My experience was for small data, thinking sphinx (or ultra) was a simple search solution for use by a developer who just wanted to run something. As soon as the data is larger than they say, one hundred thousand lines in db, both plugins show how strongly they are connected with the active record. Then you need to balance the delicate dance between when the sphinx index is executed (in cron, in deployment, etc.). How often indexing is performed, delta indexing is used and how long does it take to index. In addition, as soon as your data becomes more complicated about how and what results should be presented in the search, you begin to feel that the sphinx’s customization is limiting.
We are in the process of moving the search from the sphinx plugin to something independent of the development and deployment life cycle of our active record models. We have not decided whether we will continue to use the sphinx or switch to solr. We just want the search and indexing to be separated from the model and pull in our own service, which our application with rails speaks of.
monde source share