10474 - Where is the Marble?

[Jewel of DIU]

I tried to solve the problem but got TLE.
Here is my search algorithm
[c]
---- CUT AFTER AC -----
(I HAVE USE COUNTING SORT)
[/c]
Help me Please to improve my time. :

[alu_mathics]

I also got TLE using searching.Then i use memorization and got AC.
Count the occurence of each number using array(size of 10000 and initially all 0).Then try to figure out your desire answer from this array.
And remember its a problem of TOMAL bhai.So what looks easy actually
it is not.He really makes us to think deeper and also optimal.

[sohel]

Using linear search , O(N) will certainly get TLE.

In the real time contest ( At Daffadil ) we tried to solve it using binary search but still could not pass the time limit ( of 1 second ) .

You have to use counting sort to solve this problem.
If you are not familiar with counting sort then:
maintain a array which counts the frequecny of each number.
use another array that maintains the cumulative frequency....
then do whatever is required.

Hope it helps.

[Jewel of DIU]

I have now got WA .
Please Give me some Input and Output. So that i can find my code's bug.

[alu_mathics]

input:
10 5

1
1
3
5
7
1
3
9
120
1

1
3
120
7
1234

10 12

1111
1111
1111
5
5
5
10
11
12
10

1111
5
10
11
12
1111
134
567
246
24
12
435

output:
CASE# 1:
1 found at 1
3 found at 5
120 found at 10
7 found at 8
1234 not found
CASE# 2:
1111 found at 8
5 found at 1
10 found at 4
11 found at 6
12 found at 7
1111 found at 8
134 not found
567 not found
246 not found
24 not found
12 found at 7
435 not found

//mind it no value will be greater than 10000

[Jewel of DIU]

Thank you Shohel bhai & Aftab Bhai for your help. I have now got AC.
Thank u very very very much.

[WR]

I know there's a more effective solution and sorting's not necessary, but anyway...

Using quicksort and binary search my program received a WA in
about 0,1 seconds (forgot the exact timing, less than 0,2 s in any case).

My question is: When does the OJ return a WA? Immediately after getting a wrong result or only after evaluating all sample input?

Thank you

Oh, by the way, some test data would be welcome. There's no difference between my output and alu_mathics' output!

[Dominik Michniewski]

If your program not exceed time limit t finish first, and after that comparing your output witth judge output is done. Depends of result of comparing apropriate message is given to you
So if you got WA in less than 0.2 sec, than you'va been sure, that your program run trough all cases, but got wrong results to output ...

Best regards
DM

PS. When I try to solve this problem (on contest time) I have the same problem - I got WA or TLE,l depending on algorithm I use. But at end I manage wirth described below algorithm (with counting) and got Acc

[WR]

Thanks DM,

I wanted to know that for quite some time, but always forgot
to ask.

As to the problem, I'll try with counting sort.

bye

added 22.3.04

Accepted using counting sort!

[kiha]

I have WA, but I don't understand why ...

I used quicksort & binary search. I knew there exists a quicker algorithm using counting sort , but I would like to know why this gives me WA:

[pascal]
program miazio;

type mia = array [1..10000] of longint;

var
t:mia;
i, j, k, l:longint;
le, e, lz, z, kejs:longint;
poz, lewy, srodek, prawy:longint;


begin

kejs:=0;

while not eof do
begin

readln (le, lz);
if (le = 0) and (lz = 0) then exit;

inc (kejs);
writeln ('CASE# ', kejs, ':');

for i:=1 to le do
readln (t );

qsort (t);

for z:=1 to lz do
begin

readln (l);

lewy:=1;
prawy:=le;

repeat

srodek:=(lewy + prawy) div 2;

if t [srodek] <> l then
if t [srodek] < l then
lewy:=srodek else
prawy:=srodek;

until (t [srodek] = l) or (lewy >= prawy) or ((lewy = prawy - 1) and (lewy = srodek));

if (lewy = srodek) and (lewy = prawy - 1) then inc (srodek);

if t [srodek] <> l then
writeln (l, ' not found') else

begin

while (t [srodek - 1] = l) and (srodek > 1) do
dec (srodek);

writeln (l, ' found at ', srodek);

end;

end;

end;

end.
[/pascal]

I deleted procedure quicksort, which must be right as I cut if from Free Pascal's sample qsort.pp

[miras]

Dear Kiha

Sth is wrong with you BS
( try to rewrite it )

BTW i think that it is why u get WA...

[Salman]

got TLE on 10474

How can I improve????

[code]

code removed after AC


[/code]

[sidky]

change the whole approach. the solution can be retrived in constant time.

[Mohammad Mahmudur Rahman]

to Salman -
You can get it within time limit using binary search, too. But you must use a faster (n log n) sorting algorithm.

to Sidky -
Will you explain what is the O(1) method? I can see a O(n) solution using counting sort & AFAIK the judge solution for this problem is based on this method, too. But O(1) seems to be too much for me.

[Sedefcho]

Mohammad Mahmudur Rahman,

I suggest you read the previous posts about this problem.

What is meant by O(1) is that you can give the answer
for each Query Number in a given TEST in a O(1) time.

What I call a TEST is (A) plus (B), where:
(A) is the SET of the N numbers which are given
(B) is the SET of the Q numbers ( the SET of the query numbers )

So how it goes ?
(1) you read the N numbers which are given
(2) you do some preprocessing
(3) you start reading the Q query numbers and each time a query
number comes from the input you can give the answer for that
query number in O(1) time

The key phrase in this problem is that the numbers given are
in the interval [1,10000]. Which allows a memoization approach
to be used ( that memoization is done in the preprocessing
phase - see (2) above ).

No binary search is needed, no sorting, no complex data
structures. Two arrays of 10000 elements each will be enough.