Funcion returning an extra 0 - c++11

I have this function which calculates distance between two point, everything works fine but I get an extra zero at the end.
My input:
2 3, 12 30, 40 50, 5 1, 12 10, 3 4
float bruteForce(vector< pair <float,float> >::const_iterator brute, vector< pair <float,float> >::const_iterator end, float m)
{
float min = FLT_MAX;
for (int i = 0; i < m; i++)
{
for (int j = i+1; j < m; j++)
{
if ( dist(brute[i].first, brute[i].second, brute[j].first, brute[j].second ) < min)
{
min = dist(brute[i].first, brute[i].second, brute[j].first, brute[j].second);
cout << min << endl;
}
}
}
return min;
}
output:
1.41421,
11.4018,
34.4093,
0

A 0 indicates "no distance", so there must be two points on the same location.
When I look at your input, it appears the 3 is there twice.
So this is the shortest distance, after that, it's impossible to find anything else shorter.

Quentin was correct, it was because of the float data types I was using where I shouldn't be using them.

Related

Numbers of specified length that can be made using individual elements from an array

Let's say we have given an array of digits, A, and a positive number, B. The problem is to generate all the possible B-digit numbers combined of A's elements.
For example, if A = [0,1,2,3] and B = 2, then the output must be,
[10,11,12,13,20,21,22,23,30,31,32,33]
Generate all possible combinations​ of 2 digit numbers by multiplying and adding the elements of the array in nested for loops.
Check if the generated numbers are greater than 10 to be a valid two digit number.
`
#include<iostream>
#include<cmath>
int main () {
int A[4] = {0,1,2,3}; int B = 2; int k;
for(size_t i = 0; i < sizeof(A)/sizeof(A[0]); i++)
{
for(size_t j = 0; j < sizeof(A)/sizeof(A[0]); j++)
{
k = (A[i] * pow(10, B-1) + j);
if(k / 10 > 0)
std::cout << k << '\n';
}
}
}

Implementation of partition an array into 2 parts such that the two parts have equal average

I am implementing the approach described in this question for the same problem but I dont think this is working.
For those who dont want to go through the mathematics there, here is the algebra in gist:
Average = Sum(S1)/n(S1) = Sum(S2)/ n(S2) = Sum(Total)/n(Total)
where n() stands for the number of elements in the array & Sum() stands for the cumulative sum
S1 and S2 are mutually exclusive subsets of array Total. Thus to find the required subset where this condition will hold true, we find Sum(S1) = Sum(Total) * n(S1)/n(Total)
My approach:
#include <bits/stdc++.h>
using namespace std;
bool SubsetSum(vector<int> &A, int Sum)
{
bool dp[Sum+1][A.size()+1];
int i, j;
for(i=0; i<= A.size(); i++)
dp[0][i] = false; // When sum = 0
for(i=0; i<=Sum; i++)
dp[i][0] = 1; // When num of elements = 0
for(i = 1; i <= A.size(); i++)
{
for(j=1; j<= Sum; j++)
{
dp[i][j] = dp[i-1][j];
if(j-A[i-1] >= 0)
dp[i][j] = dp[i][j] || dp[i-1][j-A[i-1]];
}
}
return dp[Sum][A.size()];
}
void avgset(vector<int> &A) {
int total = accumulate(A.begin(), A.end(), 0); // Cumulative sum of the vector A
int ntotal = A.size(); // Total number of elements
int i;
for(i=1; i<=ntotal; i++) // Subset size can be anything between 1 to the number of elements in the total subset
{
if((total * i) % ntotal == 0)
{
if(SubsetSum(A, (total * i)/ntotal)) // Required subset sum = total * i)/ntotal
cout<<"Array can be broken into 2 arrays each with equal average of "<<(total * i)/ntotal<<endl;
}
}
}
int main()
{
vector<int> A = {1, 7, 15, 29, 11, 9};
avgset(A);
return 0;
}
This code outputs:
Array can be broken into 2 arrays each with equal average of 12
Array can be broken into 2 arrays each with equal average of 36
Array can be broken into 2 arrays each with equal average of 60
But these answers are wrong.
For example, when subset sum = 12, the corresponding elements will be {11, 1}. Then:
(11 + 1)/2 != (7 + 15 + 29 + 9)/4
Have I misunderstood something here?
Have I misunderstood something here?
Seems you did.
For given array average always is equal to 12 - total average, first subset average, second subset average.
So you have to check:
1-element subset with sum 12 - does not exist
2-element subset with sum 24 - does exist 9+15
3-element subset with sum 36 - does not exist
and there is no need to check larger sums (>n/2)
The element number of subset sum should be specified. Find element less equal than n/2 is enough. And there are other errors. Codes as below:
bool SubsetSum(vector<int> &A, int number, int Sum)
{
bool dp[Sum+1][A.size()+1];
int i, j;
for(i=0; i<= A.size(); i++)
for (j = 0; j <= Sum; j++)
dp[j][i] = false; // When sum = 0
dp[0][0] = true; // When num = 0 of 0 elements
for(i = 1; i <= A.size(); i++)
{
for(j=Sum; j>=A[i-1]; j--)
{
for (int k = A.size(); k > 0; k--)
dp[j][k] = dp[j][k] || dp[j-A[i-1]][k-1];
}
}
return dp[Sum][number];
}
void avgset(vector<int> &A) {
int total = accumulate(A.begin(), A.end(), 0); // Cumulative sum of the vector A
int ntotal = A.size(); // Total number of elements
int i;
for(i=1; i<=ntotal/2; i++) // Subset size can be anything between 1 to the number of elements in the total subset
{
if((total * i) % ntotal == 0)
{
if(SubsetSum(A, i, (total * i)/ntotal)) // Required subset sum = total * i)/ntotal
cout<<"Array can be broken into 2 arrays each with equal average of "<<(total * i)/ntotal<<endl;
}
}
}
output:
Array can be broken into 2 arrays each with equal average of 24

