How does Sphinx store null fields, and how does this affect performance?

Question

I'm thinking about using asking sphinx to index many fields (in the hundreds), many of which will be null. My question is how much having many null fields will affect performance?

This situation arises not from having incredibly denormalized data, but from requirements on the search interface and what can be searched. Basically I will be building the index config dynamically in the indexed model, and may end up with quite a few null fields from doing so.

My guess is the performance/success of this depends on what Sphinx does with null values... if it simple ignores them, then I should be fine, but if it actually stores that the field is null in its index, I could have a problem.

Answer 1

Apparently the current version of Sphinx limits to 32 text fields per index (unlimited numeric), so I would have to divide the data into several different indexes to achieve this unless all data could be represented numerically (which mine cannot). Back to the drawing board!