3 ms·
The Inside Story of How One Company Didn’t Mine Facebook
- patio11 16y agoChecking 5,000 possible candidates against a list of 7,000 known stores couldn’t happen instantly on login. That could be 35 million comparisons per user, which is a big number, no matter how well you optimize your algorithm. Some days I weep for our industry. You can do this, trivially, in one lookup per favorite-d store. Create a hash set of the stores you are interested in, grab the favorites, check if they are in the hash set. The worse case performance will be 5,000 times the cost of one hash lookup. (You can do it in the application server's memory or in memcached/redis/etc.) The typical case performance will be virtually instantaneous since most users will have virtually no favorites. In practice, the bottleneck will almost certainly be how fast you can get results from the Facebook API.