Dynamic Programming Coin Change Limited Coins

Dynamic Programming Change Problem (Limited Coins).
I'm trying to create a program that takes as INPUT:
int coinValues[]; //e.g [coin1,coin2,coin3]
int coinLimit[]; //e.g [2 coin1 available,1 coin2 available,...]
int amount; //the amount we want change for.
OUTPUT:
int DynProg[]; //of size amount+1.
And output should be an Array of size amount+1 of which each cell represents the optimal number of coins we need to give change for the amount of the cell's index.
EXAMPLE: Let's say that we have the cell of Array at index: 5 with a content of 2.
This means that in order to give change for the amount of 5(INDEX), you need 2(cell's content) coins (Optimal Solution).
Basically I need exactly the output of the first array of this video(C[p])
. It's exactly the same problem with the big DIFFERENCE of LIMITED COINS.
Link to Video.
Note: See the video to understand, ignore the 2nd array of the video, and have in mind that I don't need the combinations, but the DP array, so then I can find which coins to give as change.
Thank you.
Consider the next pseudocode:
for every coin nominal v = coinValues[i]:
loop coinLimit[i] times:
starting with k=0 entry, check for non-zero C[k]:
if C[k]+1 < C[k+v] then
replace C[k+v] with C[k]+1 and set S[k+v]=v
Is it clear?
O(nk) solution from an editorial I wrote a while ago:
We start with the basic DP solution that runs in O(k*sum(c)). We have our dp array, where dp[i][j] stores the least possible number of coins from the first i denominations that sum to j. We have the following transition: dp[i][j] = min(dp[i - 1][j - cnt * value[i]] + cnt) for cnt from 0 to j / value[i].
To optimize this to an O(nk) solution, we can use a deque to memorize the minimum values from the previous iteration and make the transitions O(1). The basic idea is that if we want to find the minimum of the last m values in some array, we can maintain an increasing deque that stores possible candidates for the minimum. At each step, we pop off values at the end of the deque greater than the current value before pushing the current value into the back deque. Since the current value is both further to the right and less than the values we popped off, we can be sure they will never be the minimum. Then, we pop off the first element in the deque if it is more than m elements away. The minimum value at each step is now simply the first element in the deque.
We can apply a similar optimization trick to this problem. For each coin type i, we compute the elements of the dp array in this order: For each possible value of j % value[i] in increasing order, we process the values of j which when divided by value[i] produces that remainder in increasing order. Now we can apply the deque optimization trick to find min(dp[i - 1][j - cnt * value[i]] + cnt) for cnt from 0 to j / value[i] in constant time.
Pseudocode:
let n = number of coin denominations
let k = amount of change needed
let v[i] = value of the ith denomination, 1 indexed
let c[i] = maximum number of coins of the ith denomination, 1 indexed
let dp[i][j] = the fewest number of coins needed to sum to j using the first i coin denominations
for i from 1 to k:
dp[0][i] = INF
for i from 1 to n:
for rem from 0 to v[i] - 1:
let d = empty double-ended-queue
for j from 0 to (k - rem) / v[i]:
let currval = rem + v[i] * j
if dp[i - 1][currval] is not INF:
while d is not empty and dp[i - 1][d.back() * v[i] + rem] + j - d.back() >= dp[i - 1][currval]:
d.pop_back()
d.push_back(j)
if d is not empty and j - d.front() > c[i]:
d.pop_front()
if d is empty:
dp[i][currval] = INF
else:
dp[i][currval] = dp[i - 1][d.front() * v[i] + rem] + j - d.front()
This is what you are looking for.
Assumptions made : Coin Values are in descending order
public class CoinChangeLimitedCoins {
public static void main(String[] args) {
int[] coins = { 5, 3, 2, 1 };
int[] counts = { 2, 1, 2, 1 };
int target = 9;
int[] nums = combine(coins, counts);
System.out.println(minCount(nums, target, 0, 0, 0));
}
private static int minCount(int[] nums, int target, int sum, int current, int count){
if(current > nums.length) return -1;
if(sum == target) return count;
if(sum + nums[current] <= target){
return minCount(nums, target, sum+nums[current], current+1, count+1);
} else {
return minCount(nums, target, sum, current+1, count);
}
}
private static int[] combine(int[] coins, int[] counts) {
int sum = 0;
for (int count : counts) {
sum += count;
}
int[] returnArray = new int[sum];
int returnArrayIndex = 0;
for (int i = 0; i < coins.length; i++) {
int count = counts[i];
while (count != 0) {
returnArray[returnArrayIndex] = coins[i];
returnArrayIndex++;
count--;
}
}
return returnArray;
}
}
You can check this question: Minimum coin change problem with limited amount of coins.
BTW, I created c++ program based above link's algorithm:
#include <iostream>
#include <map>
#include <vector>
#include <algorithm>
#include <limits>
using namespace std;
void copyVec(vector<int> from, vector<int> &to){
for(vector<int>::size_type i = 0; i < from.size(); i++)
to[i] = from[i];
}
vector<int> makeChangeWithLimited(int amount, vector<int> coins, vector<int> limits)
{
vector<int> change;
vector<vector<int>> coinsUsed( amount + 1 , vector<int>(coins.size()));
vector<int> minCoins(amount+1,numeric_limits<int>::max() - 1);
minCoins[0] = 0;
vector<int> limitsCopy(limits.size());
copy(limits.begin(), limits.end(), limitsCopy.begin());
for (vector<int>::size_type i = 0; i < coins.size(); ++i)
{
while (limitsCopy[i] > 0)
{
for (int j = amount; j >= 0; --j)
{
int currAmount = j + coins[i];
if (currAmount <= amount)
{
if (minCoins[currAmount] > minCoins[j] + 1)
{
minCoins[currAmount] = minCoins[j] + 1;
copyVec(coinsUsed[j], coinsUsed[currAmount]);
coinsUsed[currAmount][i] += 1;
}
}
}
limitsCopy[i] -= 1;
}
}
if (minCoins[amount] == numeric_limits<int>::max() - 1)
{
return change;
}
copy(coinsUsed[amount].begin(),coinsUsed[amount].end(), back_inserter(change) );
return change;
}
int main()
{
vector<int> coins;
coins.push_back(20);
coins.push_back(50);
coins.push_back(100);
coins.push_back(200);
vector<int> limits;
limits.push_back(100);
limits.push_back(100);
limits.push_back(50);
limits.push_back(20);
int amount = 0;
cin >> amount;
while(amount){
vector<int> change = makeChangeWithLimited(amount,coins,limits);
for(vector<int>::size_type i = 0; i < change.size(); i++){
cout << change[i] << "x" << coins[i] << endl;
}
if(change.empty()){
cout << "IMPOSSIBE\n";
}
cin >> amount;
}
system("pause");
return 0;
}
Code in c#
private static int MinCoinsChangeWithLimitedCoins(int[] coins, int[] counts, int sum)
{
var dp = new int[sum + 1];
Array.Fill(dp, int.MaxValue);
dp[0] = 0;
for (int i = 0; i < coins.Length; i++) // n
{
int coin = coins[i];
for (int j = 0; j < counts[i]; j++) //
{
for (int s = sum; s >= coin ; s--) // sum
{
int remainder = s - coin;
if (remainder >= 0 && dp[remainder] != int.MaxValue)
{
dp[s] = Math.Min(1 + dp[remainder], dp[s]);
}
}
}
}
return dp[sum] == int.MaxValue ? -1 : dp[sum];
}

How do we solve the given scenario efficiently?

We are given a maze in which we need to visit as many rooms as possible. The specialty of the maze is that once you enter any room it will only lead you to rooms with a higher tag in the direction you move . B and C decide to move in opposite directions trying their luck to maximize the number of rooms they search .(They can start with any room , need not be the same)
We need to find out the maximum number of rooms that can be searched.
1. Access to any room with a higher tag is allowed, not just adjacent rooms or the next room with a higher tag.
2. Tags are unique.
So given the input:
12 11 10 1 2 3 4 13 6 7 8 5 9
the answer is 12: (1,2,3,4,6,7,8,9) for B and (5,10,11,12) for C.
I thought of solving this using longest increasing sub sequence first from right and then from left.And the count of unique elements in above two sub sequence would be the answer.
But my logic seems to fail,how can this be done?
My program below computes the maximum number of rooms searched. This has time complexity of O(n^3). I modified the DP algorithm for computing the longest increasing sequence available online to solve OP's problem. This also addresses OP's concerns on arrays like {1,4,6,2,5}. I rightly get the max value as 5 for the previous example. So, I used the idea from #BeyelerStudios that we need to compute the longest increasing subsequence from both left to right and from right to left. But, there is a caveat. If we compute the Left to right max sequence, the sequence from right to left should be on the remaining elements. Example:
For the array {1, 4, 6, 2, 5}. If the forward rooms selected are {1, 4, 5 }, then the reverse longest increasing sequence should be computed on the left out elements {6, 2}.
Below is my program:
#include <iostream>
using namespace std;
// compute the max increasing sequence from right to left.
int r2lRooms (int arr[], int n)
{
int dp[n];
int i =0, j = 0;
int max = 0;
for ( i = 0; i < n; i++ ) {
dp[i] = 1;
}
for (i = n-2; i >= 0; i--) {
for ( j = n-1; j > i; j-- ) {
if ( arr[i] > arr[j] && dp[i] < dp[j] + 1) {
dp[i] = dp[j] + 1;
}
}
}
for ( i = 0; i < n; i++ ) {
if ( max < dp[i] ) {
max = dp[i];
}
}
return max;
}
// compute max rooms.
int maxRooms( int arr[], int n )
{
int dp[n], revArray[n];
int i =0, j = 0, k = 0;
int currentMax = 0;
int forwardMax = 0, reverseMax = 0;
for ( i = 0; i < n; i++ ) {
dp[i] = 1;
}
// First case is that except for first elem, all others are in revArray
for (i=1; i < n; i++, k++) {
revArray[k] = arr[i];
}
reverseMax = r2lRooms (revArray, k);
forwardMax = 1;
if (currentMax < (forwardMax + reverseMax)) {
currentMax = forwardMax + reverseMax;
}
cout << "forwardmax revmax and currentmax are: " << forwardMax << " " << reverseMax << " " << currentMax << endl;
cout << endl;
for ( i = 1; i < n; i++ ) {
k = 0;
forwardMax = 1;
reverseMax = 0;
cout << "Forward elems for arr[" << i << "]=" << arr[i] << endl;
for ( j = 0; j < i; j++ ) {
if ( arr[i] > arr[j] && dp[i] < dp[j] + 1) {
dp[i] = dp[j] + 1;
forwardMax = dp[i];
cout << arr[j] << " ";
}
else {
// element was not in DP calculation, so put in revArray.
revArray[k] = arr[j];
k++;
}
}
// copy the remaining elements in revArray.
for ( j = i+1; j < n; j++ ) {
revArray[k] = arr[j];
k++;
}
cout << endl;
reverseMax = r2lRooms (revArray, k);
if (currentMax < (forwardMax + reverseMax)) {
currentMax = forwardMax + reverseMax;
}
cout << "forwardmax revmax and currentmax are: " << forwardMax << " " << reverseMax << " " << currentMax << endl;
cout << endl;
}
cout << " Max rooms searched " << currentMax << endl;
return currentMax;
}
int main (void) {
int arr[] = {12, 11, 10, 1, 2, 3, 4, 13, 6, 7, 8, 5, 9 };
int size = sizeof(arr) / sizeof(int);
cout << maxRooms (arr, size);
}
I think the trick is at the intersection, where B and C might share a value or there's options to go around that (say the sequence is 12 11 10 1 2 3 4 <3 5> 13 6 7 8 9 The extra numbers here adds 1 to the solution, but doesn't change the result for either longest increasing sub-sequences.
So the only problem is the one room in the middle, since on both side the values chosen diverge.
What I would do is this: do the longest subsequence in one direction, figure out a solution (any solution), take out the numbers in the solution and do the longest subsequence in the other direction. This way if there's a way around the crossing room in the middle the second pass will prefer it, unless that's the chosen number is really needed. To check for that do the same thing, but build the first subsequence in the opposite direction and the second one (after removing the solution) in the direction chosen initially.
Complexity remains O(N) but with a slightly higher constant factor.

