jolt/jolt-core/clojure/core/40-lazy.clj
Yogthos 4a1a9e3aec core: lazy realization is shared across walks (once-only effects); pmap family
Every walk over a lazy seq created FRESH wrapper tables around the shared
rest-thunks (ls-rest, ls-seq/ls-count, realize-for-iteration, the printers,
reduce — each had its own make-lazy-seq loop), so independent walks re-ran
the thunks: side effects duplicated, and a doall'd seq of futures was
re-spawned serially by the deref walk. Every walker now goes through
ls-rest-cached, which memoizes the rest wrapper on its node — thunks run
exactly once, as in Clojure. Costs ~10% on walk-heavy benches (the per-node
cache get/put — Clojure's LazySeq pays the same); net still -9% vs the
pre-linear-walks baseline. Three regression rows pin once-only effects and
value stability across walks.

On top of that: pmap/pcalls/pvalues (jolt-oeu) over the real-thread futures
— spawn-all-then-deref (the once-only fix is what makes the doall actually
mean that), snapshot semantics documented, multi-coll arity via the
canonical vector-zip. System/currentTimeMillis + nanoTime land as System
statics (the realtime clock — os/time is whole seconds, which quantized
every elapsed measurement to 1000ms). Seven pmap rows incl. a generous-
margin parallelism check (4 x 200ms sleeps under 700ms after warmup).
2026-06-10 19:14:49 -04:00

191 lines
5.7 KiB
Clojure

;; clojure.core — lazy tier. Canonical CLJS-based lazy seq fns.
;; Loaded after 30-macros.clj, so lazy-seq macro is available.
;;
;; Each fn ported from CLJS core.cljs, stripped of chunked-seq branches.
;; --- distinct --- (transducer + lazy collection arity; value-based dedup)
(defn distinct
([]
(fn [rf]
(let [seen (volatile! #{})]
(fn ([] (rf)) ([result] (rf result))
([result input]
(if (contains? @seen input)
result
(do (vswap! seen conj input) (rf result input))))))))
([coll]
(let [step (fn step [xs seen]
(lazy-seq
((fn [[f :as xs] seen]
(when-let [s (seq xs)]
(if (contains? seen f)
(recur (rest s) seen)
(cons f (step (rest s) (conj seen f))))))
xs seen)))]
(step coll #{}))))
;; --- keep ---
(defn keep
([f]
(fn [rf]
(fn ([] (rf)) ([result] (rf result))
([result input]
(let [v (f input)]
(if (nil? v) result (rf result v)))))))
([f coll]
(lazy-seq
(when-let [s (seq coll)]
(let [x (f (first s))]
(if (nil? x)
(keep f (rest s))
(cons x (keep f (rest s)))))))))
;; --- keep-indexed ---
(defn keep-indexed
([f]
(fn [rf]
(let [ia (volatile! -1)]
(fn ([] (rf)) ([result] (rf result))
([result input]
(let [i (vswap! ia inc)
v (f i input)]
(if (nil? v) result (rf result v))))))))
([f coll]
(letfn [(keepi [idx coll]
(lazy-seq
(when-let [s (seq coll)]
(let [x (f idx (first s))]
(if (nil? x)
(keepi (inc idx) (rest s))
(cons x (keepi (inc idx) (rest s))))))))]
(keepi 0 coll))))
;; --- map-indexed ---
(defn map-indexed
([f]
(fn [rf]
(let [i (volatile! -1)]
(fn ([] (rf)) ([result] (rf result))
([result input] (rf result (f (vswap! i inc) input)))))))
([f coll]
(letfn [(mapi [idx coll]
(lazy-seq
(when-let [s (seq coll)]
(cons (f idx (first s)) (mapi (inc idx) (rest s))))))]
(mapi 0 coll))))
;; --- cycle ---
(defn cycle [coll]
(if-let [vals (seq coll)]
(let [n (count vals)]
(letfn [(cstep [i]
(lazy-seq
(cons (nth vals (mod i n)) (cstep (inc i)))))]
(cstep 0)))
()))
;; --- repeatedly --- ((f) throws on a non-fn; (take n …) throws on a non-number
;; count — both now enforced in the seed (jolt-call / core-take), so the canonical
;; CLJ form matches the repeatedly.cljc exception cases.)
(defn repeatedly
([f] (lazy-seq (cons (f) (repeatedly f))))
([n f] (take n (repeatedly f))))
;; --- repeat ---
(defn repeat
([x] (lazy-seq (cons x (repeat x))))
([n x] (take n (repeat x))))
;; --- iterate ---
(defn iterate [f x]
(lazy-seq (cons x (iterate f (f x)))))
;; --- partition-all --- (transducer + [n coll] + [n step coll])
;; The collection arities realize EXACTLY n per chunk via a first/rest loop and
;; continue from the advanced cursor (not a re-drop / nthrest), so they realize
;; minimally — matching the Janet pstep the §6.3 laziness counters were written
;; against. (A take/nthrest form is correct but over-realizes.)
(defn partition-all
([n]
(fn [rf]
(let [a (volatile! [])]
(fn
([] (rf))
([result]
(let [result (if (zero? (count @a))
result
(let [v @a] (vreset! a []) (unreduced (rf result v))))]
(rf result)))
([result input]
(vswap! a conj input)
(if (= n (count @a))
(let [v @a] (vreset! a []) (rf result v))
result))))))
([n coll]
(letfn [(go [s]
(lazy-seq
(when (seq s)
(loop [i 0 chunk [] cur s]
(if (and (< i n) (seq cur))
(recur (inc i) (conj chunk (first cur)) (rest cur))
(cons chunk (go cur)))))))]
(go coll)))
([n step coll]
(letfn [(go [s]
(lazy-seq
(when (seq s)
(cons (take n s) (go (nthrest s step))))))]
(go coll))))
;; --- Phase 2 leaf batch 3 (jolt-ded): canonical lazy + transducer arities ----
(defn interpose
([sep]
(fn [rf]
(let [started (volatile! false)]
(fn
([] (rf))
([result] (rf result))
([result input]
(if (deref started)
(let [sepr (rf result sep)]
(if (reduced? sepr)
sepr
(rf sepr input)))
(do (vreset! started true)
(rf result input))))))))
([sep coll]
(drop 1 (interleave (repeat sep) coll))))
(defn take-nth
([n]
(fn [rf]
(let [iv (volatile! -1)]
(fn
([] (rf))
([result] (rf result))
([result input]
(let [i (vswap! iv inc)]
(if (zero? (rem i n))
(rf result input)
result)))))))
([n coll]
(lazy-seq
(when-let [s (seq coll)]
(cons (first s) (take-nth n (drop n s)))))))
;; --- pmap family (jolt-oeu): parallel map over real-thread futures ----------
;; Each element's work runs on its own OS thread with SNAPSHOT semantics
;; (futures marshal captured state — pure fns only, mutations don't propagate
;; back). All futures are spawned up front (doall), then derefed in order:
;; coarse-grained work only, as with Clojure's pmap.
(defn pmap
([f coll]
(map deref (doall (map (fn [x] (future (f x))) coll))))
([f coll & colls]
(pmap (fn [xs] (apply f xs)) (apply map vector coll colls))))
(defn pcalls [& fns] (pmap (fn [f] (f)) fns))