> Hi, the full code is attached (I hope Google Groups will preserve it...).
Thank you for this! There is absolutely a performance gap, and I'll definitely
look over it and see if I can figure out exactly why (I think a well-built
sequence-based model should have comparable speed). I did implement an
equivalent test to the one in your blog post, which I've included at the end of
this email. While the stateful implementation does indeed perform noticeably
better, the results I got with my lazy implementation are not nearly as dire as
you seem to have described.
Specifically, the version operating on 100,000 elements and 1,000,000 elements
yielded the following times, respectively:
; cpu time: 774 real time: 774 gc time: 195
; cpu time: 7029 real time: 7045 gc time: 1814
I did not encounter any out-of-memory problems in either test.
The test I used is as follows:
#lang racket
(require alexis/collection)
(define (sum seq)
(foldl + 0 seq))
(define (test-sequence n foo)
(compose
sum
(curry take n)
(curry map foo)
(curry filter odd?)))
(time
((test-sequence
1000000
(lambda (x)
(* x
(sin (* pi (/ x 12)))
(+ 1 (exp (- (/ x 100)))))))
(in-naturals)))
--
You received this message because you are subscribed to the Google Groups
"Racket Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to racket-users+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
