Files
rose-ash/lib/minikanren/tabling.sx
giles adc8467c78
Some checks failed
Test, Build, and Deploy / test-build-deploy (push) Failing after 58s
mk: phase 7 — naive ground-arg tabling, Fibonacci canary green
`table-2` wraps a 2-arg (input, output) relation. On a ground input
walk, looks up the (string-encoded) cache key; on miss, runs the
relation, drains the answer stream, extracts walk*-output values from
each subst, stores them, and replays. On hit, replays the cached
values directly — no recomputation.

Cache lifetime: a single global mk-tab-cache (mutated via set!).
mk-tab-clear! resets between independent queries.

Canonical demo: tabled fib(25) = 75025 in ~5 seconds; the same naive
fib-o times out at 60s. Memoization collapses the exponential redundant
recomputation in the binary recursion.

Limitations (deferred to future SLG work): cyclic recursive calls with
the same ground key still diverge — naive memoization populates the
cache only AFTER computation completes, so a recursive call inside its
own computation can't see the in-progress entry. The brief's "tabled
patho on cyclic graphs" use case requires producer/consumer
scheduling and is left for a future iteration.

12 new tests, fib(0..20) + ground-term predicate + cache-replay
verification. 638/638 cumulative.
2026-05-08 22:27:10 +00:00

92 lines
2.8 KiB
Plaintext

;; lib/minikanren/tabling.sx — Phase 7 piece A: naive memoization.
;;
;; A `table-2` wrapper for 2-arg relations (input, output). Caches by
;; ground input (walked at call time). On hit, replays the cached output
;; values; on miss, runs the relation, collects all output values from
;; the answer stream, stores, then replays.
;;
;; Limitations of naive memoization (vs proper SLG / producer-consumer
;; tabling):
;; - Each call must terminate before its result enters the cache —
;; so cyclic recursive calls with the SAME ground input would still
;; diverge (not addressed here).
;; - Caching by full ground walk only; partially-ground args fall
;; through to the underlying relation.
;;
;; Despite the limitations, naive memoization is enough for the
;; canonical demo: Fibonacci goes from exponential to linear because
;; each fib(k) result is computed at most once.
;;
;; Cache lifetime: a single global mk-tab-cache. Use `(mk-tab-clear!)`
;; between independent queries.
(define mk-tab-cache {})
(define mk-tab-clear! (fn () (set! mk-tab-cache {})))
(define
mk-tab-lookup
(fn
(key)
(cond
((has-key? mk-tab-cache key) (get mk-tab-cache key))
(:else :miss))))
(define
mk-tab-store!
(fn (key vals) (set! mk-tab-cache (assoc mk-tab-cache key vals))))
(define
mk-tab-ground-term?
(fn
(t)
(cond
((is-var? t) false)
((mk-cons-cell? t)
(and
(mk-tab-ground-term? (mk-cons-head t))
(mk-tab-ground-term? (mk-cons-tail t))))
((mk-list-pair? t) (every? mk-tab-ground-term? t))
(:else true))))
(define
mk-tab-replay-vals
(fn
(vals output s)
(cond
((empty? vals) mzero)
(:else
(let
((sp (mk-unify output (first vals) s)))
(let
((this-stream (cond ((= sp nil) mzero) (:else (unit sp)))))
(mk-mplus this-stream (mk-tab-replay-vals (rest vals) output s))))))))
(define
table-2
(fn
(name rel-fn)
(fn
(input output)
(fn
(s)
(let
((winput (mk-walk* input s)))
(cond
((mk-tab-ground-term? winput)
(let
((key (str name "@" winput)))
(let
((cached (mk-tab-lookup key)))
(cond
((= cached :miss)
(let
((all-substs (stream-take -1 ((rel-fn input output) s))))
(let
((vals (map (fn (s2) (mk-walk* output s2)) all-substs)))
(begin
(mk-tab-store! key vals)
(mk-tab-replay-vals vals output s)))))
(:else (mk-tab-replay-vals cached output s))))))
(:else ((rel-fn input output) s))))))))