;; ;; Copyright (C) John Cowan 2013. All Rights Reserved. ;; ;; Permission is hereby granted, free of charge, to any person obtaining ;; a copy of this software and associated documentation ;; files (the "Software"), to deal in the Software without restriction, ;; including without limitation the rights to use, copy, modify, merge, ;; publish, distribute, sublicense, and/or sell copies of the Software, ;; and to permit persons to whom the Software is furnished to do so, ;; subject to the following conditions: ;; ;; The above copyright notice and this permission notice shall be ;; included in all copies or substantial portions of the Software. ;; ;; THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, ;; EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF ;; MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND ;; NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE ;; LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION ;; OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION ;; WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. ;; Apapted to Gauche by Shiro Kawai ;;;; Implementation of general sets and bags for SRFI 113 ;;; A "sob" object is the representation of both sets and bags. ;;; This allows each set-* and bag-* procedure to be implemented ;;; using the same code, without having to deal in ugly indirections ;;; over the field accessors. There are three fields, "sob-multi?", ;;; "sob-hash-table", and "sob-comparator." ;;; The value of "sob-multi?" is #t for bags and #f for sets. ;;; "Sob-hash-table" maps the elements of the sob to the number of times ;;; the element appears, which is always 1 for a set, any positive value ;;; for a bag. "Sob-comparator" is the comparator for the elements of ;;; the set. ;;; Note that sob-* procedures do not do type checking or (typically) the ;;; copying required for supporting pure functional update. These things ;;; are done by the set-* and bag-* procedures, which are externally ;;; exposed (but trivial and mostly uncommented below). ;;; Shim to convert from SRFI 69 to the future "intermediate hash tables" ;;; SRFI. Unfortunately, hash-table-fold is incompatible between the two ;;; and so is not usable. ;; Originally SRFI-113 (define-module scheme.set (use srfi.114) (use gauche.collection) (use gauche.generator) (export set set-unfold set? set-contains? set-empty? set-disjoint? set-member set-element-comparator set-adjoin set-adjoin! set-replace set-replace! set-delete set-delete! set-delete-all set-delete-all! set-search! set-size set-find set-count set-any? set-every? set-map set-for-each set-fold set-filter set-remove set-remove set-partition set-filter! set-remove! set-partition! set-copy set->list list->set list->set! set=? set? set<=? set>=? set-union set-intersection set-difference set-xor set-union! set-intersection! set-difference! set-xor! set-comparator bag bag-unfold bag? bag-contains? bag-empty? bag-disjoint? bag-member bag-element-comparator bag-adjoin bag-adjoin! bag-replace bag-replace! bag-delete bag-delete! bag-delete-all bag-delete-all! bag-search! bag-size bag-find bag-count bag-any? bag-every? bag-map bag-for-each bag-fold bag-filter bag-remove bag-partition bag-filter! bag-remove! bag-partition! bag-copy bag->list list->bag list->bag! bag=? bag? bag<=? bag>=? bag-union bag-intersection bag-difference bag-xor bag-union! bag-intersection! bag-difference! bag-xor! bag-comparator bag-sum bag-sum! bag-product bag-product! bag-unique-size bag-element-count bag-for-each-unique bag-fold-unique bag-increment! bag-decrement! bag->set set->bag set->bag! bag->alist alist->bag )) (select-module scheme.set) ;; Gauche adaptation (define hash-table-contains? hash-table-exists?) (define (hash-table-for-each proc hash-table) ; argument order reversed ((with-module gauche hash-table-for-each) hash-table proc)) (define make-hash-table/comparator make-hash-table) (define hash-table-size hash-table-num-entries) (define hash-table-ref/default hash-table-get) (define hash-table-set! hash-table-put!) (define hash-table-update!