Very similar to techniques used in Knowledge Based Engineering systems, where the term "dependency tracking" is used. Together with caching of nodes/tensors this reduces calculations, especially useful for large parametric 3D models. When getting a value it will recursively call the binary/dependency tree to find out which variables have changed and only recalculate them when needed. Custom python objects and properties with __set__ and __get__ methods makes this a built-in feature of an object-oriented model.

x = Tensor(3)

y = Tensor(5)

z = x + y

print(x, y) # 3, 5

print(z) # 8

x.value = 4 # when setting value nothing is recalculated

print(z) # 9 since getting value triggers recalculation of dependencies that have changed