Prospective research ideas

docs/content/todo/todo-research.md

@@ -0,0 +1,35 @@
+# Areas of possible research
+
+### Overall performance of Reactor constraints processing library
+
+There are a few points that are critical for performance of Reactor library. 
+
+1. Indexing of productions by constraint symbol and arguments;
+2. Dealing with factorial expansion when searching for a match;
+3. Matching of constraint occurrence with production’s constraint;
+4. Unification.
+
+### Incremental processing of code rules
+
+How can we have incremental type checking? 
+
+1. Can a program be analysed and code rules applied in such a way, so that the constraints program is updated incrementally, and
+2. Can we launch constraints program (provided there are some constraints in the store), so that only the necessary productions are triggered as required by the changes.
+
+### Properties of user-defined type system
+
+What properties can be asserted of a constraints production system?
+
+1. How to check if the production system is decidable, what productions are questionable and/or conflicting. 
+2. How to check if constraints program is terminating, what assertions can be made about a constraints program execution time. 
+
+### Missing features that may affect constraints processing 
+
+A few features are not implemented, need to understand their impact on the semantics of constraints processing. 
+
+1. Logical variables as indirect arguments to a constraint should reactivate the constraint on update (value/parent).
+2. The state of logical variables is not restored on frame dropping, which is what happens on switching to alternative body.
+
+Implementing these features must not affect the performance. 
+
+