Hi Hermann,
There's no array bug -- it's an aspect of its intentional behaviour.
Appending to an element of an array always causes a copy, so you get
O(N^2) complexity.
Please re-test in LiveCode 8.0.0-dp-7, because there are some large
performance optimisations in that release that may affect these results.
In particular, I expect anything that uses "put after" outside an
array to be much faster, and anything that uses sort to be faster (if I
remember correctly). This Week in LiveCode has links to the pull
requests, which explain exactly what's been made faster.
Peter
On 13/10/2015 14:42, hh wrote:
Similar things are useful for a very specialized statistical
evaluation I have to do soon, so I tested again with LC 7:
There is an interesting difference between LC versions 6 and 7.
The 'direct way' of GC reveals now as clearly fastest.
Tested with one of GC's 12 MB strings on a Mac mini(i5 2.5 GHz),
testString() as below, and a critical value of v=500000000.
The (correct) result is 499999246, used time is in seconds.
[a] LC 6.7.7
repeat-PMB 0.326 <- fastest
repeat-GC 0.350
repeat-hh 0.445
sort-GC 0.985
[b] LC 7.1.0
repeat-PMB 3.562
repeat-GC 1.483 <- fastest, the only one with a factor < 9
repeat-hh 1163.0 <- an ARRAY BUG? :-(
sort-GC 8.156
-------------------
function testString
set randomseed to 1444826515
-- so you'll get exactly my test string
repeat 1000000
put comma & random(1000000000) after L
end repeat
delete byte 1 of L
return L
end testString
function repeatPMB L,v --[1]
repeat for each item i in L
if i < v then put comma & i after outL
end repeat
return max(outL)
end repeatPMB
function repeatGC L,v --[2]
put empty into R
repeat for each item i in L
if i < v and i > R then put i into R
end repeat
return R
end repeatGC
function repeathh L,v --[3]
repeat for each item i in L
put comma & i after z[i < v]
end repeat
return max(z[true])
end repeathh
function sortGC L,v --[4]
sort items of L descending numeric
sort items of L by (each >= v)
return item 1 of L
end sortGC
_______________________________________________
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode
--
Dr Peter Brett <peter.br...@livecode.com>
LiveCode Open Source Team
LiveCode on reddit: https://reddit.com/r/livecode
_______________________________________________
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode
