Jon Bratseth

I don't understand what you mean by "sparse vectors generated during Vespa indexing" - could you explain some more?

Right, so you'd like access to a sparse vector of term -> frequency for a field in a document. Yes, that's doable, although not something that's directly available - Vespa...

Yes. We could e.g signal this by passing a default value with the operation. The alternative is to use 1 with multiplicative and 0 with additive ops.

Whether it is "multi" is a property of the function so it seems wrong to make the caller choose? Since a FunctionEvaluator is a single-use object with a life-cycle how...

Yes, so you have a goal of doing the same also for regular functions?

But why support multiple regular functions within one evaluator then?

Got it. But should we capture this in the model instead of creating an evaluator having (and always evaluating) every function of the model? We could add a List of...

By model I mean the class owning these functions, ai.vespa.models.evaluation.Model. At line 185 in that you pass the entire list of functions to the FunctionEvaluator constructor if no name is...

Notwithstanding different opinions on that, it's a feature we have and will continue to have. (The performance consideration cuts both ways, some times you solve computational problem once and that...

> I would say that in case 3, the current behavior is wrong and we should just fail. Agreed. > To handle all cases, we would need a registry of...