Great write up! I cringe whenever I see that blog post; other spatial indexing systems at Uber were and are designed more thoughtfully, and with better performance characteristics.
The article didn't really go into detail about how this works within each city.
I am surprised they're scanning each city linearly, even a crude index (such as boxing), or a binary search would greatly improve the lookup time. I guess if the data fits into the cpu cache the order these are scanned doesn't matter so much?
I am surprised they're scanning each city linearly, even a crude index (such as boxing), or a binary search would greatly improve the lookup time. I guess if the data fits into the cpu cache the order these are scanned doesn't matter so much?