# leetcode 37. Sudoku Solver 36. Valid Sudoku 数独问题 season雅宁 2016-05-04

36. Valid Sudoku

Determine if a Sudoku is valid, according to: Sudoku Puzzles - The Rules.

The Sudoku board could be partially filled, where empty cells are filled with the character `'.'`. A partially filled sudoku which is valid.

Note:
A valid Sudoku board (partially filled) is not necessarily solvable. Only the filled cells need to be validated.

Subscribe to see which companies asked this question

Idea

Just go through all you see (like "7 in row 3") and check for duplicates.

Solution 1

Using `Counter`. One logical line, seven physical lines.

``````def isValidSudoku(self, board):
return 1 == max(collections.Counter(
x
for i, row in enumerate(board)
for j, c in enumerate(row)
if c != '.'
for x in ((c, i), (j, c), (i/3, j/3, c))
).values() + )
``````

The `+ ` is only for the empty board, where `max` would get an empty list and complain. It's not necessary to get it accepted here, as the empty board isn't among the test cases, but it's good to have.

Solution 2

Using `len(set)`.

``````def isValidSudoku(self, board):
seen = sum(([(c, i), (j, c), (i/3, j/3, c)]
for i, row in enumerate(board)
for j, c in enumerate(row)
if c != '.'), [])
return len(seen) == len(set(seen))
``````

Solution 3

Using `any`.

``````def isValidSudoku(self, board):
seen = set()
return not any(x in seen or seen.add(x)
for i, row in enumerate(board)
for j, c in enumerate(row)
if c != '.'
for x in ((c, i), (j, c), (i/3, j/3, c)))
``````

Solution 4

Iterating a different way.

``````def isValidSudoku(self, board):
seen = sum(([(c, i), (j, c), (i/3, j/3, c)]
for i in range(9) for j in range(9)
for c in [board[i][j]] if c != '.'), [])
return len(seen) == len(set(seen))``````

# Clean and Easy82ms Python

402 views

class Solution(object):

``````def isValidSudoku(self, board):
"""
:type board: List[List[str]]
:rtype: bool
"""
big = set()
for i in xrange(0,9):
for j in xrange(0,9):
if board[i][j]!='.':
cur = board[i][j]
if (i,cur) in big or (cur,j) in big or (i/3,j/3,cur) in big:
return False

37. Sudoku Solver

Write a program to solve a Sudoku puzzle by filling the empty cells.

Empty cells are indicated by the character `'.'`.

You may assume that there will be only one unique solution. A sudoku puzzle... ...and its solution numbers marked in red.

Subscribe to see which companies asked this question

# Sharing my 2ms C++ solution with comments and explanations.

3,937 views

Update: there's a follow-up 0ms solution which is even more optimized

This is one of the fastest Sudoku solvers I've ever written. It is compact enough - just 150 lines of C++ code with comments. I thought it'd be interesting to share it, since it combines several techniques like reactive network update propagation and backtracking with very aggressive pruning.

The algorithm is online - it starts with an empty board and as you add numbers to it, it starts solving the Sudoku.

Unlike in other solutions where you have bitmasks of allowed/disallowed values per row/column/square, this solution track bitmask for every(!) cell, forming a set of constraints for the allowed values for each particular cell. Once a value is written into a cell, new constraints are immediately propagated to row, column and 3x3 square of the cell. If during this process a value of other cell can be unambiguously deduced - then the value is set, new constraints are propagated, so on.... You can think about this as an implicit reactive network of cells.

If we're lucky (and we'll be lucky for 19 of 20 of Sudokus published in magazines) then Sudoku is solved at the end (or even before!) processing of the input.

Otherwise, there will be empty cells which have to be resolved. Algorithm uses backtracking for this purpose. To optimize it, algorithm starts with the cell with the smallest ambiguity. This could be improved even further by using priority queue (but it's not implemented here). Backtracking is more or less standard, however, at each step we guess the number, the reactive update propagation comes back into play and it either quickly proves that the guess is unfeasible or significantly prunes the remaining search space.