/default hash-table-update!) (define-class () ((table :init-keyword :table))) (define-class () ()) (define-class () ()) (define (raw-make-sob hash-table comparator multi?) (if multi? (make :table hash-table) (make :table hash-table))) (define (set? obj) (is-a? obj )) (define (bag? obj) (is-a? obj )) ;; we assume sob is either or (define (sob-hash-table sob) (slot-ref sob 'table)) (define (sob-comparator sob) (hash-table-comparator (slot-ref sob'table))) (define (sob-multi? sob) (is-a? sob )) (define-method object-hash ((obj )) (sob-hash obj)) (define-method object-equal? ((a ) (b )) (sob=? a b)) (define-method object-equal? ((a ) (b )) (sob=? a b)) (define-method write-object ((obj ) out) (format out "#" (debug-label obj) (hash-table-num-entries (sob-hash-table obj)))) (define-method write-object ((obj ) out) (format out "#" (debug-label obj) (hash-table-num-entries (sob-hash-table obj)) (apply + (hash-table-values (sob-hash-table obj))))) (define-method call-with-iterator ((obj ) proc :allow-other-keys) ($ call-with-iterator (sob-hash-table obj) (^[end? next] (define current (if (end?) #f (next))) ; ( . ) (define (over?) (or (not current) (and (end?) (zero? (cdr current))))) (define (get) (and (pair? current) (if (zero? (cdr current)) (and (not (end?)) (begin (set! current (next)) (dec! (cdr current)) (car current))) (begin (dec! (cdr current)) (car current))))) (proc over? get)))) ;; TODO: To impelment call-with-builder, we need some way to specify ;; comparator argument for the constructor. ;; The following code is mostly intact from the reference implementation ;; I commented out some irrelevant definitions, and rewrote early break ;; pattern (by call/cc) with hash-table-find. ;;; Record definition and core typing/checking procedures ;; (define-record-type sob ;; (raw-make-sob hash-table comparator multi?) ;; sob? ;; (hash-table sob-hash-table) ;; (comparator sob-comparator) ;; (multi? sob-multi?)) ;; (define (set? obj) (and (sob? obj) (not (sob-multi? obj)))) ;; (define (bag? obj) (and (sob? obj) (sob-multi? obj))) (define (check-set obj) (if (not (set? obj)) (error "not a set" obj))) (define (check-bag obj) (if (not (bag? obj)) (error "not a bag" obj))) ;; These procedures verify that not only are their arguments all sets ;; or all bags as the case may be, but also share the same comparator. (define (check-all-sets list) (for-each (lambda (obj) (check-set obj)) list) (sob-check-comparators list)) (define (check-all-bags list) (for-each (lambda (obj) (check-bag obj)) list) (sob-check-comparators list)) (define (sob-check-comparators list) (if (not (null? list)) (for-each (lambda (sob) (check-same-comparator (car list) sob)) (cdr list)))) ;; This procedure is used directly when there are exactly two arguments. ;; [SK] Gauche's equal? is more permissive than eq? when comparing comparators. (define (check-same-comparator a b) (if (not (equal? (sob-comparator a) (sob-comparator b))) (error "different comparators" a b))) ;; This procedure defends against inserting an element ;; into a sob that violates its constructor, since ;; typical hash-table implementations don't check for us. (define (check-element sob element) (comparator-check-type (sob-comparator sob) element)) ;;; Constructors ;; Construct an arbitrary empty sob out of nothing. (define (make-sob comparator multi?) (raw-make-sob (make-hash-table/comparator comparator) comparator multi?)) ;; Copy a sob, sharing the constructor. (define (sob-copy sob) (raw-make-sob (hash-table-copy (sob-hash-table sob)) (sob-comparator sob) (sob-multi? sob))) (define (set-copy set) (check-set set) (sob-copy set)) (define (bag-copy bag) (check-bag bag) (sob-copy bag)) ;; Construct an empty sob that shares the constructor of an existing sob. (define (sob-empty-copy sob) (make-sob (sob-comparator sob) (sob-multi? sob))) ;; Construct a set or a bag and insert elements into it. These are the ;; simplest external constructors. (define (set comparator . elements) (let ((result (make-sob comparator #f))) (for-each (lambda (x) (sob-increment! result x 1)) elements) result)) (define (bag comparator . elements) (let ((result (make-sob comparator #t))) (for-each (lambda (x) (sob-increment! result x 1)) elements) result)) ;; The fundamental (as opposed to simplest) constructor: unfold the ;; results of iterating a function as a set. In line with SRFI 1, ;; we provide an opportunity to map the sequence of seeds through a ;; mapper function. (define (sob-unfold stop? mapper successor seed comparator multi?) (let ((result (make-sob comparator multi?))) (let loop ((seed seed)) (if (stop? seed) result (begin (sob-increment! result (mapper seed) 1) (loop (successor seed))))))) (define (set-unfold continue? mapper successor seed comparator) (sob-unfold continue? mapper successor seed comparator #f)) (define (bag-unfold continue? mapper successor seed comparator) (sob-unfold continue? mapper successor seed comparator #t)) ;;; Predicates ;; Just a wrapper of hash-table-contains?. (define (sob-contains? sob member) (hash-table-contains? (sob-hash-table sob) member)) (define (set-contains? set member) (check-set