20)"
"(defn h [a & more] (count more)) (h 1 2 3 4)"
# empty rest is nil (Clojure): count 0, first nil (prints "")
"(defn h [a & more] (count more)) (h 1)"
"(defn h [a & more] (first more)) (h 1)"
"(defn h [a & more] (first more)) (h 1 2 3)"
"(defn h [a & more] (reduce + a more)) (h 1 2 3 4)"
"(defn h [a & more] (reduce + a more)) (apply h [1 2 3 4])"
# self-call from one arity to another, then recur within it
"(defn f ([n] (f n 0)) ([n acc] (if (zero? n) acc (recur (- n 1) (+ acc n))))) (f 5)"
"((fn r [& xs] (if (seq xs) (+ (first xs) (apply r (rest xs))) 0)) 1 2 3)"]
(let [[code out err] (d/run-on-chez ctx src)
want (cli-oracle src)]
(ok (string "arity: " src) (and (= code 0) (= out want))
(string "chez=" out " janet=" want " | " err))))
# 4) perf signal: emitted fib(30) in-Scheme timing (excludes Chez startup), to
# track against the spike ceiling (hand-Scheme fib ~5ms). Informational — the
# jolt-truthy? wrapper (~3x) and flonum modeling are known Phase-4 levers.