I’m starting out to design a site that has some requirements that I’ve never really dealt with. Specifically, the data objects will have similar, but not exact, attributes. Yes, I could probably figure out most of the possible attributes and then just not populate the ones that don’t make sense, and therefore keep a traditional “Relational” table and column design, but I’m thinking this might be a really good time to learn NoSQL.
In addition, the user will have 1, and only 1, textbox to search, and I will need to search all data objects and their attributes to find that string.
Ideally, I’d like to have the search return in order of “importance”, meaning that if a match for the user’s entered string is found in a “name” attribute, it would be returned as a higher confidence match than if the string was matched on a sub-attribute.
Anyone have any experience in this sort of situation? What have you tried that worked or didn’t work? Am I wrong in thinking that this project is very well suited to a NoSQL type of database?
Stick with a traditional relational database such as MySQL or Postgresql. I would suggest sorting by relevance in your application code after obtaining the matching results. The size of your result set should impact your design choices, but if you will have less than 1-2k results then just keep it simple and don’t worry too much about optimization.
NoSQL is just a dumb key value store, a persistent dictionary that can be shared across multiple application instances. It can solve scalability issues, but introduces new ones since you now just have a dumb data store. Relational databases have had years of performance tuning and do a great job.
I find NoSQL to be much more suited to storing state data, like a users preferences or cache. If you are analyzing the relationship between data then you need a relational database.