Dealing with Json output

Consider the general form again: Gout@G1,G2
Both G1 and G2 are directed graphs without loops, and the @ operator is a single atomic operation, with grammar rules that guarantee a complete exploration of G1 an G2. That means any emissions from the convolution come out in nested order, and can be streamed as multi-part Json documents.  All the code is there to do this task, the emission, even from square tables, look like sequence of  |link|rowcount|key value.  There is a few days work, but then we get intelligently structure Json results from a query by example between two Jsons. Otherwise the compete system if stable, at moderate transaction rates. I have a lot of defined limits, still. This thing is going to be a personal Watson.