racket/collects/mrflow/assoc-set-list.ss

361 lines
18 KiB
Scheme

; associative sets implementation, using lists.
; - key equality based on eq? by default, uses equal? if given the 'equal flag
; - raises exn:assoc-set:key-not-found if key not in associative set when trying
; to remove a key or when trying to get a value and no default thunk is given.
; - raise exn:assoc-set:duplicate-key by default when trying to add a key to a
; set where it already exists
; - strange things might happen if you use assoc-set-union, assoc-set-intersection,
; or assoc-set-difference with two sets that don't use the same comparaison
; function: you might end up with duplicate keys in some sets.
;
; Note: lots of set! and tail-recursive loops in this code, for speed
(module assoc-set-list (lib "mrflow.ss" "mrflow")
(require
(lib "list.ss") ; for foldr
(lib "etc.ss") ; for opt-lambda
"assoc-set-exn.ss" ; no prefix so we can re-provide
)
; table = (listof (cons value value))
(define-struct assoc-set (=? cardinality table))
(provide/contract
(exn:assoc-set? (any/c . -> . boolean?))
(struct (exn:assoc-set:key-not-found exn:assoc-set) ((message (and/c string? immutable?))
(continuation-mark-set continuation-mark-set?)
(assoc-set assoc-set?)
(key any/c)))
(struct (exn:assoc-set:duplicate-key exn:assoc-set) ((message (and/c string? immutable?))
(continuation-mark-set continuation-mark-set?)
(assoc-set assoc-set?)
(key any/c)))
(assoc-set-make (() ((symbols 'equal)) . opt-> . assoc-set?))
(assoc-set-reset (assoc-set? . -> . assoc-set?))
(assoc-set? (any/c . -> . boolean?))
(assoc-set-set ((assoc-set? any/c any/c) (boolean?) . opt-> . assoc-set?))
(assoc-set-get ((assoc-set? any/c) ((-> any)) . opt-> . any))
(assoc-set-in? (assoc-set? any/c . -> . boolean?))
(assoc-set-remove ((assoc-set? any/c) (boolean?) . opt-> . assoc-set?))
(assoc-set-cardinality (assoc-set? . -> . non-negative-exact-integer?))
(assoc-set-empty? (assoc-set? . -> . boolean?))
(assoc-set-copy (assoc-set? . -> . assoc-set?))
(assoc-set-map (assoc-set? (any/c any/c . -> . any) . -> . (listof any/c)))
(assoc-set-fold (assoc-set? (any/c any/c any/c . -> . any) any/c . -> . any))
(assoc-set-for-each (assoc-set? (any/c any/c . -> . any) . -> . assoc-set?))
(assoc-set-for-each! (assoc-set? (any/c any/c . -> . any) . -> . assoc-set?))
(assoc-set-filter ((assoc-set? (any/c any/c . -> . boolean?)) ((symbols 'new 'same)) . opt-> . assoc-set?))
(assoc-set-union ((assoc-set? assoc-set? (any/c any/c . -> . any)) ((symbols 'new 'first 'second)) . opt-> . assoc-set?))
(assoc-set-intersection ((assoc-set? assoc-set? (any/c any/c . -> . any)) ((symbols 'new 'first 'second)) . opt-> . assoc-set?))
(assoc-set-difference ((assoc-set? assoc-set?) ((symbols 'new 'first 'second)) . opt-> . assoc-set?))
)
; (opt 'equal) -> assoc-set
; we test the optional argument ourselves to preserve data abstraction even in the
; presence of an exception
(define assoc-set-make
(case-lambda
[() (make-assoc-set eq? 0 '())]
[(flag) (make-assoc-set equal? 0 '())]))
; assoc-set -> assoc-set
; doesn't change =?
