Performance Improvements Using Judy Arrays

Post by jamesaguilar » Thu, May 02, 2013, 6:47 AM UTC

AaronFriel wrote: post #28

jemfinch wrote: post #14

The article's credibility is significantly reduced by referring to Judy Trie lookups as O(log n) operations. They're actually O(log w) operations: the number of operations is bounded by the array's word size , not the size of the Judy array. A million elements in a Judy32 array will cause at most 7 node hops, not the 20 you'd expect from a binary tree.

7 node hops? It should be at most 4. Judy32 has a maximum branching factor of 256, and log base 256 of 2^32 is 4.

Post by lobster_johnson » Thu, May 02, 2013, 2:58 PM UTC

lobster_johnson wrote: post #47

While this is very interesting, the real problem here is not Linguist stuff, but the fact that it's crammed into the web app. A web app involved in a site as complex as Github's should really contain only the parts required to service web pages. A service like language classification is clearly better designed as a standalone service. Aside from working around GC overhead, compositing big app from many smaller apps h…