#lang racket/base (require "grammar.rkt" "graph.rkt" racket/list racket/class) ;; Handle the LR0 automaton (provide build-lr0-automaton lr0% (struct-out trans-key) trans-key-list-remove-dups kernel-items kernel-index) ;; kernel = (make-kernel (LR1-item list) index) ;; the list must be kept sorted according to item ;; (vectorof (symbol X hashtable)) (define (build-transition-table num-states assoc) (define transitions (make-vector num-states #f)) (let loop ([i (sub1 (vector-length transitions))]) (when (>= i 0) (vector-set! transitions i (make-hasheq)) (loop (sub1 i)))) (for ([trans-key/kernel (in-list assoc)]) (define tk (car trans-key/kernel)) (hash-set! (vector-ref transitions (kernel-index (trans-key-st tk))) (gram-sym-symbol (trans-key-gs tk)) (cdr trans-key/kernel))) transitions) ;; reverse-assoc : (listof (cons/c trans-key? kernel?)) -> ;; (listof (cons/c trans-key? (listof kernel?))) (define (reverse-assoc assoc) (define reverse-hash (make-hash)) (define (hash-table-add! ht k v) (hash-set! ht k (cons v (hash-ref ht k (λ () null))))) (for ([trans-key/kernel (in-list assoc)]) (define tk (car trans-key/kernel)) (hash-table-add! reverse-hash (make-trans-key (cdr trans-key/kernel) (trans-key-gs tk)) (trans-key-st tk))) (hash-map reverse-hash cons)) ;; kernel-list-remove-duplicates ;; LR0-automaton = object of class lr0% (define lr0% (class object% (super-instantiate ()) ;; term-assoc : (listof (cons/c trans-key? kernel?)) ;; non-term-assoc : (listof (cons/c trans-key? kernel?)) ;; states : (vectorof kernel?) ;; epsilons : ??? (init-field term-assoc non-term-assoc states epsilons) (define transitions (build-transition-table (vector-length states) (append term-assoc non-term-assoc))) (define reverse-term-assoc (reverse-assoc term-assoc)) (define reverse-non-term-assoc (reverse-assoc non-term-assoc)) (define reverse-transitions (build-transition-table (vector-length states) (append reverse-term-assoc reverse-non-term-assoc))) (define mapped-non-terms (map car non-term-assoc)) (define/public (get-mapped-non-term-keys) mapped-non-terms) (define/public (get-num-states) (vector-length states)) (define/public (get-epsilon-trans) epsilons) (define/public (get-transitions) (append term-assoc non-term-assoc)) ;; for-each-state : (state ->) -> ;; Iteration over the states in an automaton (define/public (for-each-state f) (define num-states (vector-length states)) (let loop ([i 0]) (when (< i num-states) (f (vector-ref states i)) (loop (add1 i))))) ;; run-automaton: kernel? gram-sym? -> (union kernel #f) ;; returns the state reached from state k on input s, or #f when k ;; has no transition on s (define/public (run-automaton k s) (hash-ref (vector-ref transitions (kernel-index k)) (gram-sym-symbol s) (λ () #f))) ;; run-automaton-back : (listof kernel?) gram-sym? -> (listof kernel) ;; returns the list of states that can reach k by transitioning on s. (define/public (run-automaton-back k s) (for*/list ([k (in-list k)] [val (in-list (hash-ref (vector-ref reverse-transitions (kernel-index k)) (gram-sym-symbol s) (λ () null)))]) val)))) (define ((union comp (eq? a b) (define (kernel->string k) (apply string-append `("{" ,@(map (λ (i) (string-append (item->string i) ", ")) (kernel-items k)) "}"))) ;; build-LR0-automaton: grammar -> LR0-automaton ;; Constructs the kernels of the sets of LR(0) items of g (define (build-lr0-automaton grammar) ; (printf "LR(0) automaton:\n") (define epsilons (make-hash)) (define grammar-symbols (append (send grammar get-non-terms) (send grammar get-terms))) ;; first-non-term: non-term -> non-term list ;; given a non-terminal symbol C, return those non-terminal ;; symbols A s.t. C -> An for some string of terminals and ;; non-terminals n where -> means a rightmost derivation in many ;; steps. Assumes that each non-term can be