(define (assoc-set-reset assoc-set)
(set-assoc-set-table! assoc-set '())
(set-assoc-set-cardinality! assoc-set 0)
assoc-set)
; value -> boolean
; assoc-set? comes from the structure definition
; assoc-set value value (opt boolean) -> assoc-set
(define assoc-set-set
(opt-lambda (assoc-set key value (exn? #t))
(let ([=? (assoc-set-=? assoc-set)]
[original-table (assoc-set-table assoc-set)])
(set-assoc-set-table! assoc-set (let loop ([table original-table])
(if (null? table)
(begin
(set-assoc-set-cardinality! assoc-set (add1 (assoc-set-cardinality assoc-set)))
(cons (cons key value) original-table))
(let ([key-value-pair (car table)])
(if (=? (car key-value-pair) key)
(if exn?
(raise-duplicate-key-exn "assoc-set-set" assoc-set key)
(begin
; silently replace
(set-cdr! key-value-pair value)
original-table))
(loop (cdr table)))))))
assoc-set)))
; assoc-set value (-> value) -> value
(define assoc-set-get
(opt-lambda (assoc-set key (not-found-thunk (lambda () (raise-key-not-found-exn "assoc-set-get" assoc-set key))))
(let ([=? (assoc-set-=? assoc-set)])
(let loop ([table (assoc-set-table assoc-set)])
(if (null? table)
(not-found-thunk)
(let ([key-value-pair (car table)])
(if (=? (car key-value-pair) key)
(cdr key-value-pair)
(loop (cdr table)))))))))
; assoc-set value -> boolean
(define (assoc-set-in? assoc-set key)
(let ([=? (assoc-set-=? assoc-set)])
(ormap (lambda (key-value-pair)
(=? (car key-value-pair) key))
(assoc-set-table assoc-set))))
; assoc-set value (opt boolean) -> assoc-set
(define assoc-set-remove
(opt-lambda (assoc-set key (exn? #t))
(let ([=? (assoc-set-=? assoc-set)]
[original-table (assoc-set-table assoc-set)])
(set-assoc-set-table! assoc-set
(let loop ([table original-table]
[previous #f])
(if (null? table)
(if exn?
(raise-key-not-found-exn "assoc-set-remove" assoc-set key)
; silently ignore
original-table)
(let ([key-value-pair (car table)])
(if (=? (car key-value-pair) key)
(begin
(set-assoc-set-cardinality! assoc-set (sub1 (assoc-set-cardinality assoc-set)))
(if previous
(begin
; return shortened table
(set-cdr! previous (cdr table))
original-table)
(cdr original-table)))
(loop (cdr table) table)))))))
assoc-set))
; assoc-set -> exact-non-negative-integer
; assoc-set-cardinality comes from the structure definition
; assoc-set -> boolean
(define (assoc-set-empty? assoc-set)
(= 0 (assoc-set-cardinality assoc-set)))
; (listof (cons value value)) (listof (cons value value)) -> (listof (cons value value))
; creates a (reversed) copy of l1 (to prevent list sharing between sets) and prefixes l2 with it
(define (copy-reverse-and-prefix-assoc-lists l1 l2)
(let loop ([l1 l1]
[l2 l2])
(if (null? l1)
l2
(loop (cdr l1) (cons (cons (caar l1) (cdar l1)) l2)))))
; (listof (cons value value)) -> (listof (cons value value))
(define (copy-assoc-list l)
(copy-reverse-and-prefix-assoc-lists l '()))
; assoc-set -> assoc-set
(define (assoc-set-copy assoc-set)
(make-assoc-set (assoc-set-=? assoc-set)
(assoc-set-cardinality assoc-set)
(copy-assoc-list (assoc-set-table assoc-set))))
; assoc-set (value value -> value) -> (listof value)
(define (assoc-set-map assoc-set f)
(let ([unary-f (lambda (key-value-pair)
(f (car key-value-pair) (cdr key-value-pair)))])
(map unary-f (assoc-set-table assoc-set))))
; assoc-set (value value value -> value) value -> value
(define (assoc-set-fold assoc-set f acc)
(foldr (lambda (key-value-pair acc)
(f (car key-value-pair) (cdr key-value-pair) acc))
acc
(assoc-set-table assoc-set)))
; assoc-set (value value -> value) -> assoc-set
(define (assoc-set-for-each assoc-set f)
(let ([unary-f (lambda (key-value-pair)
(f (car key-value-pair) (cdr key-value-pair)))])
(for-each unary-f (assoc-set-table assoc-set)))
assoc-set)
; assoc-set (value value -> value) -> assoc-set
; we know lists are never shared between sets, so we can set-cdr!