set) (sob-contains? set member)) (define (bag-contains? bag member) (check-bag bag) (sob-contains? bag member)) ;; A sob is empty if its size is 0. (define (sob-empty? sob) (= 0 (hash-table-size (sob-hash-table sob)))) (define (set-empty? set) (check-set set) (sob-empty? set)) (define (bag-empty? bag) (check-bag bag) (sob-empty? bag)) ;; Two sobs are disjoint if, when looping through one, we can't find ;; any of its elements in the other. We have to try both ways: ;; sob-half-disjoint checks just one direction for simplicity. (define (sob-half-disjoint? a b) (let ((ha (sob-hash-table a)) (hb (sob-hash-table b))) (not (hash-table-find ha (^[k _] (hash-table-contains? hb k)))))) (define (set-disjoint? a b) (check-set a) (check-set b) (check-same-comparator a b) (and (sob-half-disjoint? a b) (sob-half-disjoint? b a))) (define (bag-disjoint? a b) (check-bag a) (check-bag b) (check-same-comparator a b) (and (sob-half-disjoint? a b) (sob-half-disjoint? b a))) ;; Accessors ;; If two objects are indistinguishable by the comparator's ;; equality procedure, only one of them will be represented in the sob. ;; This procedure lets us find out which one it is; it will return ;; the value stored in the sob that is equal to the element. ;; Note that we have to search the whole hash table item by item. ;; The default is returned if there is no such element. (define (sob-member sob element default) (define (same? a b) (=? (sob-comparator sob) a b)) (let1 r (hash-table-find (sob-hash-table sob) (^[k v] (and (same? k element) (list k)))) (if (pair? r) (car r) default))) (define (set-member set element default) (check-set set) (sob-member set element default)) (define (bag-member bag element default) (check-bag bag) (sob-member bag element default)) ;; Retrieve the comparator. (define (set-element-comparator set) (check-set set) (sob-comparator set)) (define (bag-element-comparator bag) (check-bag bag) (sob-comparator bag)) ;; Updaters (pure functional and linear update) ;; The primitive operation for adding an element to a sob. ;; There are a few cases where we bypass this for efficiency. (define (sob-increment! sob element count) (check-element sob element) (hash-table-update!/default (sob-hash-table sob) element (if (sob-multi? sob) (lambda (value) (+ value count)) (lambda (value) 1)) 0)) ;; The primitive operation for removing an element from a sob. Note this ;; procedure is incomplete: it allows the count of an element to drop below 1. ;; Therefore, whenever it is used it is necessary to call sob-cleanup! ;; to fix things up. This is done because it is unsafe to remove an ;; object from a hash table while iterating through it. (define (sob-decrement! sob element count) (hash-table-update!/default (sob-hash-table sob) element (lambda (value) (- value count)) 0)) ;; This is the cleanup procedure, which happens in two passes: it ;; iterates through the sob, deciding which elements to remove (those ;; with non-positive counts), and collecting them in a list. When the ;; iteration is done, it is safe to remove the elements using the list, ;; because we are no longer iterating over the hash table. It returns ;; its argument, because it is often tail-called at the end of some ;; procedure that wants to return the clean sob. (define (sob-cleanup! sob) (let ((ht (sob-hash-table sob))) (for-each (lambda (key) (hash-table-delete! ht key)) (nonpositive-keys ht)) sob)) (define (nonpositive-keys ht) (let ((result '())) (hash-table-for-each (lambda (key value) (when (<= value 0) (set! result (cons key result)))) ht) result)) ;; We expose these for bags but not sets. (define (bag-increment! bag element count) (check-bag bag) (sob-increment! bag element count) bag) (define (bag-decrement! bag element count) (check-bag bag) (sob-decrement! bag element count) (sob-cleanup! bag) bag) ;; The primitive operation to add elements from a list. We expose ;; this two ways: with a list argument and with multiple arguments. (define (sob-adjoin-all! sob elements) (for-each (lambda (elem) (sob-increment! sob elem 1)) elements)) (define (set-adjoin! set . elements) (check-set set) (sob-adjoin-all! set elements) set) (define (bag-adjoin! bag . elements) (check-bag bag) (sob-adjoin-all! bag elements) bag) ;; These versions copy the set or bag before adjoining. (define (set-adjoin set . elements) (check-set set) (let ((result (sob-copy set))) (sob-adjoin-all! result elements) result)) (define (bag-adjoin bag . elements) (check-bag bag) (let ((result (sob-copy bag))) (sob-adjoin-all! result elements) result)) ;; Given an element which resides in a set, this makes