;; ;; testing mutable and immutable char-set lookup speed ;; (use gauche.time) (use data.random) (use gauche.sequence) (use scheme.list) (use scheme.charset) (use srfi.42) (define *cats* '(Lu Ll Lt Lm Lo Mn Nd Nl No Po Sm Zs Cc Co)) ;; Returns a map (define (setup-char-sets) (let1 vec (vector-tabulate (length *cats*) (^_ (char-set))) (do-ec (: c 0 #x20000) (let1 ch (ucs->char c) (if-let1 j (list-index (cute eq? <> (char-general-category ch)) *cats*) (char-set-adjoin! (~ vec j) ch)))) (map (^[cs] (cons cs (char-set-freeze cs))) vec))) (define (setup-input-data) (take (generator->lseq (chars$ #[\x0000;-\x1ffff;])) 2000000)) (define (runner-1 cs input) (^[] (dolist [i input] (char-set-contains? cs i)))) ;; returns (cat ) (define (run-benchmark csss input) ($ map (^x (list (caar x) (cdar x) (cdadr x))) $ (cut slices <> 2) $ cdr $ time-these 1000 $ append-map (^[cat css] `((,cat . ,(runner-1 (car css) input)) (,#"i~cat" . ,(runner-1 (cdr css) input)))) *cats* csss)) (define (report-benchmark res) (define (row title ta tb) (format #t "~6a ~8@a ~8@a ~8@a ~8@a ~8@a ~8@a ~8@a\n" title (fmt (time-result-real ta)) (fmt (time-result-user ta)) (fmt (time-result-sys ta)) (fmt (time-result-real tb)) (fmt (time-result-user tb)) (fmt (time-result-sys tb)) (fmt (* 100 (/ (time-result-user ta) (time-result-user tb)))))) (format #t " mutable immutable\n") (format #t "category real user sys real user sys speed ratio\n") (dolist [x res] (apply row x))) (define (report-charset-stats charsets) (define (cs-chunks cs) (length (filter (^p (>= (cdr p) #x80)) ((with-module gauche.internal %char-set-ranges) cs)))) (define (cs-report-1 category css) (format #t "~6a ~4@a ranges, ~6@a chars\n" category (cs-chunks (car css)) (char-set-size (car css)))) (for-each cs-report-1 *cats* charsets)) (define (fmt num) (let* ([integral (floor->exact num)] [fraction (- num integral)] [base (* fraction 1000)] [base1 (floor->exact base)] [bfraction (- base base1)]) (when (>= bfraction 0.5) (inc! base1)) (when (>= base1 1000) (inc! integral) (dec! base1 1000)) (format "~d.~3,'0d" integral base1))) (define (main args) (let1 charsets (setup-char-sets) ; (( . ) ...) (report-benchmark (run-benchmark charsets (setup-input-data))) (print) (report-charset-stats charsets)))