Weikeng Chen

I think it would probably take some time to merge (auto-merge is not enough!).

Ok. The conflicts are now resolved. Note that additional changes may be pushed due to Marlin.

Marlin passes. I think I will let @Pratyush and @npwardberkeley do a final pass and then we would merge this, probably today or tomorrow.

(We would pause here and leave the changes in the `constraints` branch, and we will come back after the paper deadline.)

I would come back to this soon. This PR has been a little bit too big and is not easy to review. I plan to break it down. My next...

Yes. At this point CycleEngine is almost useless---it does not simplify the `where` clause of the type. ``` pub struct VerifierKeyVar< CycleE: CycleEngine, PG: PairingVar, > where ::G1Projective: MulAssign, ::G2Projective:...

Let me have a try. If so, then CycleEngine would be a useful abstraction.

I agree, it shouldn't limit itself to cycles of curves. Then, let me start simplifying the relations.

I will be back tomorrow to continue working on this. The rest is to apply previous code review comments. We might need to split the `constraints.rs` into a file for...

We should add some assertions on that or try to support very simple statements. But basically, the problem is here: the matrices of the statement are way too simple, such...