sure that the ;; specified element is represented by the form given. Thus if a ;; sob contains 2 and the equality predicate is =, then calling ;; (sob-replace! sob 2.0) will replace the 2 with 2.0. Does nothing ;; if there is no such element in the sob. (define (sob-replace! sob element) (let* ((comparator (sob-comparator sob)) (= (comparator-equality-predicate comparator)) (ht (sob-hash-table sob))) (comparator-check-type comparator element) (or (hash-table-find ht (^[k v] (and (= k element) (begin (hash-table-delete! ht k) (hash-table-set! ht element v) sob)))) sob))) (define (set-replace! set element) (check-set set) (sob-replace! set element) set) (define (bag-replace! bag element) (check-bag bag) (sob-replace! bag element) bag) ;; Non-destructive versions that copy the set first. Yes, a little ;; bit inefficient because it copies the element to be replaced before ;; actually replacing it. (define (set-replace set element) (check-set set) (let ((result (sob-copy set))) (sob-replace! result element) result)) (define (bag-replace bag element) (check-bag bag) (let ((result (sob-copy bag))) (sob-replace! result element) result)) ;; The primitive operation to delete elemnets from a list. ;; Like sob-adjoin-all!, this is exposed two ways. It calls ;; sob-cleanup! itself, so its callers don't need to (though it is safe ;; to do so.) (define (sob-delete-all! sob elements) (for-each (lambda (element) (sob-decrement! sob element 1)) elements) (sob-cleanup! sob) sob) (define (set-delete! set . elements) (check-set set) (sob-delete-all! set elements)) (define (bag-delete! bag . elements) (check-bag bag) (sob-delete-all! bag elements)) (define (set-delete-all! set elements) (check-set set) (sob-delete-all! set elements)) (define (bag-delete-all! bag elements) (check-bag bag) (sob-delete-all! bag elements)) ;; Non-destructive version copy first; this is inefficient. (define (set-delete set . elements) (check-set set) (sob-delete-all! (sob-copy set) elements)) (define (bag-delete bag . elements) (check-bag bag) (sob-delete-all! (sob-copy bag) elements)) (define (set-delete-all set elements) (check-set set) (sob-delete-all! (sob-copy set) elements)) (define (bag-delete-all bag elements) (check-bag bag) (sob-delete-all! (sob-copy bag) elements)) ;; Flag used by sob-search! to represent a missing object. (define missing (string-copy "missing")) ;; Searches and then dispatches to user-defined procedures on failure ;; and success, which in turn should reinvoke a procedure to take some ;; action on the set (insert, ignore, replace, or remove). (define (sob-search! sob element failure success) (define (insert obj) (sob-increment! sob element 1) (values sob obj)) (define (ignore obj) (values sob obj)) (define (update new-elem obj) (sob-decrement! sob element 1) (sob-increment! sob new-elem 1) (values (sob-cleanup! sob) obj)) (define (remove obj) (sob-decrement! sob element 1) (values (sob-cleanup! sob) obj)) (let ((true-element (sob-member sob element missing))) (if (eq? true-element missing) (failure insert ignore) (success true-element update remove)))) (define (set-search! set element failure success) (check-set set) (sob-search! set element failure success)) (define (bag-search! bag element failure success) (check-bag bag) (sob-search! bag element failure success)) ;; Return the size of a sob. If it's a set, we can just use the ;; number of associations in the hash table, but if it's a bag, we ;; have to add up the counts. (define (sob-size sob) (if (sob-multi? sob) (let ((result 0)) (hash-table-for-each (lambda (elem count) (set! result (+ count result))) (sob-hash-table sob)) result) (hash-table-size (sob-hash-table sob)))) (define (set-size set) (check-set set) (sob-size set)) (define (bag-size bag) (check-bag bag) (sob-size bag)) ;; Search a sob to find something that matches a predicate. You don't ;; know which element you will get, so this is not as useful as finding ;; an element in a list or other ordered container. If it's not there, ;; call the failure thunk. (define (sob-find pred sob failure) (let1 r (hash-table-find (sob-hash-table sob) (^[k _] (and (pred k) (list k)))) (if (pair? r) (car r) (failure)))) (define (set-find pred set failure) (check-set set) (sob-find pred set failure)) (define (bag-find pred bag failure) (check-bag bag) (sob-find pred bag failure)) ;; Count the number of elements in the sob that satisfy the predicate. ;; This is a special case of folding. (define (sob-count pred sob) (sob-fold (lambda (elem total) (if (pred elem) (+ total 1) total)) 0 sob)) (define (set-count pred set) (check-set