A question for tsh73

[honkytonk]

In this thread ( justbasiccom.proboards.com/thread/821/help-crazy ), you gave me the closest point calculation.

   g=1
   #w.g, "color white;backcolor darkgray"
   minDist=1e10  '10000 also works.
   for x=1 to 800
     for y=my-2 to 3 step -1
       if map(x,y)=1 then
           dist = sqr((x-mx)^2+(y-my)^2)
           'print mx, my, x, y, dist,  minDist
           if dist < minDist then
               minDist= dist
               tagx(g)=x: tagy(g)=y
           end if
        ' pointy(y-1)=0
       end if
     next y
     #w.g, "place 20 40": #w.g, "\ "; str$(x)
   next x
   #w.g, "color red;backcolor darkgray"
   for x=1 to g
     #w.g, "size 5": #w.g, "set ";tagx(x);" ";tagy(x)
   next x

I don't understand the reason for this:

minDist=1e10  '10000 also works.
'....
'....
if dist < minDist then

Could you explain it to me?

Thank you for...

[tsh73]

well,
This is standard way of searching MIN out of something:
minDist= "big enough number"
for i=...
   dist="some calculation"
   if dist < minDist then
        minDist=dist
        'also you can store "i" on which it is happened in some variable
   end if
next
Now, "big enough number" should be big enough so first calculated "dist" will trigger condition.
Since (usually?) there less then 2000 pixels on screen, 10000 is big enough.
But if I want to save some time and use not distance
sqr(x^2+y^2)
but distance squared
x^2+y^2
then suddenly 10000 is not enough.
I thought about it and put 1e10, that means 1*10^10 or 10000000000
But printing exactly 10 zeros is awkward (just now I had to split it two groups to cout, and it happened I missed one zero)
While printing 1e10 is just 4 characters, easily recognised and defnitely big enough.

That's all.
(oh. And since I am part time teacher, I used to explain this and put 1e10 as big enough number.
So why not 1e100? Well, that extra character to type. Why not 1e99?
Now, this stems from really old time then floating point numbers were restricted to 1e38)

[honkytonk]

So we cannot calculate the nearest point without going through this condition?
Why ?

[tsh73]

Getting that WHY makes me nervous (not really).
Why do you asking?

So we cannot calculate the nearest point without going through this condition?

This is condition
if dist < minDist then
and it is standart working way to search for minimum.

Now can we

calculate the nearest point without going through this condition?

I have no idea.
(actually I have a couple, but I pretty sure it would have same condition inside, may be hidden from our sight)
And since I have working way I'd rather not bother, really.
EDIT
Is something in this solution does not suit you?
Too slow may be?

[honkytonk]

This condition is very troubling;
And I would like to know his reason for being there.
Doing without it is not the point, even if one could list all the hypothenuses at the last plotted point, then sort the values and the indices (associated indices), at the same time to extract the minimum.

I understand that this is "the standart working way".
But I don't understand why.

[Rod]

It is really very simple. We are comparing two numbers to see which is lower. At the very start of the comparison we only have the first point. The number we compare it against MUST be bigger than it. So we use a very big number, as big as we can imagine, safe in the knowledge that it will be bigger than our first point and so the first point will become the lowest point. Until the next comparison where it gets compared with the next real point and may or may not stay as the lowest point.

[honkytonk]

Ah ok, now I understand.
Thanks a lot.

[tsh73]

Thanks Rod ;))