reduced to a string ;; of terms. (define first-non-term (digraph (send grammar get-non-terms) (λ (nt) (filter non-term? (map (λ (prod) (sym-at-dot (make-item prod 0))) (send grammar get-prods-for-non-term nt)))) (λ (nt) (list nt)) (union non-term LR1-item list ;; Creates a set of items containing i s.t. if A -> n.Xm is in it, ;; X -> .o is in it too. (define (LR0-closure i) (cond [(null? i) null] [else (define next-gsym (sym-at-dot (car i))) (cond [(non-term? next-gsym) (cons (car i) (append (for*/list ([non-term (in-list (first-non-term next-gsym))] [x (in-list (send grammar get-prods-for-non-term non-term))]) (make-item x 0)) (LR0-closure (cdr i))))] [else (cons (car i) (LR0-closure (cdr i)))])])) ;; maps trans-keys to kernels (define automaton-term null) (define automaton-non-term null) ;; keeps the kernels we have seen, so we can have a unique ;; list for each kernel (define kernels (make-hash)) (define counter 0) ;; goto: LR1-item list -> LR1-item list list ;; creates new kernels by moving the dot in each item in the ;; LR0-closure of kernel to the right, and grouping them by ;; the term/non-term moved over. Returns the kernels not ;; yet seen, and places the trans-keys into automaton (define (goto kernel) ;; maps a gram-syms to a list of items (define table (make-hasheq)) ;; add-item!: ;; (symbol (listof item) hashtable) item? -> ;; adds i into the table grouped with the grammar ;; symbol following its dot (define (add-item! table i) (define gs (sym-at-dot i)) (cond [gs (define already (hash-ref table (gram-sym-symbol gs) (λ () null))) (unless (member i already) (hash-set! table (gram-sym-symbol gs) (cons i already)))] ((zero? (vector-length (prod-rhs (item-prod i)))) (define current (hash-ref epsilons kernel (λ () null))) (hash-set! epsilons kernel (cons i current))))) ;; Group the items of the LR0 closure of the kernel ;; by the character after the dot (for ([item (in-list (LR0-closure (kernel-items kernel)))]) (add-item! table item)) ;; each group is a new kernel, with the dot advanced. ;; sorts the items in a kernel so kernels can be compared ;; with equal? for using the table kernels to make sure ;; only one representitive of each kernel is created (define is (let loop ([gsyms grammar-symbols]) (cond [(null? gsyms) null] [else (define items (hash-ref table (gram-sym-symbol (car gsyms)) (λ () null))) (cond [(null? items) (loop (cdr gsyms))] [else (cons (list (car gsyms) items) (loop (cdr gsyms)))])]))) (filter values (for/list ([i (in-list is)]) (define gs (car i)) (define items (cadr i)) (define new #f) (define new-kernel (sort (filter values (map move-dot-right items)) item ~a on ~a\n" (kernel->string kernel) (kernel->string unique-kernel) (gram-sym-symbol gs)) (and new unique-kernel)))) (define starts (map (λ (init-prod) (list (make-item init-prod 0))) (send grammar get-init-prods))) (define startk (for/list ([start (in-list starts)]) (define k (make-kernel start counter)) (hash-set! kernels start k) (set! counter (add1 counter)) k)) (define new-kernels (make-queue)) (let loop ([old-kernels startk] [seen-kernels null]) (cond [(and (empty-queue? new-kernels) (null? old-kernels)) (make-object lr0% automaton-term automaton-non-term (list->vector (reverse seen-kernels)) epsilons)] [(null? old-kernels) (loop (deq! new-kernels) seen-kernels)] [else (enq! new-kernels (goto (car old-kernels))) (loop (cdr old-kernels) (cons (car old-kernels) seen-kernels))]))) (define-struct q (f l) #:inspector (make-inspector) #:mutable) (define (empty-queue? q) (null? (q-f q))) (define (make-queue) (make-q null null)) (define (enq! q i) (cond [(empty-queue? q) (let ([i (mcons i null)]) (set-q-l! q i) (set-q-f! q i))] [else (set-mcdr! (q-l q) (mcons i null)) (set-q-l! q (mcdr (q-l q)))])) (define (deq! q) (begin0 (mcar (q-f q)) (set-q-f! q (mcdr (q-f q)))))