(define (assoc-set-for-each! assoc-set f)
(for-each (lambda (key-value-pair)
(set-cdr! key-value-pair (f (car key-value-pair) (cdr key-value-pair))))
(assoc-set-table assoc-set))
assoc-set)
; assoc-set (value value -> boolean) (opt (union 'new 'same)) -> assoc-set
(define assoc-set-filter
(let (; assoc-set (value value -> boolean) -> assoc-set
[filter-into-new-assoc-set
(lambda (assoc-set tester)
(let ([table '()]
[count 0])
(for-each (lambda (key value)
(when (tester key value)
(set! table (cons (cons key value) table))
(set! count (add1 count))))
(assoc-set-table assoc-set))
(make-assoc-set (assoc-set-=? assoc-set) count table)))])
(opt-lambda (assoc-set tester (which-assoc-set 'new))
(let ([new-assoc-set (filter-into-new-assoc-set assoc-set tester)])
(case which-assoc-set
[(new) new-assoc-set]
[(same)
(set-assoc-set-table! assoc-set (assoc-set-table new-assoc-set))
(set-assoc-set-cardinality! assoc-set (assoc-set-cardinality new-assoc-set))
assoc-set])))))
; assoc-set assoc-set (value value -> value) (opt (union 'new 'first 'second)) -> assoc-set
(define assoc-set-union
(opt-lambda (assoc-set1 assoc-set2 merge-values (which-assoc-set 'new))
(let* ([=? (assoc-set-=? assoc-set1)]
[new-assoc-set
(let loop ([table1 (assoc-set-table assoc-set1)]
; we shouldn't modify the original list
[table2 (copy-assoc-list (assoc-set-table assoc-set2))]
[count1 (assoc-set-cardinality assoc-set1)]
[count2 (assoc-set-cardinality assoc-set2)]
[acc '()]
[count 0])
(if (null? table1)
; we have already copied table2, so we can destructively modify it
(make-assoc-set =? (+ count count2)
(append! table2 acc))
(if (null? table2)
(make-assoc-set =? (+ count count1)
(copy-reverse-and-prefix-assoc-lists table1 acc))
(let ([key1 (caar table1)])
; search table2 for same key
(let loop-assoc-set2 ([t2 table2]
[previous #f])
(if (null? t2)
(begin
(set! acc (cons (cons key1 (cdar table1)) acc))
(set! count (add1 count))
(set! table1 (cdr table1))
(set! count1 (sub1 count1)))
(if (=? key1 (caar t2))
(begin
(set! acc (cons (cons key1 (merge-values (cdar table1) (cdar t2))) acc))
(set! count (add1 count))
(set! table1 (cdr table1))
(set! count1 (sub1 count1))
(if previous
(set-cdr! previous (cdr t2))
(set! table2 (cdr table2)))
(set! count2 (sub1 count2)))
(loop-assoc-set2 (cdr t2) t2))))
(loop table1 table2 count1 count2 acc count)))))])
(case which-assoc-set
[(new) new-assoc-set]
[(first)
(set-assoc-set-cardinality! assoc-set1 (assoc-set-cardinality new-assoc-set))
(set-assoc-set-table! assoc-set1 (assoc-set-table new-assoc-set))
assoc-set1]
[(second)
(set-assoc-set-cardinality! assoc-set2 (assoc-set-cardinality new-assoc-set))
(set-assoc-set-table! assoc-set2 (assoc-set-table new-assoc-set))
assoc-set2]))))
; assoc-set assoc-set (value value -> value) (opt (union 'new 'first 'second)) -> assoc-set
(define assoc-set-intersection
(opt-lambda (assoc-set1 assoc-set2 merge-values (which-assoc-set 'new))
