Skip to content

Tosha1409/Closest_pair_of_points

Repository files navigation

Closest_pair_of_points

I was doing Codility Rubidium challenge while and at some point it required to check algorithm for finding pair of points with smallest distance between them. I didnt used it in the end for Rubidium cause task was a bit different, but it was good to refresh memory(i did saw it before) and i decided to write my version from scratch later.
So i started to make it yesterday and when it was almost done i decided to recheck links that i saw before:
https://medium.com/@andriylazorenko/closest-pair-of-points-in-python-79e2409fc0b2
https://www.geeksforgeeks.org/closest-pair-of-points-using-divide-and-conquer-algorithm/
2nd link was in C++, and first in Python and i decided to take a closer look at python solution and realized that it is pretty much rewritten program from previous. Python is more flexible language and allow to makes more elegancy, so i got excited to see differences.

shortest_distance.py - my implementation of algoritm.
shortest_distance_without_points.py - almost same program, but without keeping points itself(and works little bit faster).
testcases.py - 5 different arrays for tests.

so i runned tests (Python -m cProfile):
mine.txt - those are results for mine.
original.txt - those are results for Andriy Lazorenko code.
so it shows basicly double difference in speed.

Other thing that bothered me was this "maximum 6 points at strip":
C++ article was saying

This is a proven fact that this loop runs at most 6 times.

and at other link it was:

In short: it is enough to check only seven points following each point on the s_y subarray. You should really look through the proof of correctness, because it explains a lot better this ‘trick’ that allows for great running speed increase.

so that left me pretty puzzled, is it running 6 times(because there is no more points) or you need check only seven points(but there can be more). So decided to experiment and replaced line number 55

  for y in range (x+1, min ((x+7),len(strip))):  

with

  for y in range (x+1, len(strip)):  

and results improved even a bit more.

shortest_distance2.py - mine code without limiting loop for checking strip.
mine2.txt - test results

so it looks like this check/choosing minimum was unnecessary and perfomance a bit improved a bit more.

About

Python implementation of algorithm for seeking "Closest pair of points" (divide and conquer).

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages