You cannot select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
br-parser-tools/br-parser-tools-lib/br-parser-tools/private-yacc/lalr.rkt

253 lines
9.3 KiB
Racket

#lang racket/base
(require "lr0.rkt"
"grammar.rkt"
racket/list
racket/class)
;; Compute LALR lookaheads from DeRemer and Pennello 1982
(provide compute-LA)
;; compute-DR: LR0-automaton * grammar -> (trans-key -> term set)
;; computes for each state, non-term transition pair, the terminals
;; which can transition out of the resulting state
;; output term set is represented in bit-vector form
(define ((compute-DR a g) tk)
(define r (send a run-automaton (trans-key-st tk) (trans-key-gs tk)))
(term-list->bit-vector
(filter (λ (term) (send a run-automaton r term)) (send g get-terms))))
;; compute-reads:
;; LR0-automaton * grammar -> (trans-key -> trans-key list)
(define (compute-reads a g)
(define nullable-non-terms (filter (λ (nt) (send g nullable-non-term? nt)) (send g get-non-terms)))
(λ (tk)
(define r (send a run-automaton (trans-key-st tk) (trans-key-gs tk)))
(for/list ([non-term (in-list nullable-non-terms)]
#:when (send a run-automaton r non-term))
(make-trans-key r non-term))))
;; compute-read: LR0-automaton * grammar -> (trans-key -> term set)
;; output term set is represented in bit-vector form
(define (compute-read a g)
(define dr (compute-DR a g))
(define reads (compute-reads a g))
(digraph-tk->terml (send a get-mapped-non-term-keys)
reads
dr
(send a get-num-states)))
;; returns the list of all k such that state k transitions to state start on the
;; transitions in rhs (in order)
(define (run-lr0-backward a rhs dot-pos start num-states)
(let loop ([states (list start)]
[i (sub1 dot-pos)])
(cond
[(< i 0) states]
[else (loop (send a run-automaton-back states (vector-ref rhs i))
(sub1 i))])))
;; prod->items-for-include: grammar * prod * non-term -> lr0-item list
;; returns the list of all (B -> beta . nt gamma) such that prod = (B -> beta nt gamma)
;; and gamma =>* epsilon
(define (prod->items-for-include g prod nt)
(define rhs (prod-rhs prod))
(define rhs-l (vector-length rhs))
(append (if (and (> rhs-l 0) (eq? nt (vector-ref rhs (sub1 rhs-l))))
(list (make-item prod (sub1 rhs-l)))
null)
(let loop ([i (sub1 rhs-l)])
(cond
[(and (> i 0)
(non-term? (vector-ref rhs i))
(send g nullable-non-term? (vector-ref rhs i)))
(if (eq? nt (vector-ref rhs (sub1 i)))
(cons (make-item prod (sub1 i))
(loop (sub1 i)))
(loop (sub1 i)))]
[else null]))))
;; prod-list->items-for-include: grammar * prod list * non-term -> lr0-item list
;; return the list of all (B -> beta . nt gamma) such that (B -> beta nt gamma) in prod-list
;; and gamma =>* epsilon
(define (prod-list->items-for-include g prod-list nt)
(apply append (map (λ (prod) (prod->items-for-include g prod nt)) prod-list)))
;; comput-includes: lr0-automaton * grammar -> (trans-key -> trans-key list)
(define (compute-includes a g)
(define num-states (send a get-num-states))
(define items-for-input-nt (make-vector (send g get-num-non-terms) null))
(for ([input-nt (in-list (send g get-non-terms))])
(vector-set! items-for-input-nt (non-term-index input-nt)
(prod-list->items-for-include g (send g get-prods) input-nt)))
(λ (tk)
(define goal-state (trans-key-st tk))
(define non-term (trans-key-gs tk))
(define items (vector-ref items-for-input-nt (non-term-index non-term)))
(trans-key-list-remove-dups
(apply append
(for/list ([item (in-list items)])
(define prod (item-prod item))
(define rhs (prod-rhs prod))
(define lhs (prod-lhs prod))
(map (λ (state) (make-trans-key state lhs))
(run-lr0-backward a
rhs
(item-dot-pos item)
goal-state
num-states)))))))
;; compute-lookback: lr0-automaton * grammar -> (kernel * proc -> trans-key list)
(define (compute-lookback a g)
(define num-states (send a get-num-states))
(λ (state prod)
(map (λ (k) (make-trans-key k (prod-lhs prod)))
(run-lr0-backward a (prod-rhs prod) (vector-length (prod-rhs prod)) state num-states))))
;; compute-follow: LR0-automaton * grammar -> (trans-key -> term set)