(let* ([=? (assoc-set-=? assoc-set1)]
[new-assoc-set
(let loop ([table1 (assoc-set-table assoc-set1)]
; we shouldn't modify the original list
[table2 (copy-assoc-list (assoc-set-table assoc-set2))]
[count1 (assoc-set-cardinality assoc-set1)]
[count2 (assoc-set-cardinality assoc-set2)]
[acc '()]
[count 0])
(if (null? table1)
(make-assoc-set =? count acc)
(if (null? table2)
(make-assoc-set =? count acc)
(let ([key1 (caar table1)])
; search table2 for same key
(let loop-assoc-set2 ([t2 table2]
[previous #f])
(if (null? t2)
(begin
(set! table1 (cdr table1))
(set! count1 (sub1 count1)))
(if (=? key1 (caar t2))
(begin
(set! acc (cons (cons key1 (merge-values (cdar table1) (cdar t2))) acc))
(set! count (add1 count))
(set! table1 (cdr table1))
(set! count1 (sub1 count1))
(if previous
(set-cdr! previous (cdr t2))
(set! table2 (cdr table2)))
(set! count2 (sub1 count2)))
(loop-assoc-set2 (cdr t2) t2))))
(loop table1 table2 count1 count2 acc count)))))])
(case which-assoc-set
[(new) new-assoc-set]
[(first)
(set-assoc-set-cardinality! assoc-set1 (assoc-set-cardinality new-assoc-set))
(set-assoc-set-table! assoc-set1 (assoc-set-table new-assoc-set))
assoc-set1]
[(second)
(set-assoc-set-cardinality! assoc-set2 (assoc-set-cardinality new-assoc-set))
(set-assoc-set-table! assoc-set2 (assoc-set-table new-assoc-set))
assoc-set2]))))
; assoc-set assoc-set (opt (union 'new 'first 'second)) -> assoc-set
(define assoc-set-difference
(opt-lambda (assoc-set1 assoc-set2 (which-assoc-set 'new))
(let* ([=? (assoc-set-=? assoc-set1)]
[new-assoc-set
(let loop ([table1 (assoc-set-table assoc-set1)]
; we shouldn't modify the original list
[table2 (copy-assoc-list (assoc-set-table assoc-set2))]
[count1 (assoc-set-cardinality assoc-set1)]
[count2 (assoc-set-cardinality assoc-set2)]
[acc '()]
[count 0])
(if (null? table1)
(make-assoc-set =? count acc)
(if (null? table2)
(make-assoc-set =? (+ count count1)
(copy-reverse-and-prefix-assoc-lists table1 acc))
(let ([key1 (caar table1)])
; search table2 for same key
(let loop-assoc-set2 ([t2 table2]
[previous #f])
(if (null? t2)
(begin
(set! acc (cons (cons key1 (cdar table1)) acc))
(set! count (add1 count))
(set! table1 (cdr table1))
(set! count1 (sub1 count1)))
(if (=? key1 (caar t2))
(begin
(set! table1 (cdr table1))
(set! count1 (sub1 count1))
(if previous
(set-cdr! previous (cdr t2))
(set! table2 (cdr table2)))
(set! count2 (sub1 count2)))
(loop-assoc-set2 (cdr t2) t2))))
(loop table1 table2 count1 count2 acc count)))))])
(case which-assoc-set
[(new) new-assoc-set]
[(first)
(set-assoc-set-cardinality! assoc-set1 (assoc-set-cardinality new-assoc-set))
(set-assoc-set-table! assoc-set1 (assoc-set-table new-assoc-set))
assoc-set1]
[(second)
(set-assoc-set-cardinality! assoc-set2 (assoc-set-cardinality new-assoc-set))
(set-assoc-set-table! assoc-set2 (assoc-set-table new-assoc-set))
assoc-set2]))))
)