On Wed, Jan 18, 2012 at 10:08 PM, john skaller <[email protected]> wrote: > > On 19/01/2012, at 1:25 PM, Martin Sustrik wrote: > >>> about? >> >> The thing is that exact matching allows for more efficient filtering >> algorithm (hash table). > Judy array is even faster.
If you're going to make high performance matching engines, do study the topic matching engine that I built for OpenAMQ: http://www.openamq.org/doc:fast-topic-matching This model can be extended for regexps quite trivially and gives extraordinarily good performance. The key is to pre-match regexp subscriptions against all known topic keys, given that topic keys are not random but highly repetitive (stock symbols is the usual case). -Pieter _______________________________________________ zeromq-dev mailing list [email protected] http://lists.zeromq.org/mailman/listinfo/zeromq-dev