;; output term set is represented in bit-vector form
(define (compute-follow a g includes)
(define read (compute-read a g))
(digraph-tk->terml (send a get-mapped-non-term-keys)
includes
read
(send a get-num-states)))
;; compute-LA: LR0-automaton * grammar -> kernel * prod -> term set
;; output term set is represented in bit-vector form
(define (compute-LA a g)
(define includes (compute-includes a g))
(define lookback (compute-lookback a g))
(define follow (compute-follow a g includes))
(λ (k p)
(define l (lookback k p))
(define f (map follow l))
(apply bitwise-ior (cons 0 f))))
(define (print-DR dr a g)
(print-input-st-sym dr "DR" a g print-output-terms))
(define (print-Read Read a g)
(print-input-st-sym Read "Read" a g print-output-terms))
(define (print-includes i a g)
(print-input-st-sym i "includes" a g print-output-st-nt))
(define (print-lookback l a g)
(print-input-st-prod l "lookback" a g print-output-st-nt))
(define (print-follow f a g)
(print-input-st-sym f "follow" a g print-output-terms))
(define (print-LA l a g)
(print-input-st-prod l "LA" a g print-output-terms))
(define (print-input-st-sym f name a g print-output)
(printf "~a:\n" name)
(send a for-each-state
(λ (state)
(for-each
(λ (non-term)
(let ([res (f (make-trans-key state non-term))])
(when (not (null? res))
(printf "~a(~a, ~a) = ~a\n"
name
state
(gram-sym-symbol non-term)
(print-output res)))))
(send g get-non-terms))))
(newline))
(define (print-input-st-prod f name a g print-output)
(printf "~a:\n" name)
(send a for-each-state
(λ (state)
(for-each
(λ (non-term)
(for-each
(λ (prod)
(let ([res (f state prod)])
(when (not (null? res))
(printf "~a(~a, ~a) = ~a\n"
name
(kernel-index state)
(prod-index prod)
(print-output res)))))
(send g get-prods-for-non-term non-term)))
(send g get-non-terms)))))
(define (print-output-terms r)
(map gram-sym-symbol r))
(define (print-output-st-nt r)
(map (λ (p) (list (kernel-index (trans-key-st p)) (gram-sym-symbol (trans-key-gs p)))) r))
;; init-tk-map : int -> (vectorof hashtable?)
(define (init-tk-map n)
(define v (make-vector n #f))
(let loop ([i (sub1 (vector-length v))])
(when (>= i 0)
(vector-set! v i (make-hasheq))
(loop (sub1 i))))
v)
;; lookup-tk-map : (vectorof (symbol? int hashtable)) -> trans-key? -> int
(define ((lookup-tk-map map) tk)
(define st (trans-key-st tk))
(define gs (trans-key-gs tk))
(hash-ref (vector-ref map (kernel-index st))
(gram-sym-symbol gs)
(λ () 0)))
;; add-tk-map : (vectorof (symbol? int hashtable)) -> trans-key int ->
(define ((add-tk-map map) tk v)
(define st (trans-key-st tk))
(define gs (trans-key-gs tk))
(hash-set! (vector-ref map (kernel-index st))
(gram-sym-symbol gs)
v))
;; digraph-tk->terml:
;; (trans-key list) * (trans-key -> trans-key list) * (trans-key -> term list) * int * int * int
;; -> (trans-key -> term list)
;; DeRemer and Pennello 1982
;; Computes (f x) = (f- x) union Union{(f y) | y in (edges x)}
;; A specialization of digraph in the file graph.rkt
(define (digraph-tk->terml nodes edges f- num-states)
;; Will map elements of trans-key to term sets represented as bit vectors
(define results (init-tk-map num-states))
;; Maps elements of trans-keys to integers.
(define N (init-tk-map num-states))
(define get-N (lookup-tk-map N))
(define set-N (add-tk-map N))
(define get-f (lookup-tk-map results))
(define set-f (add-tk-map results))
(define stack null)
(define (push x) (set! stack (cons x stack)))
(define (pop) (begin0
(car stack)
(set! stack (cdr stack))))
(define (depth) (length stack))
;; traverse: 'a ->
(define (traverse x)
(push x)
(let ([d (depth)])
(set-N x d)
(set-f x (f- x))
(for-each (λ (y)
(when (= 0 (get-N y))
(traverse y))
(set-f x (bitwise-ior (get-f x) (get-f y)))
(set-N x (min (get-N x) (get-N y))))
(edges x))
(when (= d (get-N x))
(let loop ([p (pop)])
(set-N p +inf.0)
(set-f p (get-f x))
(unless (equal? x p)
(loop (pop)))))))
(for ([x (in-list nodes)]
#:when (zero? (get-N x)))
(traverse x))
get-f)