Random sort demonstration

[Jacques Hausser]

Thanks, Dar, now I understand !

Till yesterday evening, I always used the numeric result of random(tUpperlimit) to "de-sort" chunks of text, being not aware of the use of function's combination like "sort lines of […] random(xxx)". 
… and at least my scripts worked exactly as I expected !

IMHO, the two uses of "random" arequite different and should be clearly distinguished in the language.

"put random(tUpperLimit) into tRanVar" -- OK, this is standard

and something like

"scramble lines of myVar", instead of "sort lines of myVar random(xxx)". Could be a command using the same function behind the scene, with the max integer as parameter to avoid prevalence of "the first" in the case of repetitions. And it is semantically clearer than "sort random".

Meanwhile, it should be useful to put an user note in the dictionary clarifying the process. A vous l'honneur !

Jacques



Le 23 mai 2013 à 08:08, Dar Scott <dsc at swcp.com> a écrit :

> This script demonstrates what is going on with random sorts of short lists using this:
> 
>      sort items of myVar by random( the number of items of myVar)
> 
> This is a button script on a stack with a wide field with a scrollbar.
> 
> local lineAssignments
> 
> on mouseUp
>   set the cursor to watch
>   put empty into field 1
>   put "a,b,c" into original
>   put 0 into numFirstSame
>   put 100 into trials
>   repeat trials times
>      put original into myVar
>      put empty into lineAssignments
>      sort items of myVar by randomLog( the number of items of myVar, each)
>      delete char -1 of lineAssignments
>      put "(" & original & ") is assigned (" & lineAssignments & ") and sorted as (" & myVar & ") where the" after field 1
>      if  (item 1 of original) is (item 1 of myVar) then
>         put " first is same" after field 1
>         add 1 to numFirstSame
>      else
>         put " first is different" after field 1
>      end if
>      put return after field 1
>   end repeat
>   put round( (numFirstSame/trials) * 100) into percentSame
>   put "The first was the same " & percentSame & "% of the time" after field 1
>   set the scroll  of field 1 to 999999
> end mouseUp
> 
> function randomLog pRange, pEach
>   get random(pRange)
>   put pEach & "->" & it & comma after lineAssignments
>   return it
> end randomLog
> 
> Note that this uses function randomLog() instead of random() so the assignments can be gleamed.
> 
> These are the last few lines of a typical output:
> 
> (a,b,c) is assigned (a->3,b->2,c->2) and sorted as (b,c,a) where the first is different
> (a,b,c) is assigned (a->1,b->2,c->2) and sorted as (a,b,c) where the first is same
> (a,b,c) is assigned (a->1,b->2,c->1) and sorted as (a,c,b) where the first is same
> (a,b,c) is assigned (a->2,b->2,c->3) and sorted as (a,b,c) where the first is same
> The first was the same 53% of the time
> 
> I got percentages in the range 45% to 60%.  My theory (other email) predicted 52 percent.  You can try upping the number of trials and commenting out the per-sort lines, and see what you get.
> 
> The problem is equality in the sort.  It keeps the same order in comparison of pairs of items.  For example, the items sorted in the last case above as though they were 2,2,3.  The first item is still first.  
> 
> So...
> 
> Use large values for the argument to random() in random sorts.
> 
> Dar
> 
> 
> 
> 
> 
> 
> 
> _______________________________________________
> use-livecode mailing list
> use-livecode at lists.runrev.com
> Please visit this url to subscribe, unsubscribe and manage your subscription preferences:
> http://lists.runrev.com/mailman/listinfo/use-livecode