On Dec 30, 2010, at 02:56 , Cyrus Harmon wrote:
CL-USER> (reductions #'+ '(1 2 3 4)) (1 3 6 10)
I find this a bit surprising behavior: CL-USER> (reductions #'+ '(1 2 3 4) :initial-value 10) (1 11 13 16 20) I would have expected (11 13 16 20) or (10 11 13 16 20). Here's a counter-proposal, with a similar interface as MAP: CL-USER> (left-scan 'vector #'+ 10 '(1 2 3 4)) #(11 13 16 20) ;; (vector (+ 10 1) (+ 11 2) ...) CL-USER> (right-scan 'list #'+ 0 '(1 2 3 4)) (4 7 9 10) ;; (list (+ 0 4) (+ 4 3) (+ 7 2) ...) CL-USER> (left-scan1 'vector #'+ '(1 2 3 4)) #(3 6 10) ;; (vector (+ 1 2) (+ 3 3) (+ 6 4)) (defun left-scan (type fn init seq &rest seqs) (flet ((next (&rest args) (setf init (apply fn init args)))) (apply #'map type #'next seq seqs))) (defun left-scan1 (type fn seq) (if (= 0 (length seq)) (coerce '() type) (scanl type fn (elt seq 0) (subseq seq 1)))) (defun left-scan-into (result fn init &rest seqs) (flet ((next (&rest args) (setf init (apply fn init args)))) (apply #'map-into result #'next seqs))) (defun right-scan (type fn init seq &rest seqs) (let ((rev-seqs (mapcar #'reverse (list* seq seqs)))) (apply #'left-scan type fn init rev-seqs))) (defun right-scan1 (type fn seq) (apply #'left-scan1 type fn (reverse seq))) etc. Obviously under-documented, but the intent should be identifyable. -- Cheers, Michael