I can't verify that. For some values it is faster, yes, but it seems
measurements with 1 000 000 items shows your original assumption was
correct. See below. /Erling
NB. Find the longest ascending subsequence in y
NB. y is a list of numbers
NB. Result is indexes of the values that form the la
It depends on both m and n in ?m#n
Also, note that it's better to move definition of the input outside the
time
(and space) test.
Using ?. to fix the samples, and 1 ts to save waiting on longer arguments:
ts each '#longascseq1 q';'#longascseq2 q'[q =: ?.~10NB. set outside
+---
simple rules,
think only of fork (not hook). So odd number of tines, and (0 index) odd verbs
are dyads, evens are ambivalent (same valence as whole verb)
as an exercise, look at the hook rule, and write an equivalent fork for it.
[: will look the most like explicit. in a fork, its only vali
Any ideas about how you could make a substantially faster tacit solution
to this problem or how you could make this solution substantially
faster? /Erling
On 2016-09-13 18:17, Erling Hellenäs wrote:
Hi all !
My first idea only took me down to 1 second for 10 000 numbers, 25
times longer tha
The original question is about composing functions, but if you are really
trying to count characters there are much more efficient ways. For example:
x=: 'abcdefghijklmnopqrstuvwxyz'
y=: a.{~ 1e6 ?@$ #a.
c=: <: (#x) {. #/.~ x,y
$c
26
c
4019 3829 3893 3912 3889 4020 3965 3936 3960 3
I suspect it's pretty near optimal. The wikipedia article points at
Fredman's paper:
http://www.sciencedirect.com/science/article/pii/0012365X7590103X
analysing an algorithm of Knuth's. He shows that it performs better than
n log2(n) - n log2 log2 (n) + O(n)
rather than the O(n logn) cited