Set every cell in matrix to 0 if that row or column contains a 0

Want to improve this post? Provide detailed answers to this question, including citations and an explanation of why your answer is correct. Answers without enough detail may be edited or deleted.
Given a NxN matrix with 0s and 1s. Set every row that contains a 0 to all 0s and set every column that contains a 0 to all 0s.
For example
1 0 1 1 0
0 1 1 1 0
1 1 1 1 1
1 0 1 1 1
1 1 1 1 1
results in
0 0 0 0 0
0 0 0 0 0
0 0 1 1 0
0 0 0 0 0
0 0 1 1 0
A Microsoft Engineer told me that there is a solution that involves no extra memory, just two boolean variables and one pass, so I'm looking for that answer.
BTW, imagine it is a bit matrix, therefore just 1s and 0s are allow to be in the matrix.
Ok, so I'm tired as it's 3AM here, but I have a first try inplace with exactly 2 passes on each number in the matrix, so in O(NxN) and it is linear in the size of the matrix.
I use 1rst column and first row as markers to know where are rows/cols with only 1's. Then, there are 2 variables l and c to remember if 1rst row/column are all 1's also.
So the first pass sets the markers and resets the rest to 0's.
The second pass sets 1 in places where rows and cols where marked to be 1, and resets 1st line/col depending on l and c.
I doubt strongly that I can be done in 1 pass as squares in the beginning depend on squares in the end. Maybe my 2nd pass can be made more efficient...
import pprint
m = [[1, 0, 1, 1, 0],
[0, 1, 1, 1, 0],
[1, 1, 1, 1, 1],
[1, 0, 1, 1, 1],
[1, 1, 1, 1, 1]]
N = len(m)
### pass 1
# 1 rst line/column
c = 1
for i in range(N):
c &= m[i][0]
l = 1
for i in range(1,N):
l &= m[0][i]
# other line/cols
# use line1, col1 to keep only those with 1
for i in range(1,N):
for j in range(1,N):
if m[i][j] == 0:
m[0][j] = 0
m[i][0] = 0
else:
m[i][j] = 0
### pass 2
# if line1 and col1 are ones: it is 1
for i in range(1,N):
for j in range(1,N):
if m[i][0] & m[0][j]:
m[i][j] = 1
# 1rst row and col: reset if 0
if l == 0:
for i in range(N):
m [i][0] = 0
if c == 0:
for j in range(1,N):
m [0][j] = 0
pprint.pprint(m)
This cannot be done in one pass since a single bit has an effect on bits before and after it in any ordering. IOW Whatever order you traverse the array in, you may later come accross a 0 which means you have to go back and change a previous 1 to a 0.
Update
People seem to think that by restricting N to some fixed value (say 8) you can solve this is one pass. Well that's a) missing the point and b) not the original question. I wouldn't post a question on sorting and expect an answer which started "assuming you only want to sort 8 things...".
That said, it's a reasonable approach if you know that N is in fact restricted to 8. My answer above answers the original question which has no such retriction.
So my idea is to use the values in the last row/column as a flag to indicate whether all of the values in the corresponding column/row are 1s.
Using a Zig Zag scan through the entire matrix EXCEPT the final row/column. At each element, you set the value in the final row/column as to the logical AND of itself with the value in the current element. In other words, if you hit a 0, the final row/column will be set to 0. If you it a 1, the value in the final row/column will be 1 only if it was 1 already. In any case set the current element to 0.
When you've finished, your final row/column should have 1s iff the corresponding column/row was filled with 1s.
Do a linear scan through the final row and column and looking for 1s. Set 1s in the corresponding elements in body of the matrix where the final row and column are both 1s.
Coding it will be tricky to avoid off-by-one errors etc but it should work in one pass.
I've got a solution here, it runs in a single pass, and does all processing "in place" with no extra memory (save for growing the stack).
It uses recursion to delay the writing of zeros which of course would destroy the matrix for the other rows and cols:
#include <iostream>
/**
* The idea with my algorithm is to delay the writing of zeros
* till all rows and cols can be processed. I do this using
* recursion:
* 1) Enter Recursive Function:
* 2) Check the row and col of this "corner" for zeros and store the results in bools
* 3) Send recursive function to the next corner
* 4) When the recursive function returns, use the data we stored in step 2
* to zero the the row and col conditionally
*
* The corners I talk about are just how I ensure I hit all the row's a cols,
* I progress through the matrix from (0,0) to (1,1) to (2,2) and on to (n,n).
*
* For simplicities sake, I use ints instead of individual bits. But I never store
* anything but 0 or 1 so it's still fair ;)
*/
// ================================
// Using globals just to keep function
// call syntax as straight forward as possible
int n = 5;
int m[5][5] = {
{ 1, 0, 1, 1, 0 },
{ 0, 1, 1, 1, 0 },
{ 1, 1, 1, 1, 1 },
{ 1, 0, 1, 1, 1 },
{ 1, 1, 1, 1, 1 }
};
// ================================
// Just declaring the function prototypes
void processMatrix();
void processCorner( int cornerIndex );
bool checkRow( int rowIndex );
bool checkCol( int colIndex );
void zeroRow( int rowIndex );
void zeroCol( int colIndex );
void printMatrix();
// This function primes the pump
void processMatrix() {
processCorner( 0 );
}
// Step 1) This is the heart of my recursive algorithm
void processCorner( int cornerIndex ) {
// Step 2) Do the logic processing here and store the results
bool rowZero = checkRow( cornerIndex );
bool colZero = checkCol( cornerIndex );
// Step 3) Now progress through the matrix
int nextCorner = cornerIndex + 1;
if( nextCorner < n )
processCorner( nextCorner );
// Step 4) Finially apply the changes determined earlier
if( colZero )
zeroCol( cornerIndex );
if( rowZero )
zeroRow( cornerIndex );
}
// This function returns whether or not the row contains a zero
bool checkRow( int rowIndex ) {
bool zero = false;
for( int i=0; i<n && !zero; ++i ) {
if( m[ rowIndex ][ i ] == 0 )
zero = true;
}
return zero;
}
// This is just a helper function for zeroing a row
void zeroRow( int rowIndex ) {
for( int i=0; i<n; ++i ) {
m[ rowIndex ][ i ] = 0;
}
}
// This function returns whether or not the col contains a zero
bool checkCol( int colIndex ) {
bool zero = false;
for( int i=0; i<n && !zero; ++i ) {
if( m[ i ][ colIndex ] == 0 )
zero = true;
}
return zero;
}
// This is just a helper function for zeroing a col
void zeroCol( int colIndex ) {
for( int i=0; i<n; ++i ) {
m[ i ][ colIndex ] = 0;
}
}
// Just a helper function for printing our matrix to std::out
void printMatrix() {
std::cout << std::endl;
for( int y=0; y<n; ++y ) {
for( int x=0; x<n; ++x ) {
std::cout << m[y][x] << " ";
}
std::cout << std::endl;
}
std::cout << std::endl;
}
// Execute!
int main() {
printMatrix();
processMatrix();
printMatrix();
}
I don't think it's doable. When you're on the first square and its value is 1, you have no way of knowing what the values of the other squares in the same row and column are. So you have to check those and if there's a zero, return to the first square and change its value to zero. I'll recommend doing it in two passes - the first pass gathers information about which rows and columns must be zeroed out (the information is stored in an array, so we're using some extra memory). The second pass changes the values. I know that's not the solution you're looking for, but I think it's a practical one. The constraints given by you render the problem unsolvable.
I can do it with two integer variables and two passes (up to 32 rows and columns...)
bool matrix[5][5] =
{
{1, 0, 1, 1, 0},
{0, 1, 1, 1, 0},
{1, 1, 1, 1, 1},
{1, 0, 1, 1, 1},
{1, 1, 1, 1, 1}
};
int CompleteRows = ~0;
int CompleteCols = ~0;
// Find the first 0
for (int row = 0; row < 5; ++row)
{
for (int col = 0; col < 5; ++col)
{
CompleteRows &= ~(!matrix[row][col] << row);
CompleteCols &= ~(!matrix[row][col] << col);
}
}
for (int row = 0; row < 5; ++row)
for (int col = 0; col < 5; ++col)
matrix[row][col] = (CompleteRows & (1 << row)) && (CompleteCols & (1 << col));
the problem can be solved in one pass
saving the matrix in an i X j array.
1 0 1 1 0
0 1 1 1 0
1 1 1 1 1
1 0 1 1 1
1 1 1 1 1
one each pass save the values of i and j for an element which is 0 in arrays a and b
when first row is scanned a= 1 b = 2,5
when second row is scanned a=1,2 b= 1,2,5
when third row is scanned no change
when fourth row is scanned a= 1,2,4 and b= 1,2,5
when fifth row is scanned no change .
now print all values as 0 for values of i and j saved in a and b
rest of the values are 1 ie (3,3) (3,4) (5,3) and (5,4)
Another solution that takes two passes, is to accumulate ANDs horizontally and vertically:
1 0 1 1 0 | 0
0 1 1 1 0 | 0
1 1 1 1 1 | 1
1 0 1 1 1 | 0
1 1 1 1 1 | 1
----------+
0 0 1 1 0
I thought I could design such an algorithm using parity bits, Hamming codes or dynamic programming, possibly using those two booleans as a 2-bit number, but I've had no success yet.
Can you please re-check the problem statement with your engineer and let us know? If
there is indeed a solution, I want to keep chipping away at the problem.
Keep a single variable to keep track of what all of the rows ANDed together are.
If a row is -1 (all 1s), then make the next row a reference to that variable
If a row is anything but, it's a 0. You can do everything in one pass. Psuedo-code:
foreach (my $row) rows {
$andproduct = $andproduct & $row;
if($row != -1) {
zero out the row
} else {
replace row with a reference to andproduct
}
}
That should do it, in a single pass -- but there is an assumption here that N is small enough for the CPU to do arithmetic on a single row, else you're going to need to loop over each row to determine if it's all 1s or not, I believe. But given you're asking about algos and not constraining my hardware, I would just start my answer with "Build a CPU that supports N-bit arithmetic..."
Here's one example how it can be done in C. Note I argue that values and arr taken together represent the array, and p and numproduct are my iterator and AND product variables use to implement the problem. (I could have looped over arr with pointer arithmetic to validate my work, but once was enough!)
int main() {
int values[] = { -10, 14, -1, -9, -1 }; /* From the problem spec, converted to decimal for my sanity */
int *arr[5] = { values, values+1, values+2, values+3, values+4 };
int **p;
int numproduct = 127;
for(p = arr; p < arr+5; ++p) {
numproduct = numproduct & **p;
if(**p != -1) {
**p = 0;
} else {
*p = &numproduct;
}
}
/* Print our array, this loop is just for show */
int i;
for(i = 0; i < 5; ++i) {
printf("%x\n",*arr[i]);
}
return 0;
}
This produces 0, 0, 6, 0, 6, which is the result for the given inputs.
Or in PHP, if people think my stack games in C are cheating (I suggest to you that it's not, because I should be able to store the matrix whichever way I please):
<?php
$values = array(-10, 14, -1, -9, -1);
$numproduct = 127;
for($i = 0; $i < 5; ++$i) {
$numproduct = $numproduct & $values[$i];
if($values[$i] != -1) {
$values[$i] = 0;
} else {
$values[$i] = &$numproduct;
}
}
print_r($values);
Am I missing something?
Nice challange. This solution sort of uses just two booleans created on the stack, but the booleans are created several times on the stack since the function is recursive.
typedef unsigned short WORD;
typedef unsigned char BOOL;
#define true 1
#define false 0
BYTE buffer[5][5] = {
1, 0, 1, 1, 0,
0, 1, 1, 1, 0,
1, 1, 1, 1, 1,
1, 0, 1, 1, 1,
1, 1, 1, 1, 1
};
int scan_to_end(BOOL *h,BOOL *w,WORD N,WORD pos_N)
{
WORD i;
for(i=0;i<N;i++)
{
if(!buffer[i][pos_N])
*h=false;
if(!buffer[pos_N][i])
*w=false;
}
return 0;
}
int set_line(BOOL h,BOOL w,WORD N,WORD pos_N)
{
WORD i;
if(!h)
for(i=0;i<N;i++)
buffer[i][pos_N] = false;
if(!w)
for(i=0;i<N;i++)
buffer[pos_N][i] = false;
return 0;
}
int scan(int N,int pos_N)
{
BOOL h = true;
BOOL w = true;
if(pos_N == N)
return 0;
// Do single scan
scan_to_end(&h,&w,N,pos_N);
// Scan all recursive before changeing data
scan(N,pos_N+1);
// Set the result of the scan
set_line(h,w,N,pos_N);
return 0;
}
int main(void)
{
printf("Old matrix\n");
printf( "%d,%d,%d,%d,%d \n", (WORD)buffer[0][0],(WORD)buffer[0][1],(WORD)buffer[0][2],(WORD)buffer[0][3],(WORD)buffer[0][4]);
printf( "%d,%d,%d,%d,%d \n", (WORD)buffer[1][0],(WORD)buffer[1][1],(WORD)buffer[1][2],(WORD)buffer[1][3],(WORD)buffer[1][4]);
printf( "%d,%d,%d,%d,%d \n", (WORD)buffer[2][0],(WORD)buffer[2][1],(WORD)buffer[2][2],(WORD)buffer[2][3],(WORD)buffer[2][4]);
printf( "%d,%d,%d,%d,%d \n", (WORD)buffer[3][0],(WORD)buffer[3][1],(WORD)buffer[3][2],(WORD)buffer[3][3],(WORD)buffer[3][4]);
printf( "%d,%d,%d,%d,%d \n", (WORD)buffer[4][0],(WORD)buffer[4][1],(WORD)buffer[4][2],(WORD)buffer[4][3],(WORD)buffer[4][4]);
scan(5,0);
printf("New matrix\n");
printf( "%d,%d,%d,%d,%d \n", (WORD)buffer[0][0],(WORD)buffer[0][1],(WORD)buffer[0][2],(WORD)buffer[0][3],(WORD)buffer[0][4]);
printf( "%d,%d,%d,%d,%d \n", (WORD)buffer[1][0],(WORD)buffer[1][1],(WORD)buffer[1][2],(WORD)buffer[1][3],(WORD)buffer[1][4]);
printf( "%d,%d,%d,%d,%d \n", (WORD)buffer[2][0],(WORD)buffer[2][1],(WORD)buffer[2][2],(WORD)buffer[2][3],(WORD)buffer[2][4]);
printf( "%d,%d,%d,%d,%d \n", (WORD)buffer[3][0],(WORD)buffer[3][1],(WORD)buffer[3][2],(WORD)buffer[3][3],(WORD)buffer[3][4]);
printf( "%d,%d,%d,%d,%d \n", (WORD)buffer[4][0],(WORD)buffer[4][1],(WORD)buffer[4][2],(WORD)buffer[4][3],(WORD)buffer[4][4]);
system( "pause" );
return 0;
}
This scans in a pattern like:
s,s,s,s,s
s,0,0,0,0
s,0,0,0,0
s,0,0,0,0
s,0,0,0,0
0,s,0,0,0
s,s,s,s,s
0,s,0,0,0
0,s,0,0,0
0,s,0,0,0
and so on
And then changeing the values in this pattern on return on each of the scan functions. (Bottom up):
