I think you would need to use only datomic's transactor and tx log and then build your own incrementally maintained query engine, which is basically what this p

I think you would need to use only datomic's transactor and tx log and then build your own incrementally maintained query engine, which is basically what this project declarative dataflows does. Differential dataflow and regular dataflow are both ways to implement this. Differential dataflow has the restriction that your query engine must be implemented in terms of diff-aware primitives (https://news.ycombinator.com/item?id=22095199 https://news.ycombinator.com/item?id=22095199). Regular dataflow is like Excel; it works with any functions, but is not as good at incrementally maintaining things like large lists and nesting – if you're familiar with React.js, at a high level, regular dataflow works a lot like React.js rendering which can call any function during rendering. (Note React.js is highly specialized to html view maintenance, not general purpose incremental maintenance of any computation). Someone correct me if I've said something wrong.