10784 - Diagonal

sumankar · Post by **sumankar** » Sat Mar 12, 2005 11:33 am

I am trying this for ages now ...

...
    long long int d, n;
...  
    double m;
...
/*  This is where I calculate the answer, ndia = n*(n-3)/2,
 ** frame a quadriatic, get the roots, take the +ve one */   
    m = (3.0 + sqrt( 9.0 + (d*8.0)))/2.0; 
   /* and output */
    printf(".0lf", m+0.5);

.. · Post by .. » Sat Mar 12, 2005 1:53 pm

Precision error.
I used +0.5 as theshold when I need to round a "normal" floating point number(say, 3.4, 1.2, etc.) to integer.
If a floating point number is storing a number, which should be an integer in theory, I will use much smaller theshold like +1e-10. You will get AC by trying different combination of theshold value and ceil() or floor().

neno_uci · Post by **neno_uci** » Sat Mar 12, 2005 2:09 pm

I solved this approximating the value of N via binary search..., i know this is not the best method but i am not so good at math but it worked ok...,

sumankar · Post by **sumankar** » Sat Mar 12, 2005 2:33 pm

Thanks, but

.. wrote:Precision error.
I used +0.5 as theshold when I need to round a "normal" floating point number(say, 3.4, 1.2, etc.) to integer.
If a floating point number is storing a number, which should be an integer in theory, I will use much smaller theshold like +1e10. You will get AC by trying different combination of theshold value and ceil() or floor().

+1e10 is small?

Code: Select all

#define EPS = 1e-8
m = (3.0 + sqrt( 9.0 + (d*8.0)))/2.0; 
        frac = modf(m, &whole);
        if ( frac < EPS )
            ;
        else
            whole++;
        printf("Case %d: %.0lf\n", k, whole);

doesn't work either.
Suman.

.. · Post by .. » Sat Mar 12, 2005 2:42 pm

Sorry typo, should be 1e-10.
fmod() sucks, I never use it. And I already say using ceil or floor is enough.

The problem of your code is
printf(".0lf", m+0.5);
People think it will work perfectly in rounding, but it doesn't.

sumankar · Post by **sumankar** » Sat Mar 12, 2005 2:58 pm

.. - I resign.Can you give me how to do it?

I tried everything, in my capacity(rather limited, you see...

)

Suman

.. · Post by .. » Sat Mar 12, 2005 3:05 pm

This is what I used in my AC code:

Code: Select all

 printf ("Case %d: %d\n", t++,          (int)ceil((3.0+sqrt(9.0+8.0*N))/2.0-1e-10));

sumankar · Post by **sumankar** » Sat Mar 12, 2005 3:11 pm

All of which gives a WA ...

I 'll go crazy.

Suman.

.. · Post by .. » Sat Mar 12, 2005 3:18 pm

If the line of AC code can't give you AC, then you must have some other problems.

P.S. It seems that I don't remember the detail correctly, the second method in my last post is wrong.

sumankar · Post by **sumankar** » Sat Mar 12, 2005 3:22 pm

Code: Select all

scanf() == 1 && d

was the culprit .

Here is the culprit, may be you can find something in here...
Suman.

.. · Post by .. » Sat Mar 12, 2005 3:24 pm

The mistake is obvious......
Run the sample input again........

sumankar · Post by **sumankar** » Sat Mar 12, 2005 3:26 pm

..

Thanks a zillion tonnes ... I should have killed myself by now for that stupid^inf bug.

Suman.

Debashis Maitra · Post by **Debashis Maitra** » Wed Dec 20, 2006 6:59 pm

WHY WA

Code: Select all


Removed after AC

}

Jan · Post by **Jan** » Wed Dec 20, 2006 10:56 pm

Debashis, your code is correct, except two things.

1. Case ('C' capital).
2. The problem states...

Input is terminated by a line containing a zero

So, if you get a 0 then you can finish processing.

Hope you get accepted.

Debashis Maitra · Post by **Debashis Maitra** » Fri Dec 22, 2006 6:33 pm

Thanx Jan Bhai

that was a silly mistake