0,0,0,0,c
0,0,0,0,c
0,0,0,0,c
0,0,0,0,c
c,c,c,c,c
0,0,0,c,0
0,0,0,c,0
0,0,0,c,0
c,c,c,c,c
0,0,0,c,0
and so on
Okay this is a solution that
uses just one extra long value for working storage.
uses no recursion.
one pass of only N, not even N*N.
will work for other values of N but will need new #defines.
#include <stdio.h>
#define BIT30 (1<<24)
#define COLMASK 0x108421L
#define ROWMASK 0x1fL
unsigned long long STARTGRID =
((0x10 | 0x0 | 0x4 | 0x2 | 0x0) << 20) |
((0x00 | 0x8 | 0x4 | 0x2 | 0x0) << 15) |
((0x10 | 0x8 | 0x4 | 0x2 | 0x1) << 10) |
((0x10 | 0x0 | 0x4 | 0x2 | 0x1) << 5) |
((0x10 | 0x8 | 0x4 | 0x2 | 0x1) << 0);
void dumpGrid (char* comment, unsigned long long theGrid) {
char buffer[1000];
buffer[0]='\0';
printf ("\n\n%s\n",comment);
for (int j=1;j<31; j++) {
if (j%5!=1)
printf( "%s%s", ((theGrid & BIT30)==BIT30)? "1" : "0",(((j%5)==0)?"\n" : ",") );
theGrid = theGrid << 1;
}
}
int main (int argc, const char * argv[]) {
unsigned long long rowgrid = STARTGRID;
unsigned long long colGrid = rowgrid;
unsigned long long rowmask = ROWMASK;
unsigned long long colmask = COLMASK;
dumpGrid("Initial Grid", rowgrid);
for (int i=0; i<5; i++) {
if ((rowgrid & rowmask)== rowmask) rowgrid |= rowmask;
else rowgrid &= ~rowmask;
if ((colGrid & colmask) == colmask) colmask |= colmask;
else colGrid &= ~colmask;
rowmask <<= 5;
colmask <<= 1;
}
colGrid &= rowgrid;
dumpGrid("RESULT Grid", colGrid);
return 0;
}
Actually. If you just want to run the algorithm and print out the results (i.e. not restore them, then this can easily be done in one pass. The trouble comes when you try to modify the array as you're running the algorithm.
Here is my solution It just involves ANDing the rows/columns values for a givein (i,j)'s element and printing it out.
#include <iostream>
#include "stdlib.h"
void process();
int dim = 5;
bool m[5][5]{{1,0,1,1,1},{0,1,1,0,1},{1,1,1,1,1},{1,1,1,1,1},{0,0,1,1,1}};
int main() {
process();
return 0;
}
void process() {
for(int j = 0; j < dim; j++) {
for(int i = 0; i < dim; i++) {
std::cout << (
(m[0][j] & m[1][j] & m[2][j] & m[3][j] & m[4][j]) &
(m[i][0] & m[i][1] & m[i][2] & m[i][3] & m[i][4])
);
}
std::cout << std::endl;
}
}
I tried to solve this problem in C#.
I've used two loop variables (i and j) apart from the actual matrix and n representing its dimension
Logic I tried is to:
Calculate AND for rows and cols involved in each concentric square of the matrix
Store it in its corner cells (I've stored them in anti-clockwise order)
Two bool variables are used to retain values of two corners when evaluating a particular square.
This process would end when outer loop (i) is mid way.
Evaluate results of other cells based on the corner cells (for rest of i). Skip the corner cells during this process.
When i reaches n, all cells would have its result except for the corner cells.
Update the corner cells. This is an extra iteration to length of n/2 other than the single pass constraint mentioned in the problem.
Code:
void Evaluate(bool [,] matrix, int n)
{
bool tempvar1, tempvar2;
for (var i = 0; i < n; i++)
{
tempvar1 = matrix[i, i];
tempvar2 = matrix[n - i - 1, n - i - 1];
var j = 0;
for (j = 0; j < n; j++)
{
if ((i < n/2) || (((n % 2) == 1) && (i == n/2) && (j <= i)))
{
// store the row and col & results in corner cells of concentric squares
tempvar1 &= matrix[j, i];
matrix[i, i] &= matrix[i, j];
tempvar2 &= matrix[n - j - 1, n - i - 1];
matrix[n - i - 1, n - i - 1] &= matrix[n - i - 1, n - j - 1];
}
else
{
// skip corner cells of concentric squares
if ((j == i) || (j == n - i - 1)) continue;
// calculate the & values for rest of them
matrix[i, j] = matrix[i, i] & matrix[n - j - 1, j];
matrix[n - i - 1, j] = matrix[n - i - 1, n - i - 1] & matrix[n - j - 1, j];
if ((i == n/2) && ((n % 2) == 1))
{
// if n is odd
matrix[i, n - j - 1] = matrix[i, i] & matrix[j, n - j - 1];
}
}
}
if ((i < n/2) || (((n % 2) == 1) && (i <= n/2)))
{
// transfer the values from temp variables to appropriate corner cells of its corresponding square
matrix[n - i - 1, i] = tempvar1;
matrix[i, n - i - 1] = tempvar2;
}
else if (i == n - 1)
{
// update the values of corner cells of each concentric square
for (j = n/2; j < n; j++)
{
tempvar1 = matrix[j, j];
tempvar2 = matrix[n - j - 1, n - j - 1];
matrix[j, j] &= matrix[n - j - 1, j];
matrix[n - j - 1, j] &= tempvar2;
matrix[n - j - 1, n - j - 1] &= matrix[j, n - j - 1];
matrix[j, n - j - 1] &= tempvar1;
}
}
}
}
One Pass - I have traversed the input only once but used a new array and only two extra Boolean variables.
public static void main(String[] args) {
Scanner sc = new Scanner(System.in);
int n = sc.nextInt();
sc.nextLine();
boolean rowDel = false, colDel = false;
int arr[][] = new int[n][n];
int res[][] = new int[n][n];
int i, j;
for (i = 0; i < n; i++) {
for (j = 0; j < n; j++) {
arr[i][j] = sc.nextInt();
res[i][j] = arr[i][j];
}
}
for (i = 0; i < n; i++) {
for (j = 0; j < n; j++) {
if (arr[i][j] == 0)
colDel = rowDel = true; //See if we have to delete the
//current row and column
if (rowDel == true){
res[i] = new int[n];
rowDel = false;
}
if(colDel == true){
for (int k = 0; k < n; k++) {
res[k][j] = 0;
}
colDel = false;
}
}
}
for (i = 0; i < n; i++) {
for (j = 0; j < n; j++) {
System.out.print(res[i][j]);
}
System.out.println();
}
sc.close();
}
While impossible given the constraints, the most space efficient way to do it is by traversing the matrix in an overlaping, alternating row/column fashion, which would make a pattern similar to laying bricks in a zig-zag fashion:
-----
|----
||---
|||--
||||-
Using this, you would go in each row/column, as indicated, and if you encounter a 0 at any time, set a boolean variable, and re-walk that row/column, setting the entries to 0 as you go.
This will require no extra memory, and will only use one boolean variable. Unfortunately, if the "far" edge is set to all be 0, that is the worst case and you walk the whole array twice.
create a result matrix and set all the values to 1.
go through the data matrix as soon as a 0 is encountered, set the result matrix row column to 0
At the end of the first pass, the result matrix will have the correct answer.
Looks pretty simple. Is there a trick I am missing? Are you not allowed to use a result set?
EDIT:
Looks like a F# function, but that is cheating a bit since even though you are doing a single pass, the function can be recursive.
It looks like the interviewer is trying to figure out if you know functional programming.
Well, I came up with a single-pass, in-place (non-recursive) solution using 4 bools and 2 loop counters. I've not managed to reduce it to 2 bools and 2 ints, but I wouldn't be surprised if it was possible. It does around 3 reads and 3 writes of each cell, and it should be O(N^2) ie. linear in the array size.