set) (sob-count pred set)) (define (bag-count pred bag) (check-bag bag) (sob-count pred bag)) ;; Check if any of the elements in a sob satisfy a predicate. Breaks out ;; early (with call/cc) if a success is found. (define (sob-any? pred sob) (hash-table-find (sob-hash-table sob) (^[k _] (boolean (pred k))))) (define (set-any? pred set) (check-set set) (sob-any? pred set)) (define (bag-any? pred bag) (check-bag bag) (sob-any? pred bag)) ;; Analogous to set-any?. Breaks out early if a failure is found. (define (sob-every? pred sob) (not (hash-table-find (sob-hash-table sob) (^[k _] (not (pred k)))))) (define (set-every? pred set) (check-set set) (sob-every? pred set)) (define (bag-every? pred bag) (check-bag bag) (sob-every? pred bag)) ;;; Mapping and folding ;; A utility for iterating a command n times. This is used by sob-for-each ;; to execute a procedure over the repeated elements in a bag. Because ;; of the representation of sets, it works for them too. (define (do-n-times cmd n) (let loop ((n n)) (when (> n 0) (cmd) (loop (- n 1))))) ;; Basic iterator over a sob. (define (sob-for-each proc sob) (hash-table-for-each (lambda (key value) (do-n-times (lambda () (proc key)) value)) (sob-hash-table sob))) (define (set-for-each proc set) (check-set set) (sob-for-each proc set)) (define (bag-for-each proc bag) (check-bag bag) (sob-for-each proc bag)) ;; Fundamental mapping operator. We map over the associations directly, ;; because each instance of an element in a bag will be treated identically ;; anyway; we insert them all at once with sob-increment!. (define (sob-map comparator proc sob) (let ((result (make-sob comparator (sob-multi? sob)))) (hash-table-for-each (lambda (key value) (sob-increment! result (proc key) value)) (sob-hash-table sob)) result)) (define (set-map comparator proc set) (check-set set) (sob-map comparator proc set)) (define (bag-map comparator proc bag) (check-bag bag) (sob-map comparator proc bag)) ;; The fundamental deconstructor. Note that there are no left vs. right ;; folds because there is no order. Each element in a bag is fed into ;; the fold separately. (define (sob-fold proc nil sob) (let ((result nil)) (sob-for-each (lambda (elem) (set! result (proc elem result))) sob) result)) (define (set-fold proc nil set) (check-set set) (sob-fold proc nil set)) (define (bag-fold proc nil bag) (check-bag bag) (sob-fold proc nil bag)) ;; Process every element and copy the ones that satisfy the predicate. ;; Identical elements are processed all at once. This is used for both ;; filter and remove. (define (sob-filter pred sob) (let ((result (sob-empty-copy sob))) (hash-table-for-each (lambda (key value) (if (pred key) (sob-increment! result key value))) (sob-hash-table sob)) result)) (define (set-filter pred set) (check-set set) (sob-filter pred set)) (define (bag-filter pred bag) (check-bag bag) (sob-filter pred bag)) (define (set-remove pred set) (check-set set) (sob-filter (lambda (x) (not (pred x))) set)) (define (bag-remove pred bag) (check-bag bag) (sob-filter (lambda (x) (not (pred x))) bag)) ;; Process each element and remove those that don't satisfy the filter. ;; This does its own cleanup, and is used for both filter! and remove!. (define (sob-filter! pred sob) (hash-table-for-each (lambda (key value) (if (not (pred key)) (sob-decrement! sob key value))) (sob-hash-table sob)) (sob-cleanup! sob)) (define (set-filter! pred set) (check-set set) (sob-filter! pred set)) (define (bag-filter! pred bag) (check-bag bag) (sob-filter! pred bag)) (define (set-remove! pred set) (check-set set) (sob-filter! (lambda (x) (not (pred x))) set)) (define (bag-remove! pred bag) (check-bag bag) (sob-filter! (lambda (x) (not (pred x))) bag)) ;; Create two sobs and copy the elements that satisfy the predicate into ;; one of them, all others into the other. This is more efficient than ;; filtering and removing separately. (define (sob-partition pred sob) (let ((res1 (sob-empty-copy sob)) (res2 (sob-empty-copy sob))) (hash-table-for-each (lambda (key value) (if (pred key) (sob-increment! res1 key value) (sob-increment! res2 key value))) (sob-hash-table sob)) (values res1 res2))) (define (set-partition pred set) (check-set set) (sob-partition pred set)) (define (bag-partition pred bag) (check-bag bag) (sob-partition pred bag)) ;; Create a sob and iterate through the given sob. Anything that satisfies ;; the predicate is left alone; anything that doesn't is removed from the ;; given sob and added to the new sob. (define (sob-partition! pred sob) (let ((result (sob-empty-copy sob))) (hash-table-for-each (lambda (key value) (if (not (pred key)) (begin (sob-decrement! sob key value) (sob-increment! result key value)))) (sob-hash-table sob)) (values (sob-cleanup! sob) result))) (define (set-partition! pred set) (check-set set) (sob-partition! pred set)) (define (bag-partition! pred bag) (check-bag bag) (sob-partition! pred bag)) ;;; Copying and conversion ;;; Convert a sob to a list; a special case of sob-fold. (define (sob->list sob) (sob-fold (lambda (elem list) (cons elem list)) '() sob)) (define (set->list set) (check-set set) (sob->list set)) (define (bag->list bag) (check-bag bag) (sob->list bag)) ;; Convert a list to a sob. Probably could be done using unfold, but ;; since sobs are mutable anyway, it's just as easy to add the elements ;; by side effect. (define (list->sob! sob list) (for-each (lambda (elem) (sob-increment! sob elem 1)) list) sob) (define (list->set comparator list) (list->sob! (make-sob comparator #f) list)) (define (list->bag comparator list) (list->sob! (make-sob comparator #t) list)) (define (list->set! set list) (check-set set) (list->sob! set list)) (define (list->bag! bag list) (check-bag bag) (list->sob! bag list)) ;;; Subsets ;; All of these procedures follow the same pattern. The ;; sob? procedures are case-lambdas that reduce the multi-argument ;; case to the two-argument case. As usual, the set? and ;; bag? procedures are trivial layers over the sob? procedure. ;; The dyadic-sob? procedures are where it gets interesting, so see ;; the comments on them. (define sob=? (case-lambda ((sob) #t) ((sob1 sob2) (dyadic-sob=? sob1 sob2)) ((sob1 sob2 . sobs) (and (dyadic-sob=? sob1 sob2) (apply sob=? sob2 sobs))))) (define (set=? . sets) (check-all-sets sets) (apply sob=? sets)) (define (bag=? . bags) (check-all-bags bags) (apply sob=? bags)) ;; First we check that there are the same number of entries in the ;; hashtables of the two sobs; if that's not true, they can't be equal. ;; Then we check that for each key, the values are the same (where ;; being absent counts as a value of 0). If any values aren't equal, ;; again they can't be equal. (define (sob-table-compare op ht1 ht2) (and (op (hash-table-size ht1) (hash-table-size ht2)) (not ($ hash-table-find ht1 (^[k v] (not (op v (hash-table-ref/default ht2 k 0)))))))) (define (dyadic-sob=? sob1 sob2) (sob-table-compare = (sob-hash-table sob1) (sob-hash-table sob2))) (define sob<=? (case-lambda ((sob) #t) ((sob1 sob2) (dyadic-sob<=? sob1 sob2)) ((sob1 sob2 . sobs) (and (dyadic-sob<=? sob1 sob2) (apply sob<=? sob2 sobs))))) (define (set<=? . sets) (check-all-sets sets) (apply sob<=? sets)) (define (bag<=? . bags) (check-all-bags bags) (apply sob<=? bags)) ;; This is analogous to dyadic-sob=?, except that we have to check ;; both sobs to make sure each value is <= in order to be sure ;; that we've traversed all the elements in either sob. (define (dyadic-sob<=? sob1 sob2) (sob-table-compare <= (sob-hash-table sob1) (sob-hash-table sob2))) (define sob=, since subset relationship ;; is partial order. Neither does it follow the same pattern as = and <=, (define (dyadic-sob v v2) #t] [else #f])))))]) (positive? small-count)))) (define sob>? (case-lambda ((sob) #t) ((sob1 sob2) (dyadic-sob>? sob1 sob2)) ((sob1 sob2 . sobs) (and (dyadic-sob>? sob1 sob2) (apply sob>? sob2 sobs))))) (define (set>? . sets) (check-all-sets sets) (apply sob>? sets)) (define (bag>? . bags) (check-all-bags bags) (apply sob>? bags)) ;; > is the inverse of <. this is only true dyadically. (define (dyadic-sob>? sob1 sob2) (dyadic-sob=? (case-lambda ((sob) #t) ((sob1 sob2) (dyadic-sob>=? sob1 sob2)) ((sob1 sob2 . sobs) (and (dyadic-sob>=? sob1 sob2) (apply sob>=? sob2 sobs))))) (define (set>=? . sets) (check-all-sets sets) (apply sob>=? sets)) (define (bag>=? . bags) (check-all-bags bags) (apply sob>=? bags)) ;; Finally, >= is the inverse of <=. (define (dyadic-sob>=? sob1 sob2) (dyadic-sob<=? sob2 sob1)) ;;; Set theory operations ;; A trivial helper function which upper-bounds n by one if multi? is false. (define (max-one n multi?) (if multi? n (if (> n 1) 1 n))) ;; The logic of union, intersection, difference, and sum is the same: the ;; sob-* and sob-*! procedures do the reduction to the dyadic-sob-*! ;; procedures. The difference is that the sob-* procedures allocate ;; an empty copy of the first sob to accumulate the results in, whereas ;; the sob-*! procedures work directly in the first sob. ;; Note that there is no set-sum, as it is the same as set-union. (define (sob-union sob1 . sobs) (if (null? sobs) sob1 (let ((result (sob-empty-copy sob1))) (dyadic-sob-union! result sob1 (car sobs)) (for-each (lambda (sob) (dyadic-sob-union! result result sob)) (cdr sobs)) result))) ;; For union, we take the max of the counts of each element found ;; in either sob and put that in the result. On the pass through ;; sob2, we know that the intersection is already accounted for, ;; so we just copy over things that aren't in sob1. (define (dyadic-sob-union! result sob1 sob2) (let ((sob1-ht (sob-hash-table sob1)) (sob2-ht (sob-hash-table sob2)) (result-ht (sob-hash-table result))) (hash-table-for-each (lambda (key value1) (let ((value2 (hash-table-ref/default sob2-ht key 0))) (hash-table-set! result-ht key (max value1 value2)))) sob1-ht) (hash-table-for-each (lambda (key value2) (let ((value1 (hash-table-ref/default sob1-ht key 0))) (if (= value1 0) (hash-table-set! result-ht key value2)))) sob2-ht))) (define (set-union . sets) (check-all-sets sets) (apply sob-union sets)) (define (bag-union . bags) (check-all-bags bags) (apply sob-union bags)) (define (sob-union! sob1 . sobs) (for-each (lambda (sob) (dyadic-sob-union! sob1 sob1 sob)) sobs) sob1) (define (set-union! . sets) (check-all-sets sets) (apply sob-union! sets)) (define (bag-union! . bags) (check-all-bags bags) (apply sob-union! bags)) (define (sob-intersection sob1 . sobs) (if (null? sobs) sob1 (let ((result (sob-empty-copy sob1))) (dyadic-sob-intersection! result sob1 (car sobs)) (for-each (lambda (sob) (dyadic-sob-intersection! result result sob)) (cdr sobs)) (sob-cleanup! result)))) ;; For intersection, we compute the min of the counts of each element. ;; We only have to scan sob1. We clean up the result when we are ;; done, in case it is the same as sob1. (define (dyadic-sob-intersection! result sob1 sob2) (let ((sob1-ht (sob-hash-table sob1)) (sob2-ht (sob-hash-table sob2)) (result-ht (sob-hash-table result))) (hash-table-for-each (lambda (key value1) (let ((value2 (hash-table-ref/default sob2-ht key 0))) (hash-table-set! result-ht key (min value1 value2)))) sob1-ht))) (define (set-intersection . sets) (check-all-sets sets) (apply sob-intersection sets)) (define (bag-intersection . bags) (check-all-bags bags) (apply sob-intersection bags)) (define (sob-intersection! sob1 . sobs) (for-each (lambda (sob) (dyadic-sob-intersection! sob1 sob1 sob)) sobs) (sob-cleanup! sob1)) (define (set-intersection! . sets) (check-all-sets sets) (apply sob-intersection! sets)) (define (bag-intersection! . bags) (check-all-bags bags) (apply sob-intersection! bags)) (define (sob-difference sob1 . sobs) (if (null? sobs) sob1 (let ((result (sob-empty-copy sob1))) (dyadic-sob-difference! result sob1 (car sobs)) (for-each (lambda (sob) (dyadic-sob-difference! result result sob)) (cdr sobs)) (sob-cleanup! result)))) ;; For difference, we use (big surprise) the numeric difference, bounded ;; by zero. We only need to scan sob1, but we clean up the result in ;; case it is the same as sob1. (define (dyadic-sob-difference! result sob1 sob2) (let ((sob1-ht (sob-hash-table sob1)) (sob2-ht (sob-hash-table sob2)) (result-ht (sob-hash-table result))) (hash-table-for-each (lambda (key value1) (let ((value2 (hash-table-ref/default sob2-ht key 0))) (hash-table-set! result-ht key (- value1 value2)))) sob1-ht))) (define (set-difference . sets) (check-all-sets sets) (apply sob-difference sets)) (define (bag-difference . bags) (check-all-bags bags) (apply sob-difference bags)) (define (sob-difference! sob1 . sobs) (for-each (lambda (sob) (dyadic-sob-difference! sob1 sob1 sob)) sobs) (sob-cleanup! sob1)) (define (set-difference! . sets) (check-all-sets sets) (apply sob-difference! sets)) (define (bag-difference! . bags) (check-all-bags bags) (apply sob-difference! bags)) (define (sob-sum sob1 . sobs) (if (null? sobs) sob1 (let ((result (sob-empty-copy sob1))) (dyadic-sob-sum! result sob1 (car sobs)) (for-each (lambda (sob) (dyadic-sob-sum! result result sob)) (cdr sobs)) result))) ;; Sum is just like union, except that we take the sum rather than the max. (define (dyadic-sob-sum! result sob1 sob2) (let ((sob1-ht (sob-hash-table sob1)) (sob2-ht (sob-hash-table sob2)) (result-ht (sob-hash-table result))) (hash-table-for-each (lambda (key value1) (let ((value2 (hash-table-ref/default sob2-ht key 0))) (hash-table-set! result-ht key (+ value1 value2)))) sob1-ht) (hash-table-for-each (lambda (key value2) (let ((value1 (hash-table-ref/default sob1-ht key 0))) (if (= value1 0) (hash-table-set! result-ht key value2)))) sob2-ht))) ;; Sum is defined for bags only; for sets, it is the same as union. (define (bag-sum . bags) (check-all-bags bags) (apply sob-sum bags)) (define (sob-sum! sob1 . sobs) (for-each (lambda (sob) (dyadic-sob-sum! sob1 sob1 sob)) sobs) sob1) (define (bag-sum! . bags) (check-all-bags bags) (apply sob-sum! bags)) ;; For xor exactly two arguments are required, so the above structures are ;; not necessary. This version accepts a result sob and computes the ;; absolute difference between the counts in the first sob and the ;; corresponding counts in the second. ;; We start by copying the entries in the second sob but not the first ;; into the first. Then we scan the first sob, computing the absolute ;; difference of the values and writing them back into the first sob. ;; It's essential to scan the second sob first, as we are not going to ;; damage it in the process. (Hat tip: Sam Tobin-Hochstadt.) (define (sob-xor! result sob1 sob2) (let ((sob1-ht (sob-hash-table sob1)) (sob2-ht (sob-hash-table sob2)) (result-ht (sob-hash-table result))) (hash-table-for-each (lambda (key value2) (let ((value1 (hash-table-ref/default sob1-ht key 0))) (if (= value1 0) (hash-table-set! result-ht key value2)))) sob2-ht) (hash-table-for-each (lambda (key value1) (let ((value2 (hash-table-ref/default sob2-ht key 0))) (hash-table-set! result-ht key (abs (- value1 value2))))) sob1-ht) (sob-cleanup! result))) (define (set-xor set1 set2) (check-set set1) (check-set set2) (check-same-comparator set1 set2) (sob-xor! (sob-empty-copy set1) set1 set2)) (define (bag-xor bag1 bag2) (check-bag bag1) (check-bag bag2) (check-same-comparator bag1 bag2) (sob-xor! (sob-empty-copy bag1) bag1 bag2)) (define (set-xor! set1 set2) (check-set set1) (check-set set2) (check-same-comparator set1 set2) (sob-xor! set1 set1 set2)) (define (bag-xor! bag1 bag2) (check-bag bag1) (check-bag bag2) (check-same-comparator bag1 bag2) (sob-xor! bag1 bag1 bag2)) ;;; A few bag-specific procedures (define (sob-product! n result sob) (let ((rht (sob-hash-table result))) (hash-table-for-each (lambda (elem count) (hash-table-set! rht elem (* count n))) (sob-hash-table sob)) result)) (define (valid-n n) (and (integer? n) (exact? n) (positive? n))) (define (bag-product n bag) (check-bag bag) (valid-n n) (sob-product! n (sob-empty-copy bag) bag)) (define (bag-product! n bag) (check-bag bag) (valid-n n) (sob-product! n bag bag)) (define (bag-unique-size bag) (check-bag bag) (hash-table-size (sob-hash-table bag))) (define (bag-element-count bag elem) (check-bag bag) (hash-table-ref/default (sob-hash-table bag) elem 0)) (define (bag-for-each-unique proc bag) (check-bag bag) (hash-table-for-each (lambda (key value) (proc key value)) (sob-hash-table bag))) (define (bag-fold-unique proc nil bag) (check-bag bag) (let ((result nil)) (hash-table-for-each (lambda (elem count) (set! result (proc elem count result))) (sob-hash-table bag)) result)) (define (bag->set bag) (check-bag bag) (let ((result (make-sob (sob-comparator bag) #f))) (hash-table-for-each (lambda (key value) (sob-increment! result key value)) (sob-hash-table bag)) result)) (define (set->bag set) (check-set set) (let ((result (make-sob (sob-comparator set) #t))) (hash-table-for-each (lambda (key value) (sob-increment! result key value)) (sob-hash-table set)) result)) (define (set->bag! bag set) (check-bag bag) (check-set set) (check-same-comparator set bag) (hash-table-for-each (lambda (key value) (sob-increment! bag key value)) (sob-hash-table set)) bag) (define (bag->alist bag) (check-bag bag) (bag-fold-unique (lambda (elem count list) (cons (cons elem count) list)) '() bag)) (define (alist->bag comparator alist) (let* ((result (bag comparator)) (ht (sob-hash-table result))) (for-each (lambda (assoc) (let ((element (car assoc))) (if (not (hash-table-contains? ht element)) (sob-increment! result element (cdr assoc))))) alist) result)) ;;; Comparators ;; Hash over sobs (define (sob-hash sob) (let ((hash (comparator-hash-function (sob-comparator sob)))) (sob-fold (lambda (element result) (logxor (hash element) result)) 5381 sob))) ;; Set and bag comparator (define set-comparator (make-comparator set? set=? #f sob-hash 'set-comparator)) (define bag-comparator (make-comparator bag? bag=? #f sob-hash 'bag-comparator)) ;;; Register above comparators for use by default-comparator ;; (comparator-register-default! set-comparator) ;; (comparator-register-default! bag-comparator) ;;; Set/bag printer (for debugging) ;; (define (sob-print sob port) ;; (display (if (sob-multi? sob) "&bag[" "&set[") port) ;; (sob-for-each ;; (lambda (elem) (display " " port) (write elem port)) ;; sob) ;; (display " ]" port)) ;; ;; Chicken-specific ;; (cond-expand ;; (chicken ;; (define-record-printer sob sob-print)) ;; (else))