Would there be any demand for this?
Yes, having dynamic filtering and continuation for result formats is pretty high on the wish list for result formats in general. Which is why people are so excited about the filtered format. I however have two concerns about this format:
* It seems to need specific code for each result format it supports, which leads to less versatility and a lot more maintenance effort then some generic system
* It loads "all data" and then does filter operations on it, which does not work for big datasets and for others is not always desired
(Don't get me wrong I think filtered is pretty awesome)
Your approach seems to tackle the second point. Since both approaches do not tackle the first one, it'd be best to at least share code in some way, else the already to big maintenance costs doubles. As to how to do this I cannot really advise, as I'm not familiar with the filtered code or the stuff you wrote.
Jeroen De Dauw
Don't panic. Don't be evil.