Took me a couple of hours to puzzle this one out - I wouldn't want to have to come up with it under the pressure of an interview! If I've made a booboo I'm too tired to spot it...
Um... I'm choosing to define "single-pass" as making one sweep through the matrix, rather than touching each value once! :-)
#include <stdio.h>
#include <memory.h>
#define SIZE 5
typedef unsigned char u8;
u8 g_Array[ SIZE ][ SIZE ];
void Dump()
{
for ( int nRow = 0; nRow < SIZE; ++nRow )
{
for ( int nColumn = 0; nColumn < SIZE; ++nColumn )
{
printf( "%d ", g_Array[ nRow ][ nColumn ] );
}
printf( "\n" );
}
}
void Process()
{
u8 fCarriedAlpha = true;
u8 fCarriedBeta = true;
for ( int nStep = 0; nStep < SIZE; ++nStep )
{
u8 fAlpha = (nStep > 0) ? g_Array[ nStep-1 ][ nStep ] : true;
u8 fBeta = (nStep > 0) ? g_Array[ nStep ][ nStep - 1 ] : true;
fAlpha &= g_Array[ nStep ][ nStep ];
fBeta &= g_Array[ nStep ][ nStep ];
g_Array[ nStep-1 ][ nStep ] = fCarriedBeta;
g_Array[ nStep ][ nStep-1 ] = fCarriedAlpha;
for ( int nScan = nStep + 1; nScan < SIZE; ++nScan )
{
fBeta &= g_Array[ nStep ][ nScan ];
if ( nStep > 0 )
{
g_Array[ nStep ][ nScan ] &= g_Array[ nStep-1 ][ nScan ];
g_Array[ nStep-1][ nScan ] = fCarriedBeta;
}
fAlpha &= g_Array[ nScan ][ nStep ];
if ( nStep > 0 )
{
g_Array[ nScan ][ nStep ] &= g_Array[ nScan ][ nStep-1 ];
g_Array[ nScan ][ nStep-1] = fCarriedAlpha;
}
}
g_Array[ nStep ][ nStep ] = fAlpha & fBeta;
for ( int nScan = nStep - 1; nScan >= 0; --nScan )
{
g_Array[ nScan ][ nStep ] &= fAlpha;
g_Array[ nStep ][ nScan ] &= fBeta;
}
fCarriedAlpha = fAlpha;
fCarriedBeta = fBeta;
}
}
int main()
{
memset( g_Array, 1, sizeof(g_Array) );
g_Array[0][1] = 0;
g_Array[0][4] = 0;
g_Array[1][0] = 0;
g_Array[1][4] = 0;
g_Array[3][1] = 0;
printf( "Input:\n" );
Dump();
Process();
printf( "\nOutput:\n" );
Dump();
return 0;
}
i hope you enjoy my 1-pass c# solution
you can retrieve an element with O(1) and only need
the space of one row and one column of the matrix
bool[][] matrix =
{
new[] { true, false, true, true, false }, // 10110
new[] { false, true, true, true, false }, // 01110
new[] { true, true, true, true, true }, // 11111
new[] { true, false, true, true, true }, // 10111
new[] { true, true, true, true, true } // 11111
};
int n = matrix.Length;
bool[] enabledRows = new bool[n];
bool[] enabledColumns = new bool[n];
for (int i = 0; i < n; i++)
{
enabledRows[i] = true;
enabledColumns[i] = true;
}
for (int rowIndex = 0; rowIndex < n; rowIndex++)
{
for (int columnIndex = 0; columnIndex < n; columnIndex++)
{
bool element = matrix[rowIndex][columnIndex];
enabledRows[rowIndex] &= element;
enabledColumns[columnIndex] &= element;
}
}
for (int rowIndex = 0; rowIndex < n; rowIndex++)
{
for (int columnIndex = 0; columnIndex < n; columnIndex++)
{
bool element = enabledRows[rowIndex] & enabledColumns[columnIndex];
Console.Write(Convert.ToInt32(element));
}
Console.WriteLine();
}
/*
00000
00000
00110
00000
00110
*/
1 pass, 2 booleans. I just have to assume the integer indexes in the iterations don't count.
This is not a complete solution but I can't get pass this point.
If I could only determine if a 0 is an original 0 or a 1 that was converted to a 0 then I wouldn't have to use -1's and this would work.
My output is like this:
-1 0 -1 -1 0
0 -1 -1 -1 0
-1 -1 1 1 -1
-1 0 -1 -1 -1
-1 -1 1 1 -1
The originality of my approach is using the first half of the examination of a row or column to determine if it contains a 0 and the last half to set the values - this is done by looking at x and width-x and then y and height-y in each iteration. Based on the results of the first half of the iteration, if a 0 in the row or column was found, I use the last half of the iteration to change the 1's to -1's.
I just realized this could be done with only 1 boolean leaving 1 to ...?
I'm posting this hoping someone might say, "Ah, just do this..." (And I spent way too much time on it not to post.)
Here's the code in VB:
Dim D(,) As Integer = {{1, 0, 1, 1, 1}, {0, 1, 1, 0, 1}, {1, 1, 1, 1, 1}, {1, 1, 1, 1, 1}, {0, 0, 1, 1, 1}}
Dim B1, B2 As Boolean
For y As Integer = 0 To UBound(D)
B1 = True : B2 = True
For x As Integer = 0 To UBound(D)
// Scan row for 0's at x and width - x positions. Halfway through I'll konw if there's a 0 in this row.
//If a 0 is found set my first boolean to false.
If x <= (Math.Ceiling((UBound(D) + 1) / 2) - 1) Then
If D(x, y) = 0 Or D(UBound(D) - x, y) = 0 Then B1 = False
End If
//If the boolean is false then a 0 in this row was found. Spend the last half of this loop
//updating the values. This is where I'm stuck. If I change a 1 to a 0 it will cause the column
//scan to fail. So for now I change to a -1. If there was a way to change to 0 yet later tell if
//the value had changed this would work.
If Not B1 Then
If x >= (Math.Ceiling((UBound(D) + 1) / 2) - 1) Then
If D(x, y) = 1 Then D(x, y) = -1
If D(UBound(D) - x, y) = 1 Then D(UBound(D) - x, y) = -1
End If
End If
//These 2 block do the same as the first 2 blocks but I switch x and y to do the column.
If x <= (Math.Ceiling((UBound(D) + 1) / 2) - 1) Then
If D(y, x) = 0 Or D(y, UBound(D) - x) = 0 Then B2 = False
End If
If Not B2 Then
If x >= (Math.Ceiling((UBound(D) + 1) / 2) - 1) Then
If D(y, x) = 1 Then D(y, x) = -1
If D(y, UBound(D) - x) = 1 Then D(y, UBound(D) - x) = -1
End If
End If
Next
Next
No one is using binary forms? since it's only 1 and 0. We can use binary vectors.
def set1(M, N):
'''Set 1/0s on M according to the rules.
M is a list of N integers. Each integer represents a binary array, e.g.,
000100'''
ruler = 2**N-1
for i,v in enumerate(M):
ruler = ruler & M[i]
M[i] = M[i] if M[i]==2**N-1 else 0 # set i-th row to all-0 if not all-1s
for i,v in enumerate(M):
if M[i]: M[i] = ruler
return M
Here's the test:
M = [ 0b10110,
0b01110,
0b11111,
0b10111,
0b11111 ]
print "Before..."
for i in M: print "{:0=5b}".format(i)
M = set1(M, len(M))
print "After..."
for i in M: print "{:0=5b}".format(i)
And the output:
Before...
10110
01110
11111
10111
11111
After...
00000
00000
00110
00000
00110
You can do something like this to use one pass but an input and output matrix:
output(x,y) = col(xy) & row(xy) == 2^n
where col(xy) is the bits in the column containing the point xy; row(xy) is the bits in the row containing the point xy. n is the size of the matrix.
Then just loop over the input. Possibly expandable to be more space efficient?
One matrix scan, two booleans, no recursion.
How to avoid the second pass?
The second pass is needed to clear the rows or columns when the zero appeares at their end.
However this problem can be solved, because when we scan row #i we already know the row status for the row #i-1. So, while we are scanning the row #i we can simultaneously clear the row #i-1 if it is needed.
The same solution works for columns, but we need to scan rows and columns simultaneously while the data is not changed by the next iteration.
