around creative solution for Moore Neighbourhood by veky - python coding challenges

Enable Javascript in your browser and then refresh this page, for a much enhanced experience.

around solution in Creative category for Moore Neighbourhood by veky

o=lambda s,i:s[i-bool(i):i+2]
count_neighbours=lambda G,i,j:sum(sum(o(r,j))for r in o(G,i))-G[i][j]

Oct. 4, 2014

Comments:

blabaster on Oct. 10, 2014, 9:27 a.m.
... functions and method calls are expensive (more so than in C or Java) ... /by Guido/

veky on Oct. 11, 2014, 4:28 a.m.
Yeah, so? This is not "speedy compared to C or Java" category. :-) Did you timeit?

blabaster on Oct. 11, 2014, 9:08 a.m.
Shure. This orthodox code bits your with time ratio 3:4 (CPython 3.4.1) <pre class='brush: python'>def count_neighbours(grid, row, col): lo = min(col - 1, 0) return sum(sum(l[lo:col + 2]) for l in grid[min(row - 1, 0):row + 2]) - grid[row][col] </pre> Do not confuse me ;) Curiously, just this solution on short matrix (like in task examples) 20% faster. <pre class='brush: python'>def count_neighbours(grid, row, col): return sum(sum(((0,) + l)[col:col + 3]) for l in (((0,),) + grid)[row:row + 3]) - grid[row][col] </pre>

veky on Oct. 11, 2014, 9:15 a.m.
Strange orthodoxy... min(col-1,0) extracted, while min(row-1,0) not. :-) (Yes, I know why you did so, but still, not really orthodox.:) But ok, you're right. I just meant it was faster than my other solution:-), but if there are simple significantly faster solutions, this doesn't belong here.

blabaster on Oct. 11, 2014, 9:34 a.m.
Something else: I thought before, that penalty for func call would be less if compiled. But in PyPy ratio is even better - 3:5

veky on Oct. 11, 2014, 7:48 p.m.
Guessing performance in CPython is heavy wizardry, but guessing performance in PyPy is just pure luck. Unless you wrote that part of PyPy yourself, and even then you might be surprised. :-D

blabaster on Oct. 11, 2014, 8:48 a.m.
Shure. This orthodox code bits your with time ratio 2:3 (CPython 3.4.1) <pre class='brush: python'>def count_neighbours(grid, row, col): lo = min(col - 1, 0) return sum(sum(l[lo:col + 2]) for l in grid[min(row - 1, 0):row + 2]) </pre> Do not confuse me ;) Curiously, just this solution on short matrix (like in task examples) 20% faster. <pre class='brush: python'>def count_neighbours(grid, row, col): return sum(sum(((0,) + l)[col:col + 3]) for l in (((0,),) + grid)[row:row + 3]) - grid[row][col] </pre>

MashkoorAhmed on Feb. 9, 2015, 9:32 a.m.
Creative Solution

veky on Feb. 9, 2015, 10 a.m.
...in a Creative Category. Worth -1. For you. Why?

GitHub

0 %

PyCharm The Python IDE for Professional Developers. Enjoy productive Python, web and scientific development with PyCharm. Download it now!