It's interesting to note, that in this case taking and restoring snapshots of the compact representation of the state is faster than doing backtracking rollback by "undoing the moves".

``````class Solution {
struct cell // encapsulates a single cell on a Sudoku board
{
uint8_t value; // cell value 1..9 or 0 if unset
// number of possible (unconstrained) values for the cell
uint8_t numPossibilities;
// if bitset[v] is 1 then value can't be v
bitset<10> constraints;
cell() : value(0), numPossibilities(9),constraints() {};
};
array<array<cell,9>,9> cells;

// sets the value of the cell to [v]
// the function also propagates constraints to other cells and deduce new values where possible
bool set(int i, int j, int v)
{
// updating state of the cell
cell& c = cells[i][j];
if (c.value == v)
return true;
if (c.constraints[v])
return false;
c.constraints = bitset<10>(0x3FE); // all 1s
c.constraints.reset(v);
c.numPossibilities = 1;
c.value = v;

// propagating constraints
for (int k = 0; k<9; k++) {
// to the row:
if (i != k && !updateConstraints(k, j, v))
return false;
// to the column:
if (j != k && !updateConstraints(i, k, v))
return false;
// to the 3x3 square:
int ix = (i / 3) * 3 + k / 3;
int jx = (j / 3) * 3 + k % 3;
if (ix != i && jx != j && !updateConstraints(ix, jx, v))
return false;
}
return true;
}
// update constraints of the cell i,j by excluding possibility of 'excludedValue'
// once there's one possibility left the function recurses back into set()
bool updateConstraints(int i, int j, int excludedValue)
{
cell& c = cells[i][j];
if (c.constraints[excludedValue]) {
return true;
}
if (c.value == excludedValue) {
return false;
}
c.constraints.set(excludedValue);
if (--c.numPossibilities > 1)
return true;
for (int v = 1; v <= 9; v++) {
if (!c.constraints[v]) {
return set(i, j, v);
}
}
assert(false);
}

// backtracking state - list of empty cells
vector<pair<int, int>> bt;

// find values for empty cells
bool findValuesForEmptyCells()
{
// collecting all empty cells
bt.clear();
for (int i = 0; i < 9; i++) {
for (int j = 0; j < 9; j++) {
if (!cells[i][j].value)
bt.push_back(make_pair(i, j));
}
}
// making backtracking efficient by pre-sorting empty cells by numPossibilities
sort(bt.begin(), bt.end(), [this](const pair<int, int>&a, const pair<int, int>&b) {
return cells[a.first][a.second].numPossibilities < cells[b.first][b.second].numPossibilities; });
return backtrack(0);
}

// Finds value for all empty cells with index >=k
bool backtrack(int k)
{
if (k >= bt.size())
return true;
int i = bt[k].first;
int j = bt[k].second;
// fast path - only 1 possibility
if (cells[i][j].value)
return backtrack(k + 1);
auto constraints = cells[i][j].constraints;
// slow path >1 possibility.
// making snapshot of the state
array<array<cell,9>,9> snapshot(cells);
for (int v = 1; v <= 9; v++) {
if (!constraints[v]) {
if (set(i, j, v)) {
if (backtrack(k + 1))
return true;
}
// restoring from snapshot,
// note: computationally this is cheaper
// than alternative implementation with undoing the changes
cells = snapshot;
}
}
return false;
}
public:
void solveSudoku(vector<vector<char>> &board) {
cells = array<array<cell,9>,9>(); // clear array
// Decoding input board into the internal cell matrix.
// As we do it - constraints are propagated and even additional values are set as we go
// (in the case if it is possible to unambiguously deduce them).
for (int i = 0; i < 9; i++)
{
for (int j = 0; j < 9; j++) {
if (board[i][j] != '.' && !set(i, j, board[i][j] - '0'))
return; // sudoku is either incorrect or unsolvable
}
}
// if we're lucky we've already got a solution,
// however, if we have empty cells we need to use backtracking to fill them
if (!findValuesForEmptyCells())
return; // sudoku is unsolvable

// copying the solution back to the board
for (int i = 0; i < 9; i++)
{
for (int j = 0; j < 9; j++) {
if (cells[i][j].value)
board[i][j] = cells[i][j].value + '0';
}
}
}
};``````

# Simple and Clean Solution / C++

871 views
``````bool check(vector<vector<char>> &board, int i, int j, char val)
{
int row = i - i%3, column = j - j%3;
for(int x=0; x<9; x++) if(board[x][j] == val) return false;
for(int y=0; y<9; y++) if(board[i][y] == val) return false;
for(int x=0; x<3; x++)
for(int y=0; y<3; y++)
if(board[row+x][column+y] == val) return false;
return true;
}
bool solveSudoku(vector<vector<char>> &board, int i, int j)
{
if(i==9) return true;
if(j==9) return solveSudoku(board, i+1, 0);
if(board[i][j] != '.') return solveSudoku(board, i, j+1);