Two booleans are required to store the status of first row and first column, because their values will be changed during the execution of main part of the algorithm.
To avoid adding more booleans we are storing the "clear" flag for the rows and columns in the first row and column of the matrix.
public void Run()
{
const int N = 5;
int[,] m = new int[N, N]
{{ 1, 0, 1, 1, 0 },
{ 1, 1, 1, 1, 0 },
{ 1, 1, 1, 1, 1 },
{ 1, 0, 1, 1, 1 },
{ 1, 1, 1, 1, 1 }};
bool keepFirstRow = (m[0, 0] == 1);
bool keepFirstColumn = keepFirstRow;
for (int i = 1; i < N; i++)
{
keepFirstRow = keepFirstRow && (m[0, i] == 1);
keepFirstColumn = keepFirstColumn && (m[i, 0] == 1);
}
Print(m); // show initial setup
m[0, 0] = 1; // to protect first row from clearing by "second pass"
// "second pass" is performed over i-1 row/column,
// so we use one more index just to complete "second pass" over the
// last row/column
for (int i = 1; i <= N; i++)
{
for (int j = 1; j <= N; j++)
{
// "first pass" - searcing for zeroes in row/column #i
// when i = N || j == N it is additional pass for clearing
// the previous row/column
// j >= i because cells with j < i may be already modified
// by "second pass" part
if (i < N && j < N && j >= i)
{
m[i, 0] &= m[i, j];
m[0, j] &= m[i, j];
m[0, i] &= m[j, i];
m[j, 0] &= m[j, i];
}
// "second pass" - clearing the row/column scanned
// in the previous iteration
if (m[i - 1, 0] == 0 && j < N)
{
m[i - 1, j] = 0;
}
if (m[0, i - 1] == 0 && j < N)
{
m[j, i - 1] = 0;
}
}
Print(m);
}
// Clear first row/column if needed
if (!keepFirstRow || !keepFirstColumn)
{
for (int i = 0; i < N; i++)
{
if (!keepFirstRow)
{
m[0, i] = 0;
}
if (!keepFirstColumn)
{
m[i, 0] = 0;
}
}
}
Print(m);
Console.ReadLine();
}
private static void Print(int[,] m)
{
for (int i = 0; i < m.GetLength(0); i++)
{
for (int j = 0; j < m.GetLength(1); j++)
{
Console.Write(" " + m[i, j]);
}
Console.WriteLine();
}
Console.WriteLine();
}
seems like the following works with no additional space requirements:
first note that multiplying the elements of the row times the elements of the line in which an element is, gives the desired value.
In order not to use any additional space (not making a new matrix and filling it up but instead apply changes to the matrix directly), start top left of the matrix and do the computation for any ixi matrix (that "starts" at (0,0)) before considering any element with any index > i.
hope this works (havent testet)
This is TESTED for different N in C++, and is:
ONE PASS, TWO BOOLS, NO RECURSION, NO EXTRA MEMORY, HOLDS FOR ARBITRARLY LARGE N
(So far none of the solutions here do ALL of these.)
More specifically, I'm amusing two loop counters are okay. I have two const unsigneds, which only exist rather than being computed every time for readability. The outer loop's interval is [0, N], and the inner loop's interval is [1, n - 1]. The switch statement is in the loop mostly exists to show very clearly that it really is just one pass.
Algorithm Strategy:
The first trick is to us a row and a column from the matrix itself to accumulate the content of the matrix, this memory becomes available by offloading all we really need to know from the first row and column into two booleans. The second trick is to get two passes out of one, by using the symmetry of the sub-matrix and its indices.
Algorithm Synopsis:
Scan the first row and store if they are all ones in a boolean, do the same for the first column storing the result in a second boolean.
For the sub-matrix excluding the first row and the first column: iterate through, left to right, top to bottom, as one would read a paragraph. Upon visiting each element, also visit the corresponding element that would be visited if visiting the sub-matrix in reverse. For each element visited AND its value into the where its row crosses the first column, and also AND its value into where its column crosses the first row.
Once the center of the sub-matrix is reached, continue to visit the two elements simultaneously as above. However now set the visited elements' value to the AND of where its row crosses the first column, and of where its column crosses the first row. After this, the sub-matrix is complete.
Use the two boolean variables computed at the begging to set the first row, and the first column to their correct values.
Templatized C++ Implementation:
template<unsigned n>
void SidewaysAndRowColumn(int((&m)[n])[n]) {
bool fcol = m[0][0] ? true : false;
bool frow = m[0][0] ? true : false;
for (unsigned d = 0; d <= n; ++d) {
for (unsigned i = 1; i < n; ++i) {
switch (d) {
case 0:
frow = frow && m[d][i];
fcol = fcol && m[i][d];
break;
default:
{
unsigned const rd = n - d;
unsigned const ri = n - i;
if (d * n + i < rd * n + ri)
{
m[ d][ 0] &= m[ d][ i];
m[ 0][ d] &= m[ i][ d];
m[ 0][ i] &= m[ d][ i];
m[ i][ 0] &= m[ i][ d];
m[rd][ 0] &= m[rd][ri];
m[ 0][rd] &= m[ri][rd];
m[ 0][ri] &= m[rd][ri];
m[ri][ 0] &= m[ri][rd];
}
else
{
m[ d][ i] = m[ d][0] & m[0][ i];
m[rd][ri] = m[rd][0] & m[0][ri];
}
break;
}
case n:
if (!frow)
m[0][i] = 0;
if (!fcol)
m[i][0] = 0;
};
}
}
m[0][0] = (frow && fcol) ? 1 : 0;
}
Ok, I realize that it isn't quite a match, but I got it in one pass using a bool and a byte instead of two bools... close. I also wouldn't vouch for the efficiency of it but these types of questions often require less than optimal solutions.
private static void doIt(byte[,] matrix)
{
byte zeroCols = 0;
bool zeroRow = false;
for (int row = 0; row <= matrix.GetUpperBound(0); row++)
{
zeroRow = false;
for (int col = 0; col <= matrix.GetUpperBound(1); col++)
{
if (matrix[row, col] == 0)
{
zeroRow = true;
zeroCols |= (byte)(Math.Pow(2, col));
// reset this column in previous rows
for (int innerRow = 0; innerRow < row; innerRow++)
{
matrix[innerRow, col] = 0;
}
// reset the previous columns in this row
for (int innerCol = 0; innerCol < col; innerCol++)
{
matrix[row, innerCol] = 0;
}
}
else if ((int)(zeroCols & ((byte)Math.Pow(2, col))) > 0)
{
matrix[row, col] = 0;
}
// Force the row to zero
if (zeroRow) { matrix[row, col] = 0; }
}
}
}
You can sorta do it in one pass -- if you don't count accessing the matrix in random-access order, which eliminates the benefits of doing it single-pass in the first place (cache-coherence/memory-bandwidth).
[edit: simple, but wrong solution deleted]
You should get better performance than any single-pass method by doing it in two passes: one to accumulate row/column info, and one to apply it. The array (in row-major order) is accessed coherently; for arrays exceeding the cache size (but whose rows can fit in cache), data should be read from memory twice, and stored once:
void fixmatrix2(int M[][], int rows, int cols) {
bool clearZeroRow= false;
bool clearZeroCol= false;
for(int j=0; j < cols; ++j) {
if( ! M[0][j] ) {
clearZeroRow= true;
}
}
for(int i=1; i < rows; ++i) { // scan/accumulate pass
if( ! M[i][0] ) {
clearZeroCol= true;
