3 ms·

Neat writeup, if missing the mathematical portion that speeds it up. One callout if folks think about implementing something similar. Euclidean distance on a s

• wikfwikf 4y ago
  It's not even a correctly normalized Euclidean distance.
  • piotrjaw 4y ago
    You're correct, this is just a synthetic value used to rank the distances.
  • tomrod 4y ago
    It would be marginally slower to take the square root when all you need is rank
• piotrjaw 4y ago
  That is correct, but - as pointed in the article - we work on floor plans, which are flat, so I wanted to simplify this a bit to not create too much issues to solve in one article. That said, it was probably an oversimplification, since actual geographic data (and a map view) was used, point taken.
  • tomrod 4y ago
    No worries whatsoever. It wasn't a critique of the general idea, simply to keep distance metrics in mind relative to scale. Working on the a floor plan or proof of concept? Euclidean metric absolutely works. Working on Mercator projections or legal type stuff? eh....