for(char c='1'; c<='9'; c++)
{
if(check(board, i, j, c))
{
board[i][j] = c;
if(solveSudoku(board, i, j+1)) return true;
board[i][j] = '.';
}
}

return false;
}
``````

public: void solveSudoku(vector<vector>& board) { solveSudoku(board, 0, 0); }

# c++ clear solution using dfs, beating 90% c++ coder.

690 views
``````class Solution {
public:
bool col,row,f;
bool flag = false;
void solveSudoku(vector<vector<char>>& board) {
memset(col,false,sizeof(col));
memset(row,false,sizeof(row));
memset(f,false,sizeof(f));
for(int i = 0; i < 9;i++){
for(int j = 0; j < 9;j++){
if(board[i][j] == '.')   continue;
int temp = 3*(i/3)+j/3;
int num = board[i][j]-'0';
col[j][num] = row[i][num] = f[temp][num] = true;
}
}
dfs(board,0,0);
}
void dfs(vector<vector<char>>& board,int i,int j){
if(flag == true)  return ;
if(i >= 9){
flag = true;
return ;
}
if(board[i][j] != '.'){
if(j < 8)  dfs(board,i,j+1);
else dfs(board,i+1,0);
if(flag)  return;
}

else{
int temp = 3*(i/3)+j/3;
for(int n = 1; n <= 9; n++){
if(!col[j][n] && !row[i][n] && !f[temp][n]){
board[i][j] = n + '0';
col[j][n] = row[i][n] = f[temp][n] = true;
if(j < 8)  dfs(board,i,j+1);
else dfs(board,i+1,0);
col[j][n] = row[i][n] = f[temp][n] = false;
if(flag)  return;
}
}
board[i][j] = '.';
}
}
};``````

# 13-line Python solution, dfs, beats 47.79%

+1vote
77 views
``````def solveSudoku(self, board):
def dfs():
for i, row in enumerate(board):
for j, char in enumerate(row):
if char == '.':
for x in s9 - {row[k] for k in r9} - {board[k][j] for k in r9} - \
{board[i / 3 * 3 + m][j / 3 * 3 + n] for m in r3 for n in r3}:
board[i][j] = x
if dfs(): return True
board[i][j] = '.'
return False
return True

r3, r9, s9 = range(3), range(9), {'1', '2', '3', '4', '5', '6', '7', '8', '9'}
dfs()``````

http://www.cnblogs.com/felixfang/p/3705754.html

16090人浏览
2020-07-13

<font color="red">[置顶]</font>
2869人浏览
2015-09-17 12:28:00
leetcode难度及面试频率
2285人浏览
2013-12-04 22:49:00
[LeetCode]36.Valid Sudoku
1998人浏览
2014-01-20 13:42:00
[LeetCode]--36. Valid Sudoku
649人浏览
2016-10-02 18:55:00
【LeetCode从零单排】No36 Valid Sudoku
751人浏览
2015-02-13 16:50:00
[LeetCode] Sudoku Solver
429人浏览
2015-09-08 17:35:00
[LeetCode] Sudoku Solver 求解数独
886人浏览
2017-12-02 14:20:00
[LeetCode] Valid Sudoku 验证数独
994人浏览
2017-12-02 14:23:00
LeetCode - 37. Sudoku Solver
684人浏览
2016-03-03 12:06:00
LeetCode - 36. Valid Sudoku
665人浏览
2016-03-02 18:52:00
POJ 2676 Sudoku (数独 DFS)
750人浏览
2017-09-24 17:50:00
leetCode 36. Valid Sudoku(数独) 哈希
763人浏览
2017-11-13 22:27:00
JavaScript之破解数独（附详细代码）
1102人浏览
2018-01-12 17:17:41
Django之破解数独
825人浏览
2018-01-10 17:07:35
LeetCode 36 Valid Sudoku（有效数独）
639人浏览
2015-11-30 20:58:35
0
0
964