}
for(int j=1; j < cols; ++j) {
if( ! M[i][j] ) {
M[0][j]= 0;
M[i][0]= 0;
}
}
}
for(int i=1; i < rows; ++i) { // update pass
if( M[i][0] ) {
for(int j=0; j < cols; ++j) {
if( ! M[j][0] ) {
M[i][j]= 0;
}
}
} else {
for(int j=0; j < cols; ++j) {
M[i][j]= 0;
}
}
if(clearZeroCol) {
M[i][0]= 0;
}
}
if(clearZeroRow) {
for(int j=0; j < cols; ++j) {
M[0][j]= 0;
}
}
}
The simplest solution I can think of is pasted below. The logic is to record which row and column to set zero while iterating.
import java.util.HashSet;
import java.util.Set;
public class MatrixExamples {
public static void zeroOut(int[][] myArray) {
Set<Integer> rowsToZero = new HashSet<>();
Set<Integer> columnsToZero = new HashSet<>();
for (int i = 0; i < myArray.length; i++) {
for (int j = 0; j < myArray.length; j++) {
if (myArray[i][j] == 0) {
rowsToZero.add(i);
columnsToZero.add(j);
}
}
}
for (int i : rowsToZero) {
for (int j = 0; j < myArray.length; j++) {
myArray[i][j] = 0;
}
}
for (int i : columnsToZero) {
for (int j = 0; j < myArray.length; j++) {
myArray[j][i] = 0;
}
}
for (int i = 0; i < myArray.length; i++) { // record which rows and // columns will be zeroed
for (int j = 0; j < myArray.length; j++) {
System.out.print(myArray[i][j] + ",");
if(j == myArray.length-1)
System.out.println();
}
}
}
public static void main(String[] args) {
int[][] a = { { 1, 0, 1, 1, 0 }, { 0, 1, 1, 1, 0 }, { 1, 1, 1, 1, 1 },
{ 1, 0, 1, 1, 1 }, { 1, 1, 1, 1, 1 } };
zeroOut(a);
}
}
Here is my Ruby implementation with the the test included, This would take O(MN) space. If we want a real time update (like to show the results when we find zeros rather than waiting the first loop of finding zeros) we can just create another class variable like #output and whenever we find a zero we update #output and not #input.
require "spec_helper"
class Matrix
def initialize(input)
#input = input
#zeros = []
end
def solve
#input.each_with_index do |row, i|
row.each_with_index do |element, j|
#zeros << [i,j] if element == 0
end
end
#zeros.each do |x,y|
set_h_zero(x)
set_v_zero(y)
end
#input
end
private
def set_h_zero(row)
#input[row].map!{0}
end
def set_v_zero(col)
#input.size.times do |r|
#input[r][col] = 0
end
end
end
describe "Matrix" do
it "Should set the row and column of Zero to Zeros" do
input = [[1, 3, 4, 9, 0],
[0, 3, 5, 0, 8],
[1, 9, 6, 1, 9],
[8, 3, 2, 0, 3]]
expected = [[0, 0, 0, 0, 0],
[0, 0, 0, 0, 0],
[0, 9, 6, 0, 0],
[0, 0, 0, 0, 0]]
matrix = Matrix.new(input)
expect(matrix.solve).to eq(expected)
end
end
The code below creates a matrix of size m,n. First decide the dimensions of the matrix. I wanted to fill the matrix[m][n] with randomly with numbers between 0..10. Then create another matrix of the same dimensions and fill it with -1s (final matrix). Then iterate through the initial matrix to see if you will hit 0. When you hit on location(x,y), go to the final matrix and fill the row x with 0s and column y with 0s.
At the end read through the final matrix, if the value is -1 (original value) copy the value in the same location of the initial matrix to final.
public static void main(String[] args) {
int m = 5;
int n = 4;
int[][] matrixInitial = initMatrix(m, n); // 5x4 matrix init randomly
int[][] matrixFinal = matrixNull(matrixInitial, m, n);
for (int i = 0; i < m; i++) {
System.out.println(Arrays.toString(matrixFinal[i]));
}
}
public static int[][] matrixNull(int[][] matrixInitial, int m, int n) {
int[][] matrixFinal = initFinal(m, n); // create a matrix with mxn and init it with all -1
for (int i = 0; i < m; i++) { // iterate in initial matrix
for (int j = 0; j < n; j++) {
if(matrixInitial[i][j] == 0){ // if a value is 0 make rows and columns 0
makeZeroX(matrixFinal, i, j, m, n);
}
}
}
for (int i = 0; i < m; i++) { // if value is -1 (original) copy from initial
for (int j = 0; j < n; j++) {
if(matrixFinal[i][j] == -1) {
matrixFinal[i][j] = matrixInitial[i][j];
}
}
}
return matrixFinal;
}
private static void makeZeroX(int[][] matrixFinal, int x, int y, int m, int n) {
for (int j = 0; j < n; j++) { // make all row 0
matrixFinal[x][j] = 0;
}
for(int i = 0; i < m; i++) { // make all column 0
matrixFinal[i][y] = 0;
}
}
private static int[][] initMatrix(int m, int n) {
int[][] matrix = new int[m][n];
for (int i = 0; i < m; i++) {
for (int j = 0; j < n; j++) {
Random rn = new Random();
int random = rn.nextInt(10);
matrix[i][j] = random;
}
}
for (int i = 0; i < m; i++) {
System.out.println(Arrays.toString(matrix[i]));
}
System.out.println("******");
return matrix;
}
private static int[][] initFinal(int m, int n) {
int[][] matrix = new int[m][n];
for (int i = 0; i < m; i++) {
for (int j = 0; j < n; j++) {
matrix[i][j] = -1;
}
}
return matrix;
}
// another approach
/**
* #param matrixInitial
* #param m
* #param n
* #return
*/
private static int[][] matrixNullNew(int[][] matrixInitial, int m, int n) {
List<Integer> zeroRowList = new ArrayList<>(); // Store rows with 0
List<Integer> zeroColumnList = new ArrayList<>(); // Store columns with 0
for (int i = 0; i < m; i++) { // read through the matrix when you hit 0 add the column to zeroColumnList and add
// the row to zeroRowList
for (int j = 0; j < n; j++) {
if (matrixInitial[i][j] == 0) {
if (!zeroRowList.contains(i)) {
zeroRowList.add(i);
}
if (!zeroColumnList.contains(j)) {
zeroColumnList.add(j);
}
}
}
}
for (int a = 0; a < m; a++) {
if (zeroRowList.contains(a)) { // if the row has 0
for (int b = 0; b < n; b++) {
matrixInitial[a][b] = 0; // replace all row with zero
}
}
}
for (int b = 0; b < n; b++) {
if (zeroColumnList.contains(b)) { // if the column has 0
for (int a = 0; a < m; a++) {
matrixInitial[a][b] = 0; // replace all column with zero
}
}
}
return matrixInitial;
}
here is my solution. As you can see from the code, given a M * N matrix, it sets the entire row to zero once it inspects a zero in that row.the time complexity of my solution is O(M * N) .
I am sharing the whole class which has a static populated array for testing and a display array method to see the result in the console.
public class EntireRowSetToZero {
static int arr[][] = new int[3][4];
static {
arr[0][0] = 1;
arr[0][1] = 9;
arr[0][2] = 2;
arr[0][3] = 2;
arr[1][0] = 1;
arr[1][1] = 5;
arr[1][2] = 88;
arr[1][3] = 7;
arr[2][0] = 0;
arr[2][1] = 8;
arr[2][2] = 4;
arr[2][3] = 4;
}
public static void main(String[] args) {
displayArr(EntireRowSetToZero.arr, 3, 4);
setRowToZero(EntireRowSetToZero.arr);
System.out.println("--------------");
displayArr(EntireRowSetToZero.arr, 3, 4);
}
static int[][] setRowToZero(int[][] arr) {
for (int i = 0; i < arr.length; i++) {
for (int j = 0; j < arr[i].length; j++) {
if(arr[i][j]==0){
arr[i]=new int[arr[i].length];
}
}
}
return arr;
}
static void displayArr(int[][] arr, int n, int k) {
for (int i = 0; i < n; i++) {
for (int j = 0; j < k; j++) {
System.out.print(arr[i][j] + " ");
}
System.out.println("");
}
}
}

Resources