Write a function (using as few bytes as possible) that takes a bi-dimensional array of any number of columns and rows in which:

  • 0 represents empty block,
  • 1 represents snake block.

The function must return the number of possible paths the snake traveled.

Example 1:



Output: 2

In the example above, the function will return 2 because the answer is either one of:

enter image description here

Example 2:



Output: 6

In this example the function will return 6 because the answer is either one of:

enter image description here


When assessing the input, you can assume that:

  • The arrays representing columns will always have the same sizes (so the arrays are rectangular);
  • There exists at least 1 valid path;
  • The snake cannot walk through the edges (as can happen in some versions of snake);
  • The snake will always have at least 2 blocks;
  • The snake cannot move diagonally;
  • The paths are directed. (so, two paths ending at different positions but otherwise looking exactly the same are not the same path, it’ll add up to the total)
  • 13
    \$\begingroup\$ Welcome to PPCG! Nice first challenge. \$\endgroup\$
    – Laikoni
    Commented May 18, 2018 at 8:14
  • 5
    \$\begingroup\$ Minor note: "There will always be at least one row and one column" is redundant, given that the snake will always have at least 2 blocks. \$\endgroup\$ Commented May 18, 2018 at 9:17
  • 2
    \$\begingroup\$ Suggested test cases: the one given by @StewieGriffin and [[0,0,1,1],[0,0,1,1],[0,0,1,1]]. Most answers give 16, but one gives 15. \$\endgroup\$ Commented May 18, 2018 at 11:01
  • 2
    \$\begingroup\$ It seems like everybody so far (including me) has made the assumption that 2 paths ending at different positions but otherwise looking exactly the same are not the same path. I think this needs to be explicitly specified. \$\endgroup\$
    – Arnauld
    Commented May 18, 2018 at 13:16
  • 2
    \$\begingroup\$ @Arnauld - that’s right. Two paths ending at different positions but otherwise looking exactly the same are not the same path, it’ll add up to the total. In your example the total should be 16 if I’m not mistaken - I can’t accuratelly calculate right now but you get the point \$\endgroup\$
    – Adelin
    Commented May 18, 2018 at 13:42

7 Answers 7


Wolfram Language (Mathematica), 16+83=99 bytes

Library import statement (16 bytes):


Actual function body (83 bytes):


Try it online!

Note that the question just ask for the number of Hamiltonian path in the graph.

However, (for some reason) the HamiltonianPath function doesn't really work with directed graph (example). So, I used the workaround described in this Mathematica.SE question:

  • Add an vertex (called True) that is connected to all other vertices.
  • Count the number of Hamiltonian cycle on the resulting graph.

The graph is constructed using MakeGraph (annoyingly there are no directly equivalent built-in), using the boolean function ##||Norm[#-#2]==1&, which returns True if and only if one of the arguments is True or the distance between the two vertices are 1.

Tr[1^x] cannot be used instead of Length@x, and <2 cannot be used instead of ==1.

HamiltonianPath can be used if the graph is undirected, with the function body takes 84 bytes (exactly 1 byte more than the current submission):


Try it online!


JavaScript (ES6), 154 134 bytes


Try it online!



Starting from each possible cell, we flood-fill the matrix, clearing all cells on our way. Whenever the matrix contains no more 1's, we increment the number n of possible paths.

Each valid path is counted 4 times because of the direction chosen on the last cell, which actually doesn't matter. Therefore, the final result is n / 4.

Recursive function

Instead of calling the recursive function g() from the callback of the second map() like this...


...we define the recursive function g() directly as the callback of map():


Despite the rather long formula y=1/y?y:Y which is needed to set the initial value of y, this saves 2 bytes overall.

Commented code

m =>                           // given the input matrix m[][]
  m.map((r, Y) =>              // for each row r[] at position Y in m[][]:
    r.map(g = (                //   for each entry in r[], use g() taking:
      _,                       //     - the value of the cell (ignored)
      x,                       //     - the x coord. of this cell
      y,                       //     - either the y coord. or an array (1st iteration),
                               //       in which case we'll set y to Y instead
      r = m[y = 1 / y ? y : Y] //     - r = the row we're currently located in
    ) =>                       //       (and update y if necessary)
      r && r[x] &&             //     do nothing if this cell doesn't exist or is 0
      [-1, 0, 1, 2].map(d =>   //     otherwise, for each direction d,
        r[                     //     with -1 = West, 0 = North, 1 = East, 2 = South:
          r[x] = 0,            //       clear the current cell
          /1/.test(m) ?        //       if the matrix still contains at least one '1':
            g(                 //         do a recursive call to g() with:
              _,               //           a dummy first parameter (ignored)
              x + d % 2,       //           the new value of x
              y + ~-d % 2      //           the new value of y
            )                  //         end of recursive call
          :                    //       else (we've found a valid path):
            ++n,               //         increment n
          x                    //       \_ either way,
        ] = 1                  //       /  do r[x] = 1 to restore the current cell to 1
      )                        //     end of map() over directions
    ),                         //   end of map() over the cells of the current row
    n = 0                      //   start with n = 0
  ) | n / 4                    // end of map() over the rows; return n / 4

Jelly, 12 11 bytes


Try it online!


ŒṪ               Positions of snake blocks.
  Œ!             All permutations.
                 For each permutation:
    ạƝ€             Calculate the absolute difference for each neighbor pair
       §            Vectorized sum.
                 Now we have a list of Manhattan distance between snake
                    blocks. Each one is at least 1.
        ÐṂL      Count the number of minimum values.
                    Because it's guaranteed that there exists a valid snake,
                    the minimum value is [1,1,1,...,1].
  • \$\begingroup\$ New features prove to be extremely useful. \$\endgroup\$
    Commented May 18, 2018 at 10:45
  • \$\begingroup\$ How about §ỊML instead of §ỊP€S to save a byte - I think it should work? \$\endgroup\$ Commented May 18, 2018 at 16:14
  • \$\begingroup\$ ...or §ÐṂL which is a little faster. \$\endgroup\$ Commented May 18, 2018 at 21:16
  • \$\begingroup\$ @JonathanAllan Only works if the result is nonzero. \$\endgroup\$
    Commented May 19, 2018 at 3:21
  • \$\begingroup\$ @JonathanAllan So it ends up actually works. \$\endgroup\$
    Commented May 19, 2018 at 4:06

Python 2, 257 246 241 234 233 227 214 210 bytes

lambda b:sum(g(b,i,j)for j,l in e(b)for i,_ in e(l))
def g(b,x,y):d=len(b[0])>x>-1<y<len(b);c=eval(`b`);c[d*y][d*x]=0;return d and b[y][x]and('1'not in`c`or sum(g(c,x+a,y)+g(c,x,y+a)for a in(1,-1)))

Try it online!


  • -8 bytes, thanks to Kevin Cruijssen
  • -14 bytes, thanks to user202729

Python 2, 158 bytes

g=lambda P,x,y:sum(g(P-{o},*o)for o in P if x<0 or abs(x-o[0])+abs(y-o[1])<2)+0**len(P)
lambda L:g({(x,y)for y,r in E(L)for x,e in E(r)if e},-1,0)

Try it online!


Haskell, 187 155 bytes

f x|l<-[(i,j)|i<-[0..l x-1],j<-[0..l(x!!0)-1],x!!i!!j>0]=l#l

Try it online!


APL (Dyalog Extended), 28 bytes


Try it online!

A full program that